Electronic medical record (EMR) and electronic health record (EHR) systems are being increasingly used for storing, retrieving, and managing patient health records that can be used for data-driven initiatives, such as clinical research, quality improvement (QI), and decision support tools [1]. Although both terms are commonly used interchangeably, EMR specifically refers to patient health data accumulated by patient encounters and limited to care delivery settings, enabled by the interoperability of the system by multiple providers [2,3]. These data typically consist of demographics and health notes, including history, past and current problems, immunizations and treatments, and laboratory and radiology reports. In contrast, EHR refers to larger, interconnected systems that bring together EMRs from various care sites and settings to enable robust information exchange [2,3]. EHRs are also designed for access by varied audiences, capturing far more than just clinical information and are used to support continuity and delivery of integrated and efficient health care services [2].

EMR and EHR systems are important sources of research data that can be used to improve patient care journeys, transitions between levels of health care, and delivery of services. With developments in computer systems and the use of digital health tools, EHR systems have evolved significantly since their first use in the 1960s [4], with the global market reaching US $33 billion by 2025 [5]. Epic Systems [6] (Epic Systems Corporation) was one of the early innovators in this space and is currently among the 5 leading EHR providers and the most adopted EHR worldwide [7,8]. In early 2023, Epic was independently rated as the best overall EHR suite for the 13th consecutive year and was in use in 89% of acute care hospitals in the United States [9]. Many academic and clinical organizations, such as Johns Hopkins University, have adopted Epic across their care delivery systems [10,11]. Recently, the province of Alberta, Canada, transitioned its varied EHRs to a unified Epic-based clinical information system called Connect Care, sparking great interest in the research community for the potential use of real-time clinical and health data in Canada [12].

Despite the global reach of the Epic EHR, there remains a gap in literature detailing how the system has been used to facilitate clinical and health research. Literature syntheses are available describing issues related to EHR implementation [13], the transition to the Epic EHR [8], and Epic integration for personalized care [14], whereas a plethora of Epic implementation studies are also available [15,16]. Two reviews on the use of EHRs and EMRs for research included a narrative review from 2009, capturing literature from 2000 to 2007 on health outcomes [17], and a more recent review from 2015 examined population health methods and applications [18]. Both reviews, however, are dated, do not focus on a single EHR, do not capture the breadth of the literature on areas of healthcare research, and do not follow methods for systematic literature syntheses.

To the best of our knowledge, a comprehensive literature review on the use of the Epic EHR for health care research is not available. We address this critical gap in the literature. Hence, the objective of this scoping review is to synthesize literature on use cases of the Epic EHR for research in various areas of clinical and health sciences.

We selected the scoping review method because it provides the basis for synthesizing and capturing the breadth of the current peer-reviewed and gray literature [19,20]. Scoping reviews enable researchers to comprehensively assess the breadth and depth of a topic and identify the strengths, weaknesses, and gaps in the existing research base. As the focus of our review was to explore the existing research applications of Epic, we determined that a scoping review methodology would be the best approach for achieving this objective [21]. We followed the Joanna Briggs Institute guidance [22] and foundational guidelines from Arksey and O’Malley’s [21] 2005 framework to conduct scoping reviews. Our approach was further supported by the unpacking details suggested by Levac et al [23] and Peters et al [24]. As Levac et al [23] suggested involving stakeholders, the review was conceptualized in consultation with a multidisciplinary team that included experts from health informatics, clinicians, data scientists, and health system managers. A brief review protocol was developed, but not published.

On the basis of the JBI guidelines, we used the population, concepts, and context framework to formulate the research question. Our population included all individuals with any condition of interest. The concepts included all possible research methods that involve the collection, integration, extraction, and analysis of data or tools built to support research using the Epic EHR. The context could be the use of EHRs in any setting. The review steps comprised determining the research question, identifying relevant studies, title, abstract and full-text screening, and data charting. Synthesis and reporting of the results used the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) checklist [25].

To develop the eligibility criteria, we conducted an initial search of the literature on OVID MEDLINE and reviewed the first 50 articles that presented the use of EHRs for research. As a first step, reviewers (JC, NS, NW, LR, and DR) independently screened these 50 pilot articles, followed by a meeting to discuss the conflicts and eligibility criteria. Each pilot article was independently reviewed by 2 reviewers at both stages of title, abstract, and full-text screening.

Working with a multidisciplinary team for this initial phase of the search, we defined the parameters of the review and identified the variety of ways Epic has been used for research. First, we learned that there were 2 clear categories of literature. Aligned with our objective, we focused on the use of Epic EHR for research, delineating our inquiry from research on the Epic EHR system. The latter body of literature includes the implementation, development of modules, enhancements in different versions, and architectural, legal, and security issues related to the Epic EHR. The consensus of our multidisciplinary team was to exclude literature related to research on the Epic EHR system as it diverged from our main objective.

Furthermore, our initial searches revealed major research domains within health care that included clinical and health services research (HSR), population health, clinical decision support (CDS) tools, registries, biobanks, and QI initiatives. The team reached a consensus to use these 6 domains to categorize the research literature while allowing others to emerge during the synthesis phase. The purpose of grouping the literature in this manner