research papers on healthcare

• Search by keyword
• Search by citation

Page 1 of 510

Work ability and return-to-work of patients with post-COVID-19: a systematic review and meta-analysis

In addition to several sequelae of post-COVID-19, individuals also experience significant limitations in work ability, resulting in negative consequences for the return-to-work (RTW) process. This systematic r...

• View Full Text

Knowledge, attitudes, and practice related to tooth loss and dentures among patients with dental arch deficiencies

Tooth loss is a common problem that affects many people worldwide. Exploring knowledge, attitude, and practice (KAP) among patients can identify barriers and challenges in following recommended practices, prov...

The impact of financial stress on student wellbeing in Lebanese higher education

The financial crisis has indirectly affected Lebanese university students, leading to economic distress. Accordingly, this study aimed to assess the substantial negative impact of financial stress on the menta...

Cost-effectiveness of single-pill and separate-pill administration of antihypertensive triple combination therapy: a population-based microsimulation study

Single-pill combination (SPC) of three antihypertensive drugs has been shown to improve adherence to therapy compared with free combinations, but little is known about its long-term costs and health consequenc...

HIV and gender identity expression among transfeminine people in the Western Cape, South Africa – a thematic analysis of data from the HPTN 071 (PopART) trial

Transfeminine people in South Africa have a high HIV risk due to structural, behavioural, and psychosocial factors. Transfeminine people and feminine identifying men who have sex with men (MSM) are often confl...

Faith and vaccination: a scoping review of the relationships between religious beliefs and vaccine hesitancy

Throughout history, vaccines have proven effective in addressing and preventing widespread outbreaks, leading to a decrease in the spread and fatality rates of infectious diseases. In a time where vaccine hesi...

Prevalence of anxiety symptoms in infertile men: a systematic review and meta-analysis

Infertility in men causes problems in various aspects of their lives, including personal, family and social life. One of the most important of these problems is anxiety. Anxiety in infertile men can affect the...

Understanding health literacy in men: a cross-sectional survey

Males have a shorter life expectancy than females. Men are less likely to seek the advice of a health professional or utilise preventive health services and programs. This study seeks to explore health literac...

The edutainment program on knowledge, perception, and uptake of cervical cancer screening among Muslim women in Southern Thailand: a quasi experimental study

Cervical cancer is a significant global health concern and is the third most common cancer in women. Owing to their religious beliefs, Muslim women in Thailand are less likely to be screened for cervical cancer.

Efficacy of relational agents for loneliness across age groups: a systematic review and meta-analysis

Loneliness is a serious public health concern. Although previous interventions have had some success in mitigating loneliness, the field is in search of novel, more effective, and more scalable solutions. Here...

Beyond the diagnosis of drug-resistant Tuberculosis in Norway: patients’ experiences before, during and after treatment

This study aims to explore the varied experiences of patients with drug-resistant tuberculosis in Norway. The study emphasizes challenges and implications of being diagnosed with drug-resistant tuberculosis, i...

Age-standardized incidence, prevalence, and mortality rates of autoimmune diseases in adolescents and young adults (15–39 years): an analysis based on the global burden of disease study 2021

Autoimmune diseases (ADs) present significant health challenges globally, especially among adolescents and young adults (AYAs) due to their unique developmental stages. Comprehensive analyses of their burden a...

Analysis of anatomic location of burns inpatients in China from 2009 to 2018

Burns cause serious physical and psychological harm to patients, placing a heavy burden on the global healthcare system. Our previous study detailed the epidemiological characteristics of burn injuries in Chin...

The impact of hospital saturation on non-COVID-19 hospital mortality during the pandemic in France: a national population-based cohort study

A previous study reported significant excess mortality among non-COVID-19 patients due to disrupted surgical care caused by resource prioritization for COVID-19 cases in France. The primary objective was to in...

Interpersonal violence against people with intellectual disabilities in São Paulo, Brazil: characteristics of victims, perpetrators and referrals

Interpersonal violence is a phenomenon that can occur with different people and conditions. However, people with intellectual disabilities have increased vulnerability to this problem, with potential risks to ...

Socioeconomic determinants and reasons for non-acceptance to vaccination recommendations during the 3 rd - 5 th waves of the COVID-19 pandemic in Hungary

In Hungary, although six types of vaccines were widely available, the percentage of people receiving the primary series of COVID-19 vaccination remained below the EU average. This paper investigates the reason...

Assessing the impact of COVID-19 on routine immunization in Sierra Leone

The COVID-19 pandemic had a profound impact on healthcare systems and services, including routine immunization (RI). To date, there is limited information on the effects of the COVID-19 pandemic on RI in West ...

Correction: Childcare needs as a barrier to healthcare among women in a safety-net health system

The original article was published in BMC Public Health 2024 24 :1608

Assessing the value and knowledge gains from an online tick identification and tick-borne disease management course for the Southeastern United States

Tick-borne diseases are a growing public health threat in the United States. Despite the prevalence and rising burden of tick-borne diseases, there are major gaps in baseline knowledge and surveillance efforts...

Receiving home care forms and the risk for emergency department visits in community-dwelling Dutch older adults, a retrospective cohort study using national data

Older adults receiving home care have a higher risk of visiting the emergency department (ED) than community-dwelling older adults not receiving home care. This may result from a higher incidence of comorbidit...

Factors related with lung functions among Orang Asli in Tasik Chini, Malaysia: a cross-sectional study

Orang Asli lifestyle and household setting may influence their health status especially respiratory system and lung functions. This cross-sectional study was carried out to investigate the status of lung funct...

Benchmarking for healthy food stores: protocol for a randomised controlled trial with remote Aboriginal and Torres Strait Islander communities in Australia to enhance adoption of health-enabling store policy and practice

Aboriginal and Torres Strait Islander communities in remote Australia have initiated bold policies for health-enabling stores. Benchmarking, a data-driven and facilitated ‘audit and feedback’ with action plann...

Types of leisure-time physical activity participation in childhood and adolescence, and physical activity behaviours and health outcomes in adulthood: a systematic review

Youth leisure-time physical activity participation benefits physical activity habits and health outcomes later in life. However, it is unknown if certain types of leisure-time physical activity contribute to t...

Gender-based violence and harassment at work and health and occupational outcomes. A systematic review of prospective studies

Many people experience forms of gender-based violence and harassment (GBVH) in the context of their work. This includes a wide range of experiences, from subtle expressions of hostility to physical assault, th...

Joint association of sedentary time and physical activity with abnormal heart rate recovery in young and middle-aged adults

Abnormal heart rate recovery (HRR), representing cardiac autonomic dysfunction, is an important predictor of cardiovascular disease. Prolonged sedentary time (ST) is associated with a slower HRR. However, it i...

Impact of global smoking prevalence on mortality: a study across income groups

Smoking significantly contributes to the mortality rates worldwide, particularly in non-communicable and preventable diseases such as cardiovascular ailments, respiratory conditions, stroke, and lung cancer. T...

Post-COVID syndrome prevalence: a systematic review and meta-analysis

Since the Coronavirus disease 2019 (COVID-19) pandemic began, the number of individuals recovering from COVID-19 infection have increased. Post-COVID Syndrome, or PCS, which is defined as signs and symptoms th...

A four-year assessment of the characteristics of Rwandan FDA drug recalls

A drug recall is an act of removing products from the market and/or returning them to the manufacturer for disposal or correction when they violate safety laws. Action can be initiated by the manufacturing com...

WeChat usage and preservation of cognitive functions in middle-aged and older Chinese adults: indications from a nationally representative survey, 2018–2020

To investigate the associations between the most popular social media platform WeChat usage and cognitive performance among the middle-aged and older Chinese population using data from a nationally representat...

Applying the multiphase optimization strategy to evaluate the feasibility and effectiveness of an online road safety education intervention for children and parents: a pilot study

Reports of children’s engagement in active transportation outline low participation rates in many countries despite many associated mental, physical, and social health benefits. One of the main contributors to...

Determinants of intended prevention behaviour against mosquitoes and mosquito-borne viruses in the Netherlands and Spain using the MosquitoWise survey: cross-sectional study

Recently, Europe has seen an emergence of mosquito-borne viruses (MBVs). Understanding citizens’ perceptions of and behaviours towards mosquitoes and MBVs is crucial to reduce disease risk. We investigated and...

Risk assessment and prediction of nosocomial infections based on surveillance data using machine learning methods

Nosocomial infections with heavy disease burden are becoming a major threat to the health care system around the world. Through long-term, systematic, continuous data collection and analysis, Nosocomial infect...

Influence of lifestyle patterns on depression among adults with diabetes: a mediation effect of dietary inflammatory index

Lifestyle has become a crucial modulator in the management of diabetes and is intimately linked with the development and exacerbation of comorbid depression. The study aimed to analyze lifestyle patterns and t...

A qualitative study of maternal and paternal parenting knowledge and practices in rural Mozambique

Providing nurturing care for young children is essential for promoting early child development (ECD). However, there is limited knowledge about how mothers and fathers across diverse contexts in sub-Saharan Af...

Predicting dyslipidemia incidence: unleashing machine learning algorithms on Lifestyle Promotion Project data

Dyslipidemia, characterized by variations in plasma lipid profiles, poses a global health threat linked to millions of deaths annually.

Measurement properties of the Regular Physical Exercise Adherence Scale (REPEAS) in individuals with chronic pain

To examine the measurement properties of the Regular Physical Exercise Adherence Scale (REPEAS) in Brazilians with chronic pain.

The mediating effect of bullying on parental–peer support matching and NSSI behaviour among adolescents

Being subjected to bullying is a significant risk factor for non-suicidal self-injury (NSSI) among adolescents. Parental support, peer support, and social connectedness play protective roles in mitigating NSSI...

Association between childhood family structure and longitudinal health behaviour changes in adulthood –Northern Finland birth cohort 1966 study

Childhood family structure is considered to play a role in person’s health and welfare. This study investigated the relationships between the longitudinal changes of adult health behaviours and childhood famil...

Psychometric properties of the modified Drug Abuse Screening Test Sinhala version (DAST-SL): evaluation of reliability and validity in Sri Lanka

Psychoactive drug use is an important public health issue in Sri Lanka as it causes substantial health, social and economic burden to the country. Screening for substance use disorders in people who use drugs ...

The association between shift work, shift work sleep disorders and premature ejaculation in male workers

Shift work and Shift Work Sleep Disorder (SWSD) are known to affect the secretion of several neurotransmitters and hormones associated with premature ejaculation (PE). However, their specific influence on the ...

The impact of regional disparities on the availability of meningococcal vaccines in the US

In the United States (US), three types of vaccines are available to prevent invasive meningococcal disease (IMD), a severe and potentially fatal infection: quadrivalent conjugate vaccines against serogroups A,...

Barriers and facilitators to healthy eating in disadvantaged adults living in the UK: a scoping review

In the UK people living in disadvantaged communities are less likely than those with higher socio-economic status to have a healthy diet. To address this inequality, it is crucial scientists, practitioners and...

Knowledge of Alzheimer’s disease and associated factors among adults in Zhuhai, China: a cross-sectional analysis

This study aimed to assess the public knowledge regarding Alzheimer’s Disease (AD) in Zhuhai, China, focusing on identifying knowledge gaps and the influence of demographic and health factors.

Reallocating just 10 min to moderate-to-vigorous physical activity from other components of 24-hour movement behaviors improves cardiovascular health in adults

As components of a 24-hour day, sedentary behavior (SB), physical activity (PA), and sleep are all independently linked to cardiovascular health (CVH). However, insufficient understanding of components’ mutual...

Nutritional status and dietary intake among Nigerian adolescent: a systematic review

The prevailing nutritional conditions and the triple challenge of malnutrition faced by adolescents have adverse consequences for both the present and future generations’ health and nutrition. Summarizing the ...

The effect of educational intervention based on the behavioral reasoning theory on self-management behaviors in type 2 diabetes patients: a randomized controlled trial

Diabetes self-management education is necessary to improve patient outcomes and reduce diabetes-related complications. According to the theory of behavioral reasoning, the likelihood of performing a behavior i...

Who gets treated for an eating disorder? Implications for inference based on clinical populations

The minority of people with an eating disorder receive treatment. Little is known about predictors of receiving treatment.

Mexican-origin women’s individual and collective strategies to access and share health-promoting resources in the context of exclusionary immigration and immigrant policies

A growing literature has documented the social, economic, and health impacts of exclusionary immigration and immigrant policies in the early 21st century for Latiné communities in the US, pointing to immigrati...

Improving knowledge, attitude and practice on norovirus infection diarrhea among staff of kindergartens and schools: a before-after study

Norovirus gastroenteritis outbreaks were common in schools and kindergartens and were more related to faculty knowledge, attitude, and practice level. Gastroenteritis outbreaks caused by norovirus in education...

Associations between accelerometer-measured physical activity and sedentary behaviour with physical function among older women: a cross-sectional study

This study aimed to investigate the relationships between accelerometer-measured physical activity (PA) and sedentary behaviour (SB) with physical function (PF) among older Chinese women in the community.

Important information

Editorial board

For authors

For editorial board members

For reviewers

• Manuscript editing services

Annual Journal Metrics

Citation Impact 2023 Journal Impact Factor: 3.5 5-year Journal Impact Factor: 3.9 Source Normalized Impact per Paper (SNIP): 1.386 SCImago Journal Rank (SJR): 1.253 Speed 2023 Submission to first editorial decision (median days): 7 Submission to acceptance (median days): 173 Usage 2023 Downloads: 24,332,405 Altmetric mentions: 24,308

• More about our metrics

Peer-review Terminology

The following summary describes the peer review process for this journal:

Identity transparency: Single anonymized

Reviewer interacts with: Editor

Review information published: Review reports. Reviewer Identities reviewer opt in. Author/reviewer communication

More information is available here

• Follow us on Twitter

BMC Public Health

ISSN: 1471-2458

• General enquiries: [email protected]

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• Explore content
• About the journal
• Publish with us
• Sign up for alerts

Collection  29 March 2022

2021 Top 25 Health Sciences Articles

We are pleased to share with you the 25 most downloaded  Nature Communications  articles* in health sciences published in 2021. (Please note we have a separate collection on the Top 25 COVID-19 papers .) Featuring authors from around the world, these papers highlight valuable research from an international community.

Browse all Top 25 subject area collections  here .

*Data obtained from SN Insights (based on Digital Science's Dimensions) and normalised to account for articles published later in the year.

Research highlights

Association of sleep duration in middle and old age with incidence of dementia

Sleep dysregulation has been linked to dementia, but it is unknown whether sleep duration earlier in life is associated with dementia risk. Here, the authors show higher dementia risk associated with short sleep duration (six hours or less) in a longitudinal study of middle and older age adults.

• Séverine Sabia
• Aurore Fayosse
• Archana Singh-Manoux

Longitudinal analysis of blood markers reveals progressive loss of resilience and predicts human lifespan limit

Aging is associated with an increased risk of chronic diseases and functional decline. Here, the authors investigate the fluctuations of physiological indices along aging trajectories and observed a characteristic decrease in the organism state recovery rate.

• Timothy V. Pyrkov
• Konstantin Avchaciov
• Peter O. Fedichev

Restoration of energy homeostasis by SIRT6 extends healthy lifespan

Aging is associated with increased frailty and disrupted energy homeostasis. Here, the authors show that SIRT6 overexpression extends the lifespan of male and female mice and demonstrate that SIRT6 optimizes energy homeostasis in old age, which delays frailty and preserves healthy aging.

• A. Roichman
• S. Elhanati
• H. Y. Cohen

Triptonide is a reversible non-hormonal male contraceptive agent in mice and non-human primates

No male contraceptive pills are currently available. Here, the authors use triptonide, a compound derived from a Chinese plant, to deform sperm so that they cannot move properly, thereby causing reversible infertility in male mice and monkeys.

• Zongliang Chang
• Weibing Qin

Fasting alters the gut microbiome reducing blood pressure and body weight in metabolic syndrome patients

Nutritional modification including fasting has been shown to reduce cardiometabolic risk linked to western diet. Here the authors show implementation of fasting resulted in alterations to the intestinal microbiota, and circulating immune cells, improving blood pressure and body weight in patients with metabolic syndrome.

• András Maifeld
• Hendrik Bartolomaeus
• Sofia K. Forslund

Transneuronal delivery of hyper-interleukin-6 enables functional recovery after severe spinal cord injury in mice

The CNS has limited ability to regenerate following injury, Here, the authors show that a single injection of AAV-hyper-interleukin-6 in the sensory motor cortex results in corticospinal and raphe spinal tracts regeneration in the injured spinal cord as well as functional recovery in mice.

• Marco Leibinger
• Charlotte Zeitler
• Dietmar Fischer

Spatially resolved transcriptomics reveals the architecture of the tumor-microenvironment interface

During tumor progression, cancer cells contact different neighboring cell types, but it is unclear how these interactions affect cancer cell behavior. Here, the authors use spatially resolved transcriptomics and single-cell RNA-seq to study the role of cilia at the tumormicroenvironment interface.

• Miranda V. Hunter
• Reuben Moncada
• Richard M. White

Adjuvant oncolytic virotherapy for personalized anti-cancer vaccination

Viruses expressing tumour antigens can prime and boost anti-tumour immunity but the efficiency of this approach depends on the capacity of the virus to infect the host. Here, the authors show that vaccination with oncolytic viruses co-administered with tumour antigenic peptides is as efficient as antigen-engineered oncolytic viruses.

• K. Geoffroy
• M.-C. Bourgeois-Daigneault

Detection and characterization of lung cancer using cell-free DNA fragmentomes

DNA from tumour cells can be detected in the blood of cancer patients. Here, the authors show that cell free DNA fragmentation patterns can identify lung cancer patients and when this information is further interrogated it can be used to predict lung cancer histological subtype.

• Dimitrios Mathios
• Jakob Sidenius Johansen
• Victor E. Velculescu

A randomized controlled trial of pharmacist-led therapeutic carbohydrate and energy restriction in type 2 diabetes

Community pharmacists are accessible healthcare providers with expertise in medication management. Here the authors show that a low-carbohydrate, low-energy diet implemented by community pharmacists reduced diabetes medication use and improved glucose control in people with type 2 diabetes.

• Cody Durrer
• Sean McKelvey
• Jonathan P. Little

A metabolome atlas of the aging mouse brain

Metabolites play an important role in physiology, yet the complexity of the metabolome and its interaction with disease and aging is poorly understood. Here the authors present a comprehensive atlas of the mouse brain metabolome and how it changes during aging.

• Oliver Fiehn

Investigating immune and non-immune cell interactions in head and neck tumors by single-cell RNA sequencing

The tumor microenvironment (TME) has an important role in Head and Neck Squamous Cell Carcinoma (HNSCC) progression. Here, using single-cell RNA sequencing and multiplexed imaging, the authors report the cellular complexity of the TME in patients with HNSCC, exploring inflammatory status, stromal heterogeneity and immune checkpoint receptor-ligand interactions.

• Cornelius H. L. Kürten
• Aditi Kulkarni
• Robert L. Ferris

Single-cell profiling of tumor heterogeneity and the microenvironment in advanced non-small cell lung cancer

Comprehensive profiles of tumour and microenvironment are critical to understand heterogeneity in non-small cell lung cancer (NSCLC). Here, the authors profile 42 late-stage NSCLC patients with single-cell RNA-seq, revealing immune landscapes that are associated with cancer subtype or heterogeneity.

• Fengying Wu
• Caicun Zhou

Biomimetic nanoparticles deliver mRNAs encoding costimulatory receptors and enhance T cell mediated cancer immunotherapy

Antibodies targeting OX40 or CD137, two T cell costimulatory receptors, have been shown to improve antitumor immunity. Here the authors design a phospholipid-derived nanoparticle to deliver OX40 or CD137 mRNA to T cells in vivo, improving efficacy of anti-OX40 and anti-CD137 antibody therapy in preclinical tumor models.

• Xinfu Zhang
• Yizhou Dong

Sexual dimorphism in glucose metabolism is shaped by androgen-driven gut microbiome

Male sex is a risk factor for impaired glucose metabolism and type 2 diabetes. Here the authors identify that androgen modulates the gut microbiome, which drives insulin resistance and contributes to sexual dimorphism in glucose metabolism in mice.

• Weiqing Wang

Blood n-3 fatty acid levels and total and cause-specific mortality from 17 prospective studies

Associations between of omega-3 fatty acids and mortality are not clear. Here the authors report that, based on a pooled analysis of 17 prospective cohort studies, higher blood omega-3 fatty acid levels correlate with lower risk of all-cause mortality.

• William S. Harris
• Nathan L. Tintle
• The Fatty Acids and Outcomes Research Consortium (FORCE)

Multi-omics analysis identifies therapeutic vulnerabilities in triple-negative breast cancer subtypes

Triple negative breast cancer can be divided into additional subtypes. Here, using omics analyses, the authors show that in the mesenchymal subtype expression of MHC-1 is repressed and that this can be restored by using drugs that target subunits of the epigenetic modifier PRC2.

• Brian D. Lehmann
• Antonio Colaprico
• X. Steven Chen

Daily caloric restriction limits tumor growth more effectively than caloric cycling regardless of dietary composition

Caloric restriction (CR) has been shown as an effective intervention to reduce tumorigenesis, but alternative less stringent dietary interventions have also been considered. Here, the authors show that in a murine model of breast cancer CR has a larger effect on preventing tumorigenesis and metastasis compared to periodic caloric cycling.

• Laura C. D. Pomatto-Watson
• Monica Bodogai
• Rafael de Cabo

Neoadjuvant immunotherapy with nivolumab and ipilimumab induces major pathological responses in patients with head and neck squamous cell carcinoma

Immune checkpoint blockade has become standard care for patients with recurrent metastatic head and neck squamous cell carcinoma (HNSCC). Here the authors present the results of a non-randomized phase Ib/IIa trial, reporting safety and efficacy of neoadjuvant nivolumab monotherapy and nivolumab plus ipilimumab prior to standard-of-care surgery in patients with HNSCC. .

• Joris L. Vos
• Joris B. W. Elbers
• Charlotte L. Zuur

9p21 loss confers a cold tumor immune microenvironment and primary resistance to immune checkpoint therapy

The molecular mechanisms of resistance to immune checkpoint therapy remain elusive. Here, the authors perform immunogenomic analysis of TCGA data and data from clinical trials for antiPD-1/PD-L1 therapy and highlight the association of 9p21 loss with a cold tumor microenvironment and resistance to therapy.

• Guangchun Han
• Guoliang Yang
• Linghua Wang

Gut bacteria identified in colorectal cancer patients promote tumourigenesis via butyrate secretion

Several bacteria in the gut microbiota have been associated with colorectal cancer (CRC) but it is not completely clear whether they have a role in tumourigenesis. Here, the authors show enrichment of 12 bacterial taxa in two cohorts of CRC patients and that two Porphyromonas species accelerate CRC onset through butyrate secretion.

• Shintaro Okumura
• Yusuke Konishi

Elevated circulating follistatin associates with an increased risk of type 2 diabetes

Follistatin promotes in type 2 diabetes (T2D) pathogenesis in model animals and is elevated in patients with T2D. Here the authors report that plasma follistatin associates with increased risk of incident T2D in two longitudinal cohorts, and show that follistatin regulates insulin-induced suppression lipolysis in cultured human adipocytes.

• Chuanyan Wu
• Yang De Marinis

Tau activates microglia via the PQBP1-cGAS-STING pathway to promote brain inflammation

Brain inflammation generally accelerates neurodegeneration but the mechanisms of this are not fully characterised. Here the authors show that PQBP1 in microglia is important for sensing extrinsic Tau 3 R/4 R proteins and triggers an innate immune response through cGAS and STING resulting in cognitive impairment.

• Hiroki Shiwaku
• Hitoshi Okazawa

Long-term treatment with senolytic drugs Dasatinib and Quercetin ameliorates age-dependent intervertebral disc degeneration in mice

Intervertebral disc degeneration is a leading cause of chronic back pain and disability. Here the authors show that long term treatment with senolytic compounds Dasatinib and Quercetin reduces disc senescence burden and ameliorates age-dependent degeneration in mice.

• Emanuel J. Novais
• Victoria A. Tran
• Makarand V. Risbud

DNA/RNA heteroduplex oligonucleotide technology for regulating lymphocytes in vivo

Using gene silencing to regulate lymphocyte function is a promising therapeutic approach for autommunity, inflammation and cancer. Here the authors use a heteroduplex oligonucleotide for improved potency, efficacy and longer retention times.

• Masaki Ohyagi
• Tetsuya Nagata
• Takanori Yokota

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

• Research article
• Open access
• Published: 10 April 2021

The role of artificial intelligence in healthcare: a structured literature review

• Silvana Secinaro 1 ,
• Davide Calandra 1 ,
• Aurelio Secinaro 2 ,
• Vivek Muthurangu 3 &
• Paolo Biancone 1  

BMC Medical Informatics and Decision Making volume  21 , Article number:  125 ( 2021 ) Cite this article

157k Accesses

25 Altmetric

Metrics details

Background/Introduction

Artificial intelligence (AI) in the healthcare sector is receiving attention from researchers and health professionals. Few previous studies have investigated this topic from a multi-disciplinary perspective, including accounting, business and management, decision sciences and health professions.

The structured literature review with its reliable and replicable research protocol allowed the researchers to extract 288 peer-reviewed papers from Scopus. The authors used qualitative and quantitative variables to analyse authors, journals, keywords, and collaboration networks among researchers. Additionally, the paper benefited from the Bibliometrix R software package.

The investigation showed that the literature in this field is emerging. It focuses on health services management, predictive medicine, patient data and diagnostics, and clinical decision-making. The United States, China, and the United Kingdom contributed the highest number of studies. Keyword analysis revealed that AI can support physicians in making a diagnosis, predicting the spread of diseases and customising treatment paths.

Conclusions

The literature reveals several AI applications for health services and a stream of research that has not fully been covered. For instance, AI projects require skills and data quality awareness for data-intensive analysis and knowledge-based management. Insights can help researchers and health professionals understand and address future research on AI in the healthcare field.

Peer Review reports

Artificial intelligence (AI) generally applies to computational technologies that emulate mechanisms assisted by human intelligence, such as thought, deep learning, adaptation, engagement, and sensory understanding [ 1 , 2 ]. Some devices can execute a role that typically involves human interpretation and decision-making [ 3 , 4 ]. These techniques have an interdisciplinary approach and can be applied to different fields, such as medicine and health. AI has been involved in medicine since as early as the 1950s, when physicians made the first attempts to improve their diagnoses using computer-aided programs [ 5 , 6 ]. Interest and advances in medical AI applications have surged in recent years due to the substantially enhanced computing power of modern computers and the vast amount of digital data available for collection and utilisation [ 7 ]. AI is gradually changing medical practice. There are several AI applications in medicine that can be used in a variety of medical fields, such as clinical, diagnostic, rehabilitative, surgical, and predictive practices. Another critical area of medicine where AI is making an impact is clinical decision-making and disease diagnosis. AI technologies can ingest, analyse, and report large volumes of data across different modalities to detect disease and guide clinical decisions [ 3 , 8 ]. AI applications can deal with the vast amount of data produced in medicine and find new information that would otherwise remain hidden in the mass of medical big data [ 9 , 10 , 11 ]. These technologies can also identify new drugs for health services management and patient care treatments [ 5 , 6 ].

Courage in the application of AI is visible through a search in the primary research databases. However, as Meskò et al. [ 7 ] find, the technology will potentially reduce care costs and repetitive operations by focusing the medical profession on critical thinking and clinical creativity. As Cho et al. and Doyle et al. [ 8 , 9 ] add, the AI perspective is exciting; however, new studies will be needed to establish the efficacy and applications of AI in the medical field [ 10 ].

Our paper will also concentrate on AI strategies for healthcare from the accounting, business, and management perspectives. The authors used the structured literature review (SLR) method for its reliable and replicable research protocol [ 11 ] and selected bibliometric variables as sources of investigation. Bibliometric usage enables the recognition of the main quantitative variables of the study stream [ 12 ]. This method facilitates the detection of the required details of a particular research subject, including field authors, number of publications, keywords for interaction between variables (policies, properties and governance) and country data [ 13 ]. It also allows the application of the science mapping technique [ 14 ]. Our paper adopted the Bibliometrix R package and the biblioshiny web interface as tools of analysis [ 14 ].

The investigation offers the following insights for future researchers and practitioners:

bibliometric information on 288 peer-reviewed English papers from the Scopus collection.

Identification of leading journals in this field, such as Journal of Medical Systems, Studies in Health Technology and Informatics, IEEE Journal of Biomedical and Health Informatics, and Decision Support Systems.

Qualitative and quantitative information on authors’ Lotka’s law, h-index, g-index, m-index, keyword, and citation data.

Research on specific countries to assess AI in the delivery and effectiveness of healthcare, quotes, and networks within each region.

A topic dendrogram study that identifies five research clusters: health services management, predictive medicine, patient data, diagnostics, and finally, clinical decision-making.

An in-depth discussion that develops theoretical and practical implications for future studies.

The paper is organised as follows. Section  2 lists the main bibliometric articles in this field. Section  3 elaborates on the methodology. Section  4 presents the findings of the bibliometric analysis. Section  5 discusses the main elements of AI in healthcare based on the study results. Section  6 concludes the article with future implications for research.

Related works and originality

As suggested by Zupic and Čater [ 15 ], a research stream can be evaluated with bibliometric methods that can introduce objectivity and mitigate researcher bias. For this reason, bibliometric methods are attracting increasing interest among researchers as a reliable and impersonal research analytical approach [ 16 , 17 ]. Recently, bibliometrics has been an essential method for analysing and predicting research trends [ 18 ]. Table  1 lists other research that has used a similar approach in the research stream investigated.

The scientific articles reported show substantial differences in keywords and research topics that have been previously studied. The bibliometric analysis of Huang et al. [ 19 ] describes rehabilitative medicine using virtual reality technology. According to the authors, the primary goal of rehabilitation is to enhance and restore functional ability and quality of life for patients with physical impairments or disabilities. In recent years, many healthcare disciplines have been privileged to access various technologies that provide tools for both research and clinical intervention.

Hao et al. [ 20 ] focus on text mining in medical research. As reported, text mining reveals new, previously unknown information by using a computer to automatically extract information from different text resources. Text mining methods can be regarded as an extension of data mining to text data. Text mining is playing an increasingly significant role in processing medical information. Similarly, the studies by dos Santos et al. [ 21 ] focus on applying data mining and machine learning (ML) techniques to public health problems. As stated in this research, public health may be defined as the art and science of preventing diseases, promoting health, and prolonging life. Using data mining and ML techniques, it is possible to discover new information that otherwise would be hidden. These two studies are related to another topic: medical big data. According to Liao et al. [ 22 ], big data is a typical “buzzword” in the business and research community, referring to a great mass of digital data collected from various sources. In the medical field, we can obtain a vast amount of data (i.e., medical big data). Data mining and ML techniques can help deal with this information and provide helpful insights for physicians and patients. More recently, Choudhury et al. [ 23 ] provide a systematic review on the use of ML to improve the care of elderly patients, demonstrating eligible studies primarily in psychological disorders and eye diseases.

Tran et al. [ 2 ] focus on the global evolution of AI research in medicine. Their bibliometric analysis highlights trends and topics related to AI applications and techniques. As stated in Connelly et al.’s [ 24 ] study, robot-assisted surgeries have rapidly increased in recent years. Their bibliometric analysis demonstrates how robotic-assisted surgery has gained acceptance in different medical fields, such as urological, colorectal, cardiothoracic, orthopaedic, maxillofacial and neurosurgery applications. Additionally, the bibliometric analysis of Guo et al. [ 25 ] provides an in-depth study of AI publications through December 2019. The paper focuses on tangible AI health applications, giving researchers an idea of how algorithms can help doctors and nurses. A new stream of research related to AI is also emerging. In this sense, Choudhury and Asan’s [ 26 ] scientific contribution provides a systematic review of the AI literature to identify health risks for patients. They report on 53 studies involving technology for clinical alerts, clinical reports, and drug safety. Considering the considerable interest within this research stream, this analysis differs from the current literature for several reasons. It aims to provide in-depth discussion, considering mainly the business, management, and accounting fields and not dealing only with medical and health profession publications.

Additionally, our analysis aims to provide a bibliometric analysis of variables such as authors, countries, citations and keywords to guide future research perspectives for researchers and practitioners, as similar analyses have done for several publications in other research streams [ 15 , 16 , 27 ]. In doing so, we use a different database, Scopus, that is typically adopted in social sciences fields. Finally, our analysis will propose and discuss a dominant framework of variables in this field, and our analysis will not be limited to AI application descriptions.

Methodology

This paper evaluated AI in healthcare research streams using the SLR method [ 11 ]. As suggested by Massaro et al. [ 11 ], an SLR enables the study of the scientific corpus of a research field, including the scientific rigour, reliability and replicability of operations carried out by researchers. As suggested by many scholars, the methodology allows qualitative and quantitative variables to highlight the best authors, journals and keywords and combine a systematic literature review and bibliometric analysis [ 27 , 28 , 29 , 30 ]. Despite its widespread use in business and management [ 16 , 31 ], the SLR is also used in the health sector based on the same philosophy through which it was originally conceived [ 32 , 33 ]. A methodological analysis of previously published articles reveals that the most frequently used steps are as follows [ 28 , 31 , 34 ]:

defining research questions;

writing the research protocol;

defining the research sample to be analysed;

developing codes for analysis; and

critically analysing, discussing, and identifying a future research agenda.

Considering the above premises, the authors believe that an SLR is the best method because it combines scientific validity, replicability of the research protocol and connection between multiple inputs.

As stated by the methodological paper, the first step is research question identification. For this purpose, we benefit from the analysis of Zupic and Čater [ 15 ], who provide several research questions for future researchers to link the study of authors, journals, keywords and citations. Therefore, RQ1 is “What are the most prominent authors, journal keywords and citations in the field of the research study?” Additionally, as suggested by Haleem et al. [ 35 ], new technologies, including AI, are changing the medical field in unexpected timeframes, requiring studies in multiple areas. Therefore, RQ2 is “How does artificial intelligence relate to healthcare, and what is the focus of the literature?” Then, as discussed by Massaro et al. [ 36 ], RQ3 is “What are the research applications of artificial intelligence for healthcare?”.

The first research question aims to define the qualitative and quantitative variables of the knowledge flow under investigation. The second research question seeks to determine the state of the art and applications of AI in healthcare. Finally, the third research question aims to help researchers identify practical and theoretical implications and future research ideas in this field.

The second fundamental step of the SLR is writing the research protocol [ 11 ]. Table  2 indicates the currently known literature elements, uniquely identifying the research focus, motivations and research strategy adopted and the results providing a link with the following points. Additionally, to strengthen the analysis, our investigation benefits from the PRISMA statement methodological article [ 37 ]. Although the SLR is a validated method for systematic reviews and meta-analyses, we believe that the workflow provided may benefit the replicability of the results [ 37 , 38 , 39 , 40 ]. Figure  1 summarises the researchers’ research steps, indicating that there are no results that can be referred to as a meta-analysis.

Source : Authors’ elaboration on Liberati et al. [ 37 ]

PRISMA workflow.

The third step is to specify the search strategy and search database. Our analysis is based on the search string “Artificial Intelligence” OR “AI” AND “Healthcare” with a focus on “Business, Management, and Accounting”, “Decision Sciences”, and “Health professions”. As suggested by [ 11 , 41 ] and motivated by [ 42 ], keywords can be selected through a top-down approach by identifying a large search field and then focusing on particular sub-topics. The paper uses data retrieved from the Scopus database, a multi-disciplinary database, which allowed the researchers to identify critical articles for scientific analysis [ 43 ]. Additionally, Scopus was selected based on Guo et al.’s [ 25 ] limitations, which suggest that “future studies will apply other databases, such as Scopus, to explore more potential papers” . The research focuses on articles and reviews published in peer-reviewed journals for their scientific relevance [ 11 , 16 , 17 , 29 ] and does not include the grey literature, conference proceedings or books/book chapters. Articles written in any language other than English were excluded [ 2 ]. For transparency and replicability, the analysis was conducted on 11 January 2021. Using this research strategy, the authors retrieved 288 articles. To strengthen the study's reliability, we publicly provide the full bibliometric extract on the Zenodo repository [ 44 , 45 ].

The fourth research phase is defining the code framework that initiates the analysis of the variables. The study will identify the following:

descriptive information of the research area;

source analysis [ 16 ];

author and citation analysis [ 28 ];

keywords and network analysis [ 14 ]; and

geographic distribution of the papers [ 14 ].

The final research phase is the article’s discussion and conclusion, where implications and future research trends will be identified.

At the research team level, the information is analysed with the statistical software R-Studio and the Bibliometrix package [ 15 ], which allows scientific analysis of the results obtained through the multi-disciplinary database.

The analysis of bibliometric results starts with a description of the main bibliometric statistics with the aim of answering RQ1, What are the most prominent authors, journal keywords and citations in the field of the research study?, and RQ2, How does artificial intelligence relate to healthcare, and what is the focus of the literature? Therefore, the following elements were thoroughly analysed: (1) type of document; (2) annual scientific production; (3) scientific sources; (4) source growth; (5) number of articles per author; (6) author’s dominance ranking; (7) author’s h-index, g-index, and m-index; (8) author’s productivity; (9) author’s keywords; (10) topic dendrogram; (11) a factorial map of the document with the highest contributions; (12) article citations; (13) country production; (14) country citations; (15) country collaboration map; and (16) country collaboration network.

Main information

Table  3 shows the information on 288 peer-reviewed articles published between 1992 and January 2021 extracted from the Scopus database. The number of keywords is 946 from 136 sources, and the number of keywords plus, referring to the number of keywords that frequently appear in an article’s title, was 2329. The analysis period covered 28 years and 1 month of scientific production and included an annual growth rate of 5.12%. However, the most significant increase in published articles occurred in the past three years (please see Fig.  2 ). On average, each article was written by three authors (3.56). Finally, the collaboration index (CI), which was calculated as the total number of authors of multi-authored articles/total number of multi-authored articles, was 3.97 [ 46 ].

Source : Authors’ elaboration

Annual scientific production.

Table  4 shows the top 20 sources related to the topic. The Journal of Medical Systems is the most relevant source, with twenty-one of the published articles. This journal's main issues are the foundations, functionality, interfaces, implementation, impacts, and evaluation of medical technologies. Another relevant source is Studies in Health Technology and Informatics, with eleven articles. This journal aims to extend scientific knowledge related to biomedical technologies and medical informatics research. Both journals deal with cloud computing, machine learning, and AI as a disruptive healthcare paradigm based on recent publications. The IEEE Journal of Biomedical and Health Informatics investigates technologies in health care, life sciences, and biomedicine applications from a broad perspective. The next journal, Decision Support Systems, aims to analyse how these technologies support decision-making from a multi-disciplinary view, considering business and management. Therefore, the analysis of the journals revealed that we are dealing with an interdisciplinary research field. This conclusion is confirmed, for example, by the presence of purely medical journals, journals dedicated to the technological growth of healthcare, and journals with a long-term perspective such as futures.

The distribution frequency of the articles (Fig.  3 ) indicates the journals dealing with the topic and related issues. Between 2008 and 2012, a significant growth in the number of publications on the subject is noticeable. However, the graph shows the results of the Loess regression, which includes the quantity and publication time of the journal under analysis as variables. This method allows the function to assume an unlimited distribution; that is, feature can consider values below zero if the data are close to zero. It contributes to a better visual result and highlights the discontinuity in the publication periods [ 47 ].

Source growth. Source : Authors’ elaboration

Finally, Fig.  4 provides an analytical perspective on factor analysis for the most cited papers. As indicated in the literature [ 48 , 49 ], using factor analysis to discover the most cited papers allows for a better understanding of the scientific world’s intellectual structure. For example, our research makes it possible to consider certain publications that effectively analyse subject specialisation. For instance, Santosh’s [ 50 ] article addresses the new paradigm of AI with ML algorithms for data analysis and decision support in the COVID-19 period, setting a benchmark in terms of citations by researchers. Moving on to the application, an article by Shickel et al. [ 51 ] begins with the belief that the healthcare world currently has much health and administrative data. In this context, AI and deep learning will support medical and administrative staff in extracting data, predicting outcomes, and learning medical representations. Finally, in the same line of research, Baig et al. [ 52 ], with a focus on wearable patient monitoring systems (WPMs), conclude that AI and deep learning may be landmarks for continuous patient monitoring and support for healthcare delivery.

Factorial map of the most cited documents.

This section identifies the most cited authors of articles on AI in healthcare. It also identifies the authors’ keywords, dominance factor (DF) ranking, h-index, productivity, and total number of citations. Table  5 identifies the authors and their publications in the top 20 rankings. As the table shows, Bushko R.G. has the highest number of publications: four papers. He is the editor-in-chief of Future of Health Technology, a scientific journal that aims to develop a clear vision of the future of health technology. Then, several authors each wrote three papers. For instance, Liu C. is a researcher active in the topic of ML and computer vision, and Sharma A. from Emory University Atlanta in the USA is a researcher with a clear focus on imaging and translational informatics. Some other authors have two publications each. While some authors have published as primary authors, most have published as co-authors. Hence, in the next section, we measure the contributory power of each author by investigating the DF ranking through the number of elements.

Authors’ dominance ranking

The dominance factor (DF) is a ratio measuring the fraction of multi-authored articles in which an author acts as the first author [ 53 ]. Several bibliometric studies use the DF in their analyses [ 46 , 54 ]. The DF ranking calculates an author’s dominance in producing articles. The DF is calculated by dividing the number of an author’s multi-authored papers as the first author (Nmf) by the author's total number of multi-authored papers (Nmt). This is omitted in the single-author case due to the constant value of 1 for single-authored articles. This formulation could lead to some distortions in the results, especially in fields where the first author is entered by surname alphabetical order [ 55 ].

The mathematical equation for the DF is shown as:

Table  6 lists the top 20 DF rankings. The data in the table show a low level of articles per author, either for first-authored or multi-authored articles. The results demonstrate that we are dealing with an emerging topic in the literature. Additionally, as shown in the table, Fox J. and Longoni C. are the most dominant authors in the field.

Authors’ impact

Table  7 shows the impact of authors in terms of the h-index [ 56 ] (i.e., the productivity and impact of citations of a researcher), g-index [ 57 ] (i.e., the distribution of citations received by a researcher's publications), m-index [ 58 ] (i.e., the h-index value per year), total citations, total paper and years of scientific publication. The H-index was introduced in the literature as a metric for the objective comparison of scientific results and depended on the number of publications and their impact [ 59 ]. The results show that the 20 most relevant authors have an h-index between 2 and 1. For the practical interpretation of the data, the authors considered data published by the London School of Economics [ 60 ]. In the social sciences, the analysis shows values of 7.6 for economic publications by professors and researchers who had been active for several years. Therefore, the youthfulness of the research area has attracted young researchers and professors. At the same time, new indicators have emerged over the years to diversify the logic of the h-index. For example, the g-index indicates an author's impact on citations, considering that a single article can generate these. The m-index, on the other hand, shows the cumulative value over the years.

The analysis, also considering the total number of citations, the number of papers published and the year of starting to publish, thus confirms that we are facing an expanding research flow.

Authors’ productivity

Figure  5 shows Lotka’s law. This mathematical formulation originated in 1926 to describe the publication frequency by authors in a specific research field [ 61 ]. In practice, the law states that the number of authors contributing to research in a given period is a fraction of the number who make up a single contribution [ 14 , 61 ].

Lotka’s law.

The mathematical relationship is expressed in reverse in the following way:

where y x is equal to the number of authors producing x articles in each research field. Therefore, C and n are constants that can be estimated in the calculation.

The figure's results are in line with Lotka's results, with an average of two publications per author in a given research field. In addition, the figure shows the percentage of authors. Our results lead us to state that we are dealing with a young and growing research field, even with this analysis. Approximately 70% of the authors had published only their first research article. Only approximately 20% had published two scientific papers.

Authors’ keywords

This section provides information on the relationship between the keywords artificial intelligence and healthcare . This analysis is essential to determine the research trend, identify gaps in the discussion on AI in healthcare, and identify the fields that can be interesting as research areas [ 42 , 62 ].

Table  8 highlights the total number of keywords per author in the top 20 positions. The ranking is based on the following elements: healthcare, artificial intelligence, and clinical decision support system . Keyword analysis confirms the scientific area of reference. In particular, we deduce the definition as “Artificial intelligence is the theory and development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages” [ 2 , 63 ]. Panch et al. [ 4 ] find that these technologies can be used in different business and management areas. After the first keyword, the analysis reveals AI applications and related research such as machine learning and deep learning.

Additionally, data mining and big data are a step forward in implementing exciting AI applications. According to our specific interest, if we applied AI in healthcare, we would achieve technological applications to help and support doctors and medical researchers in decision-making. The link between AI and decision-making is the reason why we find, in the seventh position, the keyword clinical decision support system . AI techniques can unlock clinically relevant information hidden in the massive amount of data that can assist clinical decision-making [ 64 ]. If we analyse the following keywords, we find other elements related to decision-making and support systems.

The TreeMap below (Fig.  6 ) highlights the combination of possible keywords representing AI and healthcare.

Keywords treemap.

The topic dendrogram in Fig.  7 represents the hierarchical order and the relationship between the keywords generated by hierarchical clustering [ 42 ]. The cut in the figure and the vertical lines facilitate an investigation and interpretation of the different clusters. As stated by Andrews [ 48 ], the figure is not intended to find the perfect level of associations between clusters. However, it aims to estimate the approximate number of clusters to facilitate further discussion.

Topic dendrogram.

The research stream of AI in healthcare is divided into two main strands. The blue strand focuses on medical information systems and the internet. Some papers are related to healthcare organisations, such as the Internet of Things, meaning that healthcare organisations use AI to support health services management and data analysis. AI applications are also used to improve diagnostic and therapeutic accuracy and the overall clinical treatment process [ 2 ]. If we consider the second block, the red one, three different clusters highlight separate aspects of the topic. The first could be explained as AI and ML predictive algorithms. Through AI applications, it is possible to obtain a predictive approach that can ensure that patients are better monitored. This also allows a better understanding of risk perception for doctors and medical researchers. In the second cluster, the most frequent words are decisions , information system , and support system . This means that AI applications can support doctors and medical researchers in decision-making. Information coming from AI technologies can be used to consider difficult problems and support a more straightforward and rapid decision-making process. In the third cluster, it is vital to highlight that the ML model can deal with vast amounts of data. From those inputs, it can return outcomes that can optimise the work of healthcare organisations and scheduling of medical activities.

Furthermore, the word cloud in Fig.  8 highlights aspects of AI in healthcare, such as decision support systems, decision-making, health services management, learning systems, ML techniques and diseases. The figure depicts how AI is linked to healthcare and how it is used in medicine.

Word cloud.

Figure  9 represents the search trends based on the keywords analysed. The research started in 2012. First, it identified research topics related to clinical decision support systems. This topic was recurrent during the following years. Interestingly, in 2018, studies investigated AI and natural language processes as possible tools to manage patients and administrative elements. Finally, a new research stream considers AI's role in fighting COVID-19 [ 65 , 66 ].

Keywords frequency.

Table  9 represents the number of citations from other articles within the top 20 rankings. The analysis allows the benchmark studies in the field to be identified [ 48 ]. For instance, Burke et al. [ 67 ] writes the most cited paper and analyses efficient nurse rostering methodologies. The paper critically evaluates tangible interdisciplinary solutions that also include AI. Immediately thereafter, Ahmed M.A.'s article proposes a data-driven optimisation methodology to determine the optimal number of healthcare staff to optimise patients' productivity [ 68 ]. Finally, the third most cited article lays the groundwork for developing deep learning by considering diverse health and administrative information [ 51 ].

This section analyses the diffusion of AI in healthcare around the world. It highlights countries to show the geographies of this research. It includes all published articles, the total number of citations, and the collaboration network. The following sub-sections start with an analysis of the total number of published articles.

Country total articles

Figure  9 and Table  10 display the countries where AI in healthcare has been considered. The USA tops the list of countries with the maximum number of articles on the topic (215). It is followed by China (83), the UK (54), India (51), Australia (54), and Canada (32). It is immediately evident that the theme has developed on different continents, highlighting a growing interest in AI in healthcare. The figure shows that many areas, such as Russia, Eastern Europe and Africa except for Algeria, Egypt, and Morocco, have still not engaged in this scientific debate.

Country publications and collaboration map

This section discusses articles on AI in healthcare in terms of single or multiple publications in each country. It also aims to observe collaboration and networking between countries. Table  11 and Fig.  10 highlight the average citations by state and show that the UK, the USA, and Kuwait have a higher average number of citations than other countries. Italy, Spain and New Zealand have the most significant number of citations.

Articles per country.

Figure  11 depicts global collaborations. The blue colour on the map represents research cooperation among nations. Additionally, the pink border linking states indicates the extent of collaboration between authors. The primary cooperation between nations is between the USA and China, with two collaborative articles. Other collaborations among nations are limited to a few papers.

Collaboration map.

Artificial intelligence for healthcare: applications

This section aims to strengthen the research scope by answering RQ3: What are the research applications of artificial intelligence for healthcare?

Benefiting from the topical dendrogram, researchers will provide a development model based on four relevant variables [ 69 , 70 ]. AI has been a disruptive innovation in healthcare [ 4 ]. With its sophisticated algorithms and several applications, AI has assisted doctors and medical professionals in the domains of health information systems, geocoding health data, epidemic and syndromic surveillance, predictive modelling and decision support, and medical imaging [ 2 , 9 , 10 , 64 ]. Furthermore, the researchers considered the bibliometric analysis to identify four macro-variables dominant in the field and used them as authors' keywords. Therefore, the following sub-sections aim to explain the debate on applications in healthcare for AI techniques. These elements are shown in Fig.  12 .

Dominant variables for AI in healthcare.

Health services management

One of the notable aspects of AI techniques is potential support for comprehensive health services management. These applications can support doctors, nurses and administrators in their work. For instance, an AI system can provide health professionals with constant, possibly real-time medical information updates from various sources, including journals, textbooks, and clinical practices [ 2 , 10 ]. These applications' strength is becoming even more critical in the COVID-19 period, during which information exchange is continually needed to properly manage the pandemic worldwide [ 71 ]. Other applications involve coordinating information tools for patients and enabling appropriate inferences for health risk alerts and health outcome prediction [ 72 ]. AI applications allow, for example, hospitals and all health services to work more efficiently for the following reasons:

Clinicians can access data immediately when they need it.

Nurses can ensure better patient safety while administering medication.

Patients can stay informed and engaged in their care by communicating with their medical teams during hospital stays.

Additionally, AI can contribute to optimising logistics processes, for instance, realising drugs and equipment in a just-in-time supply system based totally on predictive algorithms [ 73 , 74 ]. Interesting applications can also support the training of personnel working in health services. This evidence could be helpful in bridging the gap between urban and rural health services [ 75 ]. Finally, health services management could benefit from AI to leverage the multiplicity of data in electronic health records by predicting data heterogeneity across hospitals and outpatient clinics, checking for outliers, performing clinical tests on the data, unifying patient representation, improving future models that can predict diagnostic tests and analyses, and creating transparency with benchmark data for analysing services delivered [ 51 , 76 ].

Predictive medicine

Another relevant topic is AI applications for disease prediction and diagnosis treatment, outcome prediction and prognosis evaluation [ 72 , 77 ]. Because AI can identify meaningful relationships in raw data, it can support diagnostic, treatment and prediction outcomes in many medical situations [ 64 ]. It allows medical professionals to embrace the proactive management of disease onset. Additionally, predictions are possible for identifying risk factors and drivers for each patient to help target healthcare interventions for better outcomes [ 3 ]. AI techniques can also help design and develop new drugs, monitor patients and personalise patient treatment plans [ 78 ]. Doctors benefit from having more time and concise data to make better patient decisions. Automatic learning through AI could disrupt medicine, allowing prediction models to be created for drugs and exams that monitor patients over their whole lives [ 79 ].

• Clinical decision-making

One of the keyword analysis main topics is that AI applications could support doctors and medical researchers in the clinical decision-making process. According to Jiang et al. [ 64 ], AI can help physicians make better clinical decisions or even replace human judgement in healthcare-specific functional areas. According to Bennett and Hauser [ 80 ], algorithms can benefit clinical decisions by accelerating the process and the amount of care provided, positively impacting the cost of health services. Therefore, AI technologies can support medical professionals in their activities and simplify their jobs [ 4 ]. Finally, as Redondo and Sandoval [ 81 ] find, algorithmic platforms can provide virtual assistance to help doctors understand the semantics of language and learning to solve business process queries as a human being would.

Patient data and diagnostics

Another challenging topic related to AI applications is patient data and diagnostics. AI techniques can help medical researchers deal with the vast amount of data from patients (i.e., medical big data ). AI systems can manage data generated from clinical activities, such as screening, diagnosis, and treatment assignment. In this way, health personnel can learn similar subjects and associations between subject features and outcomes of interest [ 64 ].

These technologies can analyse raw data and provide helpful insights that can be used in patient treatments. They can help doctors in the diagnostic process; for example, to realise a high-speed body scan, it will be simpler to have an overall patient condition image. Then, AI technology can recreate a 3D mapping solution of a patient’s body.

In terms of data, interesting research perspectives are emerging. For instance, we observed the emergence of a stream of research on patient data management and protection related to AI applications [ 82 ].

For diagnostics, AI techniques can make a difference in rehabilitation therapy and surgery. Numerous robots have been designed to support and manage such tasks. Rehabilitation robots physically support and guide, for example, a patient’s limb during motor therapy [ 83 ]. For surgery, AI has a vast opportunity to transform surgical robotics through devices that can perform semi-automated surgical tasks with increasing efficiency. The final aim of this technology is to automate procedures to negate human error while maintaining a high level of accuracy and precision [ 84 ]. Finally, the -19 period has led to increased remote patient diagnostics through telemedicine that enables remote observation of patients and provides physicians and nurses with support tools [ 66 , 85 , 86 ].

This study aims to provide a bibliometric analysis of publications on AI in healthcare, focusing on accounting, business and management, decision sciences and health profession studies. Using the SLR method of Massaro et al. [ 11 ], we provide a reliable and replicable research protocol for future studies in this field. Additionally, we investigate the trend of scientific publications on the subject, unexplored information, future directions, and implications using the science mapping workflow. Our analysis provides interesting insights.

In terms of bibliometric variables, the four leading journals, Journal of Medical Systems , Studies in Health Technology and Informatics , IEEE Journal of Biomedical and Health Informatics , and Decision Support Systems , are optimal locations for the publication of scientific articles on this topic. These journals deal mainly with healthcare, medical information systems, and applications such as cloud computing, machine learning, and AI. Additionally, in terms of h-index, Bushko R.G. and Liu C. are the most productive and impactful authors in this research stream. Burke et al.’s [ 67 ] contribution is the most cited with an analysis of nurse rostering using new technologies such as AI. Finally, in terms of keywords, co-occurrence reveals some interesting insights. For instance, researchers have found that AI has a role in diagnostic accuracy and helps in the analysis of health data by comparing thousands of medical records, experiencing automatic learning with clinical alerts, efficient management of health services and places of care, and the possibility of reconstructing patient history using these data.

Second, this paper finds five cluster analyses in healthcare applications: health services management, predictive medicine, patient data, diagnostics, and finally, clinical decision-making. These technologies can also contribute to optimising logistics processes in health services and allowing a better allocation of resources.

Third, the authors analysing the research findings and the issues under discussion strongly support AI's role in decision support. These applications, however, are demonstrated by creating a direct link to data quality management and the technology awareness of health personnel [ 87 ].

The importance of data quality for the decision-making process

Several authors have analysed AI in the healthcare research stream, but in this case, the authors focus on other literature that includes business and decision-making processes. In this regard, the analysis of the search flow reveals a double view of the literature. On the one hand, some contributions belong to the positivist literature and embrace future applications and implications of technology for health service management, data analysis and diagnostics [ 6 , 80 , 88 ]. On the other hand, some investigations also aim to understand the darker sides of technology and its impact. For example, as Carter [ 89 ] states, the impact of AI is multi-sectoral; its development, however, calls for action to protect personal data. Similarly, Davenport and Kalakota [ 77 ] focus on the ethical implications of using AI in healthcare. According to the authors, intelligent machines raise issues of accountability, transparency, and permission, especially in automated communication with patients. Our analysis does not indicate a marked strand of the literature; therefore, we argue that the discussion of elements such as the transparency of technology for patients is essential for the development of AI applications.

A large part of our results shows that, at the application level, AI can be used to improve medical support for patients (Fig.  11 ) [ 64 , 82 ]. However, we believe that, as indicated by Kalis et al. [ 90 ] on the pages of Harvard Business Review, the management of costly back-office problems should also be addressed.

The potential of algorithms includes data analysis. There is an immense quantity of data accessible now, which carries the possibility of providing information about a wide variety of medical and healthcare activities [ 91 ]. With the advent of modern computational methods, computer learning and AI techniques, there are numerous possibilities [ 79 , 83 , 84 ]. For example, AI makes it easier to turn data into concrete and actionable observations to improve decision-making, deliver high-quality patient treatment, adapt to real-time emergencies, and save more lives on the clinical front. In addition, AI makes it easier to leverage capital to develop systems and facilities and reduce expenses at the organisational level [ 78 ]. Studying contributions to the topic, we noticed that data accuracy was included in the debate, indicating that a high standard of data will benefit decision-making practitioners [ 38 , 77 ]. AI techniques are an essential instrument for studying data and the extraction of medical insight, and they may assist medical researchers in their practices. Using computational tools, healthcare stakeholders may leverage the power of data not only to evaluate past data ( descriptive analytics ) but also to forecast potential outcomes ( predictive analytics ) and to define the best actions for the present scenario ( prescriptive analytics ) [ 78 ]. The current abundance of evidence makes it easier to provide a broad view of patient health; doctors should have access to the correct details at the right time and location to provide the proper treatment [ 92 ].

Will medical technology de-skill doctors?

Further reflection concerns the skills of doctors. Studies have shown that healthcare personnel are progressively being exposed to technology for different purposes, such as collecting patient records or diagnosis [ 71 ]. This is demonstrated by the keywords (Fig.  6 ) that focus on technology and the role of decision-making with new innovative tools. In addition, the discussion expands with Lu [ 93 ], which indicates that the excessive use of technology could hinder doctors’ skills and clinical procedures' expansion. Among the main issues arising from the literature is the possible de-skilling of healthcare staff due to reduced autonomy in decision-making concerning patients [ 94 ]. Therefore, the challenges and discussion we uncovered in Fig.  11 are expanded by also considering the ethical implications of technology and the role of skills.

Implications

Our analysis also has multiple theoretical and practical implications.

In terms of theoretical contribution, this paper extends the previous results of Connelly et al., dos Santos et al, Hao et al., Huang et al., Liao et al. and Tran et al. [ 2 , 19 , 20 , 21 , 22 , 24 ] in considering AI in terms of clinical decision-making and data management quality.

In terms of practical implications, this paper aims to create a fruitful discussion with healthcare professionals and administrative staff on how AI can be at their service to increase work quality. Furthermore, this investigation offers a broad comprehension of bibliometric variables of AI techniques in healthcare. It can contribute to advancing scientific research in this field.

Limitations

Like any other, our study has some limitations that could be addressed by more in-depth future studies. For example, using only one research database, such as Scopus, could be limiting. Further analysis could also investigate the PubMed, IEEE, and Web of Science databases individually and holistically, especially the health parts. Then, the use of search terms such as "Artificial Intelligence" OR "AI" AND "Healthcare" could be too general and exclude interesting studies. Moreover, although we analysed 288 peer-reviewed scientific papers, because the new research topic is new, the analysis of conference papers could return interesting results for future researchers. Additionally, as this is a young research area, the analysis will be subject to recurrent obsolescence as multiple new research investigations are published. Finally, although bibliometric analysis has limited the subjectivity of the analysis [ 15 ], the verification of recurring themes could lead to different results by indicating areas of significant interest not listed here.

Future research avenues

Concerning future research perspectives, researchers believe that an analysis of the overall amount that a healthcare organisation should pay for AI technologies could be helpful. If these technologies are essential for health services management and patient treatment, governments should invest and contribute to healthcare organisations' modernisation. New investment funds could be made available in the healthcare world, as in the European case with the Next Generation EU programme or national investment programmes [ 95 ]. Additionally, this should happen especially in the poorest countries around the world, where there is a lack of infrastructure and services related to health and medicine [ 96 ]. On the other hand, it might be interesting to evaluate additional profits generated by healthcare organisations with AI technologies compared to those that do not use such technologies.

Further analysis could also identify why some parts of the world have not conducted studies in this area. It would be helpful to carry out a comparative analysis between countries active in this research field and countries that are not currently involved. It would make it possible to identify variables affecting AI technologies' presence or absence in healthcare organisations. The results of collaboration between countries also present future researchers with the challenge of greater exchanges between researchers and professionals. Therefore, further research could investigate the difference in vision between professionals and academics.

In the accounting, business, and management research area, there is currently a lack of quantitative analysis of the costs and profits generated by healthcare organisations that use AI technologies. Therefore, research in this direction could further increase our understanding of the topic and the number of healthcare organisations that can access technologies based on AI. Finally, as suggested in the discussion section, more interdisciplinary studies are needed to strengthen AI links with data quality management and AI and ethics considerations in healthcare.

In pursuing the philosophy of Massaro et al.’s [ 11 ] methodological article, we have climbed on the shoulders of giants, hoping to provide a bird's-eye view of the AI literature in healthcare. We performed this study with a bibliometric analysis aimed at discovering authors, countries of publication and collaboration, and keywords and themes. We found a fast-growing, multi-disciplinary stream of research that is attracting an increasing number of authors.

The research, therefore, adopts a quantitative approach to the analysis of bibliometric variables and a qualitative approach to the study of recurring keywords, which has allowed us to demonstrate strands of literature that are not purely positive. There are currently some limitations that will affect future research potential, especially in ethics, data governance and the competencies of the health workforce.

Availability of data and materials

All the data are retrieved from public scientific platforms.

Tagliaferri SD, Angelova M, Zhao X, Owen PJ, Miller CT, Wilkin T, et al. Artificial intelligence to improve back pain outcomes and lessons learnt from clinical classification approaches: three systematic reviews. NPJ Digit Med. 2020;3(1):1–16.

Article   Google Scholar  

Tran BX, Vu GT, Ha GH, Vuong Q-H, Ho M-T, Vuong T-T, et al. Global evolution of research in artificial intelligence in health and medicine: a bibliometric study. J Clin Med. 2019;8(3):360.

Article   PubMed Central   Google Scholar  

Hamid S. The opportunities and risks of artificial intelligence in medicine and healthcare [Internet]. 2016 [cited 2020 May 29]. http://www.cuspe.org/wp-content/uploads/2016/09/Hamid_2016.pdf

Panch T, Szolovits P, Atun R. Artificial intelligence, machine learning and health systems. J Glob Health. 2018;8(2):020303.

Article   PubMed   PubMed Central   Google Scholar  

Yang X, Wang Y, Byrne R, Schneider G, Yang S. Concepts of artificial intelligence for computer-assisted drug discovery | chemical reviews. Chem Rev. 2019;119(18):10520–94.

Article   CAS   PubMed   Google Scholar  

Burton RJ, Albur M, Eberl M, Cuff SM. Using artificial intelligence to reduce diagnostic workload without compromising detection of urinary tract infections. BMC Med Inform Decis Mak. 2019;19(1):171.

Meskò B, Drobni Z, Bényei E, Gergely B, Gyorffy Z. Digital health is a cultural transformation of traditional healthcare. Mhealth. 2017;3:38.

Cho B-J, Choi YJ, Lee M-J, Kim JH, Son G-H, Park S-H, et al. Classification of cervical neoplasms on colposcopic photography using deep learning. Sci Rep. 2020;10(1):13652.

Article   CAS   PubMed   PubMed Central   Google Scholar  

Doyle OM, Leavitt N, Rigg JA. Finding undiagnosed patients with hepatitis C infection: an application of artificial intelligence to patient claims data. Sci Rep. 2020;10(1):10521.

Shortliffe EH, Sepúlveda MJ. Clinical decision support in the era of artificial intelligence. JAMA. 2018;320(21):2199–200.

Article   PubMed   Google Scholar  

Massaro M, Dumay J, Guthrie J. On the shoulders of giants: undertaking a structured literature review in accounting. Account Auditing Account J. 2016;29(5):767–801.

Junquera B, Mitre M. Value of bibliometric analysis for research policy: a case study of Spanish research into innovation and technology management. Scientometrics. 2007;71(3):443–54.

Casadesus-Masanell R, Ricart JE. How to design a winning business model. Harvard Business Review [Internet]. 2011 Jan 1 [cited 2020 Jan 8]. https://hbr.org/2011/01/how-to-design-a-winning-business-model

Aria M, Cuccurullo C. bibliometrix: an R-tool for comprehensive science mapping analysis. J Informetr. 2017;11(4):959–75.

Zupic I, Čater T. Bibliometric methods in management and organization. Organ Res Methods. 2015;1(18):429–72.

Secinaro S, Calandra D. Halal food: structured literature review and research agenda. Br Food J. 2020. https://doi.org/10.1108/BFJ-03-2020-0234 .

Rialp A, Merigó JM, Cancino CA, Urbano D. Twenty-five years (1992–2016) of the international business review: a bibliometric overview. Int Bus Rev. 2019;28(6):101587.

Zhao L, Dai T, Qiao Z, Sun P, Hao J, Yang Y. Application of artificial intelligence to wastewater treatment: a bibliometric analysis and systematic review of technology, economy, management, and wastewater reuse. Process Saf Environ Prot. 2020;1(133):169–82.

Article   CAS   Google Scholar  

Huang Y, Huang Q, Ali S, Zhai X, Bi X, Liu R. Rehabilitation using virtual reality technology: a bibliometric analysis, 1996–2015. Scientometrics. 2016;109(3):1547–59.

Hao T, Chen X, Li G, Yan J. A bibliometric analysis of text mining in medical research. Soft Comput. 2018;22(23):7875–92.

dos Santos BS, Steiner MTA, Fenerich AT, Lima RHP. Data mining and machine learning techniques applied to public health problems: a bibliometric analysis from 2009 to 2018. Comput Ind Eng. 2019;1(138):106120.

Liao H, Tang M, Luo L, Li C, Chiclana F, Zeng X-J. A bibliometric analysis and visualization of medical big data research. Sustainability. 2018;10(1):166.

Choudhury A, Renjilian E, Asan O. Use of machine learning in geriatric clinical care for chronic diseases: a systematic literature review. JAMIA Open. 2020;3(3):459–71.

Connelly TM, Malik Z, Sehgal R, Byrnes G, Coffey JC, Peirce C. The 100 most influential manuscripts in robotic surgery: a bibliometric analysis. J Robot Surg. 2020;14(1):155–65.

Guo Y, Hao Z, Zhao S, Gong J, Yang F. Artificial intelligence in health care: bibliometric analysis. J Med Internet Res. 2020;22(7):e18228.

Choudhury A, Asan O. Role of artificial intelligence in patient safety outcomes: systematic literature review. JMIR Med Inform. 2020;8(7):e18599.

Forliano C, De Bernardi P, Yahiaoui D. Entrepreneurial universities: a bibliometric analysis within the business and management domains. Technol Forecast Soc Change. 2021;1(165):120522.

Secundo G, Del Vecchio P, Mele G. Social media for entrepreneurship: myth or reality? A structured literature review and a future research agenda. Int J Entrep Behav Res. 2020;27(1):149–77.

Dal Mas F, Massaro M, Lombardi R, Garlatti A. From output to outcome measures in the public sector: a structured literature review. Int J Organ Anal. 2019;27(5):1631–56.

Google Scholar  

Baima G, Forliano C, Santoro G, Vrontis D. Intellectual capital and business model: a systematic literature review to explore their linkages. J Intellect Cap. 2020. https://doi.org/10.1108/JIC-02-2020-0055 .

Dumay J, Guthrie J, Puntillo P. IC and public sector: a structured literature review. J Intellect Cap. 2015;16(2):267–84.

Dal Mas F, Garcia-Perez A, Sousa MJ, Lopes da Costa R, Cobianchi L. Knowledge translation in the healthcare sector. A structured literature review. Electron J Knowl Manag. 2020;18(3):198–211.

Mas FD, Massaro M, Lombardi R, Biancuzzi H. La performance nel settore pubblico tra misure di out-put e di outcome. Una revisione strutturata della letteratura ejvcbp. 2020;1(3):16–29.

Dumay J, Cai L. A review and critique of content analysis as a methodology for inquiring into IC disclosure. J Intellect Cap. 2014;15(2):264–90.

Haleem A, Javaid M, Khan IH. Current status and applications of Artificial Intelligence (AI) in medical field: an overview. Curr Med Res Pract. 2019;9(6):231–7.

Paul J, Criado AR. The art of writing literature review: what do we know and what do we need to know? Int Bus Rev. 2020;29(4):101717.

Liberati A, Altman DG, Tetzlaff J, Mulrow C, Gøtzsche PC, Ioannidis JPA, et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: explanation and elaboration. PLoS Med. 2009;6(7):e1000100.

Biancone PP, Secinaro S, Brescia V, Calandra D. Data quality methods and applications in health care system: a systematic literature review. Int J Bus Manag. 2019;14(4):p35.

Secinaro S, Brescia V, Calandra D, Verardi GP, Bert F. The use of micafungin in neonates and children: a systematic review. ejvcbp. 2020;1(1):100–14.

Bert F, Gualano MR, Biancone P, Brescia V, Camussi E, Martorana M, et al. HIV screening in pregnant women: a systematic review of cost-effectiveness studies. Int J Health Plann Manag. 2018;33(1):31–50.

Levy Y, Ellis TJ. A systems approach to conduct an effective literature review in support of information systems research. Inf Sci Int J Emerg Transdiscipl. 2006;9:181–212.

Chen G, Xiao L. Selecting publication keywords for domain analysis in bibliometrics: a comparison of three methods. J Informet. 2016;10(1):212–23.

Falagas ME, Pitsouni EI, Malietzis GA, Pappas G. Comparison of PubMed, Scopus, Web of Science, and Google Scholar: strengths and weaknesses. FASEB J. 2007;22(2):338–42.

Article   PubMed   CAS   Google Scholar  

Sicilia M-A, Garcìa-Barriocanal E, Sànchez-Alonso S. Community curation in open dataset repositories: insights from zenodo. Procedia Comput Sci. 2017;1(106):54–60.

Secinaro S, Calandra D, Secinaro A, Muthurangu V, Biancone P. Artificial Intelligence for healthcare with a business, management and accounting, decision sciences, and health professions focus [Internet]. Zenodo; 2021 [cited 2021 Mar 7]. https://zenodo.org/record/4587618#.YEScpl1KiWh .

Elango B, Rajendran D. Authorship trends and collaboration pattern in the marine sciences literature: a scientometric Study. Int J Inf Dissem Technol. 2012;1(2):166–9.

Jacoby WG. Electoral inquiry section Loess: a nonparametric, graphical tool for depicting relationships between variables q. In 2000.

Andrews JE. An author co-citation analysis of medical informatics. J Med Libr Assoc. 2003;91(1):47–56.

PubMed   PubMed Central   Google Scholar  

White HD, Griffith BC. Author cocitation: a literature measure of intellectual structure. J Am Soc Inf Sci. 1981;32(3):163–71.

Santosh KC. AI-driven tools for coronavirus outbreak: need of active learning and cross-population train/test models on multitudinal/multimodal data. J Med Syst. 2020;44(5):93.

Shickel B, Tighe PJ, Bihorac A, Rashidi P. Deep EHR: a survey of recent advances in deep learning techniques for electronic health record (EHR) analysis. IEEE J Biomed Health Inform. 2018;22(5):1589–604.

Baig MM, GholamHosseini H, Moqeem AA, Mirza F, Lindén M. A systematic review of wearable patient monitoring systems—current challenges and opportunities for clinical adoption. J Med Syst. 2017;41(7):115.

Kumar S, Kumar S. Collaboration in research productivity in oil seed research institutes of India. In: Proceedings of fourth international conference on webometrics, informetrics and scientometrics. p. 28–1; 2008.

Gatto A, Drago C. A taxonomy of energy resilience. Energy Policy. 2020;136:111007.

Levitt JM, Thelwall M. Alphabetization and the skewing of first authorship towards last names early in the alphabet. J Informet. 2013;7(3):575–82.

Saad G. Exploring the h-index at the author and journal levels using bibliometric data of productive consumer scholars and business-related journals respectively. Scientometrics. 2006;69(1):117–20.

Egghe L. Theory and practise of the g-index. Scientometrics. 2006;69(1):131–52.

Schreiber M. A modification of the h-index: the hm-index accounts for multi-authored manuscripts. J Informet. 2008;2(3):211–6.

Engqvist L, Frommen JG. The h-index and self-citations. Trends Ecol Evol. 2008;23(5):250–2.

London School of Economics. 3: key measures of academic influence [Internet]. Impact of social sciences. 2010 [cited 2021 Jan 13]. https://blogs.lse.ac.uk/impactofsocialsciences/the-handbook/chapter-3-key-measures-of-academic-influence/ .

Lotka A. The frequency distribution of scientific productivity. J Wash Acad Sci. 1926;16(12):317–24.

Khan G, Wood J. Information technology management domain: emerging themes and keyword analysis. Scientometrics. 2015;9:105.

Oxford University Press. Oxford English Dictionary [Internet]. 2020. https://www.oed.com/ .

Jiang F, Jiang Y, Zhi H, Dong Y, Li H, Ma S, et al. Artificial intelligence in healthcare: past, present and future. Stroke Vasc Neurol. 2017;2(4):230–43.

Calandra D, Favareto M. Artificial Intelligence to fight COVID-19 outbreak impact: an overview. Eur J Soc Impact Circ Econ. 2020;1(3):84–104.

Bokolo Anthony Jnr. Use of telemedicine and virtual care for remote treatment in response to COVID-19 pandemic. J Med Syst. 2020;44(7):132.

Burke EK, De Causmaecker P, Berghe GV, Van Landeghem H. The state of the art of nurse rostering. J Sched. 2004;7(6):441–99.

Ahmed MA, Alkhamis TM. Simulation optimization for an emergency department healthcare unit in Kuwait. Eur J Oper Res. 2009;198(3):936–42.

Forina M, Armanino C, Raggio V. Clustering with dendrograms on interpretation variables. Anal Chim Acta. 2002;454(1):13–9.

Wartena C, Brussee R. Topic detection by clustering keywords. In: 2008 19th international workshop on database and expert systems applications. 2008. p. 54–8.

Hussain AA, Bouachir O, Al-Turjman F, Aloqaily M. AI Techniques for COVID-19. IEEE Access. 2020;8:128776–95.

Agrawal A, Gans JS, Goldfarb A. Exploring the impact of artificial intelligence: prediction versus judgment. Inf Econ Policy. 2019;1(47):1–6.

Chakradhar S. Predictable response: finding optimal drugs and doses using artificial intelligence. Nat Med. 2017;23(11):1244–7.

Fleming N. How artificial intelligence is changing drug discovery. Nature. 2018;557(7707):S55–7.

Guo J, Li B. The application of medical artificial intelligence technology in rural areas of developing countries. Health Equity. 2018;2(1):174–81.

Aisyah M, Cockcroft S. A snapshot of data quality issues in Indonesian community health. Int J Netw Virtual Organ. 2014;14(3):280–97.

Davenport T, Kalakota R. The potential for artificial intelligence in healthcare. Future Healthc J. 2019;6(2):94–8.

Mehta N, Pandit A, Shukla S. Transforming healthcare with big data analytics and artificial intelligence: a systematic mapping study. J Biomed Inform. 2019;1(100):103311.

Collins GS, Moons KGM. Reporting of artificial intelligence prediction models. Lancet. 2019;393(10181):1577–9.

Bennett CC, Hauser K. Artificial intelligence framework for simulating clinical decision-making: a Markov decision process approach. Artif Intell Med. 2013;57(1):9–19.

Redondo T, Sandoval AM. Text Analytics: the convergence of big data and artificial intelligence. Int J Interact Multimed Artif Intell. 2016;3. https://www.ijimai.org/journal/bibcite/reference/2540 .

Winter JS, Davidson E. Big data governance of personal health information and challenges to contextual integrity. Inf Soc. 2019;35(1):36–51.

Novak D, Riener R. Control strategies and artificial intelligence in rehabilitation robotics. AI Mag. 2015;36(4):23–33.

Tarassoli SP. Artificial intelligence, regenerative surgery, robotics? What is realistic for the future of surgery? Ann Med Surg (Lond). 2019;17(41):53–5.

Saha SK, Fernando B, Cuadros J, Xiao D, Kanagasingam Y. Automated quality assessment of colour fundus images for diabetic retinopathy screening in telemedicine. J Digit Imaging. 2018;31(6):869–78.

Gu D, Li T, Wang X, Yang X, Yu Z. Visualizing the intellectual structure and evolution of electronic health and telemedicine research. Int J Med Inform. 2019;130:103947.

Madnick S, Wang R, Lee Y, Zhu H. Overview and framework for data and information quality research. J Data Inf Qual. 2009;1:1.

Chen X, Liu Z, Wei L, Yan J, Hao T, Ding R. A comparative quantitative study of utilizing artificial intelligence on electronic health records in the USA and China during 2008–2017. BMC Med Inform Decis Mak. 2018;18(5):117.

Carter D. How real is the impact of artificial intelligence? Bus Inf Surv. 2018;35(3):99–115.

Kalis B, Collier M, Fu R. 10 Promising AI Applications in Health Care. 2018;5.

Biancone P, Secinaro S, Brescia V, Calandra D. Management of open innovation in healthcare for cost accounting using EHR. J Open Innov Technol Market Complex. 2019;5(4):99.

Kayyali B, Knott D, Van Kuiken S. The ‘big data’ revolution in US healthcare [Internet]. McKinsey & Company. 2013 [cited 2020 Aug 14]. https://healthcare.mckinsey.com/big-data-revolution-us-healthcare/ .

Lu J. Will medical technology deskill doctors? Int Educ Stud. 2016;9(7):130–4.

Hoff T. Deskilling and adaptation among primary care physicians using two work innovations. Health Care Manag Rev. 2011;36(4):338–48.

Picek O. Spillover effects from next generation EU. Intereconomics. 2020;55(5):325–31.

Sousa MJ, Dal Mas F, Pesqueira A, Lemos C, Verde JM, Cobianchi L. The potential of AI in health higher education to increase the students’ learning outcomes. TEM J. 2021. ( In press ).

Download references

Acknowledgements

The authors are grateful to the Editor-in-Chief for the suggestions and all the reviewers who spend a part of their time ensuring constructive feedback to our research article.

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Author information

Authors and affiliations.

Department of Management, University of Turin, Turin, Italy

Silvana Secinaro, Davide Calandra & Paolo Biancone

Ospedale Pediatrico Bambino Gesù, Rome, Italy

Aurelio Secinaro

Institute of Child Health, University College London, London, UK

Vivek Muthurangu

You can also search for this author in PubMed   Google Scholar

Contributions

SS and PB, Supervision; Validation, writing, AS and VM; Formal analysis, DC and AS; Methodology, DC; Writing; DC, SS and AS; conceptualization, VM, PB; validation, VM, PB. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Davide Calandra .

Ethics declarations

Ethical approval and consent to participate.

Not applicable.

Consent for publication

Competing interests.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ . The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/ ) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Cite this article.

Secinaro, S., Calandra, D., Secinaro, A. et al. The role of artificial intelligence in healthcare: a structured literature review. BMC Med Inform Decis Mak 21 , 125 (2021). https://doi.org/10.1186/s12911-021-01488-9

Download citation

Received : 24 December 2020

Accepted : 01 April 2021

Published : 10 April 2021

DOI : https://doi.org/10.1186/s12911-021-01488-9

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Artificial intelligence
• Patient data

BMC Medical Informatics and Decision Making

ISSN: 1472-6947

• General enquiries: [email protected]

Information

• Author Services

Initiatives

You are accessing a machine-readable page. In order to be human-readable, please install an RSS reader.

All articles published by MDPI are made immediately available worldwide under an open access license. No special permission is required to reuse all or part of the article published by MDPI, including figures and tables. For articles published under an open access Creative Common CC BY license, any part of the article may be reused without permission provided that the original article is clearly cited. For more information, please refer to https://www.mdpi.com/openaccess .

Feature papers represent the most advanced research with significant potential for high impact in the field. A Feature Paper should be a substantial original Article that involves several techniques or approaches, provides an outlook for future research directions and describes possible research applications.

Feature papers are submitted upon individual invitation or recommendation by the scientific editors and must receive positive feedback from the reviewers.

Editor’s Choice articles are based on recommendations by the scientific editors of MDPI journals from around the world. Editors select a small number of articles recently published in the journal that they believe will be particularly interesting to readers, or important in the respective research area. The aim is to provide a snapshot of some of the most exciting work published in the various research areas of the journal.

Original Submission Date Received: .

• Active Journals
• Find a Journal
• Proceedings Series
• For Authors
• For Reviewers
• For Editors
• For Librarians
• For Publishers
• For Societies
• For Conference Organizers
• Open Access Policy
• Institutional Open Access Program
• Special Issues Guidelines
• Editorial Process
• Research and Publication Ethics
• Article Processing Charges
• Testimonials
• Preprints.org
• SciProfiles
• Encyclopedia

Journal Description

• Open Access — free for readers, with article processing charges (APC) paid by authors or their institutions.
• High Visibility:  indexed within  Scopus ,  SCIE and SSCI (Web of Science) , PubMed , PMC , and other databases .
• Journal Rank: JCR - Q2 ( Health Care Sciences and Services ) / CiteScore - Q2 ( Leadership and Management )
• Rapid Publication: manuscripts are peer-reviewed and a first decision is provided to authors approximately 20.5 days after submission; acceptance to publication is undertaken in 2.7 days (median values for papers published in this journal in the first half of 2024).
• Recognition of Reviewers: reviewers who provide timely, thorough peer-review reports receive vouchers entitling them to a discount on the APC of their next publication in any MDPI journal, in appreciation of the work done.
• Companion journals for Healthcare include:  Trauma Care and European Burn Journal .

Latest Articles

Journal Menu

• Healthcare Home
• Aims & Scope
• Editorial Board
• Reviewer Board
• Topical Advisory Panel
• Instructions for Authors

Special Issues

• Sections & Collections
• Article Processing Charge
• Indexing & Archiving
• Editor’s Choice Articles
• Most Cited & Viewed
• Journal Statistics
• Journal History
• Journal Awards
• Society Collaborations

Conferences

• Editorial Office

Journal Browser

• arrow_forward_ios Forthcoming issue arrow_forward_ios Current issue
• Vol. 12 (2024)
• Vol. 11 (2023)
• Vol. 10 (2022)
• Vol. 9 (2021)
• Vol. 8 (2020)
• Vol. 7 (2019)
• Vol. 6 (2018)
• Vol. 5 (2017)
• Vol. 4 (2016)
• Vol. 3 (2015)
• Vol. 2 (2014)
• Vol. 1 (2013)

Highly Accessed Articles

Latest books, e-mail alert.

Topical Collections

Further information, mdpi initiatives, follow mdpi.

Subscribe to receive issue release notifications and newsletters from MDPI journals

Research articles

Trends in cardiovascular disease incidence among 22 million people in the uk over 20 years, colchicine in patients with acute ischaemic stroke or transient ischaemic attack, sglt-2 inhibitors, glp-1 receptor agonists, and dpp-4 inhibitors and risk of hyperkalemia among people with type 2 diabetes, nab-paclitaxel, cisplatin, and capecitabine versus cisplatin and gemcitabine as first line chemotherapy in patients with nasopharyngeal carcinoma, lateral episiotomy or in vacuum assisted delivery in nulliparous women, global burden of type 1 diabetes in adults aged 65 years and older, antiplatelet therapy after coronary artery bypass surgery, mailed feedback to primary care physicians on antibiotic prescribing, tislelizumab plus chemotherapy versus placebo plus chemotherapy as first line treatment for advanced gastric or gastro-oesophageal junction adenocarcinoma, epidural analgesia during labour and severe maternal morbidity, exposure to antibiotics during pregnancy or early infancy and risk of neurodevelopmental disorders, clinical and healthcare use outcomes after cessation of long term opioid treatment due to prescriber workforce exit, effect of the hpv vaccination programme on incidence of cervical cancer by socioeconomic deprivation in england, long acting progestogens vs combined oral contraceptive pill for preventing recurrence of endometriosis related pain, ultra-processed food consumption and all cause and cause specific mortality, comparative effectiveness of second line oral antidiabetic treatments among people with type 2 diabetes mellitus, efficacy of psilocybin for treating symptoms of depression, reverse total shoulder replacement versus anatomical total shoulder replacement for osteoarthritis, effect of combination treatment with glp-1 receptor agonists and sglt-2 inhibitors on incidence of cardiovascular and serious renal events, prenatal opioid exposure and risk of neuropsychiatric disorders in children, temporal trends in lifetime risks of atrial fibrillation and its complications, antipsychotic use in people with dementia, predicting the risks of kidney failure and death in adults with moderate to severe chronic kidney disease, impact of large scale, multicomponent intervention to reduce proton pump inhibitor overuse, esketamine after childbirth for mothers with prenatal depression, glucagon-like peptide 1 receptor agonist use and risk of thyroid cancer, use of progestogens and the risk of intracranial meningioma, delirium and incident dementia in hospital patients, derivation and external validation of a simple risk score for predicting severe acute kidney injury after intravenous cisplatin, quality and safety of artificial intelligence generated health information, large language models and the generation of health disinformation, 25 year trends in cancer incidence and mortality among adults in the uk, cervical pessary versus vaginal progesterone in women with a singleton pregnancy, comparison of prior authorization across insurers, diagnostic accuracy of magnetically guided capsule endoscopy with a detachable string for detecting oesophagogastric varices in adults with cirrhosis, ultra-processed food exposure and adverse health outcomes, added benefit and revenues of oncology drugs approved by the ema, exposure to air pollution and hospital admission for cardiovascular diseases, short term exposure to low level ambient fine particulate matter and natural cause, cardiovascular, and respiratory morbidity, optimal timing of influenza vaccination in young children, effect of exercise for depression, association of non-alcoholic fatty liver disease with cardiovascular disease and all cause death in patients with type 2 diabetes, duration of cpr and outcomes for adults with in-hospital cardiac arrest, clinical effectiveness of an online physical and mental health rehabilitation programme for post-covid-19 condition, atypia detected during breast screening and subsequent development of cancer, publishers’ and journals’ instructions to authors on use of generative ai in academic and scientific publishing, effectiveness of glp-1 receptor agonists on glycaemic control, body weight, and lipid profile for type 2 diabetes, neurological development in children born moderately or late preterm, invasive breast cancer and breast cancer death after non-screen detected ductal carcinoma in situ, all cause and cause specific mortality in obsessive-compulsive disorder, acute rehabilitation following traumatic anterior shoulder dislocation, perinatal depression and risk of mortality, undisclosed financial conflicts of interest in dsm-5-tr, effect of risk mitigation guidance opioid and stimulant dispensations on mortality and acute care visits, update to living systematic review on sars-cov-2 positivity in offspring and timing of mother-to-child transmission, perinatal depression and its health impact, christmas 2023: common healthcare related instruments subjected to magnetic attraction study, using autoregressive integrated moving average models for time series analysis of observational data, demand for morning after pill following new year holiday, christmas 2023: christmas recipes from the great british bake off, effect of a doctor working during the festive period on population health: experiment using doctor who episodes, christmas 2023: analysis of barbie medical and science career dolls, christmas 2023: effect of chair placement on physicians’ behavior and patients’ satisfaction, management of chronic pain secondary to temporomandibular disorders, christmas 2023: projecting complete redaction of clinical trial protocols, christmas 2023: a drug target for erectile dysfunction to help improve fertility, sexual activity, and wellbeing, christmas 2023: efficacy of cola ingestion for oesophageal food bolus impaction, conservative management versus laparoscopic cholecystectomy in adults with gallstone disease, social media use and health risk behaviours in young people, untreated cervical intraepithelial neoplasia grade 2 and cervical cancer, air pollution deaths attributable to fossil fuels, implementation of a high sensitivity cardiac troponin i assay and risk of myocardial infarction or death at five years, covid-19 vaccine effectiveness against post-covid-19 condition, association between patient-surgeon gender concordance and mortality after surgery, intravascular imaging guided versus coronary angiography guided percutaneous coronary intervention, treatment of lower urinary tract symptoms in men in primary care using a conservative intervention, autism intervention meta-analysis of early childhood studies, effectiveness of the live zoster vaccine during the 10 years following vaccination, effects of a multimodal intervention in primary care to reduce second line antibiotic prescriptions for urinary tract infections in women, pyrotinib versus placebo in combination with trastuzumab and docetaxel in patients with her2 positive metastatic breast cancer, association of dcis size and margin status with risk of developing breast cancer post-treatment, racial differences in low value care among older patients in the us, pharmaceutical industry payments and delivery of low value cancer drugs, rosuvastatin versus atorvastatin in adults with coronary artery disease, clinical effectiveness of septoplasty versus medical management for nasal airways obstruction, ultrasound guided lavage with corticosteroid injection versus sham lavage with and without corticosteroid injection for calcific tendinopathy of shoulder, early versus delayed antihypertensive treatment in patients with acute ischaemic stroke, mortality risks associated with floods in 761 communities worldwide, interactive effects of ambient fine particulate matter and ozone on daily mortality in 372 cities, association between changes in carbohydrate intake and long term weight changes, future-case control crossover analysis for adjusting bias in case crossover studies, association between recently raised anticholinergic burden and risk of acute cardiovascular events, suboptimal gestational weight gain and neonatal outcomes in low and middle income countries: individual participant data meta-analysis, efficacy and safety of an inactivated virus-particle vaccine for sars-cov-2, effect of invitation letter in language of origin on screening attendance: randomised controlled trial in breastscreen norway, visits by nurse practitioners and physician assistants in the usa, non-erosive gastro-oesophageal reflux disease and oesophageal adenocarcinoma, venous thromboembolism with use of hormonal contraception and nsaids, food additive emulsifiers and risk of cardiovascular disease, follow us on, content links.

• Collections
• Health in South Asia
• Women’s, children’s & adolescents’ health
• News and views
• BMJ Opinion
• Rapid responses
• Editorial staff
• BMJ in the USA
• BMJ in South Asia
• Submit your paper
• BMA members
• Subscribers
• Advertisers and sponsors

Explore BMJ

• Our company
• BMJ Careers
• BMJ Learning
• BMJ Masterclasses
• BMJ Journals
• BMJ Student
• Academic edition of The BMJ
• BMJ Best Practice
• The BMJ Awards
• Email alerts
• Activate subscription

Information

Loading metrics

Open Access

Peer-reviewed

Research Article

Assessing the impact of healthcare research: A systematic review of methodological frameworks

Roles Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Validation, Writing – original draft, Writing – review & editing

Affiliation Centre for Patient Reported Outcomes Research, Institute of Applied Health Research, College of Medical and Dental Sciences, University of Birmingham, Birmingham, United Kingdom

Roles Conceptualization, Formal analysis, Funding acquisition, Methodology, Project administration, Supervision, Validation, Writing – review & editing

* E-mail: [email protected]

Roles Data curation, Formal analysis, Methodology, Validation, Writing – review & editing

Roles Formal analysis, Methodology, Supervision, Validation, Writing – review & editing

• Samantha Cruz Rivera, 
• Derek G. Kyte, 
• Olalekan Lee Aiyegbusi, 
• Thomas J. Keeley, 
• Melanie J. Calvert

• Published: August 9, 2017
• https://doi.org/10.1371/journal.pmed.1002370
• Reader Comments

Increasingly, researchers need to demonstrate the impact of their research to their sponsors, funders, and fellow academics. However, the most appropriate way of measuring the impact of healthcare research is subject to debate. We aimed to identify the existing methodological frameworks used to measure healthcare research impact and to summarise the common themes and metrics in an impact matrix.

Methods and findings

Two independent investigators systematically searched the Medical Literature Analysis and Retrieval System Online (MEDLINE), the Excerpta Medica Database (EMBASE), the Cumulative Index to Nursing and Allied Health Literature (CINAHL+), the Health Management Information Consortium, and the Journal of Research Evaluation from inception until May 2017 for publications that presented a methodological framework for research impact. We then summarised the common concepts and themes across methodological frameworks and identified the metrics used to evaluate differing forms of impact. Twenty-four unique methodological frameworks were identified, addressing 5 broad categories of impact: (1) ‘primary research-related impact’, (2) ‘influence on policy making’, (3) ‘health and health systems impact’, (4) ‘health-related and societal impact’, and (5) ‘broader economic impact’. These categories were subdivided into 16 common impact subgroups. Authors of the included publications proposed 80 different metrics aimed at measuring impact in these areas. The main limitation of the study was the potential exclusion of relevant articles, as a consequence of the poor indexing of the databases searched.

Conclusions

The measurement of research impact is an essential exercise to help direct the allocation of limited research resources, to maximise research benefit, and to help minimise research waste. This review provides a collective summary of existing methodological frameworks for research impact, which funders may use to inform the measurement of research impact and researchers may use to inform study design decisions aimed at maximising the short-, medium-, and long-term impact of their research.

Author summary

Why was this study done.

• There is a growing interest in demonstrating the impact of research in order to minimise research waste, allocate resources efficiently, and maximise the benefit of research. However, there is no consensus on which is the most appropriate tool to measure the impact of research.
• To our knowledge, this review is the first to synthesise existing methodological frameworks for healthcare research impact, and the associated impact metrics by which various authors have proposed impact should be measured, into a unified matrix.

What did the researchers do and find?

• We conducted a systematic review identifying 24 existing methodological research impact frameworks.
• We scrutinised the sample, identifying and summarising 5 proposed impact categories, 16 impact subcategories, and over 80 metrics into an impact matrix and methodological framework.

What do these findings mean?

• This simplified consolidated methodological framework will help researchers to understand how a research study may give rise to differing forms of impact, as well as in what ways and at which time points these potential impacts might be measured.
• Incorporating these insights into the design of a study could enhance impact, optimizing the use of research resources.

Citation: Cruz Rivera S, Kyte DG, Aiyegbusi OL, Keeley TJ, Calvert MJ (2017) Assessing the impact of healthcare research: A systematic review of methodological frameworks. PLoS Med 14(8): e1002370. https://doi.org/10.1371/journal.pmed.1002370

Academic Editor: Mike Clarke, Queens University Belfast, UNITED KINGDOM

Received: February 28, 2017; Accepted: July 7, 2017; Published: August 9, 2017

Copyright: © 2017 Cruz Rivera et al. This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: All relevant data are within the paper and supporting files.

Funding: Funding was received from Consejo Nacional de Ciencia y Tecnología (CONACYT). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript ( http://www.conacyt.mx/ ).

Competing interests: I have read the journal's policy and the authors of this manuscript have the following competing interests: MJC has received consultancy fees from Astellas and Ferring pharma and travel fees from the European Society of Cardiology outside the submitted work. TJK is in full-time paid employment for PAREXEL International.

Abbreviations: AIHS, Alberta Innovates—Health Solutions; CAHS, Canadian Academy of Health Sciences; CIHR, Canadian Institutes of Health Research; CINAHL+, Cumulative Index to Nursing and Allied Health Literature; EMBASE, Excerpta Medica Database; ERA, Excellence in Research for Australia; HEFCE, Higher Education Funding Council for England; HMIC, Health Management Information Consortium; HTA, Health Technology Assessment; IOM, Impact Oriented Monitoring; MDG, Millennium Development Goal; NHS, National Health Service; MEDLINE, Medical Literature Analysis and Retrieval System Online; PHC RIS, Primary Health Care Research & Information Service; PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses; PROM, patient-reported outcome measures; QALY, quality-adjusted life year; R&D, research and development; RAE, Research Assessment Exercise; REF, Research Excellence Framework; RIF, Research Impact Framework; RQF, Research Quality Framework; SDG, Sustainable Development Goal; SIAMPI, Social Impact Assessment Methods for research and funding instruments through the study of Productive Interactions between science and society

Introduction

In 2010, approximately US$240 billion was invested in healthcare research worldwide [ 1 ]. Such research is utilised by policy makers, healthcare providers, and clinicians to make important evidence-based decisions aimed at maximising patient benefit, whilst ensuring that limited healthcare resources are used as efficiently as possible to facilitate effective and sustainable service delivery. It is therefore essential that this research is of high quality and that it is impactful—i.e., it delivers demonstrable benefits to society and the wider economy whilst minimising research waste [ 1 , 2 ]. Research impact can be defined as ‘any identifiable ‘benefit to, or positive influence on the economy, society, public policy or services, health, the environment, quality of life or academia’ (p. 26) [ 3 ].

There are many purported benefits associated with the measurement of research impact, including the ability to (1) assess the quality of the research and its subsequent benefits to society; (2) inform and influence optimal policy and funding allocation; (3) demonstrate accountability, the value of research in terms of efficiency and effectiveness to the government, stakeholders, and society; and (4) maximise impact through better understanding the concept and pathways to impact [ 4 – 7 ].

Measuring and monitoring the impact of healthcare research has become increasingly common in the United Kingdom [ 5 ], Australia [ 5 ], and Canada [ 8 ], as governments, organisations, and higher education institutions seek a framework to allocate funds to projects that are more likely to bring the most benefit to society and the economy [ 5 ]. For example, in the UK, the 2014 Research Excellence Framework (REF) has recently been used to assess the quality and impact of research in higher education institutions, through the assessment of impact cases studies and selected qualitative impact metrics [ 9 ]. This is the first initiative to allocate research funding based on the economic, societal, and cultural impact of research, although it should be noted that research impact only drives a proportion of this allocation (approximately 20%) [ 9 ].

In the UK REF, the measurement of research impact is seen as increasingly important. However, the impact element of the REF has been criticised in some quarters [ 10 , 11 ]. Critics deride the fact that REF impact is determined in a relatively simplistic way, utilising researcher-generated case studies, which commonly attempt to link a particular research outcome to an associated policy or health improvement despite the fact that the wider literature highlights great diversity in the way research impact may be demonstrated [ 12 , 13 ]. This led to the current debate about the optimal method of measuring impact in the future REF [ 10 , 14 ]. The Stern review suggested that research impact should not only focus on socioeconomic impact but should also include impact on government policy, public engagement, academic impacts outside the field, and teaching to showcase interdisciplinary collaborative impact [ 10 , 11 ]. The Higher Education Funding Council for England (HEFCE) has recently set out the proposals for the REF 2021 exercise, confirming that the measurement of such impact will continue to form an important part of the process [ 15 ].

With increasing pressure for healthcare research to lead to demonstrable health, economic, and societal impact, there is a need for researchers to understand existing methodological impact frameworks and the means by which impact may be quantified (i.e., impact metrics; see Box 1 , 'Definitions’) to better inform research activities and funding decisions. From a researcher’s perspective, understanding the optimal pathways to impact can help inform study design aimed at maximising the impact of the project. At the same time, funders need to understand which aspects of impact they should focus on when allocating awards so they can make the most of their investment and bring the greatest benefit to patients and society [ 2 , 4 , 5 , 16 , 17 ].

Box 1. Definitions

• Research impact: ‘any identifiable benefit to, or positive influence on, the economy, society, public policy or services, health, the environment, quality of life, or academia’ (p. 26) [ 3 ].
• Methodological framework: ‘a body of methods, rules and postulates employed by a particular procedure or set of procedures (i.e., framework characteristics and development)’ [ 18 ].
• Pathway: ‘a way of achieving a specified result; a course of action’ [ 19 ].
• Quantitative metrics: ‘a system or standard of [quantitative] measurement’ [ 20 ].
• Narrative metrics: ‘a spoken or written account of connected events; a story’ [ 21 ].

Whilst previous researchers have summarised existing methodological frameworks and impact case studies [ 4 , 22 – 27 ], they have not summarised the metrics for use by researchers, funders, and policy makers. The aim of this review was therefore to (1) identify the methodological frameworks used to measure healthcare research impact using systematic methods, (2) summarise common impact themes and metrics in an impact matrix, and (3) provide a simplified consolidated resource for use by funders, researchers, and policy makers.

Search strategy and selection criteria

Initially, a search strategy was developed to identify the available literature regarding the different methods to measure research impact. The following keywords: ‘Impact’, ‘Framework’, and ‘Research’, and their synonyms, were used during the search of the Medical Literature Analysis and Retrieval System Online (MEDLINE; Ovid) database, the Excerpta Medica Database (EMBASE), the Health Management Information Consortium (HMIC) database, and the Cumulative Index to Nursing and Allied Health Literature (CINAHL+) database (inception to May 2017; see S1 Appendix for the full search strategy). Additionally, the nonindexed Journal of Research Evaluation was hand searched during the same timeframe using the keyword ‘Impact’. Other relevant articles were identified through 3 Internet search engines (Google, Google Scholar, and Google Images) using the keywords ‘Impact’, ‘Framework’, and ‘Research’, with the first 50 results screened. Google Images was searched because different methodological frameworks are summarised in a single image and can easily be identified through this search engine. Finally, additional publications were sought through communication with experts.

Following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (see S1 PRISMA Checklist ), 2 independent investigators systematically screened for publications describing, evaluating, or utilising a methodological research impact framework within the context of healthcare research [ 28 ]. Papers were eligible if they included full or partial methodological frameworks or pathways to research impact; both primary research and systematic reviews fitting these criteria were included. We included any methodological framework identified (original or modified versions) at the point of first occurrence. In addition, methodological frameworks were included if they were applicable to the healthcare discipline with no need of modification within their structure. We defined ‘methodological framework’ as ‘a body of methods, rules and postulates employed by a particular procedure or set of procedures (i.e., framework characteristics and development)’ [ 18 ], whereas we defined ‘pathway’ as ‘a way of achieving a specified result; a course of action’ [ 19 ]. Studies were excluded if they presented an existing (unmodified) methodological framework previously available elsewhere, did not explicitly describe a methodological framework but rather focused on a single metric (e.g., bibliometric analysis), focused on the impact or effectiveness of interventions rather than that of the research, or presented case study data only. There were no language restrictions.

Data screening

Records were downloaded into Endnote (version X7.3.1), and duplicates were removed. Two independent investigators (SCR and OLA) conducted all screening following a pilot aimed at refining the process. The records were screened by title and abstract before full-text articles of potentially eligible publications were retrieved for evaluation. A full-text screening identified the publications included for data extraction. Discrepancies were resolved through discussion, with the involvement of a third reviewer (MJC, DGK, and TJK) when necessary.

Data extraction and analysis

Data extraction occurred after the final selection of included articles. SCR and OLA independently extracted details of impact methodological frameworks, the country of origin, and the year of publication, as well as the source, the framework description, and the methodology used to develop the framework. Information regarding the methodology used to develop each methodological framework was also extracted from framework webpages where available. Investigators also extracted details regarding each framework’s impact categories and subgroups, along with their proposed time to impact (‘short-term’, ‘mid-term’, or ‘long-term’) and the details of any metrics that had been proposed to measure impact, which are depicted in an impact matrix. The structure of the matrix was informed by the work of M. Buxton and S. Hanney [ 2 ], P. Buykx et al. [ 5 ], S. Kuruvila et al. [ 29 ], and A. Weiss [ 30 ], with the intention of mapping metrics presented in previous methodological frameworks in a concise way. A consensus meeting with MJC, DGK, and TJK was held to solve disagreements and finalise the data extraction process.

Included studies

Our original search strategy identified 359 citations from MEDLINE (Ovid), EMBASE, CINAHL+, HMIC, and the Journal of Research Evaluation, and 101 citations were returned using other sources (Google, Google Images, Google Scholar, and expert communication) (see Fig 1 ) [ 28 ]. In total, we retrieved 54 full-text articles for review. At this stage, 39 articles were excluded, as they did not propose new or modified methodological frameworks. An additional 15 articles were included following the backward and forward citation method. A total of 31 relevant articles were included in the final analysis, of which 24 were articles presenting unique frameworks and the remaining 7 were systematic reviews [ 4 , 22 – 27 ]. The search strategy was rerun on 15 May 2017. A further 19 publications were screened, and 2 were taken forward to full-text screening but were ineligible for inclusion.

• PPT PowerPoint slide
• PNG larger image
• TIFF original image

https://doi.org/10.1371/journal.pmed.1002370.g001

Methodological framework characteristics

The characteristics of the 24 included methodological frameworks are summarised in Table 1 , 'Methodological framework characteristics’. Fourteen publications proposed academic-orientated frameworks, which focused on measuring academic, societal, economic, and cultural impact using narrative and quantitative metrics [ 2 , 3 , 5 , 8 , 29 , 31 – 39 ]. Five publications focused on assessing the impact of research by focusing on the interaction process between stakeholders and researchers (‘productive interactions’), which is a requirement to achieve research impact. This approach tries to address the issue of attributing research impact to metrics [ 7 , 40 – 43 ]. Two frameworks focused on the importance of partnerships between researchers and policy makers, as a core element to accomplish research impact [ 44 , 45 ]. An additional 2 frameworks focused on evaluating the pathways to impact, i.e., linking processes between research and impact [ 30 , 46 ]. One framework assessed the ability of health technology to influence efficiency of healthcare systems [ 47 ]. Eight frameworks were developed in the UK [ 2 , 3 , 29 , 37 , 39 , 42 , 43 , 45 ], 6 in Canada [ 8 , 33 , 34 , 44 , 46 , 47 ], 4 in Australia [ 5 , 31 , 35 , 38 ], 3 in the Netherlands [ 7 , 40 , 41 ], and 2 in the United States [ 30 , 36 ], with 1 model developed with input from various countries [ 32 ].

https://doi.org/10.1371/journal.pmed.1002370.t001

Methodological framework development

The included methodological frameworks varied in their development process, but there were some common approaches employed. Most included a literature review [ 2 , 5 , 7 , 8 , 31 , 33 , 36 , 37 , 40 – 46 ], although none of them used a recognised systematic method. Most also consulted with various stakeholders [ 3 , 8 , 29 , 31 , 33 , 35 – 38 , 43 , 44 , 46 , 47 ] but used differing methods to incorporate their views, including quantitative surveys [ 32 , 35 , 43 , 46 ], face-to-face interviews [ 7 , 29 , 33 , 35 , 37 , 42 , 43 ], telephone interviews [ 31 , 46 ], consultation [ 3 , 7 , 36 ], and focus groups [ 39 , 43 ]. A range of stakeholder groups were approached across the sample, including principal investigators [ 7 , 29 , 43 ], research end users [ 7 , 42 , 43 ], academics [ 3 , 8 , 39 , 40 , 43 , 46 ], award holders [ 43 ], experts [ 33 , 38 , 39 ], sponsors [ 33 , 39 ], project coordinators [ 32 , 42 ], and chief investigators [ 31 , 35 ]. However, some authors failed to identify the stakeholders involved in the development of their frameworks [ 2 , 5 , 34 , 41 , 45 ], making it difficult to assess their appropriateness. In addition, only 4 of the included papers reported using formal analytic methods to interpret stakeholder responses. These included the Canadian Academy of Health Sciences framework, which used conceptual cluster analysis [ 33 ]. The Research Contribution [ 42 ], Research Impact [ 29 ], and Primary Health Care & Information Service [ 31 ] used a thematic analysis approach. Finally, some authors went on to pilot their framework, which shaped refinements on the methodological frameworks until approval. Methods used to pilot the frameworks included a case study approach [ 2 , 3 , 30 , 32 , 33 , 36 , 40 , 42 , 44 , 45 ], contrasting results against available literature [ 29 ], the use of stakeholders’ feedback [ 7 ], and assessment tools [ 35 , 46 ].

Major impact categories

1. primary research-related impact..

A number of methodological frameworks advocated the evaluation of ‘research-related impact’. This encompassed content related to the generation of new knowledge, knowledge dissemination, capacity building, training, leadership, and the development of research networks. These outcomes were considered the direct or primary impacts of a research project, as these are often the first evidenced returns [ 30 , 62 ].

A number of subgroups were identified within this category, with frameworks supporting the collection of impact data across the following constructs: ‘research and innovation outcomes’; ‘dissemination and knowledge transfer’; ‘capacity building, training, and leadership’; and ‘academic collaborations, research networks, and data sharing’.

1 . 1 . Research and innovation outcomes . Twenty of the 24 frameworks advocated the evaluation of ‘research and innovation outcomes’ [ 2 , 3 , 5 , 7 , 8 , 29 – 39 , 41 , 43 , 44 , 46 ]. This subgroup included the following metrics: number of publications; number of peer-reviewed articles (including journal impact factor); citation rates; requests for reprints, number of reviews, and meta-analysis; and new or changes in existing products (interventions or technology), patents, and research. Additionally, some frameworks also sought to gather information regarding ‘methods/methodological contributions’. These advocated the collection of systematic reviews and appraisals in order to identify gaps in knowledge and determine whether the knowledge generated had been assessed before being put into practice [ 29 ].

1 . 2 . Dissemination and knowledge transfer . Nineteen of the 24 frameworks advocated the assessment of ‘dissemination and knowledge transfer’ [ 2 , 3 , 5 , 7 , 29 – 32 , 34 – 43 , 46 ]. This comprised collection of the following information: number of conferences, seminars, workshops, and presentations; teaching output (i.e., number of lectures given to disseminate the research findings); number of reads for published articles; article download rate and number of journal webpage visits; and citations rates in nonjournal media such as newspapers and mass and social media (i.e., Twitter and blogs). Furthermore, this impact subgroup considered the measurement of research uptake and translatability and the adoption of research findings in technological and clinical applications and by different fields. These can be measured through patents, clinical trials, and partnerships between industry and business, government and nongovernmental organisations, and university research units and researchers [ 29 ].

1 . 3 . Capacity building , training , and leadership . Fourteen of 24 frameworks suggested the evaluation of ‘capacity building, training, and leadership’ [ 2 , 3 , 5 , 8 , 29 , 31 – 35 , 39 – 41 , 43 ]. This involved collecting information regarding the number of doctoral and postdoctoral studentships (including those generated as a result of the research findings and those appointed to conduct the research), as well as the number of researchers and research-related staff involved in the research projects. In addition, authors advocated the collection of ‘leadership’ metrics, including the number of research projects managed and coordinated and the membership of boards and funding bodies, journal editorial boards, and advisory committees [ 29 ]. Additional metrics in this category included public recognition (number of fellowships and awards for significant research achievements), academic career advancement, and subsequent grants received. Lastly, the impact metric ‘research system management’ comprised the collection of information that can lead to preserving the health of the population, such as modifying research priorities, resource allocation strategies, and linking health research to other disciplines to maximise benefits [ 29 ].

1 . 4 . Academic collaborations , research networks , and data sharing . Lastly, 10 of the 24 frameworks advocated the collection of impact data regarding ‘academic collaborations (internal and external collaborations to complete a research project), research networks, and data sharing’ [ 2 , 3 , 5 , 7 , 29 , 34 , 37 , 39 , 41 , 43 ].

2. Influence on policy making.

Methodological frameworks addressing this major impact category focused on measurable improvements within a given knowledge base and on interactions between academics and policy makers, which may influence policy-making development and implementation. The returns generated in this impact category are generally considered as intermediate or midterm (1 to 3 years). These represent an important interim stage in the process towards the final expected impacts, such as quantifiable health improvements and economic benefits, without which policy change may not occur [ 30 , 62 ]. The following impact subgroups were identified within this category: ‘type and nature of policy impact’, ‘level of policy making’, and ‘policy networks’.

2 . 1 . Type and nature of policy impact . The most common impact subgroup, mentioned in 18 of the 24 frameworks, was ‘type and nature of policy impact’ [ 2 , 7 , 29 – 38 , 41 – 43 , 45 – 47 ]. Methodological frameworks addressing this subgroup stressed the importance of collecting information regarding the influence of research on policy (i.e., changes in practice or terminology). For instance, a project looking at trafficked adolescents and women (2003) influenced the WHO guidelines (2003) on ethics regarding this particular group [ 17 , 21 , 63 ].

2 . 2 . Level of policy impact . Thirteen of 24 frameworks addressed aspects surrounding the need to record the ‘level of policy impact’ (international, national, or local) and the organisations within a level that were influenced (local policy makers, clinical commissioning groups, and health and wellbeing trusts) [ 2 , 5 , 8 , 29 , 31 , 34 , 38 , 41 , 43 – 47 ]. Authors considered it important to measure the ‘level of policy impact’ to provide evidence of collaboration, coordination, and efficiency within health organisations and between researchers and health organisations [ 29 , 31 ].

2 . 3 . Policy networks . Five methodological frameworks highlighted the need to collect information regarding collaborative research with industry and staff movement between academia and industry [ 5 , 7 , 29 , 41 , 43 ]. A policy network emphasises the relationship between policy communities, researchers, and policy makers. This relationship can influence and lead to incremental changes in policy processes [ 62 ].

3. Health and health systems impact.

A number of methodological frameworks advocated the measurement of impacts on health and healthcare systems across the following impact subgroups: ‘quality of care and service delivering’, ‘evidence-based practice’, ‘improved information and health information management’, ‘cost containment and effectiveness’, ‘resource allocation’, and ‘health workforce’.

3 . 1 . Quality of care and service delivery . Twelve of the 24 frameworks highlighted the importance of evaluating ‘quality of care and service delivery’ [ 2 , 5 , 8 , 29 – 31 , 33 – 36 , 41 , 47 ]. There were a number of suggested metrics that could be potentially used for this purpose, including health outcomes such as quality-adjusted life years (QALYs), patient-reported outcome measures (PROMs), patient satisfaction and experience surveys, and qualitative data on waiting times and service accessibility.

3 . 2 . Evidence-based practice . ‘Evidence-based practice’, mentioned in 5 of the 24 frameworks, refers to making changes in clinical diagnosis, clinical practice, treatment decisions, or decision making based on research evidence [ 5 , 8 , 29 , 31 , 33 ]. The suggested metrics to demonstrate evidence-based practice were adoption of health technologies and research outcomes to improve the healthcare systems and inform policies and guidelines [ 29 ].

3 . 3 . Improved information and health information management . This impact subcategory, mentioned in 5 of the 24 frameworks, refers to the influence of research on the provision of health services and management of the health system to prevent additional costs [ 5 , 29 , 33 , 34 , 38 ]. Methodological frameworks advocated the collection of health system financial, nonfinancial (i.e., transport and sociopolitical implications), and insurance information in order to determine constraints within a health system.

3 . 4 . Cost containment and cost-effectiveness . Six of the 24 frameworks advocated the subcategory ‘cost containment and cost-effectiveness’ [ 2 , 5 , 8 , 17 , 33 , 36 ]. ‘Cost containment’ comprised the collection of information regarding how research has influenced the provision and management of health services and its implication in healthcare resource allocation and use [ 29 ]. ‘Cost-effectiveness’ refers to information concerning economic evaluations to assess improvements in effectiveness and health outcomes—for instance, the cost-effectiveness (cost and health outcome benefits) assessment of introducing a new health technology to replace an older one [ 29 , 31 , 64 ].

3 . 5 . Resource allocation . ‘Resource allocation’, mentioned in 6frameworks, can be measured through 2 impact metrics: new funding attributed to the intervention in question and equity while allocating resources, such as improved allocation of resources at an area level; better targeting, accessibility, and utilisation; and coverage of health services [ 2 , 5 , 29 , 31 , 45 , 47 ]. The allocation of resources and targeting can be measured through health services research reports, with the utilisation of health services measured by the probability of providing an intervention when needed, the probability of requiring it again in the future, and the probability of receiving an intervention based on previous experience [ 29 , 31 ].

3 . 6 . Health workforce . Lastly, ‘health workforce’, present in 3 methodological frameworks, refers to the reduction in the days of work lost because of a particular illness [ 2 , 5 , 31 ].

4. Health-related and societal impact.

Three subgroups were included in this category: ‘health literacy’; ‘health knowledge, attitudes, and behaviours’; and ‘improved social equity, inclusion, or cohesion’.

4 . 1 . Health knowledge , attitudes , and behaviours . Eight of the 24 frameworks suggested the assessment of ‘health knowledge, attitudes, behaviours, and outcomes’, which could be measured through the evaluation of levels of public engagement with science and research (e.g., National Health Service (NHS) Choices end-user visit rate) or by using focus groups to analyse changes in knowledge, attitudes, and behaviour among society [ 2 , 5 , 29 , 33 – 35 , 38 , 43 ].

4 . 2 . Improved equity , inclusion , or cohesion and human rights . Other methodological frameworks, 4 of the 24, suggested capturing improvements in equity, inclusion, or cohesion and human rights. Authors suggested these could be using a resource like the United Nations Millennium Development Goals (MDGs) (superseded by Sustainable Development Goals [SDGs] in 2015) and human rights [ 29 , 33 , 34 , 38 ]. For instance, a cluster-randomised controlled trial in Nepal, which had female participants, has demonstrated the reduction of neonatal mortality through the introduction of maternity health care, distribution of delivery kits, and home visits. This illustrates how research can target vulnerable and disadvantaged groups. Additionally, this research has been introduced by the World Health Organisation to achieve the MDG ‘improve maternal health’ [ 16 , 29 , 65 ].

4 . 3 . Health literacy . Some methodological frameworks, 3 of the 24, focused on tracking changes in the ability of patients to make informed healthcare decisions, reduce health risks, and improve quality of life, which were demonstrably linked to a particular programme of research [ 5 , 29 , 43 ]. For example, a systematic review showed that when HIV health literacy/knowledge is spread among people living with the condition, antiretroviral adherence and quality of life improve [ 66 ].

5. Broader economic impacts.

Some methodological frameworks, 9 of 24, included aspects related to the broader economic impacts of health research—for example, the economic benefits emerging from the commercialisation of research outputs [ 2 , 5 , 29 , 31 , 33 , 35 , 36 , 38 , 67 ]. Suggested metrics included the amount of funding for research and development (R&D) that was competitively awarded by the NHS, medical charities, and overseas companies. Additional metrics were income from intellectual property, spillover effects (any secondary benefit gained as a repercussion of investing directly in a primary activity, i.e., the social and economic returns of investing on R&D) [ 33 ], patents granted, licences awarded and brought to the market, the development and sales of spinout companies, research contracts, and income from industry.

The benefits contained within the categories ‘health and health systems impact’, ‘health-related and societal impact’, and ‘broader economic impacts’ are considered the expected and final returns of the resources allocated in healthcare research [ 30 , 62 ]. These benefits commonly arise in the long term, beyond 5 years according to some authors, but there was a recognition that this could differ depending on the project and its associated research area [ 4 ].

Data synthesis

Five major impact categories were identified across the 24 included methodological frameworks: (1) ‘primary research-related impact’, (2) ‘influence on policy making’, (3) ‘health and health systems impact’, (4) ‘health-related and societal impact’, and (5) ‘broader economic impact’. These major impact categories were further subdivided into 16 impact subgroups. The included publications proposed 80 different metrics to measure research impact. This impact typology synthesis is depicted in ‘the impact matrix’ ( Fig 2 and Fig 3 ).

CIHR, Canadian Institutes of Health Research; HTA, Health Technology Assessment; PHC RIS, Primary Health Care Research & Information Service; RAE, Research Assessment Exercise; RQF, Research Quality Framework.

https://doi.org/10.1371/journal.pmed.1002370.g002

AIHS, Alberta Innovates—Health Solutions; CAHS, Canadian Institutes of Health Research; IOM, Impact Oriented Monitoring; REF, Research Excellence Framework; SIAMPI, Social Impact Assessment Methods for research and funding instruments through the study of Productive Interactions between science and society.

https://doi.org/10.1371/journal.pmed.1002370.g003

Commonality and differences across frameworks

The ‘Research Impact Framework’ and the ‘Health Services Research Impact Framework’ were the models that encompassed the largest number of the metrics extracted. The most dominant methodological framework was the Payback Framework; 7 other methodological framework models used the Payback Framework as a starting point for development [ 8 , 29 , 31 – 35 ]. Additional methodological frameworks that were commonly incorporated into other tools included the CIHR framework, the CAHS model, the AIHS framework, and the Exchange model [ 8 , 33 , 34 , 44 ]. The capture of ‘research-related impact’ was the most widely advocated concept across methodological frameworks, illustrating the importance with which primary short-term impact outcomes were viewed by the included papers. Thus, measurement of impact via number of publications, citations, and peer-reviewed articles was the most common. ‘Influence on policy making’ was the predominant midterm impact category, specifically the subgroup ‘type and nature of policy impact’, in which frameworks advocated the measurement of (i) changes to legislation, regulations, and government policy; (ii) influence and involvement in decision-making processes; and (iii) changes to clinical or healthcare training, practice, or guidelines. Within more long-term impact measurement, the evaluations of changes in the ‘quality of care and service delivery’ were commonly advocated.

In light of the commonalities and differences among the methodological frameworks, the ‘pathways to research impact’ diagram ( Fig 4 ) was developed to provide researchers, funders, and policy makers a more comprehensive and exhaustive way to measure healthcare research impact. The diagram has the advantage of assorting all the impact metrics proposed by previous frameworks and grouping them into different impact subgroups and categories. Prospectively, this global picture will help researchers, funders, and policy makers plan strategies to achieve multiple pathways to impact before carrying the research out. The analysis of the data extraction and construction of the impact matrix led to the development of the ‘pathways to research impact’ diagram ( Fig 4 ). The diagram aims to provide an exhaustive and comprehensive way of tracing research impact by combining all the impact metrics presented by the different 24 frameworks, grouping those metrics into different impact subgroups, and grouping these into broader impact categories.

NHS, National Health Service; PROM, patient-reported outcome measure; QALY, quality-adjusted life year; R&D, research and development.

https://doi.org/10.1371/journal.pmed.1002370.g004

This review has summarised existing methodological impact frameworks together for the first time using systematic methods ( Fig 4 ). It allows researchers and funders to consider pathways to impact at the design stage of a study and to understand the elements and metrics that need to be considered to facilitate prospective assessment of impact. Users do not necessarily need to cover all the aspects of the methodological framework, as every research project can impact on different categories and subgroups. This review provides information that can assist researchers to better demonstrate impact, potentially increasing the likelihood of conducting impactful research and reducing research waste. Existing reviews have not presented a methodological framework that includes different pathways to impact, health impact categories, subgroups, and metrics in a single methodological framework.

Academic-orientated frameworks included in this review advocated the measurement of impact predominantly using so-called ‘quantitative’ metrics—for example, the number of peer-reviewed articles, journal impact factor, and citation rates. This may be because they are well-established measures, relatively easy to capture and objective, and are supported by research funding systems. However, these metrics primarily measure the dissemination of research finding rather than its impact [ 30 , 68 ]. Whilst it is true that wider dissemination, especially when delivered via world-leading international journals, may well lead eventually to changes in healthcare, this is by no means certain. For instance, case studies evaluated by Flinders University of Australia demonstrated that some research projects with non-peer-reviewed publications led to significant changes in health policy, whilst the studies with peer-reviewed publications did not result in any type of impact [ 68 ]. As a result, contemporary literature has tended to advocate the collection of information regarding a variety of different potential forms of impact alongside publication/citations metrics [ 2 , 3 , 5 , 7 , 8 , 29 – 47 ], as outlined in this review.

The 2014 REF exercise adjusted UK university research funding allocation based on evidence of the wider impact of research (through case narrative studies and quantitative metrics), rather than simply according to the quality of research [ 12 ]. The intention was to ensure funds were directed to high-quality research that could demonstrate actual realised benefit. The inclusion of a mixed-method approach to the measurement of impact in the REF (narrative and quantitative metrics) reflects a widespread belief—expressed by the majority of authors of the included methodological frameworks in the review—that individual quantitative impact metrics (e.g., number of citations and publications) do not necessary capture the complexity of the relationships involved in a research project and may exclude measurement of specific aspects of the research pathway [ 10 , 12 ].

Many of the frameworks included in this review advocated the collection of a range of academic, societal, economic, and cultural impact metrics; this is consistent with recent recommendations from the Stern review [ 10 ]. However, a number of these metrics encounter research ‘lag’: i.e., the time between the point at which the research is conducted and when the actual benefits arise [ 69 ]. For instance, some cardiovascular research has taken up to 25 years to generate impact [ 70 ]. Likewise, the impact may not arise exclusively from a single piece of research. Different processes (such as networking interactions and knowledge and research translation) and multiple individuals and organisations are often involved [ 4 , 71 ]. Therefore, attributing the contribution made by each of the different actors involved in the process can be a challenge [ 4 ]. An additional problem associated to attribution is the lack of evidence to link research and impact. The outcomes of research may emerge slowly and be absorbed gradually. Consequently, it is difficult to determine the influence of research in the development of a new policy, practice, or guidelines [ 4 , 23 ].

A further problem is that impact evaluation is conducted ‘ex post’, after the research has concluded. Collecting information retrospectively can be an issue, as the data required might not be available. ‘ex ante’ assessment is vital for funding allocation, as it is necessary to determine the potential forthcoming impact before research is carried out [ 69 ]. Additionally, ex ante evaluation of potential benefit can overcome the issues regarding identifying and capturing evidence, which can be used in the future [ 4 ]. In order to conduct ex ante evaluation of potential benefit, some authors suggest the early involvement of policy makers in a research project coupled with a well-designed strategy of dissemination [ 40 , 69 ].

Providing an alternate view, the authors of methodological frameworks such as the SIAMPI, Contribution Mapping, Research Contribution, and the Exchange model suggest that the problems of attribution are a consequence of assigning the impact of research to a particular impact metric [ 7 , 40 , 42 , 44 ]. To address these issues, these authors propose focusing on the contribution of research through assessing the processes and interactions between stakeholders and researchers, which arguably take into consideration all the processes and actors involved in a research project [ 7 , 40 , 42 , 43 ]. Additionally, contributions highlight the importance of the interactions between stakeholders and researchers from an early stage in the research process, leading to a successful ex ante and ex post evaluation by setting expected impacts and determining how the research outcomes have been utilised, respectively [ 7 , 40 , 42 , 43 ]. However, contribution metrics are generally harder to measure in comparison to academic-orientated indicators [ 72 ].

Currently, there is a debate surrounding the optimal methodological impact framework, and no tool has proven superior to another. The most appropriate methodological framework for a given study will likely depend on stakeholder needs, as each employs different methodologies to assess research impact [ 4 , 37 , 41 ]. This review allows researchers to select individual existing methodological framework components to create a bespoke tool with which to facilitate optimal study design and maximise the potential for impact depending on the characteristic of their study ( Fig 2 and Fig 3 ). For instance, if researchers are interested in assessing how influential their research is on policy making, perhaps considering a suite of the appropriate metrics drawn from multiple methodological frameworks may provide a more comprehensive method than adopting a single methodological framework. In addition, research teams may wish to use a multidimensional approach to methodological framework development, adopting existing narratives and quantitative metrics, as well as elements from contribution frameworks. This approach would arguably present a more comprehensive method of impact assessment; however, further research is warranted to determine its effectiveness [ 4 , 69 , 72 , 73 ].

Finally, it became clear during this review that the included methodological frameworks had been constructed using varied methodological processes. At present, there are no guidelines or consensus around the optimal pathway that should be followed to develop a robust methodological framework. The authors believe this is an area that should be addressed by the research community, to ensure future frameworks are developed using best-practice methodology.

For instance, the Payback Framework drew upon a literature review and was refined through a case study approach. Arguably, this approach could be considered inferior to other methods that involved extensive stakeholder involvement, such as the CIHR framework [ 8 ]. Nonetheless, 7 methodological frameworks were developed based upon the Payback Framework [ 8 , 29 , 31 – 35 ].

Limitations

The present review is the first to summarise systematically existing impact methodological frameworks and metrics. The main limitation is that 50% of the included publications were found through methods other than bibliographic databases searching, indicating poor indexing. Therefore, some relevant articles may not have been included in this review if they failed to indicate the inclusion of a methodological impact framework in their title/abstract. We did, however, make every effort to try to find these potentially hard-to-reach publications, e.g., through forwards/backwards citation searching, hand searching reference lists, and expert communication. Additionally, this review only extracted information regarding the methodology followed to develop each framework from the main publication source or framework webpage. Therefore, further evaluations may not have been included, as they are beyond the scope of the current paper. A further limitation was that although our search strategy did not include language restrictions, we did not specifically search non-English language databases. Thus, we may have failed to identify potentially relevant methodological frameworks that were developed in a non-English language setting.

In conclusion, the measurement of research impact is an essential exercise to help direct the allocation of limited research resources, to maximise benefit, and to help minimise research waste. This review provides a collective summary of existing methodological impact frameworks and metrics, which funders may use to inform the measurement of research impact and researchers may use to inform study design decisions aimed at maximising the short-, medium-, and long-term impact of their research.

Supporting information

S1 appendix. search strategy..

https://doi.org/10.1371/journal.pmed.1002370.s001

S1 PRISMA Checklist. Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist.

https://doi.org/10.1371/journal.pmed.1002370.s002

Acknowledgments

We would also like to thank Mrs Susan Bayliss, Information Specialist, University of Birmingham, and Mrs Karen Biddle, Research Secretary, University of Birmingham.

• View Article
• PubMed/NCBI
• Google Scholar
• 3. HEFCE. REF 2014: Assessment framework and guidance on submissions 2011 [cited 2016 15 Feb]. Available from: http://www.ref.ac.uk/media/ref/content/pub/assessmentframeworkandguidanceonsubmissions/GOS%20including%20addendum.pdf .
• 8. Canadian Institutes of Health Research. Developing a CIHR framework to measure the impact of health research 2005 [cited 2016 26 Feb]. Available from: http://publications.gc.ca/collections/Collection/MR21-65-2005E.pdf .
• 9. HEFCE. HEFCE allocates £3.97 billion to universities and colleges in England for 2015–1 2015. Available from: http://www.hefce.ac.uk/news/newsarchive/2015/Name,103785,en.html .
• 10. Stern N. Building on Success and Learning from Experience—An Independent Review of the Research Excellence Framework 2016 [cited 2016 05 Aug]. Available from: https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/541338/ind-16-9-ref-stern-review.pdf .
• 11. Matthews D. REF sceptic to lead review into research assessment: Times Higher Education; 2015 [cited 2016 21 Apr]. Available from: https://www.timeshighereducation.com/news/ref-sceptic-lead-review-research-assessment .
• 12. HEFCE. The Metric Tide—Report of the Independent Review of the Role of Metrics in Research Assessment and Management 2015 [cited 2016 11 Aug]. Available from: http://www.hefce.ac.uk/media/HEFCE,2014/Content/Pubs/Independentresearch/2015/The,Metric,Tide/2015_metric_tide.pdf .
• 14. LSE Public Policy Group. Maximizing the impacts of your research: A handbook for social scientists. http://www.lse.ac.uk/government/research/resgroups/LSEPublicPolicy/Docs/LSE_Impact_Handbook_April_2011.pdf . London: LSE; 2011.
• 15. HEFCE. Consultation on the second Research Excellence Framework. 2016.
• 18. Merriam-Webster Dictionary 2017. Available from: https://www.merriam-webster.com/dictionary/methodology .
• 19. Oxford Dictionaries—pathway 2016 [cited 2016 19 June]. Available from: http://www.oxforddictionaries.com/definition/english/pathway .
• 20. Oxford Dictionaries—metric 2016 [cited 2016 15 Sep]. Available from: https://en.oxforddictionaries.com/definition/metric .
• 21. WHO. WHO Ethical and Safety Guidelines for Interviewing Trafficked Women 2003 [cited 2016 29 July]. Available from: http://www.who.int/mip/2003/other_documents/en/Ethical_Safety-GWH.pdf .
• 31. Kalucy L, et al. Primary Health Care Research Impact Project: Final Report Stage 1 Adelaide: Primary Health Care Research & Information Service; 2007 [cited 2016 26 Feb]. Available from: http://www.phcris.org.au/phplib/filedownload.php?file=/elib/lib/downloaded_files/publications/pdfs/phcris_pub_3338.pdf .
• 33. Canadian Academy of Health Sciences. Making an impact—A preferred framework and indicators to measure returns on investment in health research 2009 [cited 2016 26 Feb]. Available from: http://www.cahs-acss.ca/wp-content/uploads/2011/09/ROI_FullReport.pdf .
• 39. HEFCE. RAE 2008—Guidance in submissions 2005 [cited 2016 15 Feb]. Available from: http://www.rae.ac.uk/pubs/2005/03/rae0305.pdf .
• 41. Royal Netherlands Academy of Arts and Sciences. The societal impact of applied health research—Towards a quality assessment system 2002 [cited 2016 29 Feb]. Available from: https://www.knaw.nl/en/news/publications/the-societal-impact-of-applied-health-research/@@download/pdf_file/20021098.pdf .
• 48. Weiss CH. Using social research in public policy making: Lexington Books; 1977.
• 50. Kogan M, Henkel M. Government and research: the Rothschild experiment in a government department: Heinemann Educational Books; 1983.
• 51. Thomas P. The Aims and Outcomes of Social Policy Research. Croom Helm; 1985.
• 52. Bulmer M. Social Science Research and Government: Comparative Essays on Britain and the United States: Cambridge University Press; 2010.
• 53. Booth T. Developing Policy Research. Aldershot, Gower1988.
• 55. Kalucy L, et al Exploring the impact of primary health care research Stage 2 Primary Health Care Research Impact Project Adelaide: Primary Health Care Research & Information Service (PHCRIS); 2009 [cited 2016 26 Feb]. Available from: http://www.phcris.org.au/phplib/filedownload.php?file=/elib/lib/downloaded_files/publications/pdfs/phcris_pub_8108.pdf .
• 56. CHSRF. Canadian Health Services Research Foundation 2000. Health Services Research and Evidence-based Decision Making [cited 2016 February]. Available from: http://www.cfhi-fcass.ca/migrated/pdf/mythbusters/EBDM_e.pdf .
• 58. W.K. Kellogg Foundation. Logic Model Development Guide 2004 [cited 2016 19 July]. Available from: http://www.smartgivers.org/uploads/logicmodelguidepdf.pdf .
• 59. United Way of America. Measuring Program Outcomes: A Practical Approach 1996 [cited 2016 19 July]. Available from: https://www.bttop.org/sites/default/files/public/W.K.%20Kellogg%20LogicModel.pdf .
• 60. Nutley S, Percy-Smith J and Solesbury W. Models of research impact: a cross sector review of literature and practice. London: Learning and Skills Research Centre 2003.
• 61. Spaapen J, van Drooge L. SIAMPI final report [cited 2017 Jan]. Available from: http://www.siampi.eu/Content/SIAMPI_Final%20report.pdf .
• 63. LSHTM. The Health Risks and Consequences of Trafficking in Women and Adolescents—Findings from a European Study 2003 [cited 2016 29 July]. Available from: http://www.oas.org/atip/global%20reports/zimmerman%20tip%20health.pdf .
• 70. Russell G. Response to second HEFCE consultation on the Research Excellence Framework 2009 [cited 2016 04 Apr]. Available from: http://russellgroup.ac.uk/media/5262/ref-consultation-response-final-dec09.pdf .

Advertisement

Keeping pace with the healthcare transformation: a literature review and research agenda for a new decade of health information systems research

• Research Paper
• Open access
• Published: 17 July 2021
• Volume 31 , pages 901–921, ( 2021 )

Cite this article

You have full access to this open access article

• Nadine Ostern   ORCID: orcid.org/0000-0003-3867-3385 1 ,
• Guido Perscheid 2 ,
• Caroline Reelitz 2 &
• Jürgen Moormann 2  

7294 Accesses

1 Altmetric

Explore all metrics

A Correction to this article was published on 20 December 2021

This article has been updated

Accelerated by the coronavirus disease 2019 (Covid-19) pandemic, major and lasting changes are occuring in healthcare structures, impacting people's experiences and value creation in all aspects of their lives. Information systems (IS) research can support analysing and anticipating resulting effects.

The purpose of this study is to examine in what areas health information systems (HIS) researchers can assess changes in healthcare structures and, thus, be prepared to shape future developments.

A hermeneutic framework is applied to conduct a literature review and to identify the contributions that IS research makes in analysing and advancing the healthcare industry.

We draw an complexity theory by borrowing the concept of 'zooming-in and out', which provides us with a overview of the current, broad body of research in the HIS field. As a result of analysing almost 500 papers, we discovered various shortcomings of current HIS research.

Contribution

We derive future pathways and develop a research agenda that realigns IS research with the transformation of the healthcare industry already under way.

Similar content being viewed by others

Afterword: Moving Forward in Healthcare Systems in the Twenty-First Century

From Substantialist to Process Metaphysics – Exploring Shifts in IS Research

Background Literature on the Adoption of Health Information Technologies

Avoid common mistakes on your manuscript.

Introduction

Particularly since the last decade, IT has opened up new opportunities for ‘ehealth’ through telemedicine and remote patient monitoring, alongside potential improvements in the cost-effectiveness and accessibility of health care (Chiasson & Davidson, 2004 ). Accordingly, health information systems (HIS) research has come to focus on how healthcare organizations invest in and then assimilate HIS, looking in particular at the impact of digitalization on healthcare costs, healthcare quality, and patient privacy (Chen et al., 2019 ; Park, 2016 ).

Less attention has been paid to issues such as mobile health, health information interchange, digital health communities, and services that change customer expectations and may lead to major disruptions (Chen et al., 2019 ; Park, 2016 ). These topics, however, are becoming increasingly important due to the penetration of the user and health market by external players, especially tech companies, providing services such as fitness trackers, and surveillance software for patient monitoring in hospitals (Gantori et al., 2020 ). Modern IT, thus, becomes a catalyst to provide greater operational efficiency, offering new possibilities for tech companies to build new health-centred business models and services (Park, 2016 ).

The ways in which tech companies are entering the healthcare industry can be seen amid the spread of coronavirus disease 2019 (Covid-19), which is pushing healthcare systems to the edge of their capacities (Worldbank, 2020 ). In this extraordinary condition, the pandemic has provided an additional opportunity for tech companies that were hitherto not active or not allowed to enter the healthcare industry (Gantori et al., 2020 ).

We are currently seeing how entering the healthcare market is actually taking place, particularly in the USA, where tech companies are increasingly offering services to help address some of the problems associated with Covid-19. Google’s subsidiary Verily, for instance, facilitates the automation of coronavirus symptom screening and provides actionable, up-to-date information that supports community-based decision-making (Landi, 2020 ). Although the collaboration with Verily assists the US government in tracking cases to identify the spread of the virus, it is reasonable to suggest that Verily probably did not launch the screening tool out of altruism. In fact, to receive preliminary screening results via the Verily app, citizens have to log into their personal Google account (Greenwood, 2020 ). This allows Verily to gain immense value by aggregating huge, structured data sets and analyse them to come up with new health services, such as better tools for disease detection, new data infrastructures, and insurance offerings that – for better or for worse – may outplay current healthcare providers and even disrupt whole healthcare ecosystems (CB Insights, 2018 ). Similarly, Amazon has started to provide cloud space through Amazon Web Services to store health surveillance data for the Australian government’s tracing app (Tillett, 2020 ), and Amazon Care, a division initially responsible for handling internal staff care needs, now cooperates with the Bill and Melinda Gates Foundation to distribute Covid-19 testing kits to US residents (Lee & Nilsson, 2020 ).

Looking at information systems (IS) researchers’ previous assessments of state-of-the-art healthcare-related IS literature reveals that most scholars seem to have little or no concern for the beginning of those potentially long-lasting changes that are occurring in the healthcare industry (Chen et al., 2019 ). This is worrying, considering that it is already apparent that the years ahead will be marked by economic volatility and social upheaval as well as direct and indirect health consequences, including sweeping transformations in many of the world’s healthcare systems.

While it is clear that recent developments and the push of tech and platform companies into the healthcare sector can significantly improve the quality of life for billions of people around the world, it will be accompanied by serious challenges for healthcare industries, governments, and individuals (Park, 2016 ). Technological advances are, for instance, giving rise to a plethora of smart, connected products and services, combining sensors, software, data, analytics, and connectivity in all kinds of ways, which in turns leads to a restructuring of health industry boundaries and the empowerment of novel actors, especially tech and platform companies such as IBM, Google, and Amazon (Park, 2016 ).

Observing those changes, we need to develop a general understanding of long-term trends such as digitalization and blurring industry boundaries. As the pandemic is only an amplifier of longer-lasting trends, it is likely that the consequences and exogenous effects on the healthcare industry will go far beyond the time of the current pandemic. Given these observations, we wonder whether the IS research domain is ready to capture, understand, and accompany these developments, which require a holistic view of the healthcare industry, its structures, and the interdependencies between incumbents and new entrants. Thus, we argue that it is now time to develop a more comprehensive understanding of these developments and to determine the role that IS research can play by asking: How can we prepare HIS research to capture and anticipate current developments in the healthcare industry?

To find answers to this question, our paper provides a literature overview of HIS research by ‘zooming in and zooming out’ (Gaskin et al., 2014 ) and by drawing on complexity theory (Benbya et al., 2020 ). Since a healthcare system, like the industry as a whole, can be understood as a complex, digital socio-technical system (Kernick & Mitchell, 2009 ; Therrien et al., 2017 ), zooming in and zooming out is a way to view, capture, and theorize the causes, dynamics, and consequences of a system’s complexity. Complex systems are characterized by adaptiveness, openness (Cilliers, 2001 ), and the diversity of actors and their mutual dependency in the system, meaning that outcomes and research span various levels within these systems, although the boundaries of socio-technical systems are elusive. Assuming that HIS research is just as complex as the socio-technical system investigated, we first zoom in, focusing on concrete research outcomes across levels (i.e., what we can actually observe). Zooming in is followed by zooming out, which means abstracting from the concrete level and embracing the strengths and disparities of overall HIS research on a higher level in which concrete research outcomes are embedded (Benbya et al., 2020 ). Using this approach, we can capture and understand the complexity of HIS research without losing sight of concrete research issues and topics that drive research in this field.

To do this, we chose a hermeneutic framework to guide us in a thorough review and interpretation of HIS literature and lead us to the following overarching observations: (i) The literature review determines the unique contribution that IS research plays in analysing and advancing the healthcare industry. However, it also shows that we are hardly prepared to take up current developments and anticipate their consequences. (ii) The reason for this unpreparedness is that we currently neglect the ecosystem perspective and thus ignore holistic approaches to resolve the striking number of interrelated issues in HIS research. (iii) Based on the unique insights of this literature review, our paper provides a research agenda in which we use complexity theory to discuss the consequences of current developments. This theory assists IS researchers not only to better understand developments and implications thereof for the healthcare industry (and thus HIS research) but also to create a meaningful impact on the future of this industry. Since we have limited our research explicitly to the IS domain, our results may not be generally applicable to other healthcare research domains and we do not claim to provide an overview of the literature in the field of HIS research. However, while IS researchers cannot solve the pandemic directly, preparing them by providing a new research agenda will support them in developing concepts and applications, thereby helping them to overcome the negative effects of the pandemic. In our opinion, it is particularly important that IS research, and especially HIS-related research, obtains a deeper understanding of the needed transformation that is caused by digitalization and the emergence of new players catalysed by the current pandemic.

The remainder of this paper is structured as follows. The next section is concerned with the hermeneutic framework used to conduct the systematic literature review. After explaining the hermeneutic approach and the research steps, we elaborate on the key findings by zooming in; that is, we focus on the key results that emerge from analysing and interpreting the literature for each of the phases defined in the course of the literature sorting process. We then concentrate on zooming out, emphasizing the patterns and interdependencies across phases, which helps us determine the state of HIS research. The results of both parts of the literature review – i.e., zooming in and zooming out (Benbya et al., 2020 ; Gaskin et al., 2014 ) – support us in identifying strengths, as well as drawbacks, in HIS research. On this basis, we develop a research agenda that provides future directions for how HIS research can evolve to anticipate the impending transformation of the healthcare industry.

Literature review: a hermeneutic approach

To answer our research question, we conducted a literature review based on hermeneutic understanding. In particular, we followed Boell and Cecez-Kecmanovic ( 2014 ). They proposed a hermeneutic philosophy as a theoretical foundation and methodological approach that focuses on the inherently interpretive processes in which a reader engages in an ever-expanding and deepening understanding of a relevant body of literature. Adopting a comprehensive literature review approach that addresses well-known issues resulting from applying structured literature review approaches (e.g., Webster & Watson, 2002 ), we strive toward the dual purpose of hermeneutic analysis – i.e., to synthesize and critically assess the body of knowledge (Boell & Cecez-Kecmanovic, 2014 ). We would like to emphasize that the hermeneutic approach to literature reviews is not in opposition to structured approaches. Rather, it addresses the weaknesses of structured approaches (i.e., that they view engagement with the literature as a routine task rather than as a process of intellectual development) and complements them with the hermeneutic perspective to create a holistic approach for conducting literature reviews.

Theoretical underpinning and research method

A methodological means for engaging in reciprocal interpretation of a whole and its constituent elements is the hermeneutic cycle (Bleicher, 2017 ), which consists of a mutually intertwined search and acquisition circle (Circle 1 in Fig.  1 ) and the wider analysis and interpretation circle (Circle 2 in Fig.  1 ) (Boell & Cecez-Kecmanovic, 2014 ). Figure  1 depicts the steps associated with the hermeneutic literature review. The search and acquisition circle is shown on the left of the figure, while the analysis and interpretation circle containing steps of meta and content analysis is depicted on the right. The two circles should be understood as an iterative procedure, the nature of which will be explained in the following. 

Hermeneutic procedure applied to the literature review

Circle 1: Search and acquisition

The hermeneutic literature review starts with the search and acquisition circle, which is aimed at finding, acquiring, and sorting relevant publications. In line with holistic thinking, we began with the identification of a rather small set of highly relevant literature (Boell & Cecez-Kecmanovic, 2014 ) and went on to identify further literature on the basis of progressively emerging keywords. This step is central to the hermeneutic approach and addresses a criticism on structured literature reviews, namely that they downplay the importance of reading and dialogical interaction between the literature and the reader in the literature search process, reducing it to a formalistic search, stifling academic curiosity, and threatening quality and critique in scholarship and research (Boell & Cecez-Kecmanovic, 2014 ; MacLure, 2005 ). Thus, while the search process remains formalized, as in pure structured approaches, the hermeneutic approach allows us to acquire more information about the problem at hand and to identify more relevant sources of information (Boell & Cecez-Kecmanovic, 2014 ).

Given our initial research question and the scope of the review, we began by searching for papers in the Association for Information System’s (AIS’s) eLibrary over a period of 30 years (1990 to 2019). We consider this database to be a source of the most significant publications in the field of HIS research with a focus on the IS research domain. Using the keywords ‘digital health’ and ‘digital healthcare service’, we identified an initial set of 54 papers based on the title, abstract, and keyword search. Engaging in a first round of the hermeneutic search and acquisition circle, we extended and refined these keywords by identifying emerging topics within the literature, as well as using backward and forward search (Webster & Watson, 2002 ). In particular, with each additional paper identified through backward and forward search, we compared keyword references in the papers to our list of keywords and added them if there was sufficient content delimitation. The decision to add a keyword was discussed with all authors until consensus was reached. This led us to a set of 12 keywords, including ‘electronic health’, ‘ehealth’, ‘mobile health’, ‘mhealth’, ‘health apps’, ‘tech health’, ‘healthcare services’, ‘healthcare informatics’, ‘medical informatics’, and ‘health data’.

The selection of publications being considered for our research comprised all journals belonging to the AIS eLibrary, the Senior Scholars’ Basket of Eight Journals (e.g., European Journal of Information Systems, Information Systems Research , and MIS Quarterly ), well-regarded journals following the analyses of Chiasson and Davidson ( 2004 ) and Chen et al. ( 2019 ) (e.g., Business & Information Systems Engineering , Communications of the ACM, and Decision Support Systems ), and the proceedings of the major AIS conferences (e.g., Americas Conference on Information Systems (AMCIS), International Conference on Information Systems (ICIS)). An overview of the selected journals and proceedings is provided in Appendix 1 .

Using our set of keywords, we searched for each keyword individually in the AIS eLibrary and the databases of the respective journals. Subsequently, we created a dataset and filtered out the duplicates, yielding a total number of 1,789 papers to be screened in the search and acquisition circle (Circle 1 in Fig.  1 ). Figure  2 provides an overview of this process by listing the total number of articles identified for each journal individually.

Steps of the search process to create the data set

The resulting 1,789 papers progressively passed through the intertwined hermeneutic circles. Because of the large number, we divided the papers at random into four equally sized groups and assigned them to each of the authors. Each author then screened the paper in his or her group. In the course of several rounds of discussion, decisions on the inclusion of keywords and articles in the literature review were made by all authors, based on the original recommendations of the author responsible for the respective group. To ensure rigor and transparency of the analysis and results, we kept a logbook in which all decisions of the authors and steps of the literature review were recorded (Humphrey, 2011 ).

Given the abundance of topics that were already apparent from titles and abstracts, we began to sort the publications (Boell & Cecez-Kecmanovic, 2014 ). The process of sorting proved to be challenging, as HIS research is diverse and tends to be eclectic (Agarwal et al., 2010 ). This is why researchers have developed frameworks for clustering and analysing HIS research (LeRouge et al., 2007 ). So far, however, no consent on a unified framework has emerged, and sorting is often strongly influenced by the authors’ views on HIS research (Agarwal et al., 2010 ; Fichman et al., 2011 ). For instance, Agarwal et al. ( 2010 ) predetermined health IT adoption and health IT impact as major themes associated with health ITs, acknowledging that this pre-categorization of research topics made a systematic review of the growing and increasingly complex HIS literature unfeasible. Consequently, we decided to sort the articles we had identified into groups inspired by and loosely related to the phases of design science research (DSR) (Peffers et al., 2008 ), which is an essential step in hermeneutics – i.e., defining guidelines to facilitate interpretive explication (Cole & Avison, 2007 ). DSR can be understood as a cumulative endeavour and, therefore, we understood HIS research as accumulative knowledge that can be reconstructed and consolidated using DSR phases as guidance (vom Brocke et al., 2015 ; vom Brocke et al., 2009 ). In particular, this helped us to sort the articles without prejudice to expected HIS research topics and clusters (Grondin, 2016 ).

In the past, researchers have used the DSR process in the context of literature reviews to identify advances in design science-related research outcomes (Offermann et al., 2010 ). In this paper, we use the DSR phases – in the sense of a rough guideline – as a neutral lens to classify articles according to their research outcomes. We thereby assume that HIS literature can be seen as an overall process, where research results and progress are built upon each other and can be classified into phases of problem identification and research issues , definition of research objectives and possible solution space , design and development of solutions , demonstration of research effectiveness, innovativeness and acceptance , and evaluation . These phases served as a guide to achieve an outcome-oriented, first-hand sorting of articles, while this approach also gave us the opportunity to take a bird's-eye view on HIS research. Note that we intentionally omitted the last step of DSR – i.e., communication – as we regard communication as present in all published articles. Based on our initial reading, we assigned all 1,789 papers to the phases and discussed this sorting in multiple rounds until all authors agreed on the assignments.

Simultaneously, we applied criteria for the inclusion and exclusion of articles. We included full papers published in the journals and conference proceedings belonging to our selection. We excluded articles that were abstract-only papers, research-in-progress papers, panel formats, or workshop formats, as well as papers without direct thematic reference to our research objective. Additionally, during the acquisition stage we stored selected papers in a separate database whenever they fulfilled certain quality criteria (e.g., for separate studies using the same dataset, such as a conference publication and a subsequent journal publication, we only used the articles with the most comprehensive reporting of data to avoid over-representation).

The authors read the resulting 489 papers to identify new core terms and keywords that were used in subsequent searches, which not only provided the link to the analysis and interpretation circle but also informed the literature search. For this purpose, each author read the papers and kept notes in the logbook that supported us in systematically recording the review process and allowed us to shift from concentrating on particular papers to focusing on scientific concepts (Boell & Cecez-Kecmanovic, 2014 ; Webster & Watson, 2002 ).

Circle 2: Analysis and interpretation

The search and acquisition circle formed part of the iterative procedure of analysis and interpretation, whereby the reading of individual papers was the key activity linking Circle 1 to the steps of Circle 2 (Boell & Cecez-Kecmanovic, 2014 ). Through orientational reading we gained a general understanding of the literature, thus laying the foundation for the subsequent steps of analysis and interpretation (Boell & Cecez-Kecmanovic, 2014 ).

Within the analysis and interpretation circle, two types of reviews were conducted for all identified and sorted articles: in a first round a meta-review, and in a second round a content analysis of the papers was performed. Meta-reviews are a useful tool for capturing and analysing massive quantities of knowledge using systematic measures and metrics. We followed Palvia et al. ( 2015 ), who proposed a structured method that is integrated into the hermeneutic approach. In particular, having identified and sorted the relevant research articles, we applied proposed review features, including methodological approach, level of observation, sample size, and research focus (Humphrey, 2011 ; Palvia et al., 2015 ) to map, classify, and analyse the publications (Boell & Cecez-Kecmanovic, 2014 ). In doing so, we slightly adapted the classic meta-analysis by focusing on meta-synthesis, which is similar to meta-analysis but follows an interpretive rather than a deductive approach. Whereas a classic meta-analysis tries to increase certainty in cause-and-effect conclusions, meta-synthesis seeks to understand and explain the phenomena of mainly qualitative work (Walsh & Downe, 2005 ). The results of the meta-synthesis provided the basis for our subsequent critical assessment of content. Furthermore, we created a classification matrix based on particularly salient features of the meta-review (i.e., levels of observation and research foci), which facilitated and standardized the content analysis.

Within the matrix, the levels of observation comprised infrastructure (e.g., information exchange systems, electronic health records), individuals (patients and users of digital health services), professionals (e.g., nurses and general practitioners), organizations (hospitals and other medical institutions), and an ecosystem level. The latter is defined as individuals, professionals, organizations, and other stakeholders integrated via a digital infrastructure and aiming to create a digital environment for networked services and organizations with common resources and expectations (Leon et al., 2016 ). To identify the most important concepts used by researchers, we discussed a variety of approaches to the derivation of research foci – i.e., areas containing related or similar concepts that are frequently used in research on HIS. Finally, six research focus areas emerged, covering all relevant research areas. To describe the core HIS research issues addressed by these foci, we used the following questions:

HIS strategy: What are the prerequisites for configuring, implementing, using, maintaining, and finding value in HISs?

HIS creation: How are HISs composed or developed?

HIS implementation: How are HISs implemented and integrated?

HIS use and maintenance: How can HISs be used and maintained once in place?

Consequences and value of HIS: What are the consequences and the added value of HISs?

HIS theorization: What is the intellectual contribution of HIS research?

We used the classification matrix as a tool for assigning publications and finding patterns across research articles and phases. In particular, we used open, axial, and selective coding (Corbin & Strauss, 1990 ) to analyse the content of articles in a second round of the analysis and interpretation circle. Each author individually assigned open codes to text passages while reading the identified research articles, noting their thoughts in the shared digital logbook that was used for constant comparative analysis. Once all authors had agreed on the open codes, axial coding – which is the process of relating the categories and subcategories (including their properties) to each other (Wolfswinkel et al., 2013 ) – was conducted by each author and then discussed until consent on codes was reached. Next, we conducted selective coding and discussed the codes until theoretical saturation was achieved (Corbin & Strauss, 1990 ; Matavire & Brown, 2008 ). For the sake of consistent terminology, we borrowed terms from Chen et al. ( 2019 ), who used multimethod data analysis to investigate the intellectual structure of HIS research. In particular, they proposed 22 major research themes, which we assigned to the initial codes whenever possible. In two rounds of discussion in which we compared the assignment of codes, two additional codes emerged, which left us with a total of 24 theme labels (Appendix 2 ). By discussing the codes at all stages of coding, theoretical saturation emerged, which is the stage at which no additional data are being found or properties of selective codes can be developed (Glaser & Straus, 1968 ; Saunders et al., 2018 ). In fact, independent from each other, all authors saw similar instances occurring over and over again, resulting in the same codes, making us confident that we had reached theoretical saturation (Saunders et al., 2018 ).

Finally, we entered the codes into the classification matrix, which allowed us to identify patterns based on the meta and content analysis. This enabled us to provide insights into the strengths and weaknesses of current HIS research; these are presented in the following section.

Zooming-in: key findings of the phase-based literature analysis

In the following, we ‘zoom in’ (Gaskin et al., 2014 ) by presenting key findings of the literature review for each phase, illustrated by means of the classification matrices. We assigned selective codes that emerged from the content analysis to the fields of the matrices, with the numbers in brackets indicating the frequency with which codes emerged. Note that, for the sake of clarity, we displayed only the most relevant research themes in the matrices and indicated the number of further papers using the reference ‘other themes.’ A complete list of research themes for each phase can be found in the appendix (Appendix 2 ). In the following, each table shows the classification matrix and selective codes that resulted from the meta and content analysis of papers in the respective phase. The shaded areas in the matrix show focused research themes (i.e., selective codes) and characteristics of research articles that gave way to clusters (i.e., collections of themes that appear frequently and/or characteristically for the respective focus).

Phase 1: Problem identification and research issues

Within the first phase, a large body of literature was found (218 articles). This phase encompasses articles that identify problems and novel research issues as a main outcome, with the aim of pointing out shortcomings and provoking further research. For instance, besides behavioural issues such as missing user acceptances or trust in certain HISs, the design and effectiveness of national health programs and/or HIS is a frequently mentioned topic. It should be noted, however, that literature assigned to this phase is extremely diverse in terms of research foci, levels of observation, and research themes, and hardly any gaps can be identified (Table 1 ).

The first cluster (1a) encompasses the research focus of HIS strategy, spanning all levels of observation and totalling 24 publications. HIS strategy appears to be of particular relevance to the levels of organization and infrastructure. Content-wise, the theme of health information interchange is of particular interest, referring, for example, to the development of a common data infrastructure (Ure et al., 2009 ), consumer-oriented health websites (Fisher et al., 2007 ), and security risks of inter-organizational data sharing (Zhang & Pang, 2019 ). HIS productivity and HIS security are the second most salient themes, focusing, for example, on measuring the effectiveness of fitness apps (Babar et al., 2018 ) and presenting challenges with regard to the interoperability of medical devices (Sametinger et al., 2015 ).

The second cluster (1b), comprising 25 publications, represents the ecosystem level and focuses mainly on national and cross-national HIS-related issues such as the relation between ICT penetration and access to ehealth technologies across the European Union (Currie & Seddon, 2014 ), as well as on the collaboration and involvement of different stakeholders (Chang et al., 2009 ; King, 2009 ). Most important here is health information interchange – e.g., the provision, sharing, and transfer of information (Bhandari & Maheshwari, 2009 ; Blinn & Kühne, 2013 ).

Cluster 1c covers the research focus of HIS use and maintenance, as well as the consequences of HIS. Whereas most papers addressing the HIS acceptance theme focus on professionals’ or patients’ acceptance of specific technological solutions, such as telemedicine (Djamsbi et al., 2009 ) or electronic health records (Gabel et al., 2019 ), papers assigned to health information interchange focus on topics related to information disclosure, such as self-tracking applications (Gimpel et al., 2013 ). Finally, the HIS outsourcing and performance theme concentrates on financial aspects in organizations, including potential for quality improvements and cost reductions (Setia et al., 2011 ; Singh et al., 2011 ).

Finally, the fourth cluster (1d) focuses on HIS theorizing with respect to the individual and infrastructure levels of observation. Although this cluster represents a range of theme labels (15), those addressing HIS acceptance, HIS patient-centred care, as well as health analytics and data mining predominate. Papers within the theme label HIS acceptance cover a wide range of topics, such as the acceptance of telehealth (Tsai et al., 2019 ) up to the usage intentions of gamified systems (Hamari & Koivisto, 2015 ). The same applies to the papers assigned to the theme labels of health analytics and data mining. Focusing on the infrastructure level of observation, the identified papers mostly review academic research on data mining in healthcare in general (Werts & Adya, 2000 ), through to the review of articles on the usage of data mining with regard to diabetes self-management (Idrissi et al., 2019 ). Papers on HIS patient-centred care mostly address the challenges and opportunities of patient-centred ehealth applications (Sherer, 2014 ).

Apart from these clusters, quite a few research articles refer to the infrastructure level of observation, addressing information sharing in general (Li et al., 2008 ), electronic medical records (George & Kohnke, 2018 ; Wessel et al., 2017 ), and security and privacy issues (Zafar & Sneha, 2012 ).

Most common in terms of research methods within this phase are case studies (57), followed by quantitative data analyses (50), theoretical discussions (29), and literature studies (14). In particular, case studies dominate when referring to the ecosystem or infrastructure level of observation, whereas quantitative analyses are conducted when individuals or professionals are at the centre of the discussion. However, and unsurprisingly given the considerable diversity of research themes within this phase, the variety of research methods is also quite large, ranging from field studies (Paul & McDaniel, 2004 ), to interviews (Knight et al., 2008 ), to multimethod research designs (Motamarri et al., 2014 ).

Phase 2: Definition of research objectives and solution space

The second phase of HIS research yielded a lower number of articles (45) compared to the phase of problem identification and research issues. The second phase comprises articles that focus on proposing possible solutions to existing problems – i.e., introducing theory-driven, conceptual designs of health ecosystems including health information interchange, as well as scenario analyses anticipating the consequences of HIS implementation on an organizational level. Based on the research foci and levels of observation, we identified three specific thematic clusters, as shown in Table 2 .

The first cluster (2a) comprises the ecosystem level of observation and encompasses eight publications. Besides a strong tendency toward theory-driven research, health information interchange is the most common theme. We found that the need to enable cooperation within networks and to ensure accurate data input was addressed in most of the literature. While a majority of studies focus on the application of HIS in networks within specific boundaries, such as medical emergency coordination (Sujanto et al., 2008 ) or Singapore’s crisis management in the fight against the SARS outbreak in 2003 (Devadoss & Pan, 2004 ), other studies, such as that by Aanestad et al. ( 2019 ), take an overarching perspective, addressing the need to break down silo thinking and to start working in networks. Following the question of why action research fails to persist over time, Braa et al. ( 2004 ) highlighted the role of network alignment, criticizing action research projects for failing to move beyond the prototyping phase and, therefore, failing to have any real impact.

Cluster 2b, encompassing nine publications, was derived from the observation that studies within the organizational level concentrated strongly on HIS use and maintenance and the consequences of HIS research. Herein, a vast array of topics was observed, such as the potential for cost reduction through HIS (Byrd & Byrd, 2009 ), the impact of HIS on product and process innovation in European hospitals (Arvanitis & Loukis, 2014 ), and the perceived effectiveness of security risk management in healthcare (Zafar et al., 2012 ). Moreover, we found that practice-oriented methods, such as mixed-method approaches, surveys, data analyses, and case studies, are used predominantly within this cluster. Focusing on the latter, most studies analyse particular scenarios by using a rather small sample of cases, for instance, Al-Qirim ( 2003 ) analysed factors influencing telemedicine success in psychiatry and dermatology in Norway.

The third cluster (2c) was derived from analysis of the HIS creation research focus (nine publications). Although health information interchange is the most represented in this cluster, a large number of further themes can be observed. Studies within this cluster predominantly address design aspects of system interoperability, focusing on data processing and data interchange between the actors. HISs mostly serve as a tool for the development or enhancement of decision support systems, such as for real-time diagnostics combining knowledge management with specific patient information (Mitsa et al., 2007 ) or clinical learning models incorporating decision support systems in the dosing process of initial drug selection (Akcura & Ozdemir, 2008 ).

Phase 3: Design and development

The design and development phase comprises 84 research articles concerned with the creation of novel IS artefacts (e.g., theories, models, instantiations). We thereby refer to Lee et al.’s ( 2015 ) definition of the IS artefact – i.e., the information, technology, and social artefact that forms an IS artefact by interacting. We assigned to this phase papers that are explicitly concerned with developing solutions for information exchange (e.g., design of messaging systems or knowledge systems in hospitals), technological artefacts (e.g., hardware or software used for generating electronic health records), and social artefacts that relate to social objects (e.g., design of national or international institutions and policies to control specific health settings and patient-centred solutions). Within the design and development phase, the analysis revealed two clusters (Table 3 ).

The first cluster (3a) was identified in the research focus of HIS creation (31 articles). Here, the most frequent research theme is HIS innovation followed by HIS and patient-centred care, HIS productivity, and health analytics and data mining. The focus is on specific contexts, mostly medical conditions and artefacts developed for their treatment, such as in the context of mental health/psychotherapy (Neben et al., 2016 ; Patel et al., 2018 ), diabetes (Lichtenberg et al., 2019 ), or obesity (Pletikosa et al., 2014 ). Furthermore, information infrastructures or architectures – for instance, for the process of drug prescription (Rodon & Silva, 2015 ), or for communication between healthcare providers and patients (Volland et al., 2014 ) – are represented.

The second aggregation of research articles is found in cluster 3b, focusing on theoretical aspects of HIS (32 articles). Again, these studies span all levels of observation (including infrastructure, individual, professional, organization, and ecosystem). Topics in this theme are diverse, ranging from HIS on a national level (Preko et al., 2019 ), to knowledge management in healthcare (Wu & Hu, 2012 ) to security of HIS (Kenny & Connolly, 2016 ).

Beyond both clusters, it is evident that during design and development, researchers do not deal with the consequences of HIS, nor does HIS strategy play an important role. Furthermore, only in the research focus of theorization is the ecosystem level of some relevance to other levels (e.g., the individual level). It should be noted that ecosystems are mostly referred to in terms of nations or communities, without any transnational or global perspective. Furthermore, the term ‘ecosystem’ has not been used in research, and within the other research focus areas, the ecosystem level is barely represented. Moreover, articles combining different perspectives of the single levels of observation on HIS – namely individuals (i.e., patients), professionals (i.e., medical staff), and organizations (e.g., hospitals) – are rare. During design and development, potential users are not typically integrated, whereas it is quite common to derive requirements and an application design from theory, only involving users afterwards – e.g., in the form of a field experiment (e.g., Neben et al., 2016 ).

Surprisingly, theoretical papers outweigh papers on practical project work, whereby the latter mostly focus on a description of the infrastructure or artefact (e.g., Dehling & Sunyaev, 2012 ; Theobalt et al., 2013 ; Varshney, 2004 ) or are based on (mostly single) case studies (e.g., Hafermalz & Riemer, 2016 ; Klecun et al., 2019 ; Ryan et al., 2019 ). Within the design and development phase, the generation of frameworks, research models, or taxonomies is prevalent (e.g., Preko et al., 2019 ; Tokar et al., 2015 ; Yang & Varshney, 2016 ).

Phase 4: Demonstration

This phase includes 35 articles related to presenting and elaborating on proposed solutions – e.g., how HIS can be implemented organization-wide (e.g., via integration into existing hospital-wide information systems), proposed strategies and health policies, as well as novel solutions that focus on health treatment improvements. Within the demonstration phase, we identified two clusters that emerged from the meta and content analyses (Table 4 ).

Cluster 4a (10 articles) is characterized by articles that focus on HIS issues related to the infrastructure level, spanning the research foci of HIS strategy, creation, and deployment. Content-wise, the cluster deals mainly with technical feasibility and desirability of HISs, including topics such as the configuration of modular infrastructures that support a seamless exchange of HISs within and between hospitals (Dünnebeil et al., 2013 ). Moreover, papers in this cluster address HIS practicability by determining general criteria that are important for the design of health information systems (Maheshwari et al., 2006 ) or conduct HIS application tests by carrying out prototypical implementations of communication infrastructures. In particular, the latter are tested and proven to meet specific technical standards to guarantee the frictionless transmission of health information data (Schweiger et al., 2007 ). In contrast, Heine et al. ( 2003 ) upscaled existing HIS solutions and tested the infrastructure in large, realistic scenarios.

Conversely, cluster 4b (11 articles) is mainly concerned with HIS use and maintenance, spanning several levels of observation – i.e., infrastructure, individuals, professionals, and organizations. Interestingly, papers in this cluster aim at efficiency and added value when looking at the infrastructure and organizational levels, whereas researchers are more interested in acceptance when focusing on the individual and professional use of HISs. Overall, cluster 4b is primarily concerned with organizational performance (e.g., increases in efficiency due to better communication and seamless transfer of patient health information) as well as user acceptance of new HISs.

Although the two clusters constitute a diverse set of literature and themes, it is apparent that research taking an ecosystem perspective is very rarely represented. Across the papers, only three are concerned with issues related to the ecosystem level. In particular, Lebcir et al. ( 2008 ) applied computer simulations in a theoretical demonstration as a decision support system for policy and decision-makers in the healthcare ecosystem. Abouzahra and Tan ( 2014 ) used a mixed-methods approach to demonstrate a model that supports clinical health management. Findikoglu and Watson-Manheim ( 2016 ) addressed the consequences of the implementation of electronic health records (EHR) systems in developing countries.

Phase 5: Evaluation

The fifth phase includes 92 publications with a focus on assessing existing or newly introduced HIS artefacts – i.e., concepts, policies, applications, and programs – thereby proving their innovativeness, effectiveness, or user acceptance. As Table 5 shows, three clusters were identified.

The main focus of publications in the evaluation phase is on the infrastructure level, where most papers are related to HIS creation and HIS use and maintenance. Therefore, together with the publications pigeonholed to HIS deployment and consequences of HIS, these articles were summarized as the first cluster (5a, comprising 53 articles). The assessment of national HIS programs, as well as mobile health solutions, are a frequent focus (10 papers). Articles on HIS use and maintenance are largely related to the professional, organizational, and ecosystem levels and were thus grouped as cluster 5b (10 articles). A third cluster (5c – 11 articles) emerged from research articles in HIS theorization. Here, papers at all levels of observation were found. Research focusing on areas such as HIS strategy and consequences of HIS are, with a few exceptions, not covered in the evaluation phase. Methods used include interviews, focus groups, and observations (e.g., Romanow et al., 2018 ). Experiments and simulation are rarely applied (e.g., Mun & Lee, 2017 ). The number of interviews shows a huge spread, starting with 12 and reaching a maximum of 150 persons interviewed.

Under the evaluation lens, the ecosystem perspective is covered by seven articles, but only three papers look at cases, while the others focus on theorization or consequences in terms of costs. Overall, popular topics in the evaluation phase include mobile health and the fields of electronic medical records (EMR) and EHR, e.g., Huerta et al. ( 2013 ); Kim and Kwon ( 2019 ). The authors cover these themes mostly from an HIS creation perspective; thus, they deal with concrete concepts, prototypes, or even implemented systems. In the evaluation phase, just nine papers deal with HIS innovation – a good example being Bullinger et al. ( 2012 ), who investigated the adoption of open health platforms. We may conclude that, in most cases, evaluation is related to more established technologies of HIS. As expected, most articles in this phase rely on practice-oriented/empirical work (as opposed to theory-driven/conceptual work). Just two papers (Ghanvatkar & Rajan, 2019 ; Lin et al., 2017 ) deal with health analytics and data mining, one of the emerging topics of HIS.

Zooming out: key findings of the literature analysis across phases

Having elaborated on the key findings within each phase of HIS research, we now ‘zoom out’ (Benbya et al., 2020 ; Gaskin et al., 2014 ) to recognize the bigger picture. Thereby, we ‘black-box’ the concrete research themes (e.g., HIS implementation, health analytics, HIS innovation) to focus on clusters across phases, highlighting the breadth that HIS research encompasses (Leroy et al., 2013 ). In particular, while we focused on analysing the main topics within the different phases of HIS research in the zoom-in section, we now abstract from those to perform a comparative analysis of emerging clusters across those phases by zooming out. We do so by comparing the different clusters, taking into account the aspects of the level of observation and the research foci, which gave us the opportunity to identify areas of strong emphasis and potential gaps.

In particular, each author first conducted this comparative analysis on their own and then discussed and identified the potential weaknesses together. This was done in two rounds of discussion. In particular, it became obvious which areas hold immense potential for further research in healthcare (especially the penetration of new, initially non-healthcare actors, such as tech companies or other providers pushing into the industry). We summarize these potentials for research by proposing four pathways that can help HIS research to broaden its focus so that we can better understand and contribute to current developments. Notably, we expect that these insights will help to assess the state-of-the-art of HIS research and its preparedness for dealing with the consequences of Covid-19 and further pandemics, as well as for coping with associated exogenous shocks.

In zooming out, we identified discrepancies between phase 1 (problem identification and research issues) and the subsequent phases. In particular, the diversity of topics was considerably lower when it came to how researchers determined strategies; created, demonstrated, used, and maintained HISs; and coped with the consequences thereof. We observed that researchers pointed to a diverse set of issues that span all levels of observation, especially in HIS theorization, focusing on topics such as trust in HIS, data analytics, and problems associated with the carrying out of national health programs. Surprisingly, although we can assume that researchers recognized the multidimensionality of issues as a motivation to conduct HIS research, they did not seem to approach HIS research issues in a comprehensive and consistent way.

To illustrate this assertion, we point to the ‘shift of clusters’ that can be observed when comparing the single phases, from problem identification to the evaluation of HIS. We note that clusters increasingly migrate ‘downwards’ (i.e., from the ecosystem level down to the infrastructure level) and become even fewer. In line with Braa et al. ( 2004 ), we suggest that extant HIS research has identified a multitude of interrelated issues but has faced problems in translating these approaches into concrete and holistic solutions. This is reflected in the lower number of, and reduced diversity in, clusters across research themes when we move through the HIS research phases. Thus, we conclude that future HIS research can be broadened by taking into account the following pathway:

HIS research is well-prepared and able to identify and theorize on systemic problems related to the healthcare industry. Nonetheless, it has the potential to address these problems more thoroughly – i.e., to find solutions that are as diverse as the problems and, thus, suitable for coping with issues in the healthcare industry characterized by the involvement of multiple actors, such as governments, healthcare providers, tech companies, and their interactions in diverse ecosystems (pathway 1).

As we have seen, HIS research has tended to focus on important but incremental improvements to existing infrastructures, particularly in the phases of demonstration and evaluation, with the aim of presenting new IS artefacts and conceptual or practical solutions. For instance, Choi and Tulu ( 2017 ) considered improvements in user interfaces to decrease the complexity of mobile health applications using incremental interface design changes and altering touch techniques. Similarly, Roehrig and Knorr ( 2000 ) designed patient-centred access controls that can be implemented in existing infrastructures to increase the privacy and security of EHRs and avoid malicious access and misuse of patient health information by third parties.

While we sincerely acknowledge these contributions and wish to emphasize the multitude of papers that are concerned with enhancements to existing infrastructures, we would like to shift the view to the major challenges in HIS research. These challenges include combating global and fast-spreading diseases (e.g., malaria, tuberculosis, Covid-19) and tracking health statuses accurately and efficiently, especially in developing countries. All of these challenges necessitate global and comprehensive solutions, spanning individuals, organizations, and nations, and have to be embedded in a global ecosystem (Winter & Butler, 2011 ). Such grand challenges are, by nature, not easy to cope with, and the intention to develop a comprehensive solution from the perspective of IS researchers seems almost misguided. However, HIS research is currently missing the opportunity to make an impact, despite the discipline’s natural intersection with essential aspects of the healthcare industry (i.e., its infrastructures, technologies, and stakeholders, and the interdependencies between these components). Thus, we assert that:

HIS research has often focused on necessary and incremental improvements to existing IS artefacts and infrastructures. We see potential in shifting this focus to developing solutions that combine existing IS artefacts to allow for exchange of information and the creation of open systems, which will enhance support for and understanding of the emergence of ecosystems (pathway 2).

By focusing on incremental improvements, HIS research has become extraordinarily successful in solving isolated issues, especially in relation to the problems of patients and health service providers (e.g., hospitals and general practitioners). However, we observed during our analysis that spillover effects were seldom investigated. When, for example, a new decision support system in a hospital was introduced, positive consequences for patients, such as more accurate diagnoses, were rarely of interest to the research. In fact, our meta-analysis revealed that the level of observation for the majority of papers matched the level of analysed effects. While it is valid to investigate productivity and efficiency gains by introducing a hospital-wide decision support system, we are convinced that spillover effects (for instance, on patients) should also be within the focus of HIS research. Therein, we suggest that HIS research has not focused primarily on patients and their well-being but on IS infrastructures and artefacts. However, patient well-being is the ultimate direct (or indirect) goal of any HIS research (by increasing the accuracy and shortening the time of diagnosis, improving treatment success rates, etc.). Thus, we propose that:

HIS research is experienced in solving isolated issues related to the daily processes of healthcare providers; however, we see much potential in considering the value that is delivered by focusing on patient-centricity (pathway 3).

Putting the patient at the centre of HIS research implies shifting the focus of researchers to the patient’s own processes. The question remains as to how HIS researchers can support patient-centricity. While this is only possible by understanding patients’ processes, we also see the need to understand the whole system – i.e., the ecosystem in which patients’ processes are embedded. The ecosystem perspective needs to consider networked services and organizations, including resources and how they interact with stakeholders of the healthcare industry (including patients). To date, we observe, across phases the ecosystem perspective has largely been neglected. To be precise, although HIS research seems to be aware of the multilevel aspects of healthcare issues in the problem identification phase, researchers appear to stop or are hindered from developing solutions that go beyond the development of prototypes (Braa et al., 2004 ). Thus, we find that:

HIS research is capable of theorizing on an ecosystem level (i.e., capturing the complexity of the socio-technical health system), but would benefit from increasing the transfer of these insights into research so as to develop holistic solutions (pathway 4).

Looking at the strengths of HIS research, the reviewed papers accentuate the unique contribution that IS researchers can make to better understand and design IS artefacts for the healthcare context. This has been achieved by analysing empirical data and exploring contextual influences through the application and elaboration of IS theories (LeRouge et al., 2007 ). At the same time, our literature review shows the incredible diversity and high level of complexity of issues related to HISs, indicating that we need solutions characterized by holism and the inclusion of multiple actors (i.e., an integrative ecosystem perspective). So far, by concentrating on incremental improvements to existing infrastructures HIS research has widely failed to reach the necessary holistic level.

We would like to emphasize that we recognize the value of all previous approaches. Yet, it is necessary to ask whether we as IS researchers are in a position to identify current developments in the healthcare industry and to anticipate the consequences triggered by pandemics or other waves of disease. We acknowledge that this will be difficult unless we take a more holistic view and try to understand connections in the health ecosystems. Regarding whether HIS research is in a position to capture and anticipate consequences of the current push of tech companies in the healthcare industry catalysed, for example, by Covid-19, we assert that this is hardly the case, even if IS research is well-placed to interpret the expected socio-technical changes and adaptations within healthcare. Given the enormous potential for disruption caused by, for instance, pandemics and its consequences, such as the intrusion of technology companies into the market, it is now time to question and redefine the role of HIS research so that it can generate decisive impacts on the developments in this industry.

• Research agenda

To support HIS research for the transformation of the healthcare industry, we develop a research agenda that is informed by complexity theory. This theory implies that complex, socio-technical systems such as the healthcare industry can fluctuate between different states, ranging from homogenous forms of coevolution (i.e., a state where emergent structures and processes become similar to each other) to chaotic systems that are characterized by increasing levels of tension, which might result in extreme outcomes such as catastrophes or crises (Benbya et al., 2020 ).

While coevolution and chaos represent possible extreme states, the current situation – i.e., the penetration of tech companies into the healthcare industry – is best described by the dynamic process of emergence. Emergence is characterized by a disequilibrium, which implies unpredictability of outcomes that may lead to new structures, patterns, and properties within a system characterized by self-organization and bursts of amplification (Benbya et al., 2020 ; Kozlowski et al., 2013 ). Given the dynamics resulting from this, it seems impossible to predict the future; however, it is not impossible to prepare for it.

In particular, the current dynamics within the healthcare industry necessitate an understanding of exponential progress, not as the ability to foresee well-defined events in space and time, but as an anticipation of the consequences of emerging states and dynamic adaptive behaviours within the industry (Benbya et al., 2020 ). The following research agenda for HIS research is thus structured along three key issues: anticipating the range of actors’ behaviours, determining boundaries and fostering collaboration in the healthcare industry, and creating sustainable knowledge ecosystems.

According to these key issues, Table 6 offers guiding questions for HIS researchers. Addressing all issues will contribute to an understanding of the entire healthcare industry and the development of holistic solutions for a multitude of health issues by involving different actors (e.g., patients, hospitals, professionals, governments, NGOs). However, we propose approaching the agenda stepwise, in the order of the key issues, first looking at the range of behaviours and consequences of current developments for actors, then focusing on the blurring lines of the healthcare industry, and finally investigating the dissemination and sharing of knowledge, which we see as the ultimate means to connect actors and infrastructures to create a joint ecosystem. Table 6 thereby provides key guiding statements and exemplary research questions for future HIS research that support researchers in taking one of the aforementioned pathways. We structured guiding statements along three major areas of improvement. In addition, we offer exemplary research questions to these statements, as well as inspiring studies from other industries that have faced similar challenges and have been studied and supported by researchers.

Area of improvement 1: Anticipating the range of actor behaviours

As healthcare systems are becoming more open – for example, through the penetration of new market actors and the use of increasingly comprehensive and advanced health technologies – accurately determining the boundaries of an industry and its key actors is becoming more difficult. To model these systems, we must carefully model every interaction in them (Benbya et al., 2020 ), which first requires HIS researchers to identify potential actors in the ecosystem rather than predetermining assumed industry boundaries. As actors are not always evident, we follow Benbya et al. ( 2020 ) in proposing Salthe’s ( 1985 ) three-level specification, assisting researchers in identifying actors at the focal level of what is actually observed (e.g., hospitals, patients, and general practitioners) and its relations with the parts described at the lower level (e.g., administrators and legal professionals), taking into account entities or processes at a higher level in which actors at the focal level are embedded (e.g., national health system structures and supporting industries, such as the pharmaceutical or tech industries). These examples are only illustrative, and criteria for levels have to be suggested and discussed for each research endeavour.

To anticipate future developments in the healthcare industry, we also need to analyse the strategies and interests of actors for joining or staying in the healthcare industry. This is especially important because, like other complex socio-technical systems, the healthcare industry is made up of large numbers of actors that influence each other in nonlinear ways, continually adapting to internal or external tensions (Holland et al., 1996 ). If tension rises above a certain threshold, we might expect chaos or extreme outcomes. As these are not beneficial for the actors in the system, the eventual goal is to align actors’ interests and strategies across a specific range of behaviour to foster coevolution. This allows for multi-layered ecosystems that encourage joint business strategies in competitive landscapes, as well as the alignment of business processes and IT across actors (Lee et al., 2013 ).

Area of improvement 2: Determining boundaries and fostering collaboration

Actors build the cornerstones of the healthcare industry. Thus, if we want to understand and capture its blurring boundaries, there is a need to understand the complex causality of interactions among heterogeneous actors. In particular, scholars have emphasized that, in complex systems, outcomes rarely have a single cause but rather result from the interdependence of multiple conditions, implying that there exist multiple pathways from an input to an output (Benbya et al., 2020 ). To capture interaction, we follow Kozlowski et al. ( 2013 ), who envisioned a positive feedback process including bottom-up dynamic interaction among lower-level actors (upward causation), which over time manifests at higher, collective levels, while higher-level actors influence interaction at lower levels (downward causation). As these kinds of causalities shape interaction within healthcare ecosystems as well as at their boundaries, HIS researchers need to account for multi-directional causality in the form of upward, downward, and circular causality (Benbya et al., 2020 ; Kim, 1992 ).

Understanding casualties among actors in the healthcare industry is important for harnessing the advantages of the blurring of boundaries – e.g., by making use of the emergent ecosystem for launching innovation cycles (Hacklin, 2008 ). However, first, HIS researchers increasingly need to consider the ecosystem perspective by investigating interactions among actors and the role of IS infrastructures in fostering collaborative health innovations. We propose a focus on radical innovation, which is necessary to address the diversity and interdependence of issues present in the healthcare industry by putting the patient at the core of all innovation efforts. HIS researchers, however, need to break down the boundaries between different innovation phases and innovation agencies, including a higher level of unpredictability and overlap in their time horizons (Nambisan et al., 2017 ). Notably, this requires actors in the healthcare industry to discover new meaning around advanced technologies and IS infrastructures whose design needs to facilitate shared meaning among a diverse set of actors, thereby fuelling radical digital innovations (Nambisan et al., 2017 ).

Area of improvement 3: Creating sustainable knowledge ecosystems

We define knowledge dissemination and sharing as the ultimate means of connecting actors and aligning actions within common frameworks to shape an inclusive healthcare ecosystem. Paving the way for inclusive healthcare ecosystems is thus necessary to address the current shortcomings of HIS research as elaborated in the previous section.

Addressing knowledge dissemination and sharing is thereby of the utmost importance as we look at the healthcare industry in the current phase of emergence. This means that the industry might go through several transition phases in which existing actors, structures, and causal relationships dissipate and new ones emerge, resulting in a different set of causal relationships and eventually altering knowledge claims (Benbya et al., 2020 ). Creating a permeable and sustainable knowledge management system is necessary to ensure the transfer of knowledge for the best outcomes for the patient while securing the intellectual property rights and competitive advantages of diverse actors such as hospitals and other healthcare providers.

To be precise, we argue that to design sustainable knowledge management systems, HIS researchers need to implement systems with structures that create mutual benefits – i.e., encourage knowledge dissemination and sharing (e.g., open innovation) by actors in the healthcare industry. In a comprehensive and sustainable knowledge management system, however, not only corporations but also patients should be encouraged to share knowledge. Using this information, researchers and health service providers will be enabled to create optimized infrastructures, processes, and products (e.g., for predictive algorithms that improve treatment accuracy, or for assessing the likelihood of the occurrence of certain diseases and even of pandemics). At the same time, the trustworthiness of predictions and the anonymity of health information (and thus privacy) must be ensured. Bridging this duality of data sharing and knowledge dissemination, on the one hand, and protection of health information, on the other, is therefore essential for future HIS research.

This paper analyses the HIS literature within the IS research domain, prompted by the question of whether IS researchers are prepared to capture and anticipate exogenous changes and the consequences of current developments in the healthcare industry. While this review is limited to insights into the IS research domain and does not claim to offer insights into the health literature in general or related publications (e.g., governmental publications), we disclose several shortcomings and three key issues. Based on these, we provide initial guidance on how IS research can develop so that it is prepared to capture the expected large and long-lasting changes from current and possible future pandemics as well as the necessary adaptation of global healthcare industries affecting human agencies and experiences in all dimensions. Thus, while adaptations in the healthcare industry are already emerging, IS researchers have yet to develop a more comprehensive view of the healthcare industry. For this purpose, we provide a research agenda that is structured in terms of three areas of improvement: anticipating the range of actors’ behaviours, determining boundaries and fostering collaborations among actors in the healthcare industry, and creating sustainable knowledge management systems. In particular, addressing these areas will assist IS researchers in balancing the shortcomings of current HIS research with the unique contribution that IS research plays in analysing, advancing, and managing the healthcare industry. We are confident that IS research is not only capable of anticipating changes and consequences but also of actively shaping the future of the healthcare industry by promoting sustainable healthcare ecosystems, cultivating structures of mutual benefit and cooperation between actors, and realigning IS research to face the imminent transformation of the healthcare industry. IS research cannot contribute directly to solving the current pandemic problems; however, it can contribute indirectly triggering timely adaptations of novel technologies in global health systems, and proposing new processes, business models, and systematic changes that will prepare health systems to cope with increasing digitalization and emerging players whose push into the market enabled by the exogenous effects triggered by the pandemic.

While we are confident that the proposed research agenda based on the analysis of HIS literature provides fruitful arrays for being prepared in anticipating the future role of IS research for the healthcare industry, our results need to be reflected in light of their shortcomings. First and foremost, we recognize that the selection of literature, which is limited to the IS research domain, excludes other contextual factors that are not primarily considered by IS researchers. Thus, we cannot assume completeness, providing instead a broad overview of current issues in HIS research. In addition, possible biases may have arisen due to the qualitative analysis approach used. By independently coding and discussing codes to the point of theoretical saturation, we are confident that we largely eliminated biases in the thematic analysis. However, data saturation could not be achieved. This means that further insights could have emerged through the addition of other database searches and journals with a broader scope. Additionally, the initial sorting of papers into single defined phases of DSR research restricted multiple assignments that could have led to different results. However, we consider sorting as a necessary step of abstraction, especially given the large number of papers analysed.

We deliberately considered IS research, for which we have developed an agenda for potential future research avenues. For each of those avenues, researchers should go deeper into the subject matter in order to examine the complexity of the paths shown and to include them in the analysis (e.g., through in-depth case studies). However, it is also clear from the issues identified that IS researchers cannot solve current challenges by working on the pathways alone. In fact, the issues identified in the research agenda are only the starting point for further research, which should address the proposed issues step by step and in cooperation with other research disciplines. The latter is likely to generate further and deeper-rooted problems, as well as, in turn, future paths for research. Nevertheless, we are confident that this paper provides an important first step in opening up HIS research to better understand current developments in the healthcare industry. Further, by following and enhancing the proposed research pathways, we believe that HIS research can contribute to and support changes already taking place in the healthcare industry.

Change history

20 december 2021.

A Correction to this paper has been published: https://doi.org/10.1007/s12525-021-00518-8

Aanestad, M., Vassilakopoulou, P., & Ovrelid, E. (2019). Collaborative innovation in healthcare: boundary resources for peripheral actors. International Conference on Information Systems (ICIS), 1 , 1–17.

Google Scholar  

Abouzahra, M., & Tan, J. (2014). The multi-level impact of clinical decision support system: a framework and a call for mixed methods. Pacific Asia Conference on Information Systems (PACIS), 1–17.

Agarwal, R., Gao, G., DesRoches, C., & Jha, A. K. (2010). Digital transformation of healthcare: current status and the road ahead. Information Systems Research, 21 (4), 796–809. https://doi.org/10.1287/isre.1100.0327

Article   Google Scholar  

Akcura, M. T., & Ozdemir, Z. D. (2008). Physician learning and clinical decision support systems. Americas Conference on Information Systems  (AMCIS) , 1–10.

Akkerman, S. F., & Bakker, A. (2011). Boundary crossing and boundary objects. Review of Educational Research, 81 (2), 132–169. https://doi.org/10.3102/0034654311404435

Al-Qirim, N. (2003). Championing telemedicine in New Zealand: the case of utilizing video conferencing in psychiatry and dermatology. Americas Conference on Information Systems (AMCIS),  1–10.

Arvanitis, S., & Loukis, E. (2014). An empirical investigation of the impact of ICT on innovation in European hospitals. European Conference on Information Systems (ECIS),  1–13.

Babar, Y., Chan, J., & Choi, B. (2018). "Run Forrest Run!": measuring the impact of app-enabled performance and social feedback on running performance. Pacific Asia Conference on Information Systems (PACIS 2018),  160.

Benbya, H., Nan, N., Tandriverdi, H., & Yoo, Y. (2020). Complexity and information systems research in the emerging digital world. MIS Quarterly , 44 (1), 1–17. https://doi.org/10.25300/MISQ/2020/13304

Bhandari, G., & Maheshwari, B. (2009). Toward an integrated health information system for collaborative decision making and resource sharing: findings from a Canadian study.  Americas Conference on Information Systems (AMCIS),  648.

Bleicher, J. (2017). Contemporary Hermeneutics: Hermeneutics as method, philosophy, and critique (2nd ed.). Routledge.

Book   Google Scholar  

Blinn, N., & Kühne, M. (2013). Health information on the internet - state of the art and analysis. Business & Information Systems Engineering, 5 (4), 259–274. https://doi.org/10.1007/s12599-013-0274-4

Boell, S., & Cecez-Kecmanovic, D. (2014). A hermeneutic approach for conducting literature reviews and literature searches. Communication of the Association for Information Systems, 34 (12), 257–286.  https://doi.org/10.17705/1CAIS.03412

Braa, J., Monteiro, E., & Sahay, S. (2004). Networks of action: sustainable health information systems across developing countries. MIS Quarterly, 28 (3), 337–362. https://doi.org/10.2307/25148643

Bullinger, A., Rass, M., & Moeslein, K. (2012). Towards open innovation in health care. European Conference on Information Systems (ECIS),  1–13.

Byrd, L., & Byrd, T. (2009). Examining the effects of healthcare technology on operational cost. Americas Conference on Information Systems (AMCIS),  1–10.

CB Insights. (2018). How Google Plans to Use AI to Reinvent the $3 Trillion US Healthcare Industry . https://www.distilnfo.com/lifesciences/files/2018/11/CB-Insights_Google-Strategy-Healthcare.pdf

Chang, I. C., Hwang, H. G., Hung, M. C., Kuo, K. M., & Yen, D. C. (2009). Factors affecting cross-hospital exchange of electronic medical records. Information & Management, 46 (2), 109–115. https://doi.org/10.1016/j.im.2008.12.004

Chen, L., Baird, A., & Straub, D. (2019). An analysis of the evolving intellectual structure of health information systems research in the information systems discipline. Journal of the Association for Information Systems , 20 (8), 1023–1074. https://doi.org/10.17705/1jais.00561

Chiasson, M. W., & Davidson, E. (2004). Pushing the contextual envelope: developing and diffusing IS theory for health information systems research. Information and Organization, 14 (3), 155–188. https://doi.org/10.1016/j.infoandorg.2004.02.001

Choi, W., & Tulu, B. (2017). Effective use of user interfaces and user experience in an mHealth application. Hawaii International Conference on System Sciences (HICSS),  3803–3812. https://doi.org/10.24251/HICSS.2017.460

Cole, M., & Avison, D. (2007). The potential of hermeneutics in information systems research. European Journal of Information Systems, 16 (6), 820–833.  https://doi.org/10.1057/palgrave.ejis.3000725

Cilliers, P. (2001). Boundaries, hierarchies and networks in complex systems. International Journal of Innovation Management, 5 (2), 135–147. https://doi.org/10.1515/9781501502590-009

Corbin, J. M., & Strauss, A. (1990). Grounded theory research: procedures, canons, and evaluative criteria. Qualitative Sociology, 13 (1), 3–21. https://doi.org/10.1007/BF00988593

Currie, W. L., & Seddon, J. J. (2014). A cross-national analysis of eHealth in the European Union: some policy and research directions. Information & Management, 51 (6), 783–797. https://doi.org/10.1016/j.im.2014.04.004

Dehling, T., & Sunyaev, A. (2012). Architecture and design of a patient-friendly eHealth web application: patient information leaflets and supplementary services.  Americas Conference on Information Systems (AMCIS),  1–8.

Devadoss, P., & Pan, S. L. (2004). Leveraging eGovernment infrastructure for crisis management: lessons from managing SARS outbreak in Singapore. Americas Conference on Information Systems (AMCIS),  1–10.

Djamsbi, S., Fruhling, A., & Loiacono, E. (2009). The influence of affect, attitude and usefulness in the acceptance of telemedicine systems. Journal of Information Technology Theory and Application, 10 (1), 1–38.

Dünnebeil, S., Krcmar, H., Sunyaev, A., Leimeister, J., & M. (2013). Modular architecture of value-added applications for German healthcare telematics. Business & Information Systems Engineering, 5, 3–16. https://doi.org/10.1007/s12599-012-0243-3

Enkel, E., & Gassmann, O. (2010). Creative imitation: exploring the case of cross-industry innovation. R&d Management, 40 (3), 256–270. https://doi.org/10.1111/j.1467-9310.2010.00591.x

Fichman, R. G., Kohli, R., & Krishnan, R. (2011). The role of information systems in healthcare: current research and future trends. Information Systems Research, 22 (3), 419–428. https://doi.org/10.2307/23015587

Findikoglu, M., & Watson-Manheim, M. B. (2016). Linking macro-level goals to micro-level routines: EHR-enabled transformation of primary care services. Journal of Information Technology, 31 (4), 382–400. https://doi.org/10.1057/s41265-016-0023-5

Fisher, J., Burstein, F., Lynch, K., Lazarenko, K., & McKemmish, S. (2007). Health information websites: is the health consumer being well-served? Americas Conference on Information Systems (AMCIS),  1–10.

Fisman, R., Branstetter, L. G., & Foley, C. F. (2004). Do stronger intellectual property rights increase international technology transfer? Empirical evidence from US firm-level panel data . The World Bank, Washington, DC. https://doi.org/10.1596/1813-9450-3305

Gabel, M., Foege, J. N., & Nuesch, S. (2019). The (In)Effectiveness of incentives - a field experiment on the adoption of personal electronic health records.  International Conference on Information Systems (ICIS),  1–17.

Gantori, S., Issel, H., Donovan, P., Rose, B., Kane, L., Dennean, K., Ganter R., Sariyska, A., Wayne, G., Hyde, C., & Lee, A. (2020). Future of the Tech Economy . https://www.ubs.com/global/en/wealth-management/chief-investment-office/investment-opportunities/investing-in-the-future/2020/future-of-tech-economy.html

Gaskin, J., Berente, N., Lyytinen, K., & Yoo, Y. (2014). Toward generalizable sociomaterial inquiry. MIS Quarterly , 38 (3), 849–872. https://doi.org/10.25300/MISQ/2014/38.3.10

George, J. F., & Kohnke, E. (2018). Personal health record systems as boundary objects. Communication of the Association for Information Systems , 42 (1), 21–50. https://doi.org/10.17705/1CAIS.04202

Ghanvatkar, S., & Rajan, V. (2019). Deep recurrent neural networks for mortality prediction in intensive care using clinical time series at multiple resolution.  International Conference on Information Systems (ICIS),  1–18.

Gimpel, H., Nißen, M., & Görlitz, R. A. (2013). Quantifying the quantified self: a study on the motivation of patients to track their own health. International Conference on Information Systems (ICIS),  1–17.

Glaser, B. G., Strauss, A. L., & Strutzel, E. (1968). The discovery of grounded theory; strategies for qualitative research. Nursing Research, 17 (4), 364.

Greenwood, F. (2020, March). Google Want Your Data in Exchange for a Coronavirus Test. Foreign Policy. https://foreignpolicy.com/2020/03/30/google-personal-health-data-coronavirus-test-privacy-surveillance-silicon-valley/

Grondin, J. (2016). What is the hermeneutical cycle? In N. Keane & C. Lawn (Eds.), The Blackwell Companion to Hermeneutics,  299–305. Oxford, UK: Blackwell Publishing.

Hacklin, F. (2008). Management of Convergence in Innovation: Strategies and Capabilities for Value Creation Beyond Blurring Industry Boundaries . Zurich, Switzerland: Physica.

Hafermalz, E., & Riemer, K. (2016). Negotiating distance: "presencing work” in a case of remote telenursing. International Conference on Information Systems(ICIS), 1–13.

Hamari, J., & Koivisto, J. (2015). Why do people use gamification services? International Journal of Information Management, 35 (4), 419–431. https://doi.org/10.1016/j.ijinfomgt.2015.04.006

Heine, C., Herrler, R., Petsch, M., & Anhalt, C. (2003). ADAPT: Adaptive Multi-Agent Process Planning and Coordination of Clinical Trials. Americas Conference on Information Systems (AMCIS),  1823–1834.

Holland, S., Phimphachanh, C., Conn, C., & Segall, M. (1996). Impact of economic and institutional reforms on the health sector in Laos: implications for health system management. IDS Publications, Institute of Development Studies, 28 , 133–154.

Huerta, T. R., Thompson, M. A., Ford, E. W., & Ford, W. F. (2013). Electronic health record implementation and hospitals’ total factor productivity. Decision Support Systems, 55 (2), 450–458. https://doi.org/10.1016/j.dss.2012.10.004

Humphrey, S. E. (2011). What does great meta-analysis look like? Organizational Psychology Review, 1 (2), 99–103. https://doi.org/10.1177/2041386611401273

Idrissi, T. E., Idri, A., & Bakkoury, Z. (2019). Systematic map and review of predictive techniques in diabetes self-management. International Journal of Information Management, 46 , 263–277. https://doi.org/10.1016/j.ijinfomgt.2018.09.011

Kenny, C., & Connolly, R. (2016). Drivers of health information privacy concern: a comparison study.  Americas Conference on Information Systems (AMCIS),  1–10.

Kernick, D., & Mitchell, A. (2009). Working with lay people in health service research: a model of co-evolution based on complexity theory. Journal of Interprofessional Care, 24 (1), 31–40. https://doi.org/10.3109/13561820903012073

Kim, S. H., & Kwon, J. (2019). How Do EHRs and a meaningful use initiative affect breaches of patient information? Information Systems Research, 30 (4), 1107–1452. https://doi.org/10.1287/isre.2019.0858

Kim, S. (1992). Downward causation in emergentism and non-reductive physicalism. In A. Beckermann, H. Flohr, & J. Kim (Eds.). Emergence or Reduction. Essays on the Prospects of Nonreductive Physicalism (118–138). Berlin, Germany: de Gryuter.

King, N. (2009). An initial exploration of stakeholder benefit dependencies in ambulatory ePrescribing. Americas Conference on Information Systems (AMCIS),  1–10.

Klecun, E., Zhou, Y., Kankanhalli, A., Wee, Y. H., & Hibberd, R. (2019). The dynamics of institutional pressures and stakeholder behavior in national electronic health record implementations: a tale of two countries. Journal of Information Technology, 34 (4), 292–332. https://doi.org/10.1177/0268396218822478

Knight, J., Patrickson, M., & Gurd, B. (2008). Towards understanding apparent South Australian GP resistance to adopting Health Informatics systems.  Australasian Conference on Information Systems (ACIS),  492–501.

Kozlowski, S. W. J., Chao, G. T., Grand, J. A., Braun, M. T., & Kulijanin, G. (2013). Advancing multilevel research design: capturing the dynamics of emergence. Organizational Research Methods, 16 (4), 581–615. https://doi.org/10.1177/1094428113493119

Landi, H. (2020, April). Alphabet’s Verily rolls out COVID screening tool for health systems . FierceHealthcare. https://www.fiercehealthcare.com/tech/alphabet-s-verily-rolls-out-covid-screening-tool-for-health-systems

Lebcir, R. M., Choudrie, J., Atum, R. A., & Corker, R. J. (2008). Examining HIV and tuberculosis using a decision support systems computer simulation model: the case of the Russian Federation. Americas Conference on Information Systems (AMCIS),  1–18.

Lee, D., & Nilsson, P. (2020, March). Amazon auditions to be “the new Red Cross” in Covid-19 crisis . Financial Times. https://www.ft.com/content/220bf850-726c-11ea-ad98-044200cb277f

Lee, A. S., Thomas, M., & Baskerville, R. L. (2015). Going back to basics in design science: from the information technology artifact to the information systems artifact. Information Systems Journal, 25 (1), 5–21. https://doi.org/10.1111/isj.12054

Lee, C. H., Venkatraman, N., Tanriverdi, H., & Iyer, B. (2013). Complementary-based hypercompetition in the software industry. Strategic Management Journal, 31 (13), 1431–1356. https://doi.org/10.1002/smj.895

Leon, M. C., Nieto-Hipolito, J. I., Garibaldi-Beltran, J., Amaya-Parra, G., Luque-Morales, P., Magana-Espinoza, P., & Aquilar-Velazco, J. (2016). Designing a model of a digital ecosystem for healthcare and wellness using the BM canvas. Journal of Medical Systems, 40 (6), 144–154. https://doi.org/10.1007/s10916-016-0488-3

LeRouge, C., Mantzana, V., & Wilson, E. V. (2007). Healthcare information systems research, revelations and visions. European Journal of Information Systems, 16 (6), 669–671. https://doi.org/10.1057/palgrave.ejis.3000712

Leroy, J., Cova, B., & Salle, R. (2013). Zooming in VS zooming out on value co-creation: consequences for BtoB research. Industrial Marketing Management, 42 (7), 1102–1111. https://doi.org/10.1016/j.indmarman.2013.07.006

Li, L., Jeng, L., Naik, H. A., Allen, T., & Frontini, M. (2008). Creation of environmental health information system for public health service: a pilot study. Information Systems Frontiers, 10 (5), 531–542. https://doi.org/10.1007/s10796-008-9108-1

Lichtenberg, S., Greve, M., Brendel, A. B., & Kolbe, L. M. (2019). Towards the design of a mobile application to support decentralized healthcare in developing countries – The case of diabetes care in eSwatini. Americas Conference on Information Systems (AMCIS),  1–10.

Lin, Y.-K., Chen, H., Brown, R. A., Li, S.-H., & Yang, H.-J. (2017). Healthcare predictive analytics for risk profiling in chronic care: a Bayesian multitask learning approach. MIS Quarterly , 41 (2), 473–495. https://doi.org/10.25300/MISQ/2017/41.2.07

MacLure, M. (2005). Clarity bordering on stupidity: where’s the quality in systematic review? Journal of Education Policy, 20 (4), 393–416. https://doi.org/10.4324/9780203609156

Maheshwari, M., Hassan, T., & Chatterjee, S. (2006). A framework for designing healthy lifestyle management information system.  Americas Conference on Information Systems (AMCIS),  2811–2817.

Matavire, R., & Brown, I. (2008). Investigating the use of “Grounded Theory” in information systems research. Proceedings of the 2008 annual research conference of the South African Institute of Computer Scientists and Information Technologists on IT research in developing countries,  139–147.  https://doi.org/10.1145/1456659.1456676

Mitsa, T., Fortier, P. J., Shrestha, A., Yang, G., & Dluhy, N. M. (2007). Information systems and healthcare XXI: a dynamic, client-centric, point-of-care system for the novice nurse. Communication of the Association for Information Systems , 19 , 740–761. https://doi.org/10.17705/1CAIS.01936

Motamarri, S., Akter, S., Ray, P., & Tseng, C.-L. (2014). Distinguishing “mHealth” from other healthcare services in a developing country: a study from the service quality perspective. Communication of the Association for Information Systems , 34 (1), 669–692. https://doi.org/10.17705/1CAIS.03434

Mun, C., & Lee, O. (2017). Integrated supporting platform for the visually impaired: using smart devices. International Conference on Information Systems (ICIS),  1–18.

Nambisan, S., Lyytinen, K., Majchrzak, A., & Song, M. (2017). Digital innovation management: reinventing innovation management research in a digital world. MIS Quarterly , 41 (1), 223–238. https://doi.org/10.25300/MISQ/2017/41:1.03

Neben, T., Seeger, A. M., Kramer, T., Knigge, S., White, A. J., & Alpers, G. W. (2016). Make the most of waiting: theory-driven design of a pre-psychotherapy mobile health application.  Americas Conference on Information Systems (AMCIS),  1–10.

Nicholls-Nixon, C. L., & Jasinski, D. (1995). The blurring of industry boundaries: an explanatory model applied to telecommunications. Industrial and Corporate Change, 4 (4), 755–768. https://doi.org/10.1093/icc/4.4.755

Offermann, P., Blom, S., Schönherr, M., & Bub, U. (2010). Artifact types in information systems design science – a literature review. International Conference on Design Science Research in Information Systems,  77–92. https://doi.org/10.1007/978-3-642-13335-0_6

Palvia, P., Kakhki, M. D., Ghoshal, T., Uppala, V., & Wang, W. (2015). Methodological and topic trends in information systems research: a meta-analysis of IS journals. Communication of the Association for Information Systems , 37 (30), 630–650. https://doi.org/10.17705/1CAIS.03730

Park, H. A. (2016). Are we ready for the fourth industrial revolution? Yearbook of Medical Informatics , (1), 1–3. https://doi.org/10.15265/IY-2016-052

Patel, M., Shah, A., Shah, K., & Plachkinova, M. (2018). Designing a mobile app to help young adults develop and maintain mental well-being.  Americas Conference on Information Systems (AMCIS),  1–10.

Paul, L. D., & McDaniel, R. R. J. (2004). A field study of the effect of interpersonal trust on virtual collaborative relationship performance. MIS Quarterly, 28 (2), 183–227.  https://doi.org/10.2307/25148633

Peffers, K., Tuunanen, T., Rothenberger, M. A., & Chatterjee, S. (2008). A design science research methodology for information systems research. Journal of Management Information Systems, 24 (3), 45–78. https://doi.org/10.2753/MIS0742-1222240302

Pletikosa, I., Kowatsch, T., Büchter, D., Brogle, B., Dintheer, A., Wiegand, D., Durrer, D., l'Allemand-Jander, D., Schutz, Y., Maass, W. (2014). Health information system for obesity prevention and treatment of children and adolescents. European Conference on Information Systems (ECIS),  1–13.

Preko, M., Boateng, R., & Effah, J. (2019). Health informatics and brain drain mitigation in Ghana.  Americas Conference on Information Systems (AMCIS),  1–10.

Rivard, P. E., Rosen, A. K., & Carroll, J. S. (2006). Enhancing patient safety through organizational learning: are patient safety indicators a step in the right direction?  Health Services Research , 41(4p2), 1633–1653. https://doi.org/10.1111/j.1475-6773.2006.00569.x

Rodon, J., & Silva, L. (2015). Exploring the formation of a healthcare information infrastructure: hierarchy or meshwork? Journal of the Association for Information System, 16 (5), 394–417. https://doi.org/10.17705/1JAIS.00395

Roehrig, S., & Knorr, K. (2000). Toward a secure web based health care application. European Conference on Information Systems (ECIS),  1–8. https://doi.org/10.4018/978-1-930708-13-6.ch007

Romanow, D., Rai, A., & Keil, M. (2018). CPOE-Enabled Coordination: appropriation for deep structure use and impacts on patient outcomes. MIS Quarterly , 42 (1), 189–212. https://doi.org/10.25300/MISQ/2018/13275

Ryan, J., Doster, B., Daily, S., & Lewis, C. (2019). Seeking operational excellence via the digital transformation of perioperative scheduling. Americas Conference on Information Systems (AMCIS),  1–10.

Rycroft, R. W., & Kash, D. E. (2004). Self-organizing innovation networks: implications for globalization. Technovation, 24 (3), 187–197. https://doi.org/10.1016/S0166-4972(03)00092-0

Salthe, S. N. (1985). Evolving Hierarchical Systems . Columbia University Press.  https://doi.org/10.7312/salt91068

Samentinger, J., Rozenblit, J., Lysecky, R., & Ott, P. (2015). Security challenges for medical devices. Communications of the Association for Information Systems, 58 (4), 74–82.  https://doi.org/10.1145/2667218

Saunders, B., Sim, J., Kingstone, T., Baker, S., Waterfield, J., Bartlam, B., Burroughs, H., & Jinks, C. (2018). Saturation in qualitative research: exploring its conceptualization and operationalization. Quality & Quantity, 52 (4), 1893–1907. https://doi.org/10.1007/s11135-017-0574-8

Schweiger, A., Sunyaev, A., Leimeister, J. M., & Krcmar, H. (2007). Information systems and healthcare XX: toward seamless healthcare with software agents. Communications of the Association for Information Systems , 19 (33), 392–710. https://doi.org/10.17705/1CAIS.01933

Schwetschke, S., & Durugbo, C. (2018). How firms synergise: understanding motives and management of co-creation for business-to-business services. International Journal of Technology Management, 76 (3–4), 258–284. https://doi.org/10.1504/IJTM.2018.091289

Setia, P., Setia, M., Krishnan, R., & Sambamurthy, V. (2011). The effects of the assimilation and use of IT applications on financial performance in healthcare organizations. Journal of the Association for Information System , 12 (Special Issue), 274–298. https://doi.org/10.17705/1jais.0060

Shahmoradi, L., Safadari, R., & Jimma, W. (2017). Knowledge management implementation and the tools utilized in healthcare for evidence-based decision making: A systematic review. Ethiopian Journal of Health Sciences, 27 (5), 541–558. https://doi.org/10.4314/ejhs.v27i5.13

Sherer, S. A. (2014). Patients are not simply health it users or consumers: the case for “e Healthicant” applications. Communications of the Association for Information systems , 34(1), 17. https://doi.org/10.17705/1CAIS.03417 .

Singh, R., Mathiassen, L., Stachura, M. E., & Astapova, E. V. (2011). Dynamic capabilities in home health: IT-enabled transformation of post-acute care. Journal of the Association for Information System , 12 (2), 163–188. https://doi.org/10.17705/1jais.00257

Sujanto, F., F., B., Ceglowski, A., & Churilov, L. (2008). Application of domain ontology for decision support in medical emergency coordination. Americas Conference on Information Systems (AMCIS),  1–10.

Theobalt, A., Emrich, A., Werth, D., & Loos, P. (2013). A conceptual architecture for an ICT-based personal health system for cardiac rehabilitation. Americas Conference on Information Systems (AMCIS),  1–10.

Therrien, M.-C., Normandin, J.-M., & Denis, J.-L. (2017). Bridging complexity theory and resilience to develop surge capacity in health systems. Journal of Health Organisation and Management, 31 (1), 96–109. https://doi.org/10.1108/JHOM-04-2016-0067

Tillett, A. (2020, April). Amazon to store data from virus tracing app. Financial Review. https://www.afr.com/politics/federal/amazon-to-store-data-from-virus-tracing-app-20200424-p54mwq

Tokar, O., Batoroev, K., & Böhmann, T. (2015). A framework for analyzing patient-centered mobile applications for mental health. Americas Conference on Information Systems (AMCIS),  1–15.

Tsai, J. M., Cheng, M. J., Tsai, H. H., Hung, S. W., & Chen, Y. L. (2019). Acceptance and resistance of telehealth: the perspective of dual-factor concepts in technology adoption. International Journal of Information Management, 49 , 34–44. https://doi.org/10.1016/j.ijinfomgt.2019.03.003

Ure, J., Procter, R., Lin, Y. W., Hartswood, M., Anderson, S., Lloyd, S., Wardlaw, J., Gonzalez-Velez, H., & Ho, K. (2009). The development of data infrastructures for eHealth: a socio-technical perspective. Journal of the Association for Information Systems , 10 (5), 415–429. https://doi.org/10.17705/1jais.00197

Varshney, U. (2004). Using wireless networks for enhanced monitoring of patients.  Americas Conference on Information Systems (AMCIS),  1–10. https://doi.org/10.1504/IJHTM.2005.007009

Volland, D., Korak, K., & Kowatsch, T. (2014). A health information system that extends healthcare professional-patient communication.  European Conference on Information Systems (ECIS),  1–10.

vom Brocke, J., Simons, A., Reimer, K., Niehaves, B., Plattfaut, R., & Cleven, A. (2015). Standing on the shoulders of giants: challenges and recommendations of literature search in information systems research. Communications of the Association for Information Systems, 37 (1), 205–224.  https://doi.org/10.17705/1CAIS.03709

vom Brocke, J., Simons, A., Niehaves, B., Reimer, K., Plattfaut, R., & Cleven, A. (2009): Reconstructing the giant: on the importance of rigour in documenting the literature search process. European Conference on Information Systems (ECIS) , 1–12.

Walsh, D., & Downe, S. (2005). Meta-synthesis method for qualitative research: a literature review, methodological issues in nursing. Journal of Advanced Nursing, 50 (2), 204–211. https://doi.org/10.1111/j.1365-2648.2005.03380.x

Webster, J., & Watson, R. T. (2002). Analyzing the past to prepare for the future: writing a literature review. MIS Quarterly, 26 (2), 13–23. https://doi.org/10.2307/4132319

Werts, N., & Adya, M. (2000). Data mining in healthcare: issues and a research agenda. Americas Conference on Information Systems (AMCIS),  98.

Wessel, L., Gersch, M., & Harloff, E. (2017). Talking past each other - a discursive approach to the formation of societal-level information pathologies in the context of the electronic health card in Germany. Business & Information Systems Engineering, 59 (1), 23–40. https://doi.org/10.1007/s12599-016-0462-0

Winter, S. J., & Butler, B. S. (2011). Creating bigger problems: grand challenges as boundary objects and the legitimacy of the information systems field. Journal of Information Technology, 26 (2), 99–108. https://doi.org/10.1057/jit.2011.6

Wolfswinkel, J. F., Furtmueller, E., & Wilderom, C. P. M. (2013). Using grounded theory as a method for rigorously reviewing literature. European Journal of Information Systems, 22 (1), 45–55. https://doi.org/10.1057/ejis.2011.51

Malpass, D., (2020, March). Coronavirus highlights the need to strengthen health systems. Worldbank Blogs. https://blogs.worldbank.org/voices/coronavirus-covid19-highlights-need-strengthen-health-systems

Wu, I.-L., & Hu, Y.-P. (2012). Examining knowledge management enabled performance for hospital professionals: a dynamic capability view and the mediating role of process capability. Journal of the Association for Information System , 13 (12), 976–999. https://doi.org/10.17705/1jais.00319

Yang, A., & Varshney, U. (2016). A taxonomy for mobile health implementation and evaluation. Americas Conference on Information Systems (AMCIS),  1–10.

Zafar, H., & Sneha, S. (2012). Ubiquitous healthcare information system: toward crossing the security chasm. Communication of the Association for Information Systems , 31 (9), 193–206. https://doi.org/10.17705/1CAIS.03109

Zhang, L., & Pang, M. S. (2019). Does sharing make my data more insecure? An empirical study on health information exchange and data breaches. International Conference on Information Systems (ICIS),  1–14.

Download references

Open Access funding enabled and organized by Projekt DEAL.

Author information

Authors and affiliations.

Chair for Digitalization and Process Management, Philipps-University Marburg, Universitätsstraße 24, 35037, Marburg, Germany

Nadine Ostern

Frankfurt School of Finance & Management, ProcessLab, Adickesallee 32-34, 60322, Frankfurt am Main, Germany

Guido Perscheid, Caroline Reelitz & Jürgen Moormann

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Nadine Ostern .

Additional information

Responsible Editor: Shengnan Han

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

The original online version of this article was revised: In this article the affiliation of authors are incorrect. Given here are the correct affiliations.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary file1 (DOCX 195 KB)

Rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Ostern, N., Perscheid, G., Reelitz, C. et al. Keeping pace with the healthcare transformation: a literature review and research agenda for a new decade of health information systems research. Electron Markets 31 , 901–921 (2021). https://doi.org/10.1007/s12525-021-00484-1

Download citation

Received : 11 December 2020

Accepted : 26 May 2021

Published : 17 July 2021

Issue Date : December 2021

DOI : https://doi.org/10.1007/s12525-021-00484-1

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Health information systems research

JEL Classifications

• Find a journal
• Publish with us
• Track your research

• Open access
• Published: 03 July 2024

Healthcare leaders navigating complexity: a scoping review of key trends in future roles and competencies

• Samantha Spanos   ORCID: orcid.org/0000-0003-3734-3907 1 ,
• Elle Leask   ORCID: orcid.org/0000-0003-1698-9151 1 ,
• Romika Patel   ORCID: orcid.org/0009-0000-6523-8798 1 ,
• Michael Datyner 1 ,
• Erwin Loh   ORCID: orcid.org/0000-0001-7157-0826 2 &
• Jeffrey Braithwaite   ORCID: orcid.org/0000-0003-0296-4957 1  

BMC Medical Education volume  24 , Article number:  720 ( 2024 ) Cite this article

164 Accesses

Metrics details

As healthcare systems rapidly become more complex, healthcare leaders are navigating expanding role scopes and increasingly varied tasks to ensure the provision of high-quality patient care. Despite a range of leadership theories, models, and training curricula to guide leadership development, the roles and competencies required by leaders in the context of emerging healthcare challenges (e.g., disruptive technologies, ageing populations, and burnt-out workforces) have not been sufficiently well conceptualized. This scoping review aimed to examine these roles and competencies through a deep dive into the contemporary academic and targeted gray literature on future trends in healthcare leadership roles and competencies.

Three electronic databases (Business Source Premier, Medline, and Embase) were searched from January 2018 to February 2023 for peer-reviewed literature on key future trends in leadership roles and competencies. Websites of reputable healthcare- and leadership-focused organizations were also searched. Data were analyzed using descriptive statistics and thematic analysis to explore both the range and depth of literature and the key concepts underlying leadership roles and competencies.

From an initial 348 articles identified in the literature and screened for relevance, 39 articles were included in data synthesis. Future leadership roles and competencies were related to four key themes: innovation and adaptation (e.g., flexibility and vision setting), collaboration and communication (e.g., relationship and trust building), self-development and self-awareness (e.g., experiential learning and self-examination), and consumer and community focus (e.g., public health messaging). In each of these areas, a broad range of strategies and approaches contributed to effective leadership under conditions of growing complexity, and a diverse array of contexts and situations for which these roles and competencies are applicable.

Conclusions

This research highlights the inherent interdependence of leadership requirements and health system complexity. Rather than as sets of roles and competencies, effective healthcare leadership might be better conceptualized as a set of broad goals to pursue that include fostering collaboration amongst stakeholders, building cultures of capacity, and continuously innovating for improved quality of care.

Peer Review reports

Healthcare leadership has grown in scope and importance in response to the increasing complexity of healthcare delivery [ 1 ]. Healthcare systems have become increasingly multifaceted, delivering a vast array of services across multiple levels, from preventative and primary care to acute, specialized care, and long-term care, to address the care needs of a changing population [ 1 ]. As populations age, chronic diseases rise, and the epidemiology and demographics of disease shift, new models of care rapidly emerge to address the ever-expanding spectrum of patient needs [ 2 ]. Advancements in technologies, tests and treatments and personalized medicine come with regulatory and ethical implications, and a growth in workforce specializations [ 3 , 4 ]. Healthcare leaders are navigating evermore complex webs of actors in the system – doctors, nurses, technicians, administrators, insurers, and patients – striving to balance priorities, foster collaboration, and provide strategic direction toward high-quality and safe patient care [ 5 ]. At the same time as running complex services, healthcare leaders need to continually assess, implement, and govern new technologies and services, adhere to the latest regulations and guidelines, operate within the confines of budgetary allocations, and meet growing consumer expectations for affordable and accessible care [ 6 , 7 ].

Competent healthcare leadership is widely considered to be critical for improving patient safety, system performance, and the effectiveness of healthcare teams [ 8 , 9 , 10 ]. Leadership has been identified as a key shaping influence on organizational culture [ 11 ], including workplace commitment to safety [ 12 ], and on preventing workforce burnout [ 13 , 14 ]. The increased need for multidisciplinary and integrated care models has shed growing light on the leadership roles of clinicians, including physicians, nurses, and allied health practitioners [ 15 , 16 , 17 ]. Individuals with both clinical and leadership expertise have been considered vital in complex healthcare landscapes because of their ability to balance administrative needs while prioritizing safety and high-quality care provision [ 18 , 19 , 20 , 21 , 22 ]. For example, physician leaders, through their deep understanding of clinical care and their credibility and influence, have been considered best able to devise strategies that improve patient care amidst stringent financial conditions [ 23 , 24 , 25 , 26 ]. Clinical leaders, particularly physician leaders, might also be of key importance for facilitating the success of collaborative care and care integration [ 27 ].

The formalization of healthcare leadership emerged as the importance of specialized healthcare leadership skills became increasingly needed, recognized and understood [ 1 , 28 , 29 ]. Leadership in healthcare has been conceptualized in several different ways, and a multitude of theories, frameworks, and models have been proposed to explain leadership roles and responsibilities [ 30 , 31 , 32 , 33 ]. For example, the CanMEDS Framework describes the Leader Role of physicians, which is comprised of key and enabling competencies, tasks, and abilities [ 34 , 35 ], and adaptations to this Framework emphasize the varying roles that leadership comprises and the competencies that fulfill them [ 36 ]. Although these frameworks present a good starting point for articulating leadership role scopes and their associated competencies, many fall short in explaining how leaders navigate complex, dynamic, multi-dimensional, and highly variable healthcare systems [ 37 ]. This is becoming increasingly recognized; CanMEDS is due to be updated in 2025 to incorporate competencies related to complexity [ 38 ]. Meanwhile, on the front lines, lack of role clarity and ambiguity about tasks and responsibilities presents a significant barrier for healthcare leaders [ 1 , 15 ]. In complex and unpredictable systems like healthcare, leaders spend substantial time ‘sense-making’, understanding, prioritizing and responding adaptively according to the needs of the situation [ 39 , 40 ]. The latest research on future healthcare trends tells us that increasing complexity associated with digital innovation, healthcare costs, regulatory compliance, sustainability concerns and equitable resource distribution will pose challenges to all actors in health systems [ 41 , 42 , 43 , 44 , 45 ]. In the face of these emerging challenges, it is vital to understand the range and type of roles and competencies that leaders will need to fulfil in the imminent future.

The aim of this scoping review is to examine the literature on the key trends in roles and competencies required for healthcare leaders in the future. We conceptualized ‘competencies’ as the attributes, skills, and abilities that comprise the fulfilment of varying leadership roles, as informed by the CanMEDS Framework [ 34 , 36 ]. Scoping review methodology was utilized to capture a broad range of literature types and identify key themes or groupings of future trends in leadership roles and competencies. Rather than focusing on answering specific questions (as per previous systematic reviews on leadership [ 46 , 47 ]) or developing theory (by utilizing a theoretical review approach to leadership literature [ 48 , 49 ]), we sought to map and identify patterns and trends within the leadership literature [ 50 ]. To investigate trends in leadership roles and competencies, we targeted emerging perspectives from key reputable thought leaders to supplement academic research [ 51 , 52 ].

The conduct and reporting of this review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews (PRISMA-ScR) guidelines [ 53 ].

Search strategy

Comprehensive search strategies were developed, adapting search strategies utilized in a previous systematic review on physician leadership [ 26 ], and receiving input and expertise from two clinical librarians at Macquarie University (see supplementary file 1 for database search strategies). Medline, Embase, and Business Source Premier were searched from January 2018 to February 2023 to enable meaningful inferences to be made about future trends based on current perspectives. To capture key trends, patterns, shifts, and forecast changes to healthcare leadership, the Medline database search was limited to the ‘Trends’ subheading, “ used for the manner in which a subject changes, qualitatively or quantitatively, with time, whether past, present, or future. Includes “forecasting” & “futurology" ” (see supplementary file 1 ) [ 54 ]. For Embase and Business Source Premier, the ‘Trends’ subheading was not available, and instead key search terms were included to capture future trends, including “predict*”, “forecast*”, “shift*” and “transform*”. Efforts were made to locate texts that could not be retrieved, by searching Macquarie University’s digital library records and contacting authors to request the full text.

To complement the database searches, targeted searches of the Faculty of Medical Leadership and Management (FMLM; UK) website and The King’s Fund (UK) website were undertaken to identify emerging perspectives on the future roles and requirements of healthcare leaders. Targeted website searches can aid in uncovering unpublished yet relevant research identified by advocacy organizations or subject specialists, and research potentially missed by database searches [ 52 , 55 ]. Key search terms entered into the websites included ‘future healthcare’, ‘medical leader’, ‘clinical leader’, ‘medical manager’, ‘physician executive’, and ‘education and training’. We included articles that focused on leaders with a clinical background and leaders without a clinical background, to provide a comprehensive overview of leadership roles and requirements of reference to health systems [ 26 ].

Article selection

Database literature search.

References were uploaded into online data management software Rayyan [ 56 ], and duplicate records were identified and removed. Titles and abstracts of results were screened by three team members (SS, EL, RP) according to the inclusion and exclusion criteria (Table  1 ). Articles were included if they focused on future trends in the roles, competencies, attributes, or requirements of healthcare leaders, and if they reported on countries within the Organization for Economic Co-operation and Development (OECD). We limited our search to OECD countries to maximize the generalizability of findings within a developed context and enable meaningful trends to be identified. A subset of the articles was screened by all three team members to ensure that decisions were being made in a standardized manner. After this article subset was screened, the three team members discussed screening decisions, and disagreements were resolved by consensus or through discussion with JB [ 57 ]. During this process, two further exclusion criteria (#4 and #5, Table  1 ) were added to ensure that the screening process adhered to the aim of the current review. We excluded articles that focused on theories and definitions of leadership (e.g., for the purpose of developing educational or professional frameworks) without highlighting trends or changes in roles and competencies for future leadership. We also excluded articles that focused on healthcare interventions in which leaders may have been participants, but their roles or competencies were not the focus. Articles included at title and abstract screening were independently read in full and assessed for eligibility. Disagreements about inclusion were resolved through discussion, with JB available for arbitration if necessary. It was determined at this stage that if articles were conference abstracts in which the full presentation could not be accessed, the article of focus was sought and included in the analysis.

Targeted gray literature search

References were screened according to the inclusion and exclusion criteria (Table  1 ), except that articles only needed to report (rather than focus) on future leadership roles and requirements. This is because we wanted to ensure that our analysis broadly captured the most recent sources of information on healthcare leadership requirements, even if these sources did not focus exclusively on leadership.

Data charting process

Data from all records were appraised and charted simultaneously using a purpose-designed Excel data charting form designed by SS (and subsequently reviewed and endorsed by RP and EL). Multimedia records arising from targeted gray literature searches were listened to and transcribed by RP and checked by SS. Extracted data included article details (authors, year, country, text type), leadership focus (training or educational approaches, styles of leadership), and major and minor themes. Database literature were extracted first to identify and develop themes, and the targeted gray literature were extracted second to extend and embellish those themes.

Synthesis of results

Data from included articles were synthesized according to the Arksey and O’Malley framework for scoping reviews, selected for its detailed guidance on data collation, synthesis, and presentation [ 58 ]. The breadth, range, and type of data were analyzed using descriptive statistics, and underlying groups of leadership roles and competencies were analyzed using thematic analysis. First, the authorship team familiarized themselves with the articles to gain a broad overview of contexts in which leadership was discussed. An inductive approach was used to identify emerging themes of leadership roles and competencies in the database literature, where common concepts were identified, coded, and grouped together to form themes. Team discussion facilitated the final set of themes that were interpreted from the data. During this process, the extracted data were compared to the codes, groups, and resultant themes to examine the degree of consistency between the data and the interpreted findings. Where inconsistencies were identified, suggested changes (e.g., to code labels or groupings) were compared, and the most appropriate changes adopted. Targeted gray literature sources were deductively analyzed according to the identified themes.

Selection of sources of evidence

Figure  1 displays the process of identification and screening of included studies. Database searches yielded 160 records, from which 11 duplicates were removed. The remaining 149 database records were screened by title and abstract, after which a further 114 records were excluded. Of the remaining 35 that were assessed for eligibility, 22 were excluded, and 13 were included in the current review. Targeted gray literature searches yielded an additional 188 records, from which 146 were identified as duplicates and removed. The remaining 42 records were read in full and assessed for eligibility, from which a further 16 were excluded, and 26 were included in the current review. In total, 39 records were retained and synthesized.

PRISMA flowchart displaying the process of identification and selection of included articles

Characteristics of sources of evidence

The characteristics of the included records are displayed in Tables  2 and 3 . Of the database literature, most articles were published in the USA ( n  = 11), and the remaining two articles were published in Canada and Australia. Seven articles were empirical; three studies employed qualitative methods [ 59 , 60 , 61 ], three were quantitative [ 62 , 63 , 64 ], and one mixed methods [ 65 ]. Six articles were non-empirical; three were perspective pieces [ 66 , 67 , 68 ], and three were reports on training or organizational interventions [ 69 , 70 , 71 ]. Of the targeted literature, blog-type articles were most common ( n  = 11) [ 72 , 73 , 74 , 75 , 76 , 77 , 78 , 79 , 80 , 81 , 82 ], followed by news articles ( n  = 5) [ 83 , 84 , 85 , 86 , 87 ], reports ( n  = 4) [ 88 , 89 , 90 , 91 ], editorials ( n  = 2) [ 92 , 93 ], podcasts ( n  = 2) [ 94 , 95 ], and video and interview transcripts ( n  = 2) [ 96 , 97 ]. As targeted gray sources selected were The King’s Fund website and the FMLM website, the records from these websites were published in the UK.

Leadership roles and competencies

All 13 articles derived from the database searches focused on innovation and adaptation in future leadership. Two empirical articles reported on the ways in which clinical and non-clinical leaders innovated during the COVID-19 pandemic, rapidly designing new models of hospital care [ 61 ] and extending their roles to encompass the implementation of virtual leadership [ 64 ]. Qualitative investigations explored the importance of entrepreneurial leadership for implementing clinical genomics [ 59 ] and key leadership attributes for practice-level innovation and sustainability [ 60 ]. Four articles examined leadership training approaches that build physicians’ capacity to understand, adapt to, and manage change, overcome resistance, and think entrepreneurially [ 62 , 63 , 65 , 70 ]. Two reports described the necessity for healthcare leaders to be able to create a shared vision for an organization; one highlighted the importance of leaders being confident and “self-propelled to intervene” [ 69 ], and one emphasized physician leaders’ credibility as a catalyst for change management among healthcare providers [ 71 ]. The latter report also identified that visible and committed leadership that is sensitive to workplace cultures is critical for the success of change management activities [ 71 ]. Three perspective pieces discussed increasing opportunities for medical and other clinical leaders to create positive change in increasingly complex healthcare landscapes and fulfill the demands of the industry and public [ 66 , 67 , 68 ].

In the targeted gray literature, 19 of 26 records (73%) focused on innovative and adaptive leadership. Records primarily explored adaptive leadership behaviors during COVID-19, such as service redesign, introducing improved flexibility, learning mechanisms, and support platforms [ 73 , 76 , 77 , 97 ], and future innovation to manage climate change impacts [ 81 ], growing inequities [ 89 ], and emerging technologies [ 75 , 83 , 94 , 96 ]. Comfort with change, vision setting, and a desire to innovate were emphasized as key leadership attributes for future healthcare [ 82 , 83 , 88 , 96 ]. Records also explored how to best train and develop leaders for transforming health systems, including the National Health Service (NHS) [ 84 , 90 , 96 ]. New leadership training structures were proposed that foster innovation and adaptability in leaders [ 80 , 90 , 96 ] and encourage flexibility for cross-disciplinary learning.

Collaboration and communication  was a second theme that emerged across all 13 database articles. Three studies explored how collaborative leadership can foster innovation with regards to implementing genomics testing [ 59 ], creating new work models during COVID-19 [ 61 ] and developing new leadership styles via telecommunications [ 64 ]. Six articles focused on the importance of collaborating to build relationships across organisations [ 67 , 68 , 71 ] and within teams [ 65 , 69 , 70 ]. Three articles highlighted that effective communication contributes to organizational success, through fostering psychologically safe cultures [ 60 , 66 ] and generating the trust and rapport necessary for implementing technological innovations [ 71 ]. Two studies examined the impact of leadership training on physicians’ communication competencies [ 62 , 63 ].

In the targeted gray literature, 17 of 26 records (65%) focused on collaboration and communication. Records discussed specific initiatives to improve communication in clinical teams, such as staff surveys, daily huddles, and dedicated days for networking [ 75 , 77 , 80 , 95 ]. Cross-boundary collaboration and collective leadership (e.g., between clinicians and managers) [ 83 ] were advocated as a means to solve challenges [ 81 , 90 ], help build public trust [ 79 , 83 ], and improve quality of care [ 78 , 83 , 85 , 94 ]. Twelve records focused on the importance of team and leadership collaboration to create positive workplace cultures and improve staff wellbeing, through communication strategies such as openness and honesty [ 78 , 80 , 95 ], active listening and empathy [ 73 , 78 , 86 , 88 , 90 ], transparency [ 88 , 94 , 95 ], and inclusivity [ 85 , 94 ]. Three articles emphasized that encouraging staff autonomy, building trust, and demonstrating compassion facilitate better quality care than demanding and punitive leadership actions [ 73 , 74 , 88 ].

Nine of 13 database articles (69%) focused on a third theme, self-development and self-awareness in leadership. Four articles examined approaches to leadership development that incorporated self-development and self-awareness (e.g., personality testing) [ 63 , 65 , 69 , 70 ], with two articles describing these competencies as enablers for the development of other more advanced competencies (e.g., execution) [ 69 , 70 ]. Similar competencies explored included landscape awareness [ 60 ], self-organisation [ 60 ], emotional intelligence [ 64 ], and self-examination, the last of which was described as essential to gain skills beyond clinical roles [ 68 ], facilitate positive perceptions of others [ 66 ], and to remain relevant and effective in a changing healthcare environment [ 67 ]. One article also proposed strategies such as journaling, mindfulness, and feedback to encourage ongoing reflection on leadership decisions and biases [ 67 ].

In the targeted gray literature, seven of 26 records (27%) focused on self-development and self-awareness. Records examinedd the importance of continual personal leadership development, including mentoring and experiential learning, to facilitate understanding of one’s own skills [ 78 , 80 , 97 ]. Tools to facilitate self-reflection in physician leaders were advocated including the FMLM smartphone app [ 92 ] and leadership longitudinal assessments [ 91 ]. Self-care and resilience practices (e.g., meditation, social support) were also advocated for physician leaders as a means to manage “greater levels of stress and responsibility” [ 94 ].

Consumer engagement and advocacy  was a fourth theme and a focus of nine targeted gray literature records (35%). Records discussed patient and community engagement as essential for health system improvement, and examples included involving patients in health service design [ 74 , 77 ], creating channels of ongoing dialogue [ 79 , 83 ] and building stronger health system-community relationships [ 79 , 88 ]. Two records described the importance of public health messaging in improving health literacy [ 83 ] and countering misinformation [ 86 ], and two focused on the role of leaders in advocating for social justice and striving to improve equitable outcomes [ 75 , 93 ].

This scoping review identified 39 key resources that explored future trends in healthcare leadership roles and competencies. These records were derived from a combination of academic and targeted gray literature searches, juxtaposed and synthesized to build understanding of leadership to improve health systems into the future. Four themes of competencies emerged from the findings – innovation and adaptation, communication and collaboration, self-development and self-awareness, and consumer engagement and advocacy.

The competencies of healthcare leaders given the most attention in the literature over the last five years relate to innovation and adaptability . Both the academic and targeted gray literature focused on how leaders, clinical and non-clinical, demonstrated innovativeness and adapted to the demands of COVID-19, including rethinking and redesigning systems to support staff and patients [ 64 , 77 ]. The second focus of the literature on innovation and adaptability was geared toward the development of these capacities in leaders through education and training, as well as through opportunities for leaders to actualize their skills [ 70 , 90 ]. The literature indicated that as the complexity of healthcare is accelerating, leaders must both understand, and have opportunities to demonstrate, innovation amidst dynamic, variable, and demanding environments [ 59 , 60 , 71 ]. This aligns with prior research demonstrating that innovation uptake requires strong change management, and the ability to rapidly assess, understand, and apply innovative changes (e.g., medical technologies) [ 1 , 98 ]. While innovations might improve the system’s ability to deal with complex challenges in the long-term, their implementation can be challenged by a number of moving parts – including workforce changes, new rules and regulations, fluctuating resources and new patient groups – which leaders must consider and appropriately plan for [ 99 , 100 ]. Perhaps an even greater challenge for leaders to overcome when embracing innovation is the tendency for growing complexity to lock the organization into suboptimal conditions (i.e., inertia) [ 101 ]. Building awareness of the interacting components of complex systems and the flexibility required for adaptation and resilience should be a key focus of healthcare leadership education and training [ 102 ].

Competencies associated with communication and collaboration have also been a focus of the healthcare leadership literature. Academic literature dealt primarily with how collaborative structures and behaviors can help leaders innovate and build organizational cultures geared for success [ 59 , 61 , 71 ]. Targeted gray literature focused on how leaders can foster communication within teams, and the positive impacts of an open and accountable culture on staff wellbeing and productivity [ 73 , 74 ]. These findings echo research on resilient health systems emphasizing that ‘over-managing’ restricts the adaptive capacities needed by teams within dynamic healthcare environments [ 100 , 103 ]. The literature pointed to the need for leaders to strengthen communication and collaboration at varying levels – environmental, team, and organizational – to enable more efficient and better-quality healthcare delivery, and during this process they should endeavor to model the balance between autonomy and accountability [ 104 ]. Implementing regular touchpoints that engage multiple stakeholders, such as communities of practice, can help to create positive feedback loops that enable systems change [ 105 ], and overcome organizational barriers to collaboration and information sharing, such as weak relationships and inadequate communication [ 106 , 107 ].

Self-development and self-awareness  also emerged as an important aspect of leadership. Academic literature focused primarily on how these capacities are developed in leaders through structured education and training, including self-assessments and targeted educational modules [ 65 , 69 ]. Targeted gray literature discussed a range of activities outside of structured training (e.g., experiential learning) that can support leaders’ self-reflection and development, for physician leaders in particular to assess their performance and improve their leadership approaches [ 91 , 92 ]. These findings suggest that personal leadership development must go beyond formal curriculum requirements to incorporate everyday learning inputs [ 78 ], and align with other recent literature suggesting that self-regulation in leaders can be fostered through practicing self-discipline, boundary-setting, and managing disruptions, particularly in the digital age [ 108 , 109 ]. Practicing self-awareness can help leaders not only to sense-make in complex systems – to adapt to new situations and make appropriate trade-offs – but also to sense-give – to articulate and express the organization’s vision [ 40 ]. A minor theme, observed only in the targeted gray literature, was related to leaders’ roles and competencies in consumer engagement and advocacy . The importance of increasing consumer engagement in healthcare was emphasized, as well as the structures that are needed to facilitate these changes [ 79 ]. Working alongside consumers was highlighted as critical during times of changing care and need, such as during COVID-19 [ 77 , 86 ]. Although the involvement of consumers and the public in the co-production of care is increasing [ 110 ], there is limited academic literature focused on the roles of leaders in creating optimal environments for co-production. Consumer and community involvement in change efforts helps to improve care processes and outcomes [ 111 ], but leaders might face challenges understanding and operationalizing local engagement mechanisms [ 112 ]. Identifying the organizational and system levers that enable greater consumer involvement, and how leaders can advocate for these levers in their local context, is a fruitful area for future investigation.

The findings of the current review have implications for professional organizations that train healthcare leaders, such as the Australian College of Health Services Management (ACHSM) in Australia, and train clinicians to be leaders, including the UK’s FMLM. Creating a future-focused curriculum addressing the competencies related to the themes identified, in particular innovation and adaptability, is essential to prepare healthcare leaders for growing and changing scopes of responsibility. Such competencies are less amenable to formal theoretical teaching solely and require carefully crafted experiential learning programs in health settings, with supervision by experienced and effective healthcare leaders.

Strengths and limitations

A notable strength of this scoping review was the inclusion of a broad range of sources and perspectives on the future of healthcare leadership. We captured empirical studies, theoretical academic contributions (e.g., commentaries from healthcare leaders), and targeted grey literature, which is often a more useful source of information on emerging topics [ 52 ]. As a result, our findings identified key future trends in the roles and competencies of leaders, both clinical and non-clinical, across a wide range of contexts and situations. Another strength of this review was its specific focus on contemporary literature that examined future trends in leadership, to inform how leaders can prepare for upcoming challenges, rather than focusing on leadership that was effective in the past.

There are limitations to this review. Our search strategies may not have adequately captured other leadership trends applicable across contemporary healthcare settings or those faced by leaders and teams on the front lines of care [ 113 ]. Incorporating search terms related to specific settings, as well as complex systems concepts, may have enabled greater inferences to be made about how unique future challenges require new approaches to the development of healthcare leaders. To scope future-focused research and perspectives, database searches were narrowly restricted, and it is likely that key articles were missed. Targeted gray literature searches represent key thought leaders in healthcare and leadership, and while this enabled relevant information to be efficiently collected, undertaking highly focused searches may have introduced bias associated with geographical area (i.e., the UK) and particular stakeholder groups (e.g., policy-makers) [ 55 ]. Our choice to limit the current review to studies reporting in OECD countries further limited generalizability to other settings including in low-income and middle-income countries (LMICs) [ 1 ].

The roles and competencies of leaders are deeply enmeshed in, and reflective of, a complex and continuously transforming healthcare system. This research highlights the types of roles and competencies important for leaders facing a myriad of challenges, and the range of contexts and situations in which these types of roles and competencies can be applied. The ways in which roles and competencies manifest is highly contextual, dependent on both role responsibilities and the situational demands of healthcare environments.

Data availability

Data supporting these research findings are available upon reasonable request. Further inquiries can be directed to the corresponding author.

Abbreviations

Faculty of Medical Leadership and Management

National Health Service

Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews

Royal Australasian College of Medical Administrators

United States of America

United Kingdom

Figueroa CA, Harrison R, Chauhan A, et al. Priorities and challenges for health leadership and workforce management globally: a rapid review. BMC Health Serv Res. 2019;19:1–11. https://doi.org/10.1186/s12913-019-4080-7

Article   Google Scholar  

Braithwaite J, Mannion R, Matsuyama Y, et al. The future of health systems to 2030: a roadmap for global progress and sustainability. Int J Qual Health Care. 2018;30(10):823–31. https://doi.org/10.1093/intqhc/mzy242

Anklam E, Bahl MI, Ball R, et al. Emerging technologies and their impact on regulatory science. Exp Biol Med. 2022;247(1):1–75. https://doi.org/10.1177/15353702211052280

Maeda A, Socha-Dietrich K. Skills for the future health workforce: preparing health professionals for people-centred care. OECD Health Working Papers. Paris: OECD Publishing,; 2021.

Reich MR, Javadi D, Ghaffar A. Introduction to the special issue on effective leadership for health systems. Health Syst Reform. 2016;2(3):171–5. https://doi.org/10.1080/23288604.2016.1223978

Shaw J, Brewer LC, Veinot T. Recommendations for health equity and virtual care arising from the COVID-19 pandemic: narrative review. JMIR Form Res. 2021;5(4):e23233. https://doi.org/10.2196/23233

Graham R, Woodhead T. Leadership for continuous improvement in healthcare during the time of COVID-19. Clin Radiol. 2021;76(1):67–72. https://doi.org/10.1016/j.crad.2020.08.008

Parr JM, Teo S, Koziol-McLain J. A quest for quality care: exploration of a model of leadership relationships, work engagement, and patient outcomes. J Adv Nurs. 2021;77(1):207–20. https://doi.org/10.1111/jan.14583

Ree E, Wiig S. Linking transformational leadership, patient safety culture and work engagement in home care services. Nurs Open. 2020;7(1):256–64. https://doi.org/10.1002/nop2.386

Wu X, Hayter M, Lee AJ, et al. Positive spiritual climate supports transformational leadership as means to reduce nursing burnout and intent to leave. J Nurs Manag. 2020;28(4):804–13. https://doi.org/10.1111/jonm.12994

West M, Armit K, Loewenthal L, et al. Leadership and leadership development in health care: the evidence base. London: The King’s Fund; 2015.

Google Scholar  

Churruca K, Ellis LA, Pomare C, et al. Dimensions of safety culture: a systematic review of quantitative, qualitative and mixed methods for assessing safety culture in hospitals. BMJ Open. 2021;11(7):e043982. https://doi.org/10.1136/bmjopen-2020-043982

Smallwood N, Bismark M, Willis K. Burn-out in the health workforce during the COVID-19 pandemic: opportunities for workplace and leadership approaches to improve well-being. BMJ Lead. 2023;leader–2022. https://doi.org/10.1136/leader-2022-000687

Kelly RJ, Hearld LR. Burnout and leadership style in behavioral health care: a literature review. J Behav Health Serv Res. 2020;47(4):581–600. https://doi.org/10.1007/s11414-019-09679-z

Hutchinson K, Ryder T, Coleman H, et al. Determining the role and responsibilities of the community epilepsy nurse in the management of epilepsy. J Clin Nurs. 2023;32(13–14):3730–45. https://doi.org/10.1111/jocn.16582

Mianda S, Voce A. Developing and evaluating clinical leadership interventions for frontline healthcare providers: a review of the literature. BMC Health Serv Res. 2018;18(1):1–15. https://doi.org/10.1186/s12913-018-3561-4

Veronesi G, Kirkpatrick I, Vallascas F. Clinicians on the board: what difference does it make? Soc Sci Med. 2013;77:147–55. https://doi.org/10.1016/j.socscimed.2012.11.019

Ferreira TDM, de Mesquita GR, de Melo GC, et al. The influence of nursing leadership styles on the outcomes of patients, professionals and institutions: an integrative review. J Nurs Manag. 2022;30(4):936–53. https://doi.org/10.1111/jonm.13592

Boamah SA. Emergence of informal clinical leadership as a catalyst for improving patient care quality and job satisfaction. J Adv Nurs. 2019;75(5):1000–9. https://doi.org/10.1111/jan.13895

Dickinson H, Ham C. Engaging doctors in leadership: review of the literature. Birmingham: University of Birmingham; 2008.

Hershkovich O, Gilad D, Zimlichman E, et al. Effective medical leadership in times of emergency: a perspective. Disaster Mil Med. 2016;2:1–5. https://doi.org/10.1186/s40696-016-0013-8

Stoller JK. Developing physician-leaders: a call to action. J Gen Intern Med. 2009;24:876–8. https://doi.org/10.1007/s11606-009-1007-8

Warren OJ, Carnall R. Medical leadership: why it’s important, what is required, and how we develop it. Postgrad Med J. 2011;87(1023):27–32. https://doi.org/10.1136/pgmj.2009.093807

Griffith R. NHS management inquiry: report to the secretary of state for social services. London: Department of Health and Social Security; 1983.

Braithwaite J. Changing how we think about healthcare improvement. BMJ. 2018;361. https://doi.org/10.1136/bmj.k2014

Clay-Williams R, Ludlow K, Testa L, et al. Medical leadership, a systematic narrative review: do hospitals and healthcare organisations perform better when led by doctors? BMJ Open. 2017;7(9):e014474. https://doi.org/10.1136/bmjopen-2016-014474

Nieuwboer MS, van der Sande R, van der Marck MA, et al. Clinical leadership and integrated primary care: a systematic literature review. Eur J Gen Pract. 2019;25(1):7–18. https://doi.org/10.1080/13814788.2018.1515907

Dwyer AJ. Roles, attributes and career paths of medical administrators in public hospitals: survey of victorian metropolitan directors of Medical services. Aust Health Rev. 2010;34(4):506–13. https://doi.org/10.1071/AH09750

Dwyer AJ. Medical managers in contemporary healthcare organisations: a consideration of the literature. Aust Health Rev. 2010;34(4):514–22. https://doi.org/10.1071/AH09736

Health Workforce Australia. Health LEADS Australia: the Australian health leadership framework. Adelaide: Health Workforce Australia; 2013.

NHS Institute for Innovation and Improvement and Academy of Medical Royal Colleges. Medical leadership competency framework: enhancing engagement in medical leadership. Coventry: NHS Institute for Innovation and Improvement; 2010.

National Centre for Healthcare Leadership. NCHL Health Leadership Competency Model 3.0. National Centre for Healthcare Leadership. 2022. https://www.nchl.org/research/ . Accessed 28 December, 2023.

NHS Leadership Academy. Healthcare Leadership Model: the nine dimensions of leadership behaviour. Leeds: NHS Leadership Academy; 2013.

Frank JR, Snell L, Sherbino J, et al. CanMEDS 2015 Physician Competency Framework. Royal College of Physicians and Surgeons of Canada; 2015.

Dath DCM-K, Anderson G, Burke A, Razack S, Lieff S, Moineau G, Chiu A, Ellison P. Leader. In: Frank JRSL, Sherbino J, editors. CanMEDS 2015 Physician Competency Framework. Royal College of Physicians and Surgeons of Canada; 2015.

Royal Australasian College of Medical Administrators. Medical Leadership and Management Curriculum Framework. Malvern, Victoria: Royal Australasian College of Medical Administrators; 2011.

Best S, Stark Z, Brown H, et al. The leadership behaviors needed to implement clinical genomics at scale: a qualitative study. Genet Med. 2020;22(8):1384–90. https://doi.org/10.1038/s41436-020-0818-1

Thoma B, Karwowska A, Samson L, et al. Emerging concepts in the CanMEDS physician competency framework. Can Med Educ J. 2023;14(1):4–12. https://doi.org/10.36834/cmej.75591

Ahern S, Loh E. Leadership during the COVID-19 pandemic: building and sustaining trust in times of uncertainty. BMJ Lead. 2020;leader–2020. https://doi.org/10.1136/leader-2020-000271

Braithwaite J, Churruca K, Hibbert P, et al. At the heart of getting things done in complex healthcare ecosystems: leadership, strategy, sensemaking and sensegiving. In: Chambers N, editor. Research Handbook on Leadership in Healthcare. Edward Elgar Publishing; 2023. pp. 775–92.

Braithwaite J, Fisher G. Beyond the aspirational: creating the future of health care in Australia. Int Med J. 2024;54(2):342–7. https://doi.org/10.1111/imj.16286

Cernega A, Nicolescu DN, Meleșcanu Imre M, et al. Volatility, uncertainty, complexity, and ambiguity (VUCA) in healthcare. Healthc. 2024;12(7):773. https://doi.org/10.3390/healthcare12070773

Kasula BY. Revolutionizing healthcare delivery: innovations and challenges in supply chain management for improved patient care. Trans Latest Trends Health Sect. 2023;15(15).

Niaz M, Nwagwu U. Managing healthcare product demand effectively in the post-covid-19 environment: navigating demand variability and forecasting complexities. Am J Econ Manag Bus. 2023;2(8):316–30. https://doi.org/10.58631/ajemb.v2i8.55

Thirumalai S, Devaraj S. Mitigating the curse of complexity: the role of focus and the implications for costs of care. J Oper Manag. 2024;70(1):157–79. https://doi.org/10.1002/joom.1278

Savage M, Savage C, Brommels M, et al. Medical leadership: boon or barrier to organisational performance? A thematic synthesis of the literature. BMJ Open. 2020;10(7):e035542. https://doi.org/10.1136/bmjopen-2019-035542

Ahti M, Taipale-Walsh L, Kuha S, et al. Health-care leaders’ experiences of the competencies required for crisis management during COVID-19: a systematic review of qualitative studies. Leadersh Health Serv. 2023. https://doi.org/10.1108/LHS-10-2022-0104

Krasikova DV, Green SG, LeBreton JM. Destructive leadership: a theoretical review, integration, and future research agenda. J Manag. 2013;39(5):1308–38. https://doi.org/10.1177/0149206312471388

Pellegrini MM, Ciampi F, Marzi G, et al. The relationship between knowledge management and leadership: mapping the field and providing future research avenues. J Knowl Manag. 2020;24(6):1445–92. https://doi.org/10.1108/JKM-01-2020-0034

Paré G, Trudel M-C, Jaana M, et al. Synthesizing information systems knowledge: a typology of literature reviews. Inf Manag. 2015;52(2):183–99. https://doi.org/10.1016/j.im.2014.08.008

Heath A, Levay P, Tuvey D. Literature searching methods or guidance and their application to public health topics: a narrative review. Health Info Libr J. 2022;39(1):6–21. https://doi.org/10.1111/hir.12414

Adams J, Hillier-Brown FC, Moore HJ, et al. Searching and synthesising ‘grey literature’and ‘grey information’in public health: critical reflections on three case studies. Syst Rev. 2016;5(1):1–11. https://doi.org/10.1186/s13643-016-0337-y

Tricco AC, Lillie E, Zarin W, et al. PRISMA extension for scoping reviews (PRISMA-ScR): checklist and explanation. Ann Intern Med. 2018;169(7):467–73. https://doi.org/10.7326/M18-0850

Wolters Kluwer Health I. Get to know the advantages of Ovid MEDLINE ® . 2024. https://tools.ovid.com/ovidtools/medline.html . Accessed.

Stansfield C, Dickson K, Bangpan M. Exploring issues in the conduct of website searching and other online sources for systematic reviews: how can we be systematic? Syst Rev. 2016;5:1–9. https://doi.org/10.1186/s13643-016-0371-9

Ouzzani M, Hammady H, Fedorowicz Z, et al. Rayyan—a web and mobile app for systematic reviews. Syst Rev. 2016;5:1–10. https://doi.org/10.1186/s13643-016-0384-4

Peters MD, Marnie C, Tricco AC, et al. Updated methodological guidance for the conduct of scoping reviews. JBI Evid Synth. 2020;18(10):2119–26. https://doi.org/10.11124/JBIES-20-00167

Arksey H, O’Malley L. Scoping studies: towards a methodological framework. Int J Soc Res Methodol. 2005;8(1):19–32. https://doi.org/10.1080/1364557032000119616

Best S, Long J, Stark Z, et al. Clinical leadership and genomics - what works? Twin Res Hum Genet. 2019;22(5):350. 43rd Annual Scientific Meeting of the Human Genetics Society of Australasia, HGSA 2019. Wellington New Zealand.

Crabtree BF, Howard J, Miller WL, et al. Leading innovative practice: leadership attributes in LEAP practices. Milbank Q. 2020;98(2):399–445. https://doi.org/10.1111/1468-0009.12456

Woiceshyn J, Huq JL, Kannappan S, et al. We need to work differently in a crisis: peer-professional leadership to redesign physicians’ work. BMJ Lead. 2022;6(2):98–103. https://doi.org/10.1136/leader-2021-000490

Appelbaum NP, Liang C, Whitney SE, et al. What business competencies are needed for the modern physician leader and when. Physician Leadersh J. 2022;9(2):39–45.

Hopkins J, Fassiotto M, Ku MC, et al. Designing a physician leadership development program based on effective models of physician education. Health Care Manage Rev. 2018;43(4):293–302. https://doi.org/10.1097/HMR.0000000000000146

Sandberg DS, Pennington CM, Lindquist MA. Virtual leadership: CEOs and C-Level executives of healthcare organizations in the United States reimagined new roles as virtual leaders. J Leadersh Stud. 2022;16(3):61–9. https://doi.org/10.1002/jls.21827

Satiani B, Dawson K, Mehta LS, et al. Lessons learned after implementing an academic faculty leadership program over seven years. Physician Leadersh J. 2022;9(2):30–8.

Angood PB. Transitions and transformations — our choice? Physician Leadersh J. 2022;9(5):6–8.

Brace R, James D. Adapting to changing roles. J Healthc Manag. 2020;35(3):46–7.

Wright C, Pister K. New thinking leads to better transitions. Physician Leadersh J. 2018;5(1):38–40.

Ramirez B, West DJ Jr, Ramirez CL. Healthcare and leadership competencies, training methods, and venues. J Leadersh Acc Ethics. 2021;18(4):93–9.

Wain MJ, Handel DA, Williford K, et al. Transitioning from physician to hospital leader: a competency-based model. Physician Leadersh J. 2020;7(2):32–3.

Borgwardt HL, Botz CT, Doppler JM, et al. Electronic health record implementation: the people side of change. Manage Healthc. 2019;4(1):7–20.

Anandaciva S. The biggest hospital in England. The King’s Fund . 2020. https://www.kingsfund.org.uk/blog/2020/09/biggest-hospital-england . Accessed 25 February, 2023.

Bailey S, West M. Learning from staff experiences of Covid-19: let the light come streaming in. The King’s Fund . 2020. https://www.kingsfund.org.uk/blog/2020/06/learning-staff-experiences-covid-19 . Accessed 26 February, 2023.

Bailey S, West M. Naming the issue: chronic excessive workload in the NHS. The King’s Fund . 2021. https://www.kingsfund.org.uk/blog/2021/06/naming-issue-chronic-excessive-workload-nhs . Accessed 26 February, 2023.

Bhopal A, Hood G. From disaster preparedness to digital disruption: A whistle-stop tour through public health. Faculty of Medical Leadership and Management. 2018. https://www.fmlm.ac.uk/news-opinion/from-disaster-preparedness-to-digital-disruption-a-whistle-stop-tour-through-public . Accessed 27 February, 2023.

Cheetham J. How Leaders in Healthcare 2020 gave an enlightened spotlight to junior doctor leadership. Faculty of Medical Leadership and Management. 2021. https://www.fmlm.ac.uk/news-opinion/how-leaders-in-healthcare-2020-gave-an-enlightened-spotlight-to-junior-doctor . Accessed 27 February, 2023.

Collins B. Leadership and innovation during Covid-19: lessons from the Cardiff and Vale Health System. The King’s Fund . 2021. https://www.kingsfund.org.uk/blog/2021/05/leadership-innovation-covid-19-cardiff-vale . Accessed 27 February, 2023.

Cottey L. The trainee and the team. Faculty of Medical Leadership and Management. 2018. https://www.fmlm.ac.uk/news-opinion/the-trainee-and-the-team . Accessed 28 February, 2023.

Dixon M. Engaging the public in designing the future for the NHS. The King’s Fund . 2018. https://www.kingsfund.org.uk/blog/2018/01/engaging-public-designing-future-nhs . Accessed 28 February, 2023.

Leveson D. Developing clinical leaders through innovation. The King’s Fund . 2021. https://www.kingsfund.org.uk/blog/2021/03/developing-clinical-leaders-through-innovation . Accessed 25 February, 2023.

Naylor C. Getting to grips with climate change: how can the NHS get to net zero? The King’s Fund . 2020. https://www.kingsfund.org.uk/blog/2020/01/climate-change-nhs-net-zero . Accessed 19 February, 2023.

Tweedie J. Leadership - so what is it anyway? February, 2018. https://www.fmlm.ac.uk/news-opinion/blog/leadership-so-what-is-it-anyway

Faculty of Medical Leadership and Management. FMLM response to HEE’s Call for Evidence on the Future Doctor. Faculty of Medical Leadership and Management. 2019. https://www.fmlm.ac.uk/news-opinion/fmlm-response-to-hees-call-for-evidence-on-the-future-doctor . Accessed 23 February, 2023.

Faculty of Medical Leadership and Management. Supporting and developing a leading resource for the future. Faculty of Medical Leadership and Management. 2021. https://www.fmlm.ac.uk/news-opinion/supporting-and-developing-a-leading-resource-for-the-future . Accessed 27 February, 2023.

Faculty of Medical Leadership and Management. Support, wellbeing and leadership. Faculty of Medical Leadership and Management. 2021. https://www.fmlm.ac.uk/news-opinion/support-wellbeing-and-leadership . Accessed 27 February, 2023.

Faculty of Medical Leadership and Management. Reflections from the International Healthcare Leadership Conference. 2022. Faculty of Medical Leadership and Management. 2022. https://www.fmlm.ac.uk/news-opinion/reflections-from-the-international-healthcare-leadership-conference-2022 . Accessed 27 February, 2023.

Lakhani M, Osuji N, Lees P et al. FMLM statement: racism and good leadership are incompatible. Faculty of Medical Leadership and Management. 2020. https://www.fmlm.ac.uk/news-opinion/fmlm-statement-racism-and-good-leadership-are-incompatible . Accessed 28 February, 2023.

Dougall D, Ross S. Transformational change in health and care: reports from the field. 2018. Accessed February, 2023. The King’s Fund.

Fogden R, Buck D, Franklin B et al. Poverty and the health and care system: the role of data and partnership in bringing change. The King’s Fund . 2022. https://www.kingsfund.org.uk/publications/poverty-health-care-system-data-partnership . Accessed 27 February, 2023.

Ham C, Murray R. The NHS 10-year plan: how should the extra funding be spent. The King’s Fund . 2018. https://www.kingsfund.org.uk/publications/nhs-10-year-plan . Accessed 28 February, 2023.

Swanwick T, McKimm J. The assessment of leadership development in the medical undergraduate curriculum: a consensus statement. Faculty of Medical Leadership and Management. 2020. https://www.fmlm.ac.uk/resources/the-assessment-of-leadership-development-in-the-medical-undergraduate-curriculum-a . Accessed 22 February, 2023.

Evans P. A radical upward shift to benefit members and fellows. Faculty of Medical Leadership and Management. 2020. https://www.fmlm.ac.uk/news-opinion/a-radical-upward-shift-to-benefit-members-and-fellows . Accessed 22 February, 2023.

Lakhani M. Black History Month and why it matters. Faculty of Medical Leadership and Management. 2021. https://www.fmlm.ac.uk/news-opinion/black-history-month-and-why-it-matters . Accessed 28 February, 2023.

Moralee S, Exworthy M, Rajasingam D et al. What next for medical leaders? Faculty of Medical Leadership and Management. 2018. https://www.fmlm.ac.uk/resources/what-next-for-medical-leaders . Accessed 25 February, 2023.

Warren S, Laverty A. Leading with compassion: supporting the health and wellbeing of NHS staff. The King’s Fund . 2022. https://www.kingsfund.org.uk/audio-video/podcast/leading-compassion-health-wellbeing-nhs-staff . Accessed 22 February, 2023.

Hancock M. Matt Hancock at The King’s Fund annual conference 2018. 2018. https://www.kingsfund.org.uk/audio-video/matt-hancock-kings-fund-annual-conference-2018 . Accessed.

Olsson-Brown A, Interview with Dr Anna Olsson-Brown - Chair of the Academy of Medical Royal Colleges Trainee Group. Faculty of Medical Leadership and Management. 2020. https://www.fmlm.ac.uk/news-opinion/interview-with-dr-anna-olsson-brown-chair-of-the-academy-of-medical-royal-colleges . Accessed 19 February, 2023.

Denicolai S, Previtali P. Innovation strategy and digital transformation execution in healthcare: the role of the general manager. Technovation. 2023;121:102555. https://doi.org/10.1016/j.technovation.2022.102555

Borges do Nascimento IJ, Abdulazeem H, Vasanthan LT, et al. Barriers and facilitators to utilizing digital health technologies by healthcare professionals. NPJ Digit Med. 2023;6(1):161. https://doi.org/10.1038/s41746-023-00899-4

Braithwaite J, Churruca K, Ellis LA, et al. Complexity science in healthcare: aspirations, approaches, applications, and accomplishments. A white paper. Sydney: Australian Institute of Health Innovation, Macquarie University; 2017.

Coiera E. Why system inertia makes health reform so difficult. BMJ. 2011;342 https://doi.org/10.1136/bmj.d3693

Clay-Williams R, Travaglia J, Hibbert P, et al. Clinical Governance Framework. A Report prepared for the Royal Australasian College of Medical Administrators (RACMA). Melbourne: RACMA; 2017.

Plsek PE, Wilson T. Complexity, leadership, and management in healthcare organisations. BMJ. 2001;323(7315):746–9. https://doi.org/10.1136/bmj.323.7315.746

van Diggele C, Burgess A, Roberts C, et al. Leadership in healthcare education. BMC Med Educ. 2020;20:1–6. https://doi.org/10.1186/s12909-020-02288-x

Golden RE, Klap R, Carney DV, et al. Promoting learning health system feedback loops: experience with a VA practice-based research network card study. Healthc. 2021;8:100484. https://doi.org/10.1016/j.hjdsi.2020.100484

Smith CL, Fisher G, Dharmayani PNA, et al. Progress with the Learning Health System 2.0: a rapid review of Learning Health Systems’ responses to pandemics and climate change. BMC Med. 2024;22(1):131. https://doi.org/10.1186/s12916-024-03345-8

Hunt RC, Struminger BB, Redd JT, et al. Virtual peer-to-peer learning to enhance and accelerate the health system response to COVID-19: the HHS ASPR Project ECHO COVID-19 Clinical rounds Initiative. Ann Emerg Med. 2021;78(2):223–8. https://doi.org/10.1016/j.annemergmed.2021.03.035

Mander R, Hellert U, Antoni CH. Self-leadership strategies for coping with flexibility requirements of digital work with a high degree of latitude for time, place and scope for action—a qualitative study. Gr Interakt Org. 2021;52:163–71. https://doi.org/10.1007/s11612-021-00560-2

Vay C, Steinherr V. Leadership in a digitalized and crisis-ridden world: towards a comprehensive overview of relevant competencies for leaders. University of Hawaiʻi at Mānoa; 2023.

Churruca K, Pomare C, Ellis LA, et al. Patient-reported outcome measures (PROMs): a review of generic and condition‐specific measures and a discussion of trends and issues. Health Expect. 2021;24(4):1015–24. https://doi.org/10.1111/hex.13254

Best A, Greenhalgh T, Lewis S, et al. Large-system transformation in health care: a realist review. Milbank Q. 2012;90(3):421–56. https://doi.org/10.1111/j.1468-0009.2012.00670.x

Bolden R, Kars-Unluoglu S, Jarvis C, et al. Paradoxes of multi-level leadership: insights from an integrated care system. J Change Manag. 2023;23(4):337–57. https://doi.org/10.1080/14697017.2023.2234388

Braithwaite J, Dammery G, Spanos S, et al. Learning health systems 2.0: future-proofing healthcare against pandemics and climate change. A White Paper. Sydney: Australian Institute of Health Innovation, Macquarie University; 2023.

Download references

Acknowledgements

The authors thank Mary Simons and Jeremy Cullis for their specialist guidance on database searches.

This work was funded in part by RACMA. RACMA contributed to the conceptualization and design of the research. JB is funded and supported by an NHMRC Leadership Investigator Award (1176620).

Author information

Authors and affiliations.

Centre for Healthcare Resilience and Implementation Science, Australian Institute of Health Innovation, Macquarie University, 75 Talavera Rd, North Ryde, Sydney, NSW, 2109, Australia

Samantha Spanos, Elle Leask, Romika Patel, Michael Datyner & Jeffrey Braithwaite

Royal Australasian College of Medical Administrators, Melbourne, VIC, Australia

You can also search for this author in PubMed   Google Scholar

Contributions

JB and SS led the research conception and design. SS, ELe, and RP executed the research, and acquired, and analyzed the data. JB, SS, and MD significantly contributed to the interpretation of the data. SS wrote the original manuscript draft which was critically reviewed by JB, ELe, RP, MD and ELo. SS led the manuscript revisions with significant contribution from JB, and input from ELe, RP, MD, and ELo. All authors agree on the journal to which the article will be submitted, have reviewed, and approved the manuscript, and agree to be personally accountable for its contents.

Corresponding author

Correspondence to Samantha Spanos .

Ethics declarations

Ethics approval and consent to participate.

Not applicable.

Consent for publication

Competing interests.

The authors declare no competing interests.

Additional information

Publisher’s note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Supplementary Material 1

Rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ . The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/ ) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Cite this article.

Spanos, S., Leask, E., Patel, R. et al. Healthcare leaders navigating complexity: a scoping review of key trends in future roles and competencies. BMC Med Educ 24 , 720 (2024). https://doi.org/10.1186/s12909-024-05689-4

Download citation

Received : 13 February 2024

Accepted : 21 June 2024

Published : 03 July 2024

DOI : https://doi.org/10.1186/s12909-024-05689-4

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Physician leadership
• Future health care
• Future healthcare
• Complex systems

BMC Medical Education

ISSN: 1472-6920

• Submission enquiries: [email protected]
• General enquiries: [email protected]

• Open access
• Published: 03 March 2015

Health research improves healthcare: now we have the evidence and the chance to help the WHO spread such benefits globally

• Stephen R Hanney 1 &
• Miguel A González-Block 2  

Health Research Policy and Systems volume  13 , Article number:  12 ( 2015 ) Cite this article

12k Accesses

16 Altmetric

Metrics details

There has been a dramatic increase in the body of evidence demonstrating the benefits that come from health research. In 2014, the funding bodies for higher education in the UK conducted an assessment of research using an approach termed the Research Excellence Framework (REF). As one element of the REF, universities and medical schools in the UK submitted 1,621 case studies claiming to show the impact of their health and other life sciences research conducted over the last 20 years. The recently published results show many case studies were judged positively as providing examples of the wide range and extensive nature of the benefits from such research, including the development of new treatments and screening programmes that resulted in considerable reductions in mortality and morbidity.

Analysis of specific case studies yet again illustrates the international dimension of progress in health research; however, as has also long been argued, not all populations fully share the benefits. In recognition of this, in May 2013 the World Health Assembly requested the World Health Organization (WHO) to establish a Global Observatory on Health Research and Development (R&D) as part of a strategic work-plan to promote innovation, build capacity, improve access, and mobilise resources to address diseases that disproportionately affect the world’s poorest countries.

As editors of Health Research Policy and Systems ( HARPS ), we are delighted that our journal has been invited to help inform the establishment of the WHO Global Observatory through a Call for Papers covering a range of topics relevant to the Observatory, including topics on which HARPS has published articles over the last few months, such as approaches to assessing research results, measuring expenditure data with a focus on R&D, and landscape analyses of platforms for implementing R&D. Topics related to research capacity building may also be considered. The task of establishing a Global Observatory on Health R&D to achieve the specified objectives will not be easy; nevertheless, this Call for Papers is well timed – it comes just at the point where the evidence of the benefits from health research has been considerably strengthened.

The start of 2015 sees a dramatic increase in the body of evidence demonstrating the benefits arising from health research. Throughout 2014, the higher education funding bodies in the UK conducted an assessment of research, termed the Research Excellence Framework (REF), in which, for the first time, account was taken of the impact on society of the research undertaken. As part of this, UK universities and medical schools produced 1,621 case studies that aimed to show the benefits, such as improved healthcare, arising from examples of their health and other life sciences research conducted over the last 20 years. Panels of experts, including leading academics from many countries, published their assessments of these case studies in December 2014 [ 1 ], with the full case studies and an analysis of the results being made public in January 2015 [ 2 , 3 ].

As we recently anticipated [ 4 ], the expert panels concluded that the case studies did indeed overwhelmingly illustrate the wide range and extensive nature of the benefits from health research. Main Panel A covered the range of life sciences and its overview report states: “ MPA [Main Panel A] believes that the collection of impact case studies provide a unique and powerful illustration of the outstanding contribution that research in the fields covered by this panel is making to health, wellbeing, wealth creation and society within and beyond the UK ” [ 3 ], p. 1. The section of the report covering public health and health services research also notes that: “ Outstanding examples included cases focused on national screening programmes for the selection and early diagnosis of conditions ” [ 3 ], p. 30. In their section of the report, the international experts say of the REF2014: “ It is the boldest, largest, and most comprehensive exercise of its kind of any country’s assessment of its science ” [ 3 ], p. 20.

The REF2014 is therefore attracting wide international attention. Indeed, some of the methods used are already informing studies in other countries, including, for example, an innovative assessment recently published in Health Research Policy and Systems ( HARPS ) identifying the beneficial effects made on healthcare policies and practice in Australia by intervention studies funded by the National Health and Medical Research Council [ 5 ].

The REF also illustrates that, even when focusing on the research from one country, there are examples of studies in which there has been international collaboration and which have built on research conducted elsewhere. For example, one REF case study on screening describes how a major UK randomised controlled trial of screening for abdominal aortic aneurysms (AAA) involving 67,800 men [ 6 , 7 ] was the most significant trial globally. The trial provided the main evidence for the policy to introduce national screening programmes for AAA for men reaching 65 throughout the UK [ 2 ]. The importance of this trial lay partly in its size, given that it accounted for over 50% of the men included in the meta-analyses performed in the 2007 Cochrane review [ 8 ] and the 2009 practice guideline from the US Society for Vascular Surgery [ 9 ]. Nevertheless, two of the three smaller studies that were also included in these two meta-analyses came from outside the UK, specifically from Denmark [ 10 ] and Australia [ 11 ].

Moreover, a recent paper published in HARPS also included descriptions of how the research contributing to new interventions often comes from more than one country. These accounts are included in a separate set of seven extensive case studies constructed to illustrate innovative ways to measure the time that can elapse between research being conducted and its translation into improved health [ 12 ]. While being a separate set of case studies, one of them does, nevertheless, explore the international timelines involved in research on screening for AAA, and, in addition to highlighting the key role of the UK research, it also highlights that the pioneering first screening study using ultrasound had been conducted in 1983 on 73 patients in a US Army medical base [ 13 ].

These case studies therefore further reinforce the well-established argument that health research progress often involves contributions from various countries. However, as has long been argued, not all populations fully share the benefits. In recognition of this, in May 2013, the World Health Assembly requested the World Health Organization (WHO), in its resolution 66.22, to establish a Global Observatory on Health Research and Development as part of a strategic work-plan to promote innovation, build capacity, improve access, and mobilise resources to address diseases that disproportionately affect the world’s poorest countries [ 14 ].

As editors of HARPS , we are delighted that our journal has been invited to help inform the establishment of the WHO Global Observatory by publishing a series of papers whose publication costs will be funded by the WHO. In support of this WHO initiative, Taghreed Adam, John-Arne Røttingen, and Marie-Paule Kieny recently published a Call for Papers for this series [ 15 ], which can be accessed through the HARPS webpage.

The aim of the series is “ to contribute state-of-the-art knowledge and innovative approaches to analyse, interpret, and report on health R&D information… [and] to serve as a key resource to inform the future WHO-convened coordination mechanism, which will be utilized to generate evidence-informed priorities for new R&D investments to be financed through a proposed new global financing and coordination mechanism for health R&D ” [ 15 ], p. 1. The Call for Papers covers a range of topics relevant to the aims of the Global Observatory. These include ones on which HARPS has published articles in the last few months, such as approaches to assessing research results, as seen in the Australian article described above [ 5 ]; papers measuring expenditure data with a focus on R&D, as described in a recent Commentary by Young et al. [ 16 ]; and landscape analyses of platforms for implementing R&D, as described in the article by Ongolo-Zogo et al. [ 17 ], analysing knowledge translation platforms in Cameroon and Uganda, and partially in the article by Yazdizadeh et al. [ 18 ], relaying lessons learnt from knowledge networks in Iran.

Adam et al. also make clear that the topics listed in the Call for Papers are examples and that the series editors are also willing to consider other areas [ 15 ]. Indeed, in the Introduction to the Call for Papers, the importance of capacity building is highlighted. This, too, is a topic described in recent papers in HARPS , such as those by Ager and Zarowsky [ 19 ], analysing the experiences of the Health Research Capacity Strengthening initiative’s Global Learning program of work across sub-Saharan Africa, and by Hunter et al. [ 20 ], describing needs assessment to strengthen capacity in water and sanitation research in Africa.

Finally, as we noted in our earlier editorial [ 4 ], the World Health Report 2013: Health Research for Universal Coverage showed how the demonstration of the benefits from health research could be a strong motivation for further funding of such research. As the Report states, “ adding impetus to do more research is a growing body of evidence on the returns on investments … there is mounting quantitative proof of the benefits of research to health, society and the economy ” [ 21 ]. We noted, too, that since the Report’s publication in 2013, there had been further examples from many countries of the benefits from medical research. The REF2014 in the UK signifies an additional major boost to the evidence that a wide range of health research does contribute to improved health and other social benefits. The results of such evaluations highlight the appropriateness of the WHO’s actions in attempting to ensure all populations share the benefits of health research endeavours by creating the Global Observatory on Health Research and Development. This will not be an easy task, but we welcome the opportunity afforded by the current Call for Papers for researchers and other stakeholders to engage with this process and influence it [ 15 ].

Abbreviations

Abdominal aortic aneurysms

Health Research Policy and Systems

Main Panel A

Research and development

Research Excellence Framework

• World Health Organization

Higher Education Funding Council for England. Research Excellence Framework 2014: The results. 2014. http://www.ref.ac.uk/pubs/201401/ . Accessed 20 Feb 2015.

Higher Education Funding Council for England. Research Excellence Framework 2014: Results and submissions. 2015. http://results.ref.ac.uk/ . Accessed 20 Feb 2015.

Higher Education Funding Council for England. REF 2014 Panel overview reports: Main Panel A and sub-panels 1–6. 2015. http://www.ref.ac.uk/media/ref/content/expanel/member/Main%20Panel%20A%20overview%20report.pdf . Accessed 20 Feb 2015.

Hanney SR, González-Block MA. Four centuries on from Bacon: progress in building health research systems to improve health systems? Health Res Policy Syst. 2014;12:56.

Article   PubMed   PubMed Central   Google Scholar  

Cohen G, Schroeder J, Newson R, King L, Rychetnik L, Milat AJ, et al. Does health intervention research have real world policy and practice impacts: testing a new impact assessment tool. Health Res Policy Syst. 2015;13:3.

Scott RAP, Ashton HA, Buxton M, Day NE, Kim LG, Marteau TM, et al. The Multicentre Aneurysm Screening Study (MASS) into the effect of abdominal aortic aneurysm screening on mortality in men: a randomised controlled trial. Lancet. 2002;360:1531–9.

Article   PubMed   Google Scholar  

Buxton M, Ashton H, Campbell H, Day NE, Kim LG, Marteau TM, et al. Multicentre aneurysm screening study (MASS): cost effectiveness analysis of screening for abdominal aortic aneurysms based on four year results from randomised controlled trial. BMJ. 2002;325:1135–8.

Article   Google Scholar  

Cosford PA, Leng GC, Thomas J: Screening for abdominal aortic aneurysm. Cochrane Database Syst Rev 2007; (2)CD002945.

Chaikof EL, Brewster DC, Dalman RL, Makaroun MS, Illig KA, Sicard GA, et al. The care of patients with an abdominal aortic aneurysm: the Society for Vascular Surgery practice guidelines. J Vasc Surg. 2009;50(4 Suppl):S2–49.

Lindholt JS, Juul S, Fasting H, Henneberg EW. Hospital costs and benefits of screening for abdominal aortic aneurysms. Results from a randomised population screening trial. Eur J Vasc Endovasc Surg. 2002;23:55–60.

Article   CAS   PubMed   Google Scholar  

Norman PE, Jamrozik K, Lawrence-Brown MM, Le MT, Spencer CA, Tuohy RJ, et al. Population based randomised controlled trial on impact of screening on mortality from abdominal aortic aneurysm. BMJ. 2004;329(7477):1259–64.

Hanney SR, Castle-Clarke S, Grant J, Guthrie S, Henshall C, Mestre-Ferrandiz J, et al. How long does biomedical research take? Studying the time taken between biomedical and health research and its translation into products, policy, and practice. Health Res Policy Syst. 2015;13:1.

Cabellon S, Moncrief CL, Pierre DR, Cavenaugh DG. Incidence of abdominal aortic aneurysms in patients with artheromatous arterial disease. Am J Surg. 1983;146:575–6.

World Health Organization. WHA resolution 66.22: Follow up of the report of the Consultative Expert Working Group on Research and Development: financing and coordination. Geneva: WHO; 2013. http://www.who.int/phi/resolution_WHA-66.22.pdf . Accessed 20 Feb 2015.

Google Scholar  

Adam T, Røttingen JA, Kieny MP. Informing the establishment of the WHO Global Observatory on Health Research and Development: a call for papers. Health Res Policy Syst. 2015;13:9.

Young AJ, Terry RF, Røttingen JA, Viergever RF. Global trends in health research and development expenditures – the challenge of making reliable estimates for international comparison. Health Res Policy Syst. 2015;13:7.

Ongolo-Zogo P, Lavis JN, Tomson G, Sewankambo NK. Climate for evidence informed health system policymaking in Cameroon and Uganda before and after the introduction of knowledge translation platforms: a structured review of governmental policy documents. Health Res Policy Syst. 2015;13:2.

Yazdizadeh B, Majdzadeh R, Alami A, Amrolalaei S. How can we establish more successful knowledge networks in developing countries? Lessons learnt from knowledge networks in Iran. Health Res Policy Syst. 2014;12:63.

Ager A, Zarowsky C. Balancing the personal, local, institutional, and global: multiple case study and multidimensional scaling analysis of African experiences in addressing complexity and political economy in health research capacity strengthening. Health Res Policy Syst. 2015;13:5.

Hunter PR, Abdelrahman SH, Antwi-Agyei P, Awuah E, Cairncross S, Chappell E, et al. Needs assessment to strengthen capacity in water and sanitation research in Africa: experiences of the African SNOWS consortium. Health Res Policy Syst. 2014;12:68.

The World Health Organization. The World Health Report 2013: Research for Universal Health Coverage. Geneva: WHO; 2013.

Download references

Acknowledgements

The authors thank Bryony Soper for most helpful comments on an earlier draft.

Author information

Authors and affiliations.

Health Economics Research Group, Brunel University London, Kingston Lane, Uxbridge, UB8 3PH, UK

Stephen R Hanney

Universidad Anáhuac, Av. Universidad Anáhuac 46, Lomas Anáhuac, 52786 Huixquilucan, Mexico City, Mexico

Miguel A González-Block

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Stephen R Hanney .

Additional information

Competing interests.

The authors are co-Chief Editors of Health Research Policy and Systems.

Rights and permissions

This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License ( http://creativecommons.org/licenses/by/4.0 ), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/ ) applies to the data made available in this article, unless otherwise stated.

Reprints and permissions

About this article

Cite this article.

Hanney, S.R., González-Block, M.A. Health research improves healthcare: now we have the evidence and the chance to help the WHO spread such benefits globally. Health Res Policy Sys 13 , 12 (2015). https://doi.org/10.1186/s12961-015-0006-y

Download citation

Received : 20 February 2015

Accepted : 20 February 2015

Published : 03 March 2015

DOI : https://doi.org/10.1186/s12961-015-0006-y

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Assessing research benefits
• Capacity building
• Diseases of poorest countries
• Global health
• Global Observatory
• Platforms for research implementation
• Research expenditure
• Screening for abdominal aortic aneurysms
• World Health Report 2013

ISSN: 1478-4505

• Submission enquiries: Access here and click Contact Us
• General enquiries: [email protected]

Healthcare Research Paper Topics

In this page, we provide a comprehensive list of healthcare research paper topics , expert advice on selecting compelling topics, guidance on writing an impactful research paper, and information about iResearchNet’s writing services. By exploring these resources, students in the health sciences field can choose relevant and significant healthcare research paper topics, develop their papers effectively, and access professional writing assistance to excel in their academic endeavors.

100 Healthcare Research Paper Topics

The field of healthcare research encompasses a vast array of topics that are crucial for understanding, improving, and transforming healthcare practices. As students in the health sciences, you have the opportunity to explore these diverse areas and contribute to the knowledge base of healthcare research. This comprehensive list aims to inspire and guide you in selecting healthcare research paper topics that align with your interests and academic goals. The topics are divided into ten distinct categories, each containing ten thought-provoking and relevant research ideas. Let this list serve as a springboard for your exploration and a catalyst for impactful research in the dynamic field of healthcare.

Academic Writing, Editing, Proofreading, And Problem Solving Services

Get 10% off with 24start discount code.

1. Healthcare Policy and Management

• The Impact of Health Policies on Access to Care
• Assessing the Effectiveness of Health Insurance Programs
• Analyzing the Role of Healthcare Leadership in Quality Improvement
• Exploring Strategies for Healthcare Cost Containment
• Investigating the Relationship Between Healthcare Regulations and Patient Outcomes
• Evaluating the Impact of Electronic Health Records on Healthcare Delivery
• Examining the Role of Public-Private Partnerships in Healthcare
• Analyzing the Influence of Political Factors on Healthcare Decision-Making
• Assessing the Ethical Implications of Resource Allocation in Healthcare
• Investigating the Effectiveness of Health Promotion Programs in Primary Care Settings

2. Healthcare Ethics and Legal Issues

• Analyzing the Ethical Challenges of Healthcare Research Involving Human Subjects
• Exploring the Impact of Cultural and Religious Beliefs on Healthcare Decision-Making
• Examining Legal Issues in End-of-Life Care and Advance Directives
• Investigating the Ethical Implications of Genetic Testing and Personalized Medicine
• Assessing the Ethical Dilemmas in Access to Experimental Treatments
• Exploring the Role of Ethics Committees in Healthcare Organizations
• Analyzing the Intersection of Healthcare Ethics and Artificial Intelligence
• Evaluating the Legal and Ethical Implications of Telemedicine
• Investigating the Ethics of Healthcare Resource Allocation during Public Health Emergencies
• Examining the Legal and Ethical Issues of Patient Privacy in the Digital Age

3. Healthcare Technology and Innovation

• Assessing the Impact of Artificial Intelligence in Healthcare Diagnostics
• Exploring the Potential of Wearable Devices for Remote Patient Monitoring
• Investigating the Role of Big Data Analytics in Healthcare Decision-Making
• Analyzing the Use of Robotics in Surgery and Patient Care
• Examining the Impact of Telehealth on Healthcare Access and Delivery
• Evaluating the Benefits and Challenges of Electronic Health Records Implementation
• Exploring the Applications of Virtual Reality in Healthcare Education and Training
• Investigating the Role of Mobile Health Applications in Health Behavior Change
• Assessing the Potential of Blockchain Technology in Healthcare Data Security
• Analyzing the Ethical and Social Implications of Genetic Engineering in Healthcare

4. Healthcare Quality and Patient Safety

• Evaluating the Impact of Patient-Centered Care on Health Outcomes
• Analyzing the Role of Quality Improvement Initiatives in Reducing Medical Errors
• Assessing the Effectiveness of Medication Safety Practices in Healthcare Settings
• Exploring Strategies to Improve Healthcare Communication and Interprofessional Collaboration
• Investigating the Relationship Between Nursing Workforce and Patient Safety
• Examining the Impact of Clinical Practice Guidelines on Healthcare Quality
• Analyzing the Role of Patient Engagement in Enhancing Healthcare Quality
• Evaluating the Effectiveness of Lean Six Sigma in Healthcare Process Improvement
• Exploring the Role of Health Information Technology in Enhancing Patient Safety
• Investigating the Influence of Organizational Culture on Healthcare Quality and Safety

5. Mental Health and Psychological Well-being

• Analyzing the Impact of Stigma on Mental Health Help-Seeking Behavior
• Exploring the Effectiveness of Psychotherapy Approaches in Treating Mental Health Disorders
• Assessing the Role of Early Intervention in Preventing Mental Health Disorders
• Investigating the Relationship Between Adverse Childhood Experiences and Mental Health Outcomes
• Examining the Intersection of Mental Health and Substance Abuse Disorders
• Evaluating the Impact of Mindfulness-Based Interventions on Psychological Well-being
• Exploring the Role of Social Support in Mental Health Recovery
• Analyzing the Effectiveness of Mental Health Awareness Campaigns
• Investigating the Influence of Cultural Factors on Mental Health Help-Seeking Behavior
• Examining the Mental Health Needs and Challenges among Specific Populations (e.g., LGBTQ+, Veterans, Refugees)

6. Chronic Diseases and their Management

• Assessing the Impact of Lifestyle Factors on Chronic Disease Prevention and Management
• Exploring the Role of Community-Based Interventions in Chronic Disease Control
• Investigating the Relationship Between Social Determinants of Health and Chronic Disease Burden
• Analyzing the Use of Digital Health Technologies in Chronic Disease Management
• Examining the Impact of Health Literacy on Chronic Disease Outcomes
• Evaluating the Effectiveness of Self-Management Programs for Chronic Conditions
• Exploring the Role of Healthcare Providers in Chronic Disease Prevention and Management
• Analyzing the Impact of Health Policies on Chronic Disease Prevention Efforts
• Investigating the Relationship Between Mental Health and Chronic Disease Management
• Examining the Disparities in Access to Chronic Disease Care and Treatment

7. Healthcare Disparities and Access to Care

• Analyzing Racial and Ethnic Disparities in Healthcare Access and Quality
• Exploring the Role of Socioeconomic Factors in Healthcare Disparities
• Assessing the Impact of Geographic Location on Healthcare Access and Health Outcomes
• Investigating Gender Disparities in Healthcare Utilization and Treatment
• Examining the Influence of Health Insurance Status on Healthcare Disparities
• Evaluating the Effectiveness of Culturally Competent Care in Reducing Disparities
• Exploring the Relationship Between Language Barriers and Healthcare Access
• Analyzing the Impact of Implicit Bias on Healthcare Disparities
• Investigating the Role of Health Literacy in Healthcare Disparities
• Examining the Disparities in Mental Health Services and Access to Mental Healthcare

8. Healthcare Education and Training

• Assessing the Effectiveness of Simulation-Based Training in Healthcare Education
• Exploring the Role of Interprofessional Education in Improving Collaborative Practice
• Investigating the Impact of Technology-Enhanced Learning in Healthcare Education
• Analyzing the Use of Gamification in Healthcare Training and Skill Development
• Examining the Role of Continuing Education in Enhancing Healthcare Providers’ Competence
• Evaluating the Effectiveness of Mentorship Programs in Healthcare Education
• Exploring Strategies to Address Cultural Competence in Healthcare Education
• Analyzing the Role of Reflective Practice in Healthcare Professional Development
• Investigating the Use of Team-Based Learning in Healthcare Education
• Examining the Impact of Experiential Learning in Healthcare Training Programs

9. Public Health and Preventive Medicine

• Assessing the Impact of Vaccination Programs on Public Health Outcomes
• Exploring the Role of Health Promotion Campaigns in Preventing Non-communicable Diseases
• Investigating the Effectiveness of Community-Based Interventions in Disease Prevention
• Analyzing the Impact of Environmental Factors on Public Health
• Examining the Role of Social Determinants of Health in Health Disparities
• Evaluating the Effectiveness of Public Health Policies in Tobacco Control
• Exploring Strategies for Preventing and Managing Infectious Diseases
• Analyzing the Role of Health Education in Promoting Healthy Lifestyles
• Investigating the Influence of Media on Public Health Perceptions and Behaviors
• Examining the Challenges and Opportunities in Global Health Initiatives

10. Emerging Topics in Healthcare Research

• Assessing the Implications of Artificial Intelligence in Healthcare
• Exploring the Role of Precision Medicine in Personalized Healthcare
• Investigating the Impact of Genomic Research on Healthcare Delivery
• Analyzing the Use of Telemedicine in Rural and Underserved Areas
• Examining the Integration of Traditional and Complementary Medicine in Healthcare
• Evaluating the Potential of Digital Therapeutics in Disease Management
• Exploring the Ethical Considerations of Gene Editing Technologies in Healthcare
• Analyzing the Influence of Social Media on Healthcare Decision-Making
• Investigating the Role of Health Information Exchange in Coordinated Care
• Examining the Implications of Health Equity in Healthcare Research and Practice

This comprehensive list of healthcare research paper topics encompasses a wide range of areas within the healthcare field. Each category offers diverse research ideas that can inspire students in the health sciences to explore pressing issues, propose innovative solutions, and contribute to the advancement of healthcare knowledge. Whether you are interested in healthcare policy, ethics, technology, mental health, chronic diseases, healthcare disparities, education, public health, or emerging healthcare research paper topics, this list serves as a valuable resource to kickstart your research journey. Choose a topic that resonates with you, aligns with your academic goals, and enables you to make a meaningful impact in the field of healthcare research. Remember, the pursuit of knowledge and the drive to improve healthcare practices are at the heart of your journey as a student in the health sciences.

Choosing Healthcare Research Paper Topics

Choosing the right healthcare research paper topic is a crucial step in conducting a successful and impactful study. With the vast array of healthcare issues and areas to explore, it can be challenging to narrow down your focus. To help you navigate this process effectively, we have compiled expert advice and ten essential tips for selecting compelling healthcare research paper topics. Consider these insights as you embark on your research journey in the dynamic field of healthcare:

• Follow Your Passion : Choose a topic that genuinely interests you. Passion and enthusiasm will drive your motivation, ensuring that you remain engaged throughout the research process.
• Stay Informed : Keep up with the latest healthcare trends, emerging issues, and ongoing debates. Stay informed through reputable sources, academic journals, conferences, and professional networks to identify current and relevant research gaps.
• Identify a Research Gap : Conduct a thorough literature review to identify areas where there is a need for further research. Look for unanswered questions, controversies, or gaps in knowledge that you can address in your study.
• Consider Relevance and Significance : Choose a topic that is relevant to current healthcare challenges or contributes to improving healthcare practices, policies, or patient outcomes. Aim for a topic that has real-world implications and societal impact.
• Delve into Specific Areas : Narrow down your focus by selecting a specific aspect or subtopic within the broad field of healthcare. This allows for a more focused and in-depth analysis of the chosen area.
• Consult with Your Advisor or Faculty : Seek guidance from your research advisor or faculty members who specialize in healthcare research. They can provide valuable insights, help you refine your topic, and direct you to relevant literature and resources.
• Brainstorm with Peers : Engage in discussions with your peers and classmates to explore different perspectives and gain inspiration. Collaborative brainstorming sessions can generate new ideas and offer fresh insights.
• Consider Ethical Considerations : Take ethical considerations into account when selecting a healthcare research topic. Ensure that your research adheres to ethical guidelines and respects the rights and privacy of participants, especially in studies involving human subjects.
• Think Interdisciplinary : Consider interdisciplinary approaches to healthcare research. Explore how other disciplines, such as sociology, psychology, economics, or technology, intersect with healthcare, providing a broader perspective and enhancing the depth of your research.
• Feasibility and Available Resources : Assess the feasibility of your chosen topic, considering the resources, time, and data availability required for your research. Ensure that you have access to relevant data sources, research tools, and necessary support to carry out your study effectively.

By following these expert tips, you will be equipped to choose a healthcare research paper topic that aligns with your interests, is relevant to current healthcare challenges, and has the potential to make a meaningful impact in the field. Remember, selecting the right topic sets the foundation for a successful research endeavor, allowing you to contribute to the advancement of healthcare knowledge and practices.

How to Write a Healthcare Research Paper

Writing a healthcare research paper requires careful planning, organization, and attention to detail. To help you navigate the intricacies of the writing process, we have compiled ten essential tips to guide you towards crafting a well-written and impactful healthcare research paper. Follow these expert recommendations to enhance the quality and effectiveness of your research paper:

• Develop a Clear Research Question : Start by formulating a clear and concise research question that will serve as the central focus of your paper. Ensure that your question is specific, measurable, achievable, relevant, and time-bound (SMART).
• Conduct a Thorough Literature Review : Before diving into your research, conduct a comprehensive literature review to familiarize yourself with existing knowledge on the topic. Identify key theories, concepts, methodologies, and gaps in the literature that your research aims to address.
• Create a Solid Research Design : Design a robust research plan that aligns with your research question. Define your study population, sampling strategy, data collection methods, and statistical analyses. A well-designed research plan enhances the validity and reliability of your findings.
• Collect and Analyze Data : Implement your data collection methods, ensuring ethical considerations and adherence to research protocols. Once collected, analyze the data using appropriate statistical techniques and tools. Provide a clear description of your analytical methods.
• Structure your Paper Effectively : Organize your research paper into logical sections, including an introduction, literature review, methodology, results, discussion, and conclusion. Use headings and subheadings to enhance readability and guide the reader through your paper.
• Write a Compelling Introduction : Start your paper with a strong introduction that captures the reader’s attention and provides a concise overview of the research topic, objectives, and significance. Clearly state your research question and the rationale for your study.
• Present Clear and Concise Results : Present your research findings in a clear and concise manner. Use tables, graphs, and figures where appropriate to enhance the readability of your results. Provide a comprehensive interpretation of the results, highlighting key findings and their implications.
• Engage in Critical Analysis and Discussion : Analyze and interpret your findings in the context of existing literature. Discuss the strengths and limitations of your study, addressing potential biases or confounders. Consider alternative explanations and provide a thoughtful discussion of the implications of your findings.
• Follow Proper Citation and Referencing Guidelines : Adhere to the appropriate citation style (such as APA, MLA, or Chicago) consistently throughout your paper. Cite all sources accurately and include a comprehensive list of references at the end of your paper.
• Revise and Edit : Before finalizing your research paper, revise and edit it thoroughly. Pay attention to clarity, coherence, grammar, spelling, and punctuation. Ensure that your arguments flow logically and that your paper is well-structured and cohesive.

By following these tips, you will be well-equipped to write a high-quality healthcare research paper that effectively communicates your findings, contributes to the existing knowledge in the field, and engages readers with your insights and conclusions. Remember to seek feedback from your peers, professors, or research advisors to further refine your paper and ensure its overall excellence.

iResearchNet’s Custom Writing Services

At iResearchNet, we understand the challenges students face when it comes to writing healthcare research papers. To support you in your academic journey and ensure the highest quality of your work, we offer a comprehensive range of writing services. With a team of expert degree-holding writers and a commitment to excellence, we are dedicated to providing customized solutions tailored to your specific needs. Here are the features that set our writing services apart:

• Expert Degree-Holding Writers : Our team consists of highly qualified writers with advanced degrees in healthcare and related fields. They possess in-depth knowledge and expertise in various areas of healthcare, ensuring that your research paper is handled by professionals with subject matter expertise.
• Custom Written Works : We understand the importance of originality and uniqueness in academic writing. Our writers craft each research paper from scratch, tailoring it to your specific requirements and ensuring that it is entirely original and plagiarism-free.
• In-Depth Research : Our writers are skilled in conducting extensive research using reputable sources. They delve deep into the literature to gather the most relevant and up-to-date information, providing a solid foundation for your research paper.
• Custom Formatting : We offer custom formatting options to meet the specific guidelines of your institution and chosen citation style. Whether it’s APA, MLA, Chicago/Turabian, Harvard, or any other formatting style, our writers are well-versed in the intricacies of each.
• Top Quality : We are committed to delivering research papers of the highest quality. Our writers follow strict quality control measures to ensure that your paper meets the academic standards, including proper structure, clarity of writing, and logical flow of ideas.
• Customized Solutions : We recognize that every research paper is unique. Our writers work closely with you to understand your research objectives, guidelines, and preferences. They tailor their approach to ensure that your research paper reflects your vision and academic goals.
• Flexible Pricing : We offer flexible pricing options to accommodate students’ budgets. We understand the financial constraints students often face, and we strive to provide competitive and affordable pricing for our writing services.
• Short Deadlines : We understand that time is often a critical factor. We offer short turnaround times, allowing you to meet tight deadlines without compromising the quality of your research paper. With our dedicated team, we can handle urgent requests efficiently.
• Timely Delivery : We prioritize timely delivery to ensure that you receive your research paper well before your deadline. We understand the importance of submitting your work on time and offer our commitment to punctuality.
• 24/7 Support : Our customer support team is available 24/7 to assist you with any inquiries or concerns you may have. We are here to provide prompt and helpful assistance at any stage of the writing process.
• Absolute Privacy : We value your privacy and confidentiality. We have strict measures in place to protect your personal information and ensure that your identity remains anonymous throughout the process.
• Easy Order Tracking : We provide a user-friendly platform that allows you to track the progress of your order. You can stay updated on the status of your research paper and communicate directly with your assigned writer.
• Money-Back Guarantee : We are confident in the quality of our writing services. In the rare event that you are not satisfied with the final product, we offer a money-back guarantee, ensuring your peace of mind and commitment to your satisfaction.

At iResearchNet, we are dedicated to your success. We strive to exceed your expectations and provide you with a seamless and exceptional experience. Trust us with your healthcare research paper and let our expert writers bring your ideas to life with professionalism, accuracy, and academic excellence.

Customized Solutions for Your Research Needs

Are you a health sciences student in search of professional assistance for your healthcare research paper? Look no further than iResearchNet. We are here to empower your academic journey and help you excel in your research endeavors. With our comprehensive writing services and commitment to excellence, we provide the necessary tools and expert guidance to ensure your success.

At iResearchNet, we understand the unique challenges that come with writing healthcare research papers. Our team of expert degree-holding writers specializes in the health sciences field, allowing us to deliver customized solutions tailored to your specific research needs. Whether you need assistance in selecting a research topic, conducting in-depth literature reviews, analyzing data, or crafting a well-structured paper, we have the expertise to guide you every step of the way.

Take the next step in your healthcare research journey and unlock your academic potential with iResearchNet. Order your custom healthcare research paper today and let our expert writers bring your ideas to life with professionalism, accuracy, and academic excellence. Trust us to provide you with the guidance and support you need to achieve your research goals and make a meaningful impact in the field of healthcare.

ORDER HIGH QUALITY CUSTOM PAPER

• Health Tech
• Health Insurance
• Medical Devices
• Gene Therapy
• Neuroscience
• H5N1 Bird Flu
• Health Disparities
• Infectious Disease
• Mental Health
• Cardiovascular Disease
• Chronic Disease
• Alzheimer's
• Coercive Care
• The Obesity Revolution
• The War on Recovery
• Adam Feuerstein
• Matthew Herper
• Jennifer Adaeze Okwerekwu
• Ed Silverman
• CRISPR Tracker
• Breakthrough Device Tracker
• Generative AI Tracker
• Obesity Drug Tracker
• 2024 STAT Summit
• Wunderkinds Nomination
• STAT Madness
• STAT Brand Studio

Don't miss out

Subscribe to STAT+ today, for the best life sciences journalism in the industry

‘Visionary’ study finds inflammation, evidence of Covid virus years after infection

By Isabella Cueto July 3, 2024

R emember when we thought Covid was a two-week illness? So does Michael Peluso, assistant professor of medicine at the University of California, San Francisco. 

He recalls the rush to study acute Covid infection, and the crush of resulting papers. But Peluso, an HIV researcher, knew what his team excelled at: following people over the long term. 

advertisement

So they adapted their HIV research infrastructure to study Covid patients. The LIINC program, short for “Long-term Impact of Infection with Novel Coronavirus,” started in San Francisco at the very beginning of the pandemic. By April 2020, the team was already seeing patients come in with lingering illness and effects of Covid — in those early days still unnamed and unpublicized as long Covid. They planned to follow people’s progress for three months after they were infected with the virus.

By the fall, the investigators had rewritten their plans. Some people’s symptoms were so persistent, Peluso realized they had to follow patients for longer. Research published Wednesday in Science Translational Medicine builds on years of that data. In some cases, the team followed patients up to 900 days, making it one of the longest studies of long Covid (most studies launched in 2021 or 2022, including the NIH-funded RECOVER program).

Investigators found long-lasting immune activation months and even years after infection. And, even more concerning, they report what looked like lingering SARS-CoV-2 virus in participants’ guts. Even those who’d had Covid but no continuing symptoms had different results than those who’d never been infected. 

Related: Listen: Why Long Covid can feel scarier than a gun to the head

The team’s big idea — hypothesizing in early 2020 that, contrary to the popular narrative, Covid would last in the body — was “visionary,” long Covid researcher Ziyad Al-Aly said. “A lot of people don’t think like that.” Al-Aly was not involved with the study, but has published other long-term studies of Covid patients. He is chief of research and development at the VA Saint Louis Healthcare System. 

The research makes use of novel technology developed by the paper’s senior authors, Henry Vanbrocklin, professor in the department of radiology at UCSF, and associate professor of medicine Timothy Henrich. They figured out in the last several years they could use an antibody that bound to HIV’s code protein as a guide to see viral reservoirs. The HIV antibody, labeled with radioactive isotopes, could be tracked with imaging as it moved through the body and migrated to infected tissues. 

There were no antibodies to latch onto early in the coronavirus pandemic. Vanbrocklin instead used a chemical agent, called F-AraG, that binds to activated T cells — immune cells that flood into infected tissues. They injected F-AraG into patients, and into a scan they went. 

Tissues full of activated T cells glowed in the resulting image. Researchers found more glowing sites of immune activation in people who had been infected with Covid than in those who had not, including: the brain stem, spinal cord, cardiopulmonary tissues, bone marrow, upper pharynx, chest lymph nodes, and gut wall. 

In people with long Covid symptoms, like brain fog and fatigue, the study found the gut wall and spinal cord lit up more than in other participants. People with continuing pulmonary symptoms showed greater immune activation in their lungs. Gut biopsies in five participants revealed what appears to be persistent virus, said Peluso, who is part of the LongCovid Research Consortium of the PolyBio Research Foundation (which helped fund the study). 

Related: ‘Concern is real’ about long Covid’s impact on Americans and disability claims, report says

“The data are striking,” said Akiko Iwasaki, a professor of immunobiology and long Covid researcher at Yale University. Iwasaki was not involved in the study but is also part of PolyBio’s long Covid research group. 

Researchers used pre-pandemic scans as a control group, “the cleanest comparison that there is, before anybody on the planet could’ve possibly had this virus,” Peluso said. There were 30 participants in total (24 who’d had Covid, and six controls). Uninfected participants showed some T cell activation, but it showed up in parts of the body that help clear inflammation, like the kidney and liver. In the post-Covid group, immune activation was widespread, even in those who report that they are back to their normal health. 

The data don’t explain what exactly T cells are reacting to. As Iwasaki noted, activated T cells can be responding to persistent SARS-CoV-2 antigens or autoantigens found in people with autoimmune disease. The immune response could also be to antigens coming from other pathogens, like the common Epstein-Barr Virus. This piece requires more study, she said. 

In the gut, the researchers found what they think is RNA that encodes the virus’s signature spike protein. Other studies have found similar pieces of virus in autopsies, or within a couple of months after infection. Peluso’s work suggests the virus may stay in the body much longer — up to years after infection.

The researchers don’t know if what they’re seeing is “fossilized” leftover virus or active, productive virus. But they found double-stranded RNA in the guts of some patients who underwent biopsy. That should technically only be there if a virus is still alive, going through its life cycle, Peluso said. 

Related: Long Covid research gets a big-time funding boost

Scientists and patient advocates have been suspicious for a while of the gut reservoir post-Covid. This new data may add fuel to the idea that SARS-CoV-2 stays in some people’s guts for a long time and could actually be driving long Covid. Or, on the other hand, it could mean our immune response is failing to clear the virus and leaving behind little pieces (which might not be harmful). There are still a lot of questions, Peluso admitted. But the paper undermines the paradigm that declares Covid infection disappears after two weeks, and long Covid is just residual damage. 

The findings also suggest a need for more aggressive evaluation of immunomodulating therapies, and treatments that target leftover virus. 

Most researchers hunting for a long Covid biomarker have turned to the blood or small pieces of tissue as surrogates for what’s happening inside a patient. With the new imaging technique, Peluso and his team can see a full person on their screen — a patient’s phantom figure and gauzy organs covered in splotches of light. “It’s really striking,” he said. “‘Oh, my goodness, this is happening in someone’s spinal cord, or their GI tract, or their heart wall, or their lungs.’” 

For patients like Ezra Spier, a member of the LIINC cohort who’s had imaging done after the period captured in this latest study, the experience was validating. Finally, the life-changing experience of long Covid had become visible. “ I can now see with my own eyes the kind of dysfunction going on throughout my own body,” said Spier, who created a website for long Covid patients to more easily find clinical trials near them. 

Most participants had been infected with a pre-Omicron variant of the virus, and one person had repeat infections throughout the study period. Two participants had been hospitalized during their initial bout of Covid, but neither one received intensive care. A half-dozen patients in the study reported zero long Covid symptoms, but still showed elevated levels of immune activation. 

Related: Could long Covid’s signs of immune dysregulation in the blood lead to a diagnostic test?

The paper does not explain what the sites of infection mean for symptoms, and immune activation in a particular organ doesn’t correspond to symptoms (for example, a gut full of T cells doesn’t necessarily match with GI problems). More studies are needed to figure out what the glowing spots mean for patients’ experience of long Covid. 

And the scans don’t work as a diagnostic. In other words, patients shouldn’t rush to San Francisco (Peluso’s group only accepts study participants from the area). The imaging technique isn’t available to the general public, either. F-AraG is still being studied in this context.

But Peluso and Vanbrocklin said imaging could be a major tool in figuring out long Covid. They’ve expanded their research program to do imaging on about 50 additional patients. They are also scanning people before and after they receive different long Covid clinical trial interventions to see if there’s a change in immune activity.

About the Author Reprints

Isabella cueto.

Chronic Disease Reporter

Isabella Cueto covers the leading causes of death and disability: chronic diseases. Her focus includes autoimmune conditions and diseases of the lungs, kidneys, liver (and more). She writes about intriguing research, the promises and pitfalls of treatment, and what can be done about the burden of disease.

STAT encourages you to share your voice. We welcome your commentary, criticism, and expertise on our subscriber-only platform, STAT+ Connect

To submit a correction request, please visit our Contact Us page .

Recommended

Recommended Stories

First Opinion readers respond to psychiatry and social contracts, Medicare and weight loss meds, and more

STAT Plus: GLP-1s lower risk of 10 obesity-associated cancers, study shows

STAT Plus: Ascension is racing to unload hospitals as execs work to stem losses

STAT Plus: How controversial was the decision by FDA’s Peter Marks to approve Sarepta’s gene therapy? Check its footnotes

STAT Plus: The biotech scorecard for the third quarter: 19 stock-moving events to watch

MIT Technology Review

• Newsletters

Google helped make an exquisitely detailed map of a tiny piece of the human brain

A small brain sample was sliced into 5,000 pieces, and machine learning helped stitch it back together.

• Cassandra Willyard archive page

A team led by scientists from Harvard and Google has created a 3D, nanoscale-resolution map of a single cubic millimeter of the human brain. Although the map covers just a fraction of the organ—a whole brain is a million times larger—that piece contains roughly 57,000 cells, about 230 millimeters of blood vessels, and nearly 150 million synapses. It is currently the highest-resolution picture of the human brain ever created.

To make a map this finely detailed, the team had to cut the tissue sample into 5,000 slices and scan them with a high-speed electron microscope. Then they used a machine-learning model to help electronically stitch the slices back together and label the features. The raw data set alone took up 1.4 petabytes. “It’s probably the most computer-intensive work in all of neuroscience,” says Michael Hawrylycz, a computational neuroscientist at the Allen Institute for Brain Science, who was not involved in the research. “There is a Herculean amount of work involved.”

Many other brain atlases exist, but most provide much lower-resolution data. At the nanoscale, researchers can trace the brain’s wiring one neuron at a time to the synapses, the places where they connect. “To really understand how the human brain works, how it processes information, how it stores memories, we will ultimately need a map that’s at that resolution,” says Viren Jain, a senior research scientist at Google and coauthor on the paper, published in Science on May 9 . The data set itself and a preprint version of this paper were released in 2021 .

Brain atlases come in many forms. Some reveal how the cells are organized. Others cover gene expression. This one focuses on connections between cells, a field called “connectomics.” The outermost layer of the brain contains roughly 16 billion neurons that link up with each other to form trillions of connections. A single neuron might receive information from hundreds or even thousands of other neurons and send information to a similar number. That makes tracing these connections an exceedingly complex task, even in just a small piece of the brain..  

To create this map, the team faced a number of hurdles. The first problem was finding a sample of brain tissue. The brain deteriorates quickly after death, so cadaver tissue doesn’t work. Instead, the team used a piece of tissue removed from a woman with epilepsy during brain surgery that was meant to help control her seizures.

Once the researchers had the sample, they had to carefully preserve it in resin so that it could be cut into slices, each about a thousandth the thickness of a human hair. Then they imaged the sections using a high-speed electron microscope designed specifically for this project. 

Next came the computational challenge. “You have all of these wires traversing everywhere in three dimensions, making all kinds of different connections,” Jain says. The team at Google used a machine-learning model to stitch the slices back together, align each one with the next, color-code the wiring, and find the connections. This is harder than it might seem. “If you make a single mistake, then all of the connections attached to that wire are now incorrect,” Jain says. 

“The ability to get this deep a reconstruction of any human brain sample is an important advance,” says Seth Ament, a neuroscientist at the University of Maryland. The map is “the closest to the  ground truth that we can get right now.” But he also cautions that it’s a single brain specimen taken from a single individual. 

The map, which is freely available at a web platform called Neuroglancer , is meant to be a resource other researchers can use to make their own discoveries. “Now anybody who’s interested in studying the human cortex in this level of detail can go into the data themselves. They can proofread certain structures to make sure everything is correct, and then publish their own findings,” Jain says. (The preprint has already been cited at least 136 times .) 

The team has already identified some surprises. For example, some of the long tendrils that carry signals from one neuron to the next formed “whorls,” spots where they twirled around themselves. Axons typically form a single synapse to transmit information to the next cell. The team identified single axons that formed repeated connections—in some cases, 50 separate synapses. Why that might be isn’t yet clear, but the strong bonds could help facilitate very quick or strong reactions to certain stimuli, Jain says. “It’s a very simple finding about the organization of the human cortex,” he says. But “we didn’t know this before because we didn’t have maps at this resolution.”

The data set was full of surprises, says Jeff Lichtman, a neuroscientist at Harvard University who helped lead the research. “There were just so many things in it that were incompatible with what you would read in a textbook.” The researchers may not have explanations for what they’re seeing, but they have plenty of new questions: “That’s the way science moves forward.” 

Biotechnology and health

This grim but revolutionary dna technology is changing how we respond to mass disasters.

After hundreds went missing in Maui’s deadly fires, rapid DNA analysis helped identify victims within just a few hours and bring families some closure more quickly than ever before. But it also previews a dark future marked by increasingly frequent catastrophic events.

• Erika Hayasaki archive page

What’s next for bird flu vaccines

If we want our vaccine production process to be more robust and faster, we’ll have to stop relying on chicken eggs.

The messy quest to replace drugs with electricity

“Electroceuticals” promised the post-pharma future for medicine. But the exclusive focus on the nervous system is seeming less and less warranted.

• Sally Adee archive page

That viral video showing a head transplant is a fake. But it might be real someday. 

BrainBridge is best understood as the first public billboard for a hugely controversial scheme to defeat death.

• Antonio Regalado archive page

Stay connected

Get the latest updates from mit technology review.

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at [email protected] with a list of newsletters you’d like to receive.

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• Healthcare (Basel)

A Systematic Review on Healthcare Analytics: Application and Theoretical Perspective of Data Mining

Md saiful islam.

1 Mechanical and Industrial Engineering, Northeastern University, Boston, MA 02115, USA; [email protected] (M.S.I.); [email protected] (M.M.H.); [email protected] (X.W.); [email protected] (H.D.G.)

Md Mahmudul Hasan

Xiaoyi wang, hayley d. germack.

2 National Clinician Scholars Program, Yale University School of Medicine, New Haven, CT 06511, USA

3 Bouvé College of Health Sciences, Northeastern University, Boston, MA 02115, USA

Md Noor-E-Alam

Associated data.

The growing healthcare industry is generating a large volume of useful data on patient demographics, treatment plans, payment, and insurance coverage—attracting the attention of clinicians and scientists alike. In recent years, a number of peer-reviewed articles have addressed different dimensions of data mining application in healthcare. However, the lack of a comprehensive and systematic narrative motivated us to construct a literature review on this topic. In this paper, we present a review of the literature on healthcare analytics using data mining and big data. Following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, we conducted a database search between 2005 and 2016. Critical elements of the selected studies—healthcare sub-areas, data mining techniques, types of analytics, data, and data sources—were extracted to provide a systematic view of development in this field and possible future directions. We found that the existing literature mostly examines analytics in clinical and administrative decision-making. Use of human-generated data is predominant considering the wide adoption of Electronic Medical Record in clinical care. However, analytics based on website and social media data has been increasing in recent years. Lack of prescriptive analytics in practice and integration of domain expert knowledge in the decision-making process emphasizes the necessity of future research.

1. Introduction

Healthcare is a booming sector of the economy in many countries [ 1 ]. With its growth, come challenges including rising costs, inefficiencies, poor quality, and increasing complexity [ 2 ]. U.S. healthcare expenditures increased by 123% between 2010 and 2015—from $2.6 trillion to $3.2 trillion [ 3 ]. Inefficient—non-value added tasks (e.g., readmissions, inappropriate use of antibiotics, and fraud)—constitutes 21–47% of this enormous expenditure [ 4 ]. Some of these costs were associated with low quality care—researchers found that approximately 251,454 patients in the U.S. die each year due to medical errors [ 5 ]. Better decision-making based on available information could mitigate these challenges and facilitate the transition to a value-based healthcare industry [ 4 ]. Healthcare institutions are adopting information technology in their management system [ 6 ]. A large volume of data is collected through this system on a regular basis. Analytics provides tools and techniques to extract information from this complex and voluminous data [ 2 ] and translate it into information to assist decision-making in healthcare.

Analytics is the way of developing insights through the efficient use of data and application of quantitative and qualitative analysis [ 7 ]. It can generate fact-based decisions for “planning, management, measurement, and learning” purposes [ 2 ]. For instance, the Centers for Medicare and Medicaid Services (CMS) used analytics to reduce hospital readmission rates and avert $115 million in fraudulent payment [ 8 ]. Use of analytics—including data mining, text mining, and big data analytics—is assisting healthcare professionals in disease prediction, diagnosis, and treatment, resulting in an improvement in service quality and reduction in cost [ 9 ]. According to some estimates, application of data mining can save $450 billion each year from the U.S. healthcare system [ 10 ]. In the past ten years, researchers have studied data mining and big data analytics from both applied (e.g., applied to pharmacovigilance or mental health) and theoretical (e.g., reflecting on the methodological or philosophical challenges of data mining) perspectives.

In this review, we systematically organize and summarize the published peer-reviewed literature related to the applied and theoretical perspectives of data mining. We classify the literature by types of analytics (e.g., descriptive, predictive, prescriptive), healthcare application areas (i.e., clinical decision support, mental health), and data mining techniques (i.e., classification, sequential pattern mining); and we report the data source used in each review paper which, to our best knowledge, has never done before.

Motivation and Scope

There is a large body of recently published review/conceptual studies on healthcare and data mining. We outline the characteristics of these studies—e.g., scope/healthcare sub-area, timeframe, and number of papers reviewed—in Table 1 . For example, one study reviewed awareness effect in type 2 diabetes published between 2001 and 2005, identifying 18 papers [ 11 ]. This current review literature is limited—most of the papers listed in Table 1 did not report the timeframe and/or number of papers reviewed (expressed as N/A).

Characteristics of existing review/conceptual studies on the related topics.

N/A represents Not Reported.

There is no comprehensive review available which presents the complete picture of data mining application in the healthcare industry. The existing reviews (16 out of 21) are either focused on a specific area of healthcare, such as clinical medicine (three reviews) [ 16 , 17 , 19 ], adverse drug reaction signal detection (two reviews) [ 25 , 26 ], big data analytics (four reviews) [ 8 , 10 , 22 , 24 ], or the application and performance of data mining algorithms (five reviews) [ 9 , 13 , 14 , 20 , 21 ]. Two studies focused on specific diseases (diabetes [ 11 ], skin diseases [ 18 ]). To the best of our knowledge, none of these studies present the universe of research that has been done in this field. These studies are also limited in the rigor of their methodology except for four articles [ 11 , 16 , 22 , 25 ], which provide key insights including the timeframe covered in the study, database search, and literature inclusion or exclusion criteria, but they are limited in their scope of topics covered (see Table 1 ).

Beyond condensing the applied literature, our review also adds to the body of theoretical reviews in the analytics literature. Current theoretical reviews are limited to methodological challenges and techniques to overcome those challenges [ 15 , 16 , 27 ] and application and impact of big data analytics in healthcare [ 23 ]. In summary, the current reviews listed in Table 1 lacks in (1) width of coverage in terms of application areas, (2) breadth of data mining techniques, (3) assessment of literature quality, and (4) systematic selection and analysis of papers. In this review, we aim to fill the above-mentioned gaps. We add to this literature by covering the applied and theoretical perspective of data mining and big data analytics in healthcare with a more comprehensive and systematic approach.

2. Methodology

The methodology of our review followed the checklist proposed by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [ 28 ]. We assessed the quality of the selected articles using JBI Critical Appraisal Checklist for analytical cross sectional studies [ 29 ] and Critical Appraisal Skills Programme (CASP) qualitative research checklist [ 30 ].

2.1. Input Literature

Selected literature and their selection process for the review are described in this section. Initially a two phase advance keyword search was conducted on the database Web of Science and one phase (Phase 2) search in PubMed and Google Scholar with time filter 1 January 2005 to 31 December 2016 in “All Fields”. Journal articles written in English was added as additional filters. Keywords listed in Table 2 were used in different phases. The complete search procedure was conducted using the following procedure:

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow chart [ 28 ] illustrating the literature search process.

• Exclusion criteria: This included articles reporting on results of: qualitative study, survey, focus group study, feasibility study, monitoring device, team relationship measurement, job satisfaction, work environment, “what-if” analysis, data collection technique, editorials or short report, merely mention data mining, and articles not published in international journals . Duplicates were removed (33 articles). Finally, 117 articles were retained for the review. Figure 1 provides a PRISMA [ 28 ] flow diagram of the review process and Supplementary Information File S1 (Table S1) provides the PRISMA checklist.

Keywords for database search.

1 A logical operator used between the keywords during database search. 2 Cancer was listed independently because other dominant associations have the word “disease” associated with them (i.e., heart disease, skin disease, mental disease etc.).

2.2. Quality Assessment and Processing Steps

The full text of each of the 117 articles was reviewed separately by two researchers to eliminate bias [ 28 ]. To assess the quality of the cross sectional studies, we applied the JBI Critical Appraisal Checklist for Analytical Cross Sectional Studies [ 29 ]. For theoretical papers, we applied the Critical Appraisal Skills Programme (CASP) qualitative research checklist [ 30 ]. We modified the checklist items, as not all items specified in the JBI or CASP checklists were applicable to studies on healthcare analytics ( Supplementary Materials Table S2 ). We evaluated each article’s quality based on inclusion of: (1) clear objective and inclusion criteria; (2) detailed description of sample population and variables; (3) data source (e.g., hospital, database, survey) and format (e.g., structured Electronic Medical Record (EMR), International Classification of Diseases code, unstructured text, survey response); (4) valid and reliable data collection; (5) consideration of ethical issues; (6) detailed discussion of findings and implications; (7) valid and reliable measurement of outcomes; and (8) use of an appropriate data mining tool for cross-sectional studies and (1) clear statement of aims; (2) appropriateness of qualitative methodology; (3) appropriateness of research design; (4) clearly stated findings; and (5) value of research for the theoretical papers. Summary characteristics from any study fulfilling these criteria were included in the final data aggregation ( Supplementary Materials Table S3 ).

To summarize the body of knowledge, we adopted the three-step processing methodology outlined by Levy and Ellis [ 31 ] and Webster and Watson [ 32 ] ( Figure 2 ). During the review process, information was extracted by identifying and defining the problem, understanding the solution process and listing the important findings (“Know the literature”). We summarized and compared each article with the articles associated with the similar problems (“Comprehend the literature”). This simultaneously ensured that any irrelevant information was not considered for the analysis. The summarized information was stored in a spreadsheet in the form of a concept matrix as described by Webster and Watson [ 32 ]. We updated the concept matrix periodically, after completing every 20% of the articles which is approximately 23 articles, to include new findings (“Apply”). Based on the concept matrix, we developed a classification scheme (see Figure 3 ) for further comparison and contrast. We established an operational definition (see Table 3 ) for each class and same class articles were separated from the pool (“Analyze and Synthesis”). We compared classifications between researchers and we resolved disagreements (on six articles) by discussion. The final classification provided distinguished groups of articles with summary, facts, and remarks made by the reviewers (“Evaluate”).

Three stages of effective literature review process, adapted from Levy and Ellis [ 31 ].

Classification scheme of the literature.

Operational definition of the classes.

* Most of the definitions listed in this table are well established in literature and well know. Therefore, we did not use any specific reference. However, for some classes, specifically for types of analytics and data, varying definitions are available in the literature. We cited the sources of those definitions.

2.3. Results

The network diagram of selected articles and the keywords listed by authors in Figure 4 represents the outcome of the methodological review process. We elaborate on the resulting output in the subsequent sections using the structure of the developed classification scheme ( Figure 3 ). We also report the potential future research areas.

Visualization of high-frequency keywords of the reviewed papers. The white circles symbolize the articles and the blue circles represent keywords. The keywords that occurred only once are eliminated as well as the corresponding articles. The size of the blue circles and the texts represent how often that keyword is found. The size of the white circles is proportional to the number of keywords used in that article. The links represents the connections between the keywords and the articles. For example, if a blue circle has three links (e.g., Decision-Making) that means that keyword was used in three articles. The diagram is created with the open source software Gephi [ 34 ].

2.3.1. Methodological Quality of the Studies

Out of 117 papers included in this review, 92 applied analytics and 25 were qualitative/conceptual. The methodological quality of the analytical studies (92 out of 117) were evaluated by a modified version of 8 yes/no questions suggested in JBI Critical Appraisal Checklist for Analytical Cross Sectional Studies [ 29 ]. Each question contains 1 point (1 if the answer is Yes or 0 for No). The score achieved by each paper is provided in the final column of Supplementary Materials Table S3 . On average, each paper applying analytics scored 7.6 out of 8, with a range of 6–8 points. Major drawbacks were the absence of data source and performance measure of data mining algorithms. Out of 92 papers, 23 did not evaluate or mention the performance of the applied algorithms and eight did not mention the source of the data. However, all the papers in healthcare analytics had a clear objective and a detailed discussion of sample population and variables. Data used in each paper was either de-identified/anonymized or approved by institute’s ethical committee to ensure patient confidentiality.

We applied the Critical Appraisal Skills Programme (CASP) qualitative research checklist [ 30 ] to evaluate the quality of the 25 theoretical papers. Five questions (out of ten) in that checklist were not applicable to the theoretical studies. Therefore, we evaluated the papers in this section in a five-point scale (1 if the answer is Yes or 0 for No). Papers included in this review showed high methodological quality as 21 papers (out of 25) scored 5. The last column in the Supplementary Materials Table S3 provides the score achieved by individual papers.

2.3.2. Distribution by Publication Year

The distribution of articles published related to data mining and big data analytics in healthcare across the timeline of the study (2005–2016) is presented in Figure 5 . The distribution shows an upward trend with at least two articles in each year and more than ten articles in the last four years. Additionally, this trend represents the growing interest of government agencies, healthcare practitioners, and academicians in this interdisciplinary field of research. We anticipate that the use of analytics will continue in the coming years to address rising healthcare costs and need of improved quality of care.

Distribution of publication by year (117 articles).

2.3.3. Distribution by Journal

Articles published in 74 different journals were included in this study. Table 4 lists the top ten journals in terms of number of papers published. Expert System with Application was the dominant source of literature on data mining application in healthcare with 7 of the 117 articles. Journals were interdisciplinary in nature and spanned computational journals like IEEE Transection on Information Technology in Biomedicine to policy focused journal like Health Affairs . Articles published in Expert System with Application, Journal of Medical Systems, Journal of the American Medical Informatics Association, Healthcare Informatics Research were mostly related to analytics applied in clinical decision-making and healthcare administration. On the other hand, articles published in Health Affairs were predominantly conceptual in nature addressing policy issues, challenges, and potential of this field.

Top 10 journals on application of data mining in healthcare.

3. Healthcare Analytics

Out of 117 articles, 92 applied analytics for decision-making in healthcare. We discuss the types of analytics, the application area, the data, and the data mining techniques used in these articles and summarize them in Supplementary Materials Table S4 .

3.1. Types of Analytics

We identified three types of analytics in the literature: descriptive (i.e., exploration and discovery of information in the dataset), predictive (i.e., prediction of upcoming events based on historical data) and prescriptive (i.e., utilization of scenarios to provide decision support). Five of the 92 studies employed both descriptive and predictive analytics. In Figure 6 , which displays the percentage of healthcare articles using each analytics type, we show that descriptive analytics is the most commonly used in healthcare (48%). Descriptive analytics was dominant in all the application areas except in clinical decision support. Among the application areas, pharmacovigilance studies only used descriptive analytics as this application area is focused on identifying an association between adverse drug effects with medication. Predictive analytics was used in 43% articles. Among application areas, clinical decision support had the highest application of predictive analytics as many studies in this area are involved in risk and morbidity prediction of chest pain, heart attack, and other diseases. In contrast, use of prescriptive analytics was very uncommon (only 9%) as most of these studies were focused on either a specific population base or a specific disease scenario. However, some evidence of prescriptive analytics was found in public healthcare, administration, and mental health (see Supplementary Materials Table S4 ). These studies create a data repository and/or analytical platform to facilitate decision-making for different scenarios.

Types of analytics used in literature. ( a ) Percentage of analytics type; ( b ) Analytics type by application area.

3.2. Types of Data

To identify types of data, we adopted the classification scheme identified by Raghupathi and Raghupathi [ 23 ] which takes into account the nature (i.e., text, image, number, electronic signal), source, and collection method of data together. Table 3 provides the operational definitions of taxonomy adopted in this paper. Figure 7 a presents the percentage of data type used and Figure 7 b, the number of usage by application area. As expected, human generated (HG) data, including EMR, Electronic Health Record (HER), and Electronic Patient Record (EPR), is the most commonly (77%) used form. Web or Social media (WS) data is the second dominant (11%) type of data, as increasingly more people are using social media now and ongoing digital revolution in the healthcare sector [ 35 ]. In addition, recent development in Natural Language Processing (NLP) techniques is making the use of WS data easier than before [ 36 ]. The other three types of data (SD, BT, and BM) consist of only about 12% of total data usage, but popularity and market growth of wearable personal health tracking devices [ 37 ] may increase the use of SD and BM data.

Percentage of data type used ( a ) and type of data used by application area ( b ).

3.3. Data Mining Techniques

Data mining techniques used in the articles reviewed include classification, clustering, association, anomaly detection, sequential pattern mining, regression, and data warehousing. While elaborate description of each technique and available algorithms is out of scope of this review, we report the frequency of each technique and its sector wise distribution in Figure 8 a,b, respectively. Among the articles included in the review, 57 used classification techniques to analyze data. Association and clustering were used in 21 and 18 articles, respectively. Use of other techniques was less frequent.

Utilization of data mining techniques, ( a ) by percentage and ( b ) by application area.

A high proportion (8 out of 9) of pharmacovigilance papers used association. Use of classification was dominant in every sector except pharmacovigilance ( Figure 8 b). Data warehousing was mostly used in healthcare administration ( Figure 8 b).

We delved deeper into classification as it was utilized in the majority (57 out of 92) of the papers. There are a number of algorithms used for classification, which we present in a word cloud in Figure 9 . Support Vector Machine (SVM), Artificial Neural Network (ANN), Logistic Regression (LR), Decision Tree (DT), and DT based algorithms were the most commonly used. Random Forest (RF), Bayesian Network and Fuzzy-based algorithms were also often used. Some papers (three papers) introduced novel algorithms for specific applications. For example, Yeh et al. [ 38 ] developed discrete particle swarm optimization based classification algorithm to classify breast cancer patients from a pool of general population. Self-organizing maps and K-means were the most commonly used clustering algorithm in healthcare. Performance (e.g., accuracy, sensitivity, specificity, area under the ROC curve, positive predictive value, negative predictive value etc.) of each of these algorithms varied by application and data type. We recommend applying multiple algorithms and choosing the one which achieves the best accuracy.

Word cloud [ 39 ] with classification algorithms.

4. Application of Analytics in Healthcare

Table 3 provides the operational definitions of the six application areas (i.e., clinical decision support, healthcare administration, privacy and fraud detection, mental health, public health, and pharmacovigilance) identified in this review. Figure 10 shows the percentage of articles in each area. Among different classes in healthcare analytics, data mining application is mostly applied in clinical decision support (42%) and administrative purposes (32%). This section discusses the application of data mining in these areas and identifies the main aims of these studies, performance gaps, and key features.

Percentage of papers utilized healthcare analytics by application area (92 articles out of 117).

4.1. Clinical Decision Support

Clinical decision support consists of descriptive and/or predictive analysis mostly related to cardiovascular disease (CVD), cancer, diabetes, and emergency/critical care unit patients. Some studies developed novel data mining algorithms which we review. Table 5 describes the topics investigated and data sources used by papers using clinical decision-making, organized by major diseases category.

Topics and data sources of papers using clinical decision-making, organized by major disease category.

4.1.1. Cardiovascular Disease (CVD)

CVD is one of the most common causes of death globally [ 45 , 77 ]. Its public health relevance is reflected in the literature—it was addressed by seven articles (18% of articles in clinical decision support).

Risk factors related to Coronary Heart Disease (CHD) were distilled into a decision tree based classification system by researchers [ 40 ]. The authors investigated three events: Coronary Artery Bypass Graft Surgery (CABG), Percutaneous Coronary Intervention (PCI), and Myocardial Infarction (MI). They developed three models: CABG vs. non-CABG, PCI vs. non-PCI, and MI VS non-MI. The risk factors for each event were divided into four groups in two stages. The risk factors were separated into before and after the event at the 1st stage and modifiable (e.g., smoking habit or blood pressure) and non-modifiable (e.g., age or sex) at the 2nd stage for each group. After classification, the most important risk factors were identified by extracting the classification rules. The Framingham equation [ 78 ]—which is widely used to calculate global risk for CHD was used to calculate the risk for each event. The most important risk factors identified were age, smoking habit, history of hypertension, family history, and history of diabetes. Other studies on CHD show similar results [ 79 , 80 , 81 ]. This study had implications for healthcare providers and patients by identifying risk factors to specifically target, identify and in the case of modifiable factors, reduce CHD risk [ 40 ].

Data mining has also been applied to diagnose Coronary Artery Disease (CAD) [ 41 ]. Researchers showed that in lieu of existing diagnostic methods (i.e., Coronary Angiography (CA))—which are costly and require high technical skill—data mining using existing data like demographics, medical history, simple physical examination, blood tests, and noninvasive simple investigations (e.g., heart rate, glucose level, body mass index, creatinine level, cholesterol level, arterial stiffness) is simple, less costly, and can be used to achieve a similar level of accuracy. Researchers used a four-step classification process: (1) Decision tree was used to classify the data; (2) Crisp classification rules were generated; (3) A fuzzy model was created by fuzzifying the crisp classifier rules; and (4) Fuzzy model parameters were optimized and the final classification was made. The proposed optimized fuzzy model achieved 73% of prediction accuracy and improved upon an existing Artificial Neural Network (ANN) by providing better interpretability.

Traditional data mining and machine learning algorithms (e.g., probabilistic neural networks and SVM) may not be advanced enough to handle the data used for CVD diagnosis, which is often uncertain and highly dimensional in nature. To tackle this issue, researchers [ 42 ] proposed a Fuzzy standard additive model (SAM) for classification. They used adaptive vector quantization clustering to generate unsupervised fuzzy rules which were later optimized (minimized the number of rules) by Genetic Algorithm (GA). They then used the incremental form of a supervised technique, Gradient Descent, to fine tune the rules. Considering the highly time consuming process of the fuzzy system given large number of features in the data, the number of features was reduced with wavelet transformation. The proposed algorithm achieved better accuracy (78.78%) than the probabilistic neural network (73.80%), SVM (74.27%), fuzzy ARTMAP (63.46%), and adaptive neuro-fuzzy inference system (74.90%). Another common issue in cardiovascular event risk prediction is the censorship of data (i.e., the patient’s condition is not followed up after they leave hospital and until a new event occurs; the available data becomes right-censored). Elimination and exclusion of the censored data create bias in prediction results. To address the censorship of the data in their study on CVD event risk prediction after time, two studies [ 43 , 44 ] used Inverse Probability Censoring Weighting (IPCW). IPCW is a pre-processing step used to calculate the weights on data which are later classified using Bayesian Network. One of these studies [ 43 ] provided an IPCW based system which is compatible with any machine learning algorithm.

Electrocardiography (ECG)—non-invasive measurement of the electrical activity of the heartbeat—is the most commonly used medical studies in the assessment of CVD. Machine learning offers potential optimization of traditional ECG assessment which requires decompressing before making any diagnosis. This process takes time and large space in computers. In one study, researchers [ 45 ] developed a framework for real-time diagnosis of cardiovascular abnormalities based on compressed ECG. To reduce diagnosis time—which is critical for clinical decision-making regarding appropriate and timely treatment—they proposed and tested a mobile based framework and applied it to wireless monitoring of the patient. The ECG was sent to the hospital server where the ECG signals were divided into normal and abnormal clusters. The system detected cardiac abnormality with 97% accuracy. The cluster information was sent to patient’s mobile phone; and if any life-threatening abnormality was detected, the mobile phone alerted the hospital or the emergency personnel.

Data analytics have also been applied to more rare CVDs. One study [ 46 ] developed an intervention prediction model for Hypoplastic Left Heart Syndrome (HLHS). HLHS is a rare form of fatal heart disease in infants, which requires surgery. Post-surgical evaluation is critical as patient condition can shift very quickly. Indicators of wellness of the patients are not easily or directly measurable, but inferences can be made based on measurable physiological parameters including pulse, heart rhythm, systemic blood pressure, common atrial filling pressure, urine output, physical exam, and systemic and mixed venous oxygen saturations. A subtle physiological shift can cause death if not noticed and intervened upon. To help healthcare providers in decision-making, the researchers developed a prediction model by identifying the correlation between physiological parameters and interventions. They collected 19,134 records of 17 patients in Pediatric Intensive Care Units (PICU). Each record contained different physiological parameters measured by devices and noted by nurses. For each record, a wellness score was calculated by the domain experts. After classifying the data using a rough set algorithm, decision rules were extracted for each wellness score to aid in making intervention plans. A new measure for feature selection—Combined Classification Quality (CCQ)—was developed by considering the effect of variations in a feature values and distinct outcome each feature value leads to. Authors showed that higher value of CCQ leads to higher classification accuracy which is not always true for commonly used measure classification quality (CQ). For example, two features with CQ value of 1 leads to very different classification accuracy—35.5% and 75%. Same two features had CCQ value 0.25 and 0.40, features with 0.40 CCQ produced 75% classification accuracy. By using CCQ instead of CQ, researchers can avoid such inconsistency.

4.1.2. Diabetes

The disease burden related to diabetes is high and rising in every country. According to the World Health Organization’s (WHO) prediction, it will become the seventh leading cause of death by 2030 [ 82 ]. Data mining has been applied to identify rare forms of diabetes, identify the important factors to control diabetes, and explore patient history to extract knowledge. We reviewed 7 studies that applied healthcare analytics to diabetes.

Researchers extracted knowledge about diabetes treatment pathways and identified rare forms and complications of diabetes using a three level clustering framework from examination history of diabetic patients [ 48 ]. In this three-level clustering framework, the first level clustered patients who went through regular tests for monitoring purposes (e.g., checkup visit, glucose level, urine test) or to diagnose diabetes-related complications (e.g., eye tests for diabetic retinopathy). The second level explored patients who went through diagnosis for specific or different diabetic complications only (e.g., cardiovascular, eye, liver, and kidney related complications). These two level produced 2939 outliers out of 6380 patients. At the third level, authors clustered these outlier patients to gain insight about rare form of diabetes or rare complications. A density based clustering algorithm, DBSCAN, was used for clustering as it doesn’t require to specify the number of clusters apriori and is less sensitive to noise and outliers. This framework for grouping patients by treatment pathway can be utilized to evaluate treatment plans and costs. Another group of researchers [ 49 ] investigated the important factors related to type 2 diabetes control. They used feature selection via supervised model construction (FSSMC) to select the important factors with rank/order. They applied naïve bayes, IB1 and C4.5 algorithm with FSSMC technique to classify patients having poor or good diabetes control and evaluate the classification efficiency for different subsets of features. Experiments performed with physiological and laboratory information collected from 3857 patients showed that the classifier algorithms performed best (1–3% increase in accuracy) with the features selected by FSSMC. Age, diagnosis duration, and Insulin treatment were the top three important factors.

Data analytics have also been applied to identify patients with type 2 diabetes. In one study [ 52 ], using fragmented data from two different healthcare centers, researchers evaluated the effect of data fragmentation on a high throughput clinical phenotyping (HTCP) algorithm to identify patients at risk of developing type 2 diabetes. When a patient visits multiple healthcare centers during a study period, his/her data is stored in different EMRs and is called fragmented. In such cases, using HTPC algorithm can lead to improper classification. An experiment performed in a rural setting showed that using data from two healthcare centers instead of one decreased the false negative rate from 32.9% to 0%. In another study, researchers [ 51 ] utilized sparse logistic regression to predict type 2 diabetes risk from insurance claims data. They developed a model that outperformed the traditional risk prediction methods for large data sets and data sets with missing value cases by increasing the AUC value from 0.75 to 0.80. The dataset contained more than 500 features including demography, specific medical conditions, and comorbidity. And in another study, researchers [ 53 ] developed prediction and risk diagnosis model using a hybrid system with SVM. Using features like blood pressure, fasting blood sugar, two-hour post-glucose tolerance, cholesterol level along with other demographic and anthropometric features, the SVM algorithm was able to predict diabetes risk with 97% accuracy. One reason for achieving high accuracy compared to the study using insurance claims data [ 51 ] is the structured nature of the data which came from a cross-sectional survey on diabetes.

Different statistical and machine learning algorithms are available for classification purposes. Researchers [ 50 ] compared the performance of two statistical method (LR and Fisher linear discriminant analysis) and four machine learning algorithms (SVM (using radial basis function kernel), ANN, Random Forest, and Fuzzy C-mean) for predicting diabetes diagnosis. Ten features (age, gender, BMI, waist circumference, smoking, job, hypertension, residential region (rural/urban), physical activity, and family history of diabetes) were used to test the classification performance (diabetes or no diabetes). Parameters for ANN and SVM were optimized through Greedy search. SVM showed best performance in all performance measures. SVM was at least 5% more accurate than other classification techniques. Statistical methods performed similar to the other machine learning algorithms. This study was limited by a low prevalence of diabetes in the dataset, however, which can cause poor classification performance. Researchers [ 47 ] also proposed a novel pattern recognition algorithm by using convolutional nonnegative matrix factorization. They considered a patient as an entity and each of patients’ visit to the doctor, prescriptions, test result, and diagnosis are considered as an event over time. Finding such patterns can be helpful to group similar patients, identify their treatment pathway as well as patient management. Though they did not compare the pattern recognition accuracy with existing methods like single value decomposition (SVD), the matrix-like representation makes it intuitive.

4.1.3. Cancer

Cancer is another major threat to public health [ 83 ]. Machine learning has been applied to cancer patients to predict survival, and diagnosis. We reviewed five studies that applied healthcare analytics to cancer.

Despite many advances in treatment, accurate prediction of survival in patients with cancer remains challenging considering the heterogeneity of cancer complexity, treatment options, and patient population. Survival of prostate cancer patients has been predicted using a classification model [ 54 ]. The model used a public database-SEER (Surveillance, Epidemiology, and End Result) and applied a stratified ten-fold sampling approach. Survival prediction among prostate cancer patients was made using DT, ANN and SVM algorithm. SVM outperformed other algorithms with 92.85% classification accuracy wherein DT and ANN achieved 90% and 91.07% accuracy respectively. This same database has been used to predict survival of lung cancer patients [ 56 ]. After preprocessing the 11 features available in the data set, authors identified two features (1. removed and examined regional lymph node count and 2. malignant/in-situ tumor count) which had the strongest predictive power. They used several supervised classification methods on the preprocessed data; ensemble voting of five decision tree based classifiers and meta-classifiers (J48 DT, RF, LogitBoost, Random Subspace, and Alternating DT) provided the best performance—74% for 6 months, 75% for 9 months, 77% for 1 year, 86% for 2 years, and 92% for 5 years survival. Using this technique, they developed an online lung cancer outcome calculator to estimate the risk of mortality after 6 months, 9 months, 1 year, 2 years and 5 years of diagnosis.

In addition to predicting survival, machine learning techniques have also been used to identify patients with cancer. Among patients with breast cancer, researchers [ 38 ] have proposed a new hybrid algorithm to classify breast cancer patient from patients who do not have breast cancer. They used correlation and regression to select the significant features at the first stage. Then, at the second stage, they used discrete Particle Swarm Optimization (PSO) to classify the data. This hybrid algorithm was applied to Wisconsin Breast Cancer Data set available at UCI machine learning repository. It achieved better accuracy (98.71%) compared to a genetic algorithm (GA) (96.14%) [ 84 ] and another PSO-based algorithm (93.4%) [ 85 ].

Machine learning has also been used to identify the nature of cancer (benign or malignant) and to understand demographics related to cancer. Among patients with breast cancer, researchers [ 42 ] applied the Fuzzy standard additive model (SAM) with GA (discussed earlier in relation to CVD)-predicting the nature of breast cancer (benign or malignant). They used a UCI machine learning repository which was capable of classifying uncertain and high dimensional data with greater accuracy (by 1–2%). Researchers have also used big data [ 55 ] to create a visualization tool to provide a dynamic view of cancer statistics (e.g., trend, association with other diseases), and how they are associated with different demographic variables (e.g., age, sex) and other diseases (e.g., diabetes, kidney infection). Use of data mining provided a better understanding of cancer patients both at demographic and outcome level which in terms provides an opportunity of early identification and intervention.

4.1.4. Emergency Care

The Emergency department (ED) is the primary route to hospital admission [ 58 ]. In 2011, 20% of US population had at least one or more visits to the ED [ 86 ]. EDs are experiencing significant financial pressure to increase efficiency and throughput of patients. Discrete event simulation (i.e., modeling system operations with sequence of isolated events) is a useful tool to understand and improve ED operations by simulating the behavior and performance of EDs. Certain features of the ED (e.g., different types of patients, treatments, urgency, and uncertainty) can complicate simulation. One way to handle the complexity is to group the patients according to required treatment. Previously, the “casemix” principle, which was developed by expert clinicians to groups of similar patients in case-specific settings (e.g., telemetry or nephrology units), was used, but it has limitations in the ED setting [ 58 ]. Researchers applied [ 58 ] data mining (clustering) to the ED setting to group the patients based on treatment pattern (e.g., full ward test, head injury observation, ECG, blood glucose, CT scan, X-ray). The clustering model was verified and validated by ED clinicians. These grouping data were then used in discrete event simulation to understand and improve ED operations (mainly length of stay) and process flows for each group.

Chest pain admissions to the ED have also been examined using decision-making framework. Researchers [ 57 ] proposed a three stage decision-making framework for classifying severity of chest pain as: AMI, angina pectoris, or other. At the first stage, lab tests and diagnoses were collected and the association between them were extracted. In the second stage, experts developed association rules between lab tests diagnosis to help physicians make quick diagnostic decisions based diagnostic tests and avoid further unnecessary lab tests. In the third stage, authors developed a classification tree to classify the chest pain diagnosis based on selected lab test, diagnosis and medical record. This hybrid model was applied to the emergency department at one hospital. They developed the classification system using 327 association rules to selected lab tests using C5.0, Neural Network (NN) and SVM. C5.0 algorithm achieved 94.18% accuracy whereas NN and SVM achieved 88.89% and 85.19% accuracy respectively.

4.1.5. Intensive Care

Intensive care units cater to patients with severe and life-threatening illness and injury which require constant, close monitoring and support to ensure normal bodily function. Death is a much more common event in an ICU compared to a general medical unit—one study showed that 22.4% of total death in hospitals occurred in the ICU [ 87 ]. Survival predictions and identification of important factors related to mortality can help healthcare providers plan care. We identified two papers [ 59 , 60 ] that developed prediction models for ICU mortality rate prediction. Using a large amount of ICU patient data (specifically from the first 24 h of the stay) collected from University of Kentucky Hospital from 1998 to 2007 (38,474 admissions), one group of researchers identified 15 out of 40 significant features using Pearson’s Chi-square test (for categorical variables) and Student-t test (for continuous variable) [ 59 ]. The mortality rate was predicted by DT, ANN, SVM and APACHE III, a logistic regression based approach. Compared to the other methods applied, DT’s AUC value was higher by 0.02. The study was limited, however, by only considering the first 24 h of admission to the ICU, which may not be enough to make prediction on mortality rate. Another team of researchers [ 60 ] applied a similarity metric to predict 30-day mortality prediction in 17,152 ICU admissions data extracted from MIMIC-II database [ 88 ]. Their analysis concluded that a large group of similar patient data (e.g., vital sign, laboratory test result) instead of all patient data would lead to slightly better prediction accuracy. The logistic regression model for mortality prediction achieved 0.83 AUC value when 5000 similar patients were used for training but, its performance declined to 0.81 AUC when all the available patient data were used.

4.1.6. Other Applications

In addition to CVD, diabetes, cancer, emergency care, and ICU care, data mining has been applied to various clinical decision-making problems like pressure ulcer risk prediction, general problem lists, and personalized medical care. To predict pressure ulcer formation (localized skin and tissue damage because of shear, friction, pressure or any combination of these factors), researchers [ 62 ] developed two classification-based predictive models. One included all 14 features (including age, sex, course, Anesthesia, body position during operation, and skin status) and another, reduced model, including significant features only (5 in DT model, 7 in SVM, LR and Mahalanobis Taguchi System model). Mahalanobis Taguchi System (MTS), SVM, DT, and LR were used for both classification and feature selection (in the second model only) purposes. LR and SVM performed slightly better when all the features were included, but MTS achieved better sensitivity and specificity in the reduced model (+10% to +15%). These machine learning techniques can provide better assistance in pressure ulcer risk prediction than the traditional Norton and Braden medical scale [ 62 ]. Though the study provides the advantages of using data mining algorithms, the data set used here was imbalanced as it only had 8 cases of pressure ulcer in 168 patients. Also among patients with pressure ulcers, another team of researchers [ 63 ] recommended a data mining based alternative to the Braden scale for prediction. They applied data mining algorithms to four years of longitudinal patient data to identify the most important factors related to pressure ulcer prediction (i.e., days of stay in the hospital, serum albumin, and age). In terms of C-statistics, RF (0.83) provided highest predictive accuracy over DT (0.63), LR (0.82), and multivariate adaptive regression splines (0.78).

For data mining algorithms, which often show poor performance with imbalanced (i.e., low occurrence of one class compared to other classes) data, researchers [ 70 ] developed a sub-sampling technique. They designed two experiments, one considered sub-sampling technique and another one did not. For a highly imbalanced data set, Random Forest (RF), SVM, and Bagging and Boosting achieved better classification accuracy with this sub-sampling technique in classifying eight diseases (male genital disease, testis cancer, encephalitis, aneurysm, breast cancer, peripheral atherosclerosis, and diabetes mellitus) that had less than 5% occurrences in the National Inpatient Sample (NIS) data of Healthcare Cost and Utilization Project (HCUP). Surprisingly, possibly due to balancing the dataset through sub-sampling, RF slightly outperformed (+0.01 AUC) the other two methods.

The patient problem list is a vital component of clinical medicine. It enables decision support and quality measurement. But, it is often incomplete. Researchers have [ 64 ] suggested that a complete list of problems leads to better quality treatment in terms of final outcome [ 64 ]. Complete problem lists enable clinicians to get a better understanding of the issue and influence diagnostic reasoning. One group of researchers proposed a data mining model to find an association between patient problems and prescribed medications and laboratory tests which can act as a support to clinical decision-making [ 64 ]. Currently, domain experts spend a large amount of time for this purpose but, association rule mining can save both time and other resources. Additionally, consideration of unstructured data like doctor’s and/or nurse’s written comments and notes can provide additional information. These association rules can aid clinicians in preventing errors in diagnosis and reduce treatment complexity. For example, a set of problems and medications can co-occur frequently. If a clinician has knowledge about this relation, he/she can prescribe similar medications when faced with a similar set of problems. One group of researchers [ 61 ] developed an approach which achieved 90% accuracy in finding association between medications and problems, and 55% accuracy between laboratory tests and problems. Among outpatients diagnosed with respiratory infection, 92.79% were treated with drugs. Physicians could choose any of the 100,013 drugs available in the inventory. Moreover, in an attempt to examine the treatment plan patterns, they identified the 78 most commonly used drugs which could be prescribed, regardless of patient’s complaints and demography. The classification model used to identify the most common drugs achieved 74.73% accuracy and most importantly found variables like age, race, gender, and complaints of patients were insignificant.

Personalized medicine—tailored treatment based on a patient’s predicted response or risk of disease—is another venue for data mining algorithms. One group of researchers [ 66 ] used a big data framework to create personalized care system. One patient’s medical history is compared with other available patient data. Based on that comparison, possibility of a disease of an individual was calculated. All the possible diseases were ranked from high risk to low risk diseases. This approach is very similar to how online giants Netflix and Amazon suggest movies and books to the customer [ 66 ]. Another group of researchers [ 67 ] used the Electronic Patient Records (EPR), which contains structured data (e.g., disease code) and unstructured data (e.g., notes and comments made by doctors and nurses at different stages of treatment) to develop personalized care. From the unstructured text data, the researchers extracted clinical terms and mapped them to an ontology. Using this mapped codes and existing structured data (disease code), they created a phenotypic profile for each patient. The patients were divided into different clusters (with 87.78% precision) based on the similarity of their phenotypic profile. Correlation of diseases were captured by counting the occurrences of two or more diseases in patient phenotype. Then, the protein/gene structure associated with the diseases was identified and a protein network was created. From the sharing of specific protein structure by the diseases, correlation was identified.

Among patients with asthma, researchers [ 65 ] used environmental and patient physiological data to develop a prediction model for asthma attack to give doctors and patients a chance for prevention. They used data from a home-care institute where patients input their physical condition online; and environmental data (air pollutant and weather data). Their data mining model involved feature selection through sequential pattern mining and risk prediction using DT and association rule mining. This model can make asthma attack risk prediction with 86.89% accuracy. Real implementation showed that patients found risk prediction helpful to avoid severe asthma attacks.

Among patients with Parkinson’s disease, researchers [ 73 ] introduced a comprehensive end-to-end protocol for complex and heterogeneous data characterization, manipulation, processing, cleaning, analysis and validation. Specifically, the researchers used a Synthetic Minority Over-sampling Technique (SMOTE) to rebalance the data set. Rebalancing the dataset using SMOTE improved SVM’s classification accuracy from 76% to 96% and AdaBoost’s classification accuracy from 96% to 99%. Moreover, the study found that traditional statistical classification approaches (e.g., generalized linear model) failed to generate reliable predictions but machine learning-based classification methods performed very well in terms of predictive precision and reliability.

Among patients with kidney disease, researchers [ 71 ] developed a prediction model to forecast survival. Data collected from four facilities of University of Iowa Hospital and Clinics contains 188 patients with over 707 visits and features like blood pressure measures, demographic variables, and dialysis solution contents. Data was transformed using functional relation (i.e., the similarity between two or more features when two features have same values for a set of patients, they are combined to form a single feature) between the features. The data set was randomly divided into eight sub-sets. Sixteen classification rules were generated for the eight sub-sets using two classification algorithms—Rough Set (RS) and DT. Classes represented survival beyond three years, less than three years and undetermined. To make predictions, each classification rule (out of 16) had one vote and the majority vote decided the final predictive class. Transformed data increased predictive accuracy by 11% than raw data and DT (67% accuracy) performed better than RS (56% accuracy). The researchers suggested that this type of predictive analysis can be helpful in personalized treatment selection, resource allocation for patients, and designing clinical study. Among patients on kidney dialysis, another group of researchers [ 74 ] applied temporal pattern mining to predict hospitalization using biochemical data. Their result showed that amount of albumin—a type of protein float in blood—is the most important predictor of hospitalization due to kidney disease.

Among patients over 50 years of age, researchers [ 75 ] developed a data mining model to predict five years mortality using the EHR of 7463 patients. They used Ensemble Rotating Forest algorithm with alternating decision tree to classify the patients into two classes of life expectancy: (1) less than five years and (2) equal or greater than five years. Age, comorbidity count, previous record of hospitalization record, and blood urea nitrogen were a few of the significant features selected by correlation feature selection along with greedy stepwise search method. Accuracy achieved by this approach (AUC 0.86) was greater than the standard modified Charlson Index (AUC 0.81) and modified Walter Index (AUC 0.78). Their study showed that age, hospitalization prior the visit, and highest blood urea nitrogen were the most important factors for predicting five years morbidity. This five-year morbidity prediction model can be very helpful to optimally use resources like cancer screening for those patients who are more likely to be benefit from the resources.

Another group of researchers [ 76 ] addressed the limitations of existing software technology for disease diagnosis and prognosis, such as inability to handle data stream (DT), impractical for complex and large systems (Bayesian Network), exhaustive training process (NN). To overcome these restriction, authors proposed a decision tree based algorithm called “Very Fast Decision Tree (VFDT)”. Comparison with a similar system developed by IBM showed that VFDT utilizes lesser amount of system resources and it can perform real time classification.

Researchers have also used data mining to optimize the glaucoma diagnosis process [ 68 ]. Traditional approaches including Optical Coherence Tomography, Scanning Laser Polarimetry (SLP), and Heidelberg Retina Tomography (HRT) scanning methods are costly. This group used Fundus image data which is less costly and classified patient as either normal or glaucoma patient using SVM classifier. Before classification, authors selected significant features by using Higher Order Spectra (HOS) and Discrete Wavelet Transform (DWT) method combined and separately. Several kernel functions for SVM—all delivering similar levels of accuracy—were applied. Their approach produced 95% accuracy in glaucoma prediction. For diagnostic evaluation of chest imaging for suspicion for malignancy, researchers [ 69 ] designed trigger criteria to identify potential follow-up delays. The developed trigger predicted the patients who didn’t require follow-up evaluation. The analysis of the experiment result indicated that the algorithm to identify patients’ delays in follow-up of abnormal imaging is effective with 99% sensitivity and 38% specificity.

Data mining has also been applied to [ 72 ] compare three metrics to identify health care associated infections—Catheter Associated Bloodstream Infections, Catheter Associated Urinary Tract Infections and Ventilator Associated Pneumonia. Researchers compared traditional surveillance using National Healthcare Safety Network methodology to data mining using MedMined Data Mining Surveillance (CareFusion Corporation, San Diego, CA, USA), and administrative coding using ICD-9-CM. Traditional surveillance proved to be superior than data mining in terms of sensitivity, positive predictive value and rate estimation.

Data mining has been used in 38 studies of clinical decision-making CVD (7 articles), diabetes (seven articles), cancer (five articles), emergency care (two articles), intensive care (two articles), and other applications (16 articles). Most of the studies developed predictive models to facilitate decision-making and some developed decision support system or tools. Authors often tested their models with multiple algorithms; SVM was at the top of that list and often outperformed other algorithms. However, 15 [ 38 , 40 , 42 , 45 , 47 , 51 , 54 , 56 , 58 , 60 , 61 , 66 , 73 , 74 , 76 ] of the studies did not incorporate expert opinion from doctors, clinician, or appropriate healthcare personals in building models and interpreting results (see the study characteristics in Supplementary Materials Table S3 ). We also noted that there is an absence of follow-up studies on the predictive models, and specifically, how the models performed in dynamic decision-making situations, if doctors and healthcare professionals comfortable in using these predictive models, and what are the challenges in implementing the models if any exist? Existing literature does not focus on these salient issues.

4.2. Healthcare Administration

Data mining was applied to administrative purposes in healthcare in 32% (29 articles) of the articles reviewed. Researchers have applied data mining to: data warehousing and cloud computing; quality improvement; cost reduction; resource utilization; patient management; and other areas. Table 6 provides a list of these articles with major focus areas, problems analyzed and the data source.

Problem analyzed and data sources in healthcare administration.

4.2.1. Data Warehousing and Cloud Computing

Data warehousing [ 90 ] and cloud computing are used to securely and cost-effectively store the growing volume of electronic patient data [ 1 ] and to improve hospital outcomes including readmissions. To identify cause of readmission, researchers [ 89 ] developed an open source software—Analytic Information Warehouse (AIW). Users can design a virtual data model (VDM) using this software. Required data to test the model can be extracted in terms of a temporal ontology from the data warehouse and analysis can be performed using any standard analyzing tool. Another group of researchers took a similar approach to develop a Clinical Data Warehouse (CDW) for traditional Chinese medicine (TCM). The warehouse contains clinical information (e.g., symptoms, disease, and treatment) for 20,000 inpatients and 20,000 outpatients. Data was collected in a structured way using pre-specified ontology in electronic form. CDW provides an interface for online data mining, online analytical processing (OLAP) and network analysis to discover knowledge and provide clinical decision support. Using these tools, classification, association and network analysis between symptoms, diseases and medications (i.e., herbs) can be performed.

Apart from clinical purposes, data warehouses can be used for research, training, education, and quality control purposes. Such a data repository was created using the basic idea of Google search engine [ 92 ]. Users can pull the radiology report files by searching keywords like a simple google search following the predefined patient privacy protocol. Another data repository was created as a part of collaborative study between IBM and University of Virginia and its partner, Virginia Commonwealth University Health System was created [ 93 ]. The repository contains 667,000 patient record with 208 attributes. HealthMiner—a data mining package for healthcare created by IBM—was used to perform unsupervised analysis like finding associations, pattern and knowledge discovery. This study also showed the research benefits of this type of large data repository. Researchers [ 91 ] proposed a framework based on cloud computing and big data to unify data collected from different sources like public databases and personal health devices. The architecture was divided into 3 layers. The first layer unified heterogeneous data from different sources, the second layer provided storage support and facilitated data processing and analytics access, and the third layer provided result of analysis and platform for professionals to develop analytical tools. Some researchers [ 94 ] used mobile devices to collect personal health data. Users took part in a survey on their mobile devices and got a diagnosis report based on their health parameters input in the survey. Each survey data were saved in a cloud-based interface for effective storage and management. From user input stored in cloud, interactive geo-spatial maps were developed to provide effective data visualization facility.

4.2.2. Healthcare Cost, Quality and Resource Utilization

Ten articles applied data mining to cost reduction, quality improvement and resource utilization issues. One group of researchers predicted healthcare costs using an algorithmic approach [ 96 ]. They used medical claim data of 800,000 people collected by an insurance company over the period of 2004–2007. The data included diagnoses, procedures, and drugs. They used classification and clustering algorithms and found that these data mining algorithms improve the absolute prediction error more than 16%. Two prediction models were developed, one using both cost and medical information and the other used only cost information. Both models had similar accuracy on predicting healthcare costs but performed better than traditional regression methods. The study also showed that including medical information does not improve cost prediction accuracy. Risk-adjusted health care cost predictions, with diagnostic groups and demographic variables as inputs, have also been assessed using regression tree boosting [ 100 ]. Boosted regression tree and main effects linear models were used and fitted to predict current (2001) and prospective (2002) total health care costs per patient. The authors concluded that the combination of regression tree boosting and a diagnostic grouping scheme are a competitive alternative to commonly used risk-adjustment systems.

A sizable amount ($37.6 billion) of healthcare costs is attributable to medical errors, 45% of which stems from preventable errors [ 95 ]. To aid in physician decision-making and reduce medical errors, researchers [ 95 ] proposed a data mining-based framework-Sequential Clustering Algorithm. They identified patterns of treatment plans, tests, medication types and dosages prescribed for specific diseases, and other services provided to treat a patient throughout his/her stay in the hospital. The proposed framework was based on cloud computing so that the knowledge extracted from the data could be shared among hospitals without sharing the actual record. They proposed to share models using Virtual Machine (VM) images to facilitate collaboration among international institutions and prevent the threat of data leakage. This model was implemented in two hospitals, one in Taiwan and another in Mongolia. To identify best practices for specific diseases and prevent medical errors, another group of researchers [ 101 ] proposed a decision support system using information extraction from online documents through text and data mining. They focused on evidence based management, quality control, and best practice recommendations for medical prescriptions.

Length of Stay (LOS) is another important indicator of cost and quality of care. Accurate prediction of LOS can lead to efficient management of hospital beds and resources. To predict LOS for CAD patients, researchers [ 98 ] compared multiple models—SVM, ANN, DT and an ensemble algorithm, combing SVM, C5.0, and ANN. Ensemble algorithm and SVM produced highest accuracy, 95.9% and 96.4% respectively. In contrast, ANN was least accurate with 53.9% accuracy wherein DT achieved 83.5% accuracy. Anticoagulant drugs, nitrate drugs, and diagnosis were the top three predictors along with diastolic blood pressure, marital status, sex, presence of comorbidity, and insurance status.

To predict healthcare quality, researchers [ 104 ] used sentiment analysis (computationally categorizing opinions into categories like positive, negative and neutral) on patients’ online comments about their experience. They found above 80% agreement between sentiment analysis from online forums and traditional paper based surveys on quality prediction (e.g., cleanliness, good behavior, recommendation). Proposed approach can be an inexpensive alternative to traditional surveys and reports to measure healthcare quality.

Identification of influential factors in insurance coverage using data mining can aid insurance providers and regulators to design targeted service, additional service or proper allocation of resources to increase coverage rates. To develop a classification model to identify health insurance coverage, researchers [ 103 ] used data mining techniques. Based on 23 socio-economic, lifestyle and demographic factors, they developed a classification model with two classes, Insured and uninsured. The model was solved by ANN and DT. ANN provided 4% more accuracy than DT in predicting health insurance coverage. Among the factors, income, employment status, education, and marital status were the most important predictive factors of insurance coverage.

Among patients with lung cancer, researchers [ 97 ] investigated healthcare resource utilization (i.e., the number of visits to the medical oncologists) characteristics. They used DT, ANN and LR separately and an ensemble algorithm combining DT and ANN which resulted in the greatest accuracy (60% predictive accuracy). DT was employed to identify the important predictive features (among demographics, diagnosis, and other medical information) and ANN for classification. Data mining revealed that the utilization of healthcare resources by lung cancer patients is “supply-sensitive and patient sensitive” where supply represents availability of resources in certain region and patient represents patient preference and comorbidity. A resource allocation monitoring model for better management of primary healthcare network has also been developed [ 99 ]. Researchers considered the primary-care network as a collection of hierarchically connected modules given that patients could visit multiple physicians and physicians could have multiple care location, which is an indication of imbalanced resource distribution (e.g., number of physicians, care locations). The first level of the hierarchy consisted of three modules: health activities, population, and health resources. The second level monitored the healthcare provider availability and dispersion. The third level considered the actual visits, physicians and their availability, accessibility, and unlisted (i.e., without any assigned physician) patients. The top level of this network conducted an overall assessment of the network and made allocation accordingly. This hierarchical model was developed for a specific region in Slovenia, however, it could be easily adapted for any other region.

Overuse of screening and tests by physicians also contributes to inefficiencies and excess costs [ 102 ]. Current practice in pathology diagnosis is limited by disease focus. As an alternative to disease based system, researchers [ 102 ] used data mining in cooperation with case-based reasoning to develop an evidence based decision support system to decrease the use of unnecessary tests and reduce costs.

4.2.3. Patient Management

Patient management involves activities related to efficient scheduling and providing care to patients during their stay in a healthcare institute. Researchers [ 105 ] developed an efficient scheduling system for a rural free clinic in the United States. They proposed a hybrid system where data mining was used to classify the patients and association rule mining was used to assign a “no-show” probability. Results obtained from data mining were used to simulate and evaluate different scheduling techniques. On the other hand, these schedules could be divided into visits with administrative purposes and medical purposes. Researchers [ 108 ] suggested that patients who visit the health center for administrative purposes take less time than the patients with medical reasons. They proposed a predictive model to forecast the number of visits for administrative purposes. Their model improved the scheduling system with time saving of 21.73% (660,538 min). In contrast to administrative information/task seeking patients, some patients come for medical care very frequently and consume a large percentage of clinical workload [ 107 ]. Identifying the risk factors for frequent visit to health centers can help in reducing cost and resource utilization. A study among 85 working age “frequent attenders” identified the primary risk factors using Bayesian classification technique. The risk factors are, “high body mass index, alcohol abstinence, irritable bowel syndrome, low patient satisfaction, and fear of death” [ 107 ].

Improving publicly reported patient safety outcomes is also critical to healthcare institutions. Falls are one such outcome and are the most common and costly source of injury during hospitalization [ 110 ]. Researchers [ 109 ] analyzed the important factors related to patient falls during hospitalization. First, the authors selected significant features by Chi-square test (10 features out of 72 fall related variables were selected) and then applied ANN to develop a predictive model which achieves 0.77 AUC value. Stepwise logistic regression achieved 0.42 AUC value with 3 important variables. Both models showed that the fall assessment by nurses and use of anti-psychotic medication are associated with a lower risk of falls, and the use of diuretics is associated with an increased risk of falls. Another group of researchers [ 110 ] used fall related injury data to validate the structured information in EMR from clinical notes with the help of text mining. A group of nurses manually reviewed the electronic records to separate the correct documents from the erroneous ones which was considered as the basis of comparison. Authors employed both supervised (using a portion of manually labeled files as training set) and unsupervised technique (without considering the file labels) to classify and cluster the records. The unsupervised technique failed to separate the fare documents from the erroneous ones, wherein supervised technique performed better with 86% of fare documents in one cluster. This method can be applicable to semi-automate the EMR entry system.

4.2.4. Other Applications

Data mining has beed applied [ 111 ] to investigate the relationship between physician’s training at specific schools, procedures performed, and costs of the procedure. Researchers explored this relationship at three level: (1) they explored the distribution of procedures performed; (2) the relationship between procedures performed by physician and their alma mater—the institute that a doctor attended or got his/her degree from; and (3) geographic distribution of amount billed and payment received. This study suggested that medical school training does relate to practice in terms of procedures performed and bill charged. Patients can also provide useful information about physicians and their performance. Another group of researchers [ 112 ] used topic modeling algorithm—Latent Dirichlet Allocation (LDA)—to understand patients’ review of physicians and their concerns.

Data mining has also been applied [ 115 ] to analyze the information seeking behavior of health care professionals, and to assess the feasibility of measuring drug safety alert response from the usage logs of online medical information resources. Researchers analyzed two years of user log-in data in UpToDate website to measure the volume of searches associated with medical conditions and the seasonal distribution of those searches. In addition, they used a large collection of online media articles and web log posts as they characterized food and drug alert through the changes in UpToDate search activity compared to the general media activity. Some researchers [ 113 ] examined changes of key performance indicators (KPIs) and clinical workload indicators in Greek National Health System (NHS) hospitals with the help of data mining. They found significant changes in KPIs when necessary adjustments (e.g., workload) were made according to the diagnostic related group. The results remained for general hospitals like cancer hospitals, cardiac surgery as well as small health centers and regional hospitals. Their findings suggested that the assessment methodology of Greek NHS hospitals should be re-evaluated in order to identify the weaknesses in the system, and improve overall performance. And in home healthcare, another group of researchers [ 116 ] reviewed why traditional statistical analysis fails to evaluate the performance of home healthcare agencies. The authors proposed to use data mining to identify the drivers of home healthcare service among patients with heart failure, hip replacement, and chronic obstructive pulmonary disease using length of stay and discharge destination.

The relationship between epidemiological and genetic evidence and post market medical device performance has been evaluated using HCUPNet data [ 114 ]. This feasibility study explored the potential of using publicly accessible data for identifying genetic evidence (e.g., comorbidity of genetic factors like race, sex, body structure, and pneumothorax or fibrosis) related to devices. It focused on the ventilation-associated iatrogenic pneumothorax outcome in discharge of mechanical ventilation and continuous positive airway pressure (CPAP). The results demonstrated that genetic evidence-based epidemiologic analysis could lead to both cost and time efficient identification of predictive features. The literature of data mining applications in healthcare administration encompasses efficient patient management, healthcare cost reduction, quality of care, and data warehousing to facilitate analytics. We identified four studies that used cloud-based computing and analytical platforms. Most of the research proposed promising ideas, however, they do not provide the results and/or challenges during and after implementation. An ideal example of implementation could be the study of efficient appointment scheduling of patients [ 108 ].

4.3. Healthcare Privacy and Fraud Detection

Health data privacy and medical fraud are issues of prominent importance [ 118 ]. We reviewed four articles—displayed and described in Table 7 —that discussed healthcare privacy and fraud detection.

List of papers in healthcare privacy and fraud detection.

The challenges of privacy protection have been addressed by a group of researchers [ 122 ] who proposed a new anonymization algorithm for both distributed and centralized anonymization. Their proposed model performed better than K-anonymization model in terms of retaining data utility without losing much data privacy (for K = 20, the discernibility ratio—a normalized measure of data quality—of the proposed approach and traditional K-anonymization method were 0.1 and 0.4 respectively). Moreover, their proposed algorithm could handle large scale, high dimensional datasets. To address the limitations of today’s healthcare information systems—EHR data systems limited by lack of inter-operability, data size, and security—a mobile cloud computing-based big data framework has been proposed [ 119 ]. This novel cloud-based framework proposed storing EHR data from different healthcare providers in an Internet provider’s facility, offering providers and patients different levels of access and authority. Security would be ensured by using encryption algorithms, one-time passwords, or 2-factor authentication. Big data analytics would be handled using Google big query or MapReduce software. This framework could reduce cost, increase efficiency, and ensure security compared to the traditional technique which uses de-identification or anonymization technique. This traditional technique leaves healthcare data vulnerable to re-identification. In a case study, researchers demonstrated that hackers can make association between small pieces of information and can identify patients [ 120 ]. The case study made use of personal information provided in two Medicare social networking sites, MedHelp and Mp and Th1 to identify an individual.

Detection of fraud and abuse (i.e., suspicious care activity, intentional misrepresentation of information, and unnecessary repetitive visits) uses big data analytics. Using gynecological hospital data, researchers [ 121 ] developed a framework from two domain experts manually identifying features of fraudulent cases from a data pool of treatment plans doctors frequently follow. They applied this framework to Bureau of National Health Insurance (BNHI) data from Taiwan; their proposed framework detected 69% of the fraudulent cases, which improved the existing model that detected 63% of the fraudulent cases.

In summary, patient data privacy and fraud detection are of major concern given increasing use of social media and people’s tendency to put personal information on social media. Existing data anonymization or de-identification techniques can become less effective if they are not designed considering the fact that a large portion of our personal information is now available on social media.

4.4. Mental Health

Mental illness is a global and national concern [ 123 ]. According to the National Survey on Drug Use and Health (NSDUH) data from 2010 to 2012, 52.2% of U.S. population had either mental illness, or substance abuse/dependence [ 124 ]. Additionally, nearly 30 million people in the U.S. suffer from anxiety disorders [ 125 ]. Table 8 summarizes the four articles we reviewed that apply data mining in analyzing, diagnosing, and treating mental health issues.

List of data mining application in mental health with data sources.

To classify developmental delays of children based on illness, researchers [ 126 ] examined the association between illness diagnosis and delays by building a decision tree and finding association between cognitive, language, motor, and social emotional developmental delays. This study has implications for healthcare professionals to identify and intervene on delays at an early stage. To assist physicians in monitoring anxiety disorder, another group of researchers [ 125 ] developed a data mining based personalized treatment. The researchers used Context Awareness Information including static (personal information like, age, sex, family status etc.) and dynamic (stress, environmental, and symptoms context) information to build static and dynamic user models. The static model contained personal information and the dynamic model contained four treatment-supportive services (i.e., lifestyle and habits pattern detection service, context and stress level pattern detection service, symptoms and stress level pattern detection service, and stress level prediction service). Relations between different dynamic parameters were identified in first three services and the last service was used for stress level prediction under different scenarios. The model was validated using data from 27 volunteers who were selected by anxiety measuring test.

To predict early diagnosis for mental disorders (e.g., insomnia, dementia), researchers developed a model detecting abnormal physical activity recorded by a wearable device [ 127 ]. They performed two experiments to compare the development of a reference model using historical user physical movement data. In the first experiment, users wore the watch for one day and based on that day, a reference behavior model was developed. After 22 days, the same user used it again for a day and abnormality was detected if the user’s activities were significantly different from the reference model. In the second experiment, users used the watch regularly for one month. Abnormality was detected with a fuzzy valuation function and validated with user’s reported activity level. In both experiments, users manually reported their activity level, which was used as a validating point, only two out of 26 abnormal events were undetected. Through these two experiments, the researchers claimed that their model could be useful for both online and offline abnormal behavior detection as the model was able to detect 92% of the unusual events.

To classify schizophrenia, another study [ 128 ] used free speech (transcribed text) written or verbalized by psychiatric patients. In a pool of patients with schizophrenia and control subjects, using supervised algorithms (SVM and DT), they discriminated between patients with schizophrenia and normal control patients. SVM achieved 77% classification accuracy whereas DT achieved 78% accuracy. When they added patients with mania to the pool, they were unable to differentiate patients with schizophrenia.

Use of data analytics in diagnosing, analyzing, or treating mental health patients is quite different than applying analytics to predict cancer or diabetes. Context of data (static, dynamic, or unobservable environment) seemed more important than volume in this case [ 125 ], however, this is not always adopted in literature. A model without situational awareness (a context independent model) may lose predictive accuracy due to the confounding effect of surrounding environment [ 129 ].

4.5. Public Health

Seven articles addressed issues that were not limited to any specific disease or a demographic group, which we classified as public health problems. Table 9 contains the list of papers considering public health problems with data sources.

List of data mining application in public health with data sources.

To make data mining accessible to non-expert users, specifically public health decision makers who manage public cancer treatment programs in Brazil, researchers [ 134 ] developed a framework for an automated data mining system. This system performed a descriptive analysis (i.e., identifying relationships between demography, expenditure, and tumor or cancer type) for public decision makers with little or no technical knowledge. The automation process was done by creating pre-processed database, ontology, analytical platform and user interface.

Analysis of disease outbreaks has also applied data analytics. [ 131 , 133 ] Influenza, a highly contagious disease, is associated with seasonal outbreaks. The ability to predict peak outbreaks in advance would allow for anticipatory public health planning and interventions to lessen the effect of the outbreaks. To predict peak influenza visits to U.S. military health centers, researchers [ 131 ] developed a method to create models using environmental and epidemiological data. They compared six classification algorithms—One-Classifier 1, One-Classifier 2 [ 137 ], a fusion of the One-Classifiers, DT, RF, and SVM. Among them, One-Classifier 1 was the most efficient with F-score 0.672 and SVM was second best with F-score 0.652. To examine the factors that drive public and professional search patterns for infectious disease outbreaks another group of researchers [ 133 ] used online behavior records and media coverage. They identified distinct factors that drive professional and layperson search patterns with implications for tailored messaging during outbreaks and emergencies for public health agencies.

To store and integrate multidimensional and heterogeneous data (e.g., diabetes, food, nutrients) applied to diabetes management, but generalizable to other diseases researchers [ 130 ] proposed an intelligent information management framework. Their proposed methodology is a robust back-end application for web-based patient-doctor consultation and e-Health care management systems with implications for cost savings.

A real-time medical emergency response system using the Internet of Things (networking of devices to facilitate data flow) based body area networks (BANs)—a wireless network of wearable computing devices was proposed by researchers [ 136 ]. The system consists of “Intelligent Building”—a data analysis model which processes the data collected from the sensors for analysis and decision. Though the author claims that the proposed system had the capability of efficiently processing wireless BAN data from millions of users to provide real-time response for emergencies, they did not provide any comparison with the state-of-the-art methods.

Decision support tools for regional health institutes in Slovenia [ 135 ] have been developed using descriptive data mining methods and visualization techniques. These visualization methods could analyze resource availability, utilization and aid to assist in future planning of public health service.

To build better customer relations management at an Iranian hospital, researchers [ 132 ] applied data mining techniques on demographic and transactions information. The authors extended the traditional Recency, Frequency, and Monetary (RFM) model by adapting a new parameter “Length” to estimate the customer life time value (CLV) of each patient. Patients were separated into classes according to estimated CLV with a combination of clustering and classification algorithms. Both DT and ANN performed similarly in classification with approximately 90% accuracy. This type of stratification of patient groups with CLV values would help hospitals to introduce new marketing strategies to attract new customers and retain existing ones.

The application of data mining to public health decision-making has become increasingly common. Researchers utilized data mining to design healthcare programs and emergency response, to identify resource utilization, patient satisfaction as well as to develop automated analytics tool for non-expert users. Continuation of this effort could lead to a patient-centered, robust healthcare system.

4.6. Pharmacovigilance

Pharmacovigilance involves post-marketing monitoring and detection of adverse drug reactions (ADRs) to ensure patient safety [ 138 ]. The estimated annual social cost of ADR events exceeds one billion dollars, making it an important part of healthcare system [ 139 ]. Characteristics of the nine papers addressing pharmacovigilance are displayed in Table 10 .

List of data mining application in pharmacovigilance with data sources.

Researchers considered muscular and renal AEs caused by pravastatin, simvastatin, atorvastatin, and rosuvastatin by applying data mining techniques to the FDA’s Adverse Event Reporting System (FAERS) database reports from 2004 to 2009 [ 143 ]. They found that all statins except simvastatin were associated with muscular AE; rosuvastatin had the strongest association. All statins, besides atorvastatin, were associated with acute renal failure. The criteria used to identify significant association were: proportional reporting ratio (PRR), reporting odds ratio (ROR), information component (IC), and empirical Bayes geometric mean (EBGM). In another study of AEs related to statin family, researchers used a Korean claims database [ 145 ] and showed that a relative risk-based data-mining approach successfully detected signals for rosuvastatin.

Three more studies used the FDA’s AERS report database. In an examination of ADR “hypersensitivity” to six anticancer agents [ 142 ] data mining results showed that Paclitaxel is associated with mild to lethal reaction wherein Docetaxel is associated to lethal reaction, and the other four drugs were not associated to hypersensitivity [ 142 ]. Another researcher [ 139 ] argued that AEs can be caused not only by a single drug, but also by a combination of drugs [ 140 ]. They showed that that 84% of the AERs reports contain an association between at least one drug and two AEs or two drugs and one AE. Another group [ 138 ] increased precision in detecting ADRs by considering multiple data sources together. They achieved 31% (on average) improvement in identification by using publicly available EHRs in combination with the FDA’s AERS reports.

Furthermore, dose-dependent ADRs have been identified by researchers using models developed from structured and unstructured EHR data [ 141 ]. Among the top five drugs associated with ADRs, four were found to be related to dose [ 141 ]. Pharmacovigilance activity has also been prioritized using unstructured text data in EHRs [ 144 ]. In traditional pharmacovigilance, ADRs are unknown. While looking for association between a drug and any possible ADR, it is possible to get false signals. Such false signals can be avoided if a list of possible ADRs is already known. Researchers [ 144 ] developed an ordered list of 23 ADRs which can be very helpful for future pharmacovigilance activities. To detect unexpected and rare ADRs in real-world healthcare administrative databases, another group of researchers [ 146 ] designed an algorithm—Unexpected Temporal Association Rules (UTARs)—that performs more effectively than existing techniques.

We identified one study that used data outside of adverse event reports or HER data. For early detection of ADR, one group of researchers used online forums [ 140 ]. They identified the side effect of a specific drug called “Erlotinib” used for lung cancer. Sentiment analysis—a technique of categorizing opinions—on data collected from different cancer discussion forums showed that 70% of users had a positive experience after using this drug. Users most frequently reported were acne and rash. Apart from pharmacovigilance, this type of analysis can be very helpful for the pharmaceutical companies to analyze customer feedback. Researchers can take advantage of the popularity of social media and online forums for identifying adverse events. These sources can provide signals of AEs quicker than FDA database as it takes time to update the database. By the time AE reports are available in the FDA database, there could already be significant damage to patient and society. Moreover, it can help to avoid the limitations of FDA AERS database like biased reporting and underreporting [ 141 ].

5. Theoretical Study

Twenty-five of the articles we reviewed focus on the theoretical aspects of the application of data mining in healthcare including designing the database framework, data collection, and management to algorithmic development. These intellectual contributions extend beyond the analytical perspective of data—descriptive, predictive or prescriptive analytics—to the sectors and problems highlighted in Table 11 .

Problem analyzed in theoretical studies.

The existing theoretical literature on disease control highlighted the current state of epidemics, cancer and mental health. To help physicians make real-time decisions about patient care, one group of researchers [ 147 ] proposed a real-time EMR data mining based clinical decision support system. They emphasized the need to have an anonymized EMR database which can be explored by using a search engine similar to web search engine. In addition, they focused on designing a framework for next generation EMR-based database that can facilitate the clinical decision-making process, and is also capable of updating a central population database once patients’ recent (new) clinical records are available. Another researcher [ 148 ] forecasted future challenges in infection control that entails the importance of having timely surveillance system and prevention programs in place. To that end, they necessitate the formation, control and utilization of fully computerized patient record and data-mining-derived epidemiology. Finally, they recommended performance feedback to caregivers, wide accessibility of infection prevention tools, and access to documents like lessons learned and evidence-based best practices to strengthen the infection control, surveillance, and prevention scheme. Authors in [ 150 ] addressed the activities executed by national Institute of Mental Health (NIMH) in collaboration with other state organizations (e.g., Substance Abuse and Mental Health Service Administration (SAMSHSA), Center for Mental Health Service (CMHS) to promote optimal collection, pooling/aggregation, and use of big data to support ongoing and future researches of mental health practices. The outcome summary showcased that effective pooling/aggregation of state-level data from different sources can be used as a dashboard to set priorities to improve service qualities, measure system performance and to gain specific context-based insights that are generalizable and scalable across other systems, leading to a successful learning-based mental health care system. Another group of researchers [ 150 ] outlined the barriers and potential benefits of using big data from CancerLinQ (a quality and measurement reporting system as an initiative of the American Society of Clinical Oncology (ASCO) that collects information from EHRs of cancer patients for oncologists to improve the outcome and quality of care they provide to their patients). However, the authors also mentioned that these benefits are contingent upon the confidence of the patients, encouraging them to share their data out of the belief that their health records would be used appropriately as a knowledge base to improve the quality of the health care of others, as it is for themselves. This motivated ASCO to ensure that proper policies and procedures are in place to deal with the data quality, data security and data access, and adopt a comprehensive regulatory framework to ensure patients’ data privacy and security.

Another group of researchers [ 151 ] data quality and database management to quantify, and consequentially understand the inherent uncertainty originating from radiology reporting system. They discussed the necessity of having a structured reporting system and emphasized the use of standardize language, leading to Natural Language Processing (NLP). Furthermore, they also indicated the need for creating a Knowledge Discovery Database (KDD) which will be consistent to facilitate the data-driven and automated decision support technologies to help improving the care provided to patients based on enhanced diagnosis quality and clinical outcome. A group of authors in [ 152 ] pointed that the success derived from the current trend of big-data analytics largely depends on how better the quality of the data collected from variety of sources are ensured. Their findings imply that the data quality should be assessed across the entire lifecycle of health data by considering the errors and inaccuracies stemmed from multiple of sources, and should also quantify the impact that data collection purpose on the knowledge and insights derived from the big data analytics. For that to ensure, they recommend that enterprises who deal with healthcare big data should develop a systematic framework including custom software or data quality rule engines, leading to an effective management of specific data-quality related problems. Researchers in [ 155 ] uncovered the lack of connection between phenomenological and mechanistic models in computational biomedicines. They emphasized the importance of big data which, when successfully extracted and analyzed, followed by the combination with Virtual Physiological Human (VPH)—an initiative to encourage personalized healthcare—can afford with effective and robust medicine solutions. In order for that to happen, they mentioned some challenges (e.g., confidentiality, volume and complexity of big data; integration of bioinformatics, systems biology and phenomics data; efficient storage of partial or complete data within organization to maximize the performance of overall predictive analytics) and concluded that these need to be addressed for successful development of big data technologies in computational medicines, enabling their adoption in clinical settings. Even though big data can generate significant value in modern healthcare system, researchers in [ 154 ] stated that without a set of proper IT infrastructures, analytical and visualization tools, and interactive interfaces to represent the work flows, the insights generated from big data will not be able to reach its full potential. To overcome this, they recommended that health care organizations engaging in data sharing devise new policies to protect patients’ data against potential data breaches.

Three papers [ 155 , 156 , 157 ] considered health care policies and ethical and legal issues. One [ 155 ] outlined a national action plan to incorporate sharable and comparable nursing data beyond documentation of care into quality reporting and translational research. The plan advocates for standardized nursing terminologies, common data models, and information structures within EHRs. Another paper [ 157 ] analyzed the major policy, ethical, and legal challenges of performing predictive analytics on health care big data. Their proposed recommendations for overcoming challenges raised in the four-phase life cycle of a predictive analytics model (i.e., data acquisition, model formulation and validation, testing in real-world setting and implementation and use in broader scale) included developing a governance structure at the earliest phase of model development to guide patients and participating stakeholders across the process (from data acquisition to model implementation). They also recommended that model developers strictly comply with the federal laws and regulations in concert with human subject research and patients information privacy when using patients’ data. And another paper [ 156 ] explored four central questions regarding: (i) aspects of big-data most relevant to health care, (ii) policy implications, (iii) potential obstacles in achieving policy objectives, and (iv) availability of policy levers, particularly for policy makers to consider when developing public policy for using big data in healthcare. They discussed barriers (including ensuring transparency among patients and health care providers during data collection) to achieve policy objectives based on a recent UK policy experiment, and argued for providing real-life examples of ways in which data sharing can improve healthcare.

Three papers [ 158 , 159 , 160 ] offered examples of realistic ways such as establishing policy leadership and risk management framework combining commercial and health care entities to recognize existing privacy related problem and devise pragmatic and actionable strategies of maintaining patient privacy in big data analytics. One paper [ 158 ] provided a policy overview of health care and data analytics, outlined the utility of health care data from a policy perspective, reviewed a variety of methods for data collection from public and private sources, mobile devices and social media, examined laws and regulations that protect data and patients’ privacy, and discussed a dynamic interplay among three aspects of today’s big data driven personal health care—policy goals to tackle both cost, population health problem and eliminate disparity in patient care while maintaining their privacy. Another study [ 159 ] proposed a Secure and Privacy Preserving Opportunistic Computing (SPOC) framework to be used in healthcare emergencies focused on collecting intensive personal health information (through mobile devices like smart phone or wireless sensors) with minimal privacy disclosure. The premise of this framework is that when a user of this system (called medical user) faces any emergency, other users in the vicinity with similar disease or symptom (if available) can come to help that user before professional help arrives. It is assumed that two persons with similar disease are skilled enough to help each other and the threshold of similarity is controlled by the user. And in physician prescribing—another paper [ 160 ] identified strategies for data mining from physicians’ prescriptions while maintaining patient privacy.

Theoretical research on personalized-health care services—treatment plans designed for someone based on the susceptibility of his/her genomic structure to a disease—also emerged from the literature review. One study [ 161 ] highlighted the potential of powerful analytical tools to open an avenue for predictive, preventive, participatory, and personalized (P4) medicine. They suggested a more nuanced understanding of the human systems to design an accurate computational model for P4 medicine. Reviewing the research paradgims of current person-centered approaches and traditions, another study [ 162 ] advocated a transdisciplinary and complex systems approach to improve the field. They synthesized the emerging aproaches and methodologies and highlighted the gaps between academic research and accessibility of evaluation, informatics, and big data from health information systems. Another paper [ 163 ] reviewed the availability of big data and the role of biomedical informatics in personalized medicine, emphasizing the ethical concerns related to personalized medicines and health equity. Personalized medicine has a potential to reduce healthcare cost, however, the researchers think it can create race, income, and educational disparity. Certain socioeconomic and demographic groups currently have less or no access to healthcare and data driven personalized medicine will exclude those groups, increasing disparities. They also highlighted the impact of EHRs and CDWs on the field of personalized medicine through acclerated research and decreased the delivery time of new technologies.

A myriad of extant theoretical points has also been identified in the literature. These topics range from exploiting big data to: study the paradigm shift in healthcare policy and management from prioritizing volume to value [ 164 , 167 ]; aid medical device consumers in their decision-making [ 166 ]; improve emergency departments [ 169 ]; perform command surveillance and policy analysis for Army leadership [ 170 ]; to comparing different simulation methods (i.e., systems dynamics, discrete event simulation and agent based modeling) for specific health care system problems like resource allocation, length of stay [ 165 ]; to the ethical challenges of security, management, and ownership [ 170 ]. Another researcher outlined the challenges the E.U. is facing in data mining given numerous historical, technical, legal, and political barriers [ 168 ].

6. Future Research and Challenges

Data mining has been applied in many fields including finance, marketing, and manufacturing [ 172 ]. Its application in healthcare is becoming increasingly popular [ 173 ]. A growing literature addresses the challenges of data mining including noisy data, heterogeneity, high dimensionality, dynamic nature, computational time. In this section, we focus on future research applications including personalized care, information loss in preprocessing, collecting healthcare data for research purposes, automation for non-experts, interdisciplinarity of study and domain expert knowledge, integration into the healthcare system, and prediction-specific to data mining application and integration in healthcare.

• Personalized care

The EMR is increasingly used to document demographic and clinician patient information [ 1 ]. EMR data can be utilized to develop personalized care plans, enhancing patient experience [ 162 ] and improving care quality.

• Loss of information in pre-processing

Pre-processing of data, including handling missing data, is the most time-consuming and costly part of data mining. The most common method used in the papers reviewed was deletion or elimination of missing data. In one study, approximately 46.5% of the data and 363 of 410 features were eliminated due to missing values [ 49 ]. In another, researchers [ 98 ] were only able to use 2064 of 4948 observations (42%) [ 98 ]. By eliminating missing value cases and outliers, we are losing a significant amount of information. Future research should focus on finding a better method of missing value estimation than elimination. Moreover, data collection techniques should be developed or modified to avoid this issue.

Similar to missing data, deletion or elimination is a common way to handle outliers [ 174 ]. However, as illustrated in one of the studies we reviewed [ 48 ], outliers can be used to gain information about rare forms of diseases. Instead of neglecting the outliers, future research should analyze them to gain insight.

• Collecting healthcare data for research purpose

Traditionally, the primary objective of data collection in healthcare is documentation of patient condition and care planning [ 109 ]. Including research objectives in the data collection process through structured fields could yield more structured data with fewer cases of error and missing values [ 64 ]. A successful example of data collection for research purpose is the Study of Health in Pomerania (SHIP) [ 175 ]. The objective of SHIP was to identify common diseases, population level risk factors, and overall health of people living in the north-east region of Germany. This study only suffered from one “mistake” for every 1000 data entries [ 175 ] which ensures a structured form of data with high reliability, less noise and fewer missing values. We can take advantage of current documentation processes (EMR or EHR) by modifying them to collect more reliable and structured data. Long-term vision and planning is required to introduce research purpose in healthcare data collection.

• Automation of data mining process for non-expert users

The end users of data mining in healthcare are doctors, nurses, and healthcare professionals with limited training in analytics. One solution for this problem is to develop an automated (i.e., without human supervision) system for the end users [ 134 ]. A cloud-based automated structure to prevent medical errors could also be developed [ 95 ]; but the task would be challenging as it involves different application areas and one algorithm will not have similar accuracy for all applications [ 134 ].

• Interdisciplinary nature of study and domain expert knowledge

Healthcare analytics is an interdisciplinary research field [ 134 ]. As a form of analytics, data mining should be used in combination with expert opinion from specific domains—healthcare and problem specific (i.e., oncologist for cancer study, cardiologist for CVD) [ 106 ]. Approximately 32% of the articles in analytics did not utilize expert opinion in any form. Future research should include members from different disciplines including healthcare.

• Integration in healthcare system

Very few articles reviewed made an effort to integrate the data mining process into the actual decision-making framework. The impact of knowledge discovery through data mining on healthcare professional’s workload and time is unclear. Future studies should consider the integration of the developed system and explore the effect on work environments.

• Prediction error and “The Black Swan” effect

In healthcare, it is better not to predict than making an erroneous prediction [ 46 ]. A little under half of the literature we identified in analytics is dedicated to prediction but, none of the articles discussed the consequence of a prediction error. High prediction accuracy for cancer or any other disease does not ensure an accurate application to decision-making.

Moreover, prediction models may be better at predicting commonplace events than rare ones [ 176 ]. Researchers should develop more sophisticated models to address the unpredictable, “The Black Swan” [ 176 ]. One study [ 101 ] addressed a similar issue in evidence based recommendations for medical prescriptions. Their concern was, how much evidence should be sufficient to make a recommendation. Many of the studies in this review do not address these salient issues. Future research should address the implementation challenges of predictive models, especially how the decision-making process should adapt in case of errors and unpredictable incidents.

7. Conclusions

The development of an informed decision-making framework stems from the growing concern of ensuring a high value and patient-focused health care system. Concurrently, the availability of big data has created a promising research avenue for academicians and practitioners. As highlighted in our review, the increased number of publications in recent years corroborates the importance of health care analytics to build improved health care systems world-wide. The ultimate goal is to facilitate coordinated and well-informed health care systems capable of ensuring maximum patient satisfaction.

This paper adds to the literature on healthcare and data mining ( Table 1 ) as it is the first, to our knowledge, to take a comprehensive review approach and offer a holistic picture of health care analytics and data mining. The comprehensive and methodologically rigorous approach we took covers the application and theoretical perspective of analytics and data mining in healthcare. Our systematic approach starting with the review process and categorizing the output as analytics or theoretical provides readers with a more widespread review with reference to specific fields.

We also shed light on some promising recommendations for future areas of research including integration of domain-expert knowledge, approaches to decrease prediction error, and integration of predictive models in actual work environments. Future research should recommend ways so that the analytic decision can effectively adapt with the predictive model subject to errors and unpredictable incidents. Regardless of these insightful outcomes, we are not constrained to mention some limitations of our proposed review approach. The sole consideration of academic journals and exclusion of conference papers, which may have some good coverage in this sector is the prime limitation of this review. In addition to this, the search span was narrowed to three databases for 12 years which may have ignored some prior works in this area, albeit the increasing trend since 2005 and less number of publications before 2008 can minimize this limitation. The omission of articles published in languages other than English can also restrict the scope of this review as related papers written in other languages might be evident in the literature. Moreover, we did not conduct forward (reviewing the papers which cited the selected paper) and backward (reviewing the references in the selected paper and authors’ prior works) search as suggested by Levy and Ellis [ 31 ].

Despite these limitations, the systematic methodology followed in this review can be used in the universe of healthcare areas.

Supplementary Materials

The following are available online at http://www.mdpi.com/2227-9032/6/2/54/s1 , Table S1: PRISMA checklist, Table S2: Modified checklists and comparison, Table S3: Study characteristics, Table S4: Classification of reviewed papers by analytics type, application area, data type, and data mining techniques.

Author Contributions

Contribution of the authors can be summarized in following manner. Conceptualization: M.S.I., M.N.-E.-A.; Formal analysis: M.S.I., M.M.H., X.W.; Investigation: M.S.I., M.M.H., X.W.; Methodology: M.S.I.; Project administration: M.S.I., M.N.-E.-A.; Supervision: M.N.-E.-A.; Visualization: M.S.I., X.W.; Writing—draft: M.S.I., M.M.H., H.D.G.; Writing—review and editing: M.S.I., M.M.H., H.D.G., M.N.-E.-A.

Germack is supported by CTSA Grant Number TL1 TR001864 from the National Center for Advancing Translational Science (NCATS), a component of the National Institutes of Health (NIH). The content is solely the responsibility of the authors and does not necessarily represent the official views of this organization.

Conflicts of Interest

The authors declare no conflict of interest.

• Artificial intelligence in healthcare

DrAfter123/DigitalVision Vectors

10 high-value use cases for predictive analytics in healthcare

Predictive analytics can support population health management, financial success, and better outcomes across the value-based care continuum..

• Editorial Staff

As healthcare organizations pursue improved care delivery and increased operational efficiency, digital transformation remains a key strategy to help achieve these goals. Many health systems’ digital transformation journey involves identifying the value of their data and capitalizing on that value through big data analytics.

Of the four types of healthcare data analytics , predictive analytics currently has some of the highest potential for value generation. This type of analytics goes beyond showing stakeholders what happened and why, allowing users to gain insight into what’s likely to happen based on historical data trends.

Being able to forecast potential future patterns has game-changing potential as healthcare organizations aim to move from reactive to proactive, but those looking to leverage predictive analytics must first define relevant use cases.

In this primer, HealthITAnalytics will outline 10 predictive analytics use cases, in alphabetical order, that health systems can pursue as part of a successful predictive analytics strategy .

1. CARE COORDINATION

Improved care coordination can bolster patient outcomes and satisfaction, and predictive analytics is one way healthcare organizations can enhance these efforts. Predictive analytics is beneficial in hospital settings, where care coordination staff are trying to prevent outcomes like patient deterioration and readmission while optimizing patient flow.

Some healthcare organizations are already beginning to see success after deploying advanced analytics to reduce hospital readmissions .

In June, a research team from New York University (NYU) Grossman School of Medicine successfully built a large language model (LLM) known as NYUTron to predict multiple outcomes, including readmissions and length of stay.

The tool, detailed in a Nature study , can accurately forecast 30-day all-cause readmission, in-hospital mortality, comorbidity index, length of stay, and insurance denials using unaltered electronic health record (EHR) data. At the time of the study’s publication, NYUTron could predict 80 percent of all-cause readmissions, a five percent improvement over existing models.

According to a December 2023 NEJM Catalyst study , predictive models deployed at Corewell Health have seen similar success, keeping 200 patients from being readmitted and resulting in a $5 million cost savings.

In a 2022 interview with HealthITAnalytics , leadership from Children’s of Alabama discussed how real-time risk prediction allows the health system to tackle patient deterioration and pursue intensive care unit (ICU) liberation.

Alongside its applications for inpatient care management, predictive analytics is particularly useful for other preventive care uses, such as disease detection.

2. EARLY DISEASE DETECTION

Effective disease management is vital to improving patient outcomes, but capturing and analyzing the necessary data only became plausible with the advent of predictive analytics.

Using predictive analytics for disease management requires healthcare organizations to pool extensive patient data — including EHRs, genomics, social determinants of health (SDOH), and other information — to identify relevant trends. These insights can then be used as a starting point to guide early disease detection and diagnosis efforts, anticipate disease progression, flag high-risk patients, and optimize treatment plans and resource allocation.

The promise of big data and predictive analytics is valuable in infectious disease monitoring.

In a February 2024 PLOS One study , researchers from the University of Virginia detailed the development of an online big data dashboard to track enteric infectious disease burden in low- and middle-income countries.

The dashboard is part of the Planetary Child Health & Enterics Observatory (Plan-EO) initiative, which aims to provide an evidence base to help geographically target child health interventions.

The dashboard will pull data from various sources to map transmission hotspots and predict outbreaks of diarrheal diseases, which public health stakeholders can use to better understand disease burden and guide decision-making.

The impacts of infectious disease are often inequitable, which may lead some to question the role that predictive analytics plays in concerns about health equity. Like any advanced data analytics approach, these tools must be used responsibly to avoid perpetuating health disparities, but when used responsibly, predictive tools can positively impact equity efforts.

3. HEALTH EQUITY

Care disparities, bias and health inequity are rampant in the United States healthcare system. Researchers and clinicians are on the front lines of efforts to ensure that patients receive equitable care, but doing so requires healthcare stakeholders to gain a deep, nuanced understanding of how factors like SDOH impact patients .

Predictive analytics can help draw a wealth of information from the large, complex data needed to guide these efforts.

The health of those in marginalized communities is disproportionately impacted by housing, care access, social isolation and loneliness , food insecurity, and other issues. Effectively capturing data on these phenomena and designing interventions to address them is challenging, but predictive analytics has already bolstered these efforts.

Recently, researchers from Cleveland Clinic and MetroHealth were awarded over $3 million from the National Institutes of Health (NIH) to develop a digital twin-based, neighborhood-focused model to reduce disparities.

The Digital Twin Neighborhoods project uses de-identified EHR data to design digital replicas of real communities served by both organizations. Experts on the project indicated that by pulling geographic, biological, and SDOH information, researchers can better understand place-based health disparities.

Models developed using these data can simulate life course outcomes in a community. Tools that accurately predict the outcomes observed within a population’s EHRs can inform health equity interventions.

In 2021, United Healthcare launched a predictive analytics-based advocacy program to help address SDOH and improve care for its members. The system uses machine learning to identify individuals who may need social services support.

These insights are incorporated into an agent dashboard that member advocates can use, alongside more traditional tools like questionnaires, to gather more information from the patient about their situation. If necessary, the advocate connects the individual with support mechanisms.

Efforts like these also demonstrate the utility of predictive analytics tools in patient and member engagement.

4. PATIENT ENGAGEMENT

Patient engagement plays a vital role in enhancing healthcare delivery. The advent of big data analytics in healthcare provides many opportunities for stakeholders to actively involve patients in their care.

Predictive analytics has shown promise in allowing health systems to proactively address barriers to patient engagement, such as appointment no-shows and medication adherence.

In a 2021 interview with HealthITAnalytics , Community Health Network leadership detailed how the health system bolsters its engagement efforts by using predictive analytics to reduce appointment no-shows and conduct post-discharge outreach.

A key aspect of this strategy is meeting patients where they are to effectively individualize their care journeys and improve their outcomes.

Appointment no-shows present a significant hurdle to achieving these aims, leading Community Health Network to implement automated, text message-based appointment reminders, with plans to deploy a two-way communication system to streamline the appointment scheduling process further.

The health system took a similar approach to post-discharge outreach, successfully deploying an automated solution during the COVID-19 pandemic.

To further enhance these systems, Community Health Network turned to predictive analytics.  By integrating a predictive algorithm into existing workflows, the health system could personalize outreach for appointment no-shows. Patients at low risk for no-shows may receive only one text message, but those at higher risk receive additional support, including outreach to determine whether unmet needs that the health system can help address are preventing them from making it to appointments.

Data analytics can also support medication adherence strategies by identifying non-adherence or predicting poor adherence.

One 2020 study published in Psychiatry Research showed that machine learning models can “accurately predict rates of medication adherence of [greater than or equal to 80 percent] across a clinical trial, adherence over the subsequent week, and adherence the subsequent day” among a large cohort of participants with a variety of conditions.

Research published in the March 2020 issue of BMJ Open Diabetes Research & Care found that a machine learning model tasked with identifying type 2 diabetes patients at high risk of medication nonadherence was accurate and sensitive, achieving good performance.

Outside the clinical sphere, predictive analytics is also useful for helping organizations like payers meet their strategic goals.

5. PAYER FORECASTING

Payers are an integral part of the US healthcare system. As payer organizations work with providers to guide members' care journeys, they generate a wealth of data that provides insights into healthcare utilization, costs, and outcomes.

Predictive analytics can help transform these data and inform efforts to improve payer forecasting . With historical data, payers can use predictive modeling to identify care management trends, forecast membership shifts, project enrollment churn, and pinpoint changes in service demand, among other uses.

In June 2023, leaders from Elevance Health discussed how the payer’s emphasis on predictive analytics is key to improving member outcomes.

Elevance utilizes a predictive algorithm to personalize member experience by addressing diabetes management and fall risk. The predictive model pulls clinical indicators like demographics, comorbidities, and A1C levels to forecast future A1C patterns and identify individuals with uncontrolled or poorly controlled diabetes.

From there, the payer can help members manage their condition through at-home lab A1C test kits and increased member and care team engagement.

The second predictive tool incorporates data points — including past diagnoses, procedures, and medications, the presence of musculoskeletal-related conditions and connective tissue disorders, analgesic or opioid drug usage, and frailty indicators — to flag women over the age of 65 at higher risk of fracture from a fall.

Elevance then conducts outreach to these individuals to recommend bone density scans and other interventions to improve outcomes.

These efforts are one example of how predictive analytics can improve the health of specific populations, but these tools can also be applied to population health more broadly.

6. POPULATION HEALTH

While much of healthcare is concerned with improving individual patients’ well-being, advancing the health of populations is extremely valuable for boosting health outcomes on a large scale. To that end, many healthcare organizations are pursuing data-driven population health management .

Predictive analytics tools can enhance these initiatives by guiding large-scale efforts in chronic disease management and population-wide care coordination.

In one 2021 American Journal of Preventive Medicine study , a research team from New York University’s School of Global Public Health and Tandon School showed that machine learning-driven models incorporating SDOH data can accurately predict cardiovascular disease burden. Further, insights from these tools can guide treatment recommendations.

The early identification of chronic disease risk is also helpful in informing preventive care interventions and flagging gaps in care .

Being closely related to population health , public health can also benefit from applying predictive analytics.

Researchers from the Center for Neighborhood Knowledge at UCLA Luskin, writing in the International Journal of Environmental Health in 2021, detailed how a predictive model successfully helped them identify which neighborhoods in Los Angeles County were at the greatest risk for COVID-19 infections.

The tool mapped the county on a neighborhood-by-neighborhood basis to evaluate residents’ vulnerability to infection using four indicators: barriers to accessing health care, socioeconomic challenges, built-environment characteristics, and preexisting medical conditions.

The model allowed stakeholders to harness existing local data to guide public health decision-making, prioritize vulnerable populations for vaccination, and prevent new COVID-19 infections.

Alongside large-scale initiatives like these, predictive modeling can also support the advancement of precision medicine.

7. PRECISION MEDICINE

The emergence of genomics and big data analytics has opened new doors in the realm of tailored health interventions. Precision and personalized medicine rely on individual patients’ data points to guide their care and improve their well-being.

From cancer to genetic conditions, predictive analytics is a crucial aspect of precision medicine.

In 2021, a meta-analysis presented at the American Society for Radiation Oncology (ASTRO) Annual Meeting showed that a genetic biomarker test could accurately predict treatment response in men with high-risk prostate cancer.

The test analyzes gene activity in prostate tumors to generate a score to represent the aggressiveness of a patient’s cancer. These insights can be used to personalize treatment plans that balance survival risk with quality of life.

Researchers from Arizona State University (ASU) revealed in a 2024 Cell Systems paper that they developed a machine learning model to predict how a patient’s immune system will respond to foreign pathogens.

The tool uses information on individualized molecular interactions to characterize how major histocompatibility complex-1 (MHC-1) proteins — key players in the body’s ability to recognize foreign cells — impact immune response.

MHC-1s exist on the cell surface and bind foreign peptides to present to the immune system for recognition and attack. These proteins also come in thousands of varieties across the human genome, making it difficult to forecast how various MHC-1s interact with a given pathogen.

The ASU research addressed this by analyzing just under 6,000 MHC-1 alleles, shedding light on how these molecules interact with peptides and revealing that individuals with a diverse range of MHC-1s were more likely to survive cancer treatment.

Using the model, providers could potentially forecast pathological outcomes for patients, bolstering treatment planning and clinical decision-making.

In addition to these successes at the microscopic level, predictive analytics is also useful on the macro level in healthcare.

8. RESOURCE ALLOCATION AND SUPPLY CHAIN

Optimization of the supply chain and resource allocation ensures that providers and patients receive the equipment, medications, and other tools that they need to support positive outcomes. Data analytics plays a massive role in this, as supply chain management and resource use rely heavily on accurately recording and tracking resources as they move from the assembly line into the clinical setting.

Predictive analytics takes this one step further by helping stakeholders anticipate and address supply chain issues before they arise while optimizing resource use.

Seattle Children's Hospital is using predictive modeling in the form of digital twins to help the health system streamline hospital operations , particularly resource allocation.

By using digital twin simulation to “clone” the hospital, stakeholders can model how certain events, strategies, or policies might impact operational efficiency. This capability was critical in the wake of COVID-19, as it allowed the health system to identify how rapidly its personal protective equipment (PPE) supplies would diminish, forecast bed capacity, and generate insights around labor resources.

Predictive analytics can also be used by distinct parts of the supply chain to help prevent shortages.

The 2022 infant formula shortage is one example of how supply chain disruptions can significantly impact health.

One potential way for parents to deal with the formula shortage was to turn to human breast milk banks, which distribute donated milk to vulnerable babies and their families. However, accomplishing this vital work requires milk banks to effectively screen donors, accept donations, process and test them to ensure they’re safe, and dispense them.

In an interview with HealthITAnalytics , stakeholders from Mothers' Milk Bank at WakeMed Health & Hospitals described how data analytics can help optimize aspects of this process.

A crucial part of ensuring that milk is available to those who need it is tracking milk waste. Milk can be wasted for various reasons, but the presence of bacteria is one of the primary causes. To address this, the milk bank began analyzing donor records to determine what factors may make a batch of milk more likely to test positive for bacillus .

The milk bank can then use the insights generated from the analysis to predict which donors may be at high risk for having bacillus in their milk, allowing milk from these individuals to be tested separately. This removes any bacillus -positive samples before the milk is pooled for processing.

Predictive analytics is also helpful in assessing and managing risks in clinical settings.

9. RISK STRATIFICATION

Patient risk scores have the potential to improve care management initiatives, as they allow providers to formulate improved prevention strategies to eliminate or reduce adverse outcomes. Risk scores are used to help understand what characteristics may make a patient more susceptible to various conditions.

From there, the scores can inform risk stratification efforts, which enables health systems to categorize patients based on whether they are low-, medium- or high-risk. These data can show how one or more factors increase a patient's risk.

Risk stratification is one of the most valuable use cases for predictive analytics because of its ability to prevent adverse outcomes.

In February 2024, leaders from Parkland Health & Hospital System (PHHS) and Parkland Center for Clinical Innovation (PCCI) in Dallas, Texas, detailed one of these high-value use cases.

Parkland’s Universal Suicide Screening Program is an initiative designed to flag patients at risk of suicide who may have flown under the health system’s radar through proactive screening of all Parkland patients aged 10 or older, regardless of the reason for the clinical encounter.

During the encounter, nursing staff ask the patient a set of standardized, validated questions to assess their suicide risk. This information is then incorporated into the EHR for risk stratification.

These data are useful for stakeholders looking to better understand patients’ stories, including factors like healthcare utilization before suicide. Coupling these insights with state mortality could help predict and prevent suicide in the future.

Risk stratification is also crucial for improving outcomes for some of the youngest, most vulnerable patients: newborns.

Parkland also runs an initiative that uses SDOH data to identify at-risk pregnant patients and enable early interventions to help reduce preterm births .

The program’s risk prediction model and text message-based patient education program have been invaluable in understanding the nuances of preterm birth risk for Parkland patients. Major risk factors like cervical length and history of spontaneous preterm delivery may not be easy to determine for some patients. Further, many preterm births appear to be associated with additional risk factors outside of these – like prenatal visit attendance.

Using these additional factors to forecast risk, Parkland has developed clinical- and population-level interventions that have resulted in a 20 percent reduction in preterm births.

These use cases, among other things, demonstrate the key role predictive analytics can play in advancing value-based care.

10. VALUE-BASED CARE SUCCESS

Value-based care incentivizes healthcare providers to improve care quality and delivery by linking reimbursement to patient outcomes. To achieve value-based care success, providers rely on a host of tools: health information exchange (HIE), data analytics, artificial intelligence (AI) and machine learning (ML), population health management solutions, and price transparency technologies.

Predictive analytics can be utilized alongside these tools to drive long-term success for healthcare organizations pursuing value-based care.

Accountable care organizations (ACOs) are significant players in the value-based care space, and predictive modeling has already helped some achieve their goals in this area.

Buena Vida y Salud ACO partnered with the Health Data Analytics Institute (HDAI) in 2023 to explore how predictive analytics could help the organization keep patients healthy at home.

At the outset of the collaboration, the ACO’s leadership team was presented with multiple potential use cases in which data analysis could help with unplanned admissions, worsening heart failure, pneumonia development, and more.

However, providers were overwhelmed when given risk-stratified patient lists for multiple use cases. Upon working with its providers, the ACO found that allowing clinicians to choose the use cases or patient cohorts they wanted to focus on was much more successful.

The approach has helped the ACO engage its providers and enhance care management efforts through predictive modeling and digital twins. These tools provide fine-grain insights into the drivers of outcomes like pneumonia-related hospitalization, which guide the development of care management interventions.

These 10 use cases are just the beginning of predictive analytics' potential to transform healthcare. As data analytics technologies like AI, ML and digital twins continue to advance, the value of predictive analytics is likely to increase exponentially.

What Are the Benefits of Predictive Analytics in Healthcare?

• How Can Predictive Analytics Help ACOs Boost Value-Based Care Delivery?
• Putting the Pieces Together for a Successful Predictive Analytics Strategy

Dig Deeper on Artificial intelligence in healthcare

Exploring the role of AI in healthcare risk stratification

Top Data Analytics Tools for Population Health Management

Food Subsidies Key to Food Security Programs, Access to Nutritious Food

UW Health nurses are piloting a generative AI tool that drafts responses to patient messages to improve clinical efficiency ...

Industry stakeholders are working with DirectTrust to create a data standard for secure cloud fax to address health data exchange...

A new HHS final rule outlines information blocking disincentives for healthcare providers across several CMS programs, including ...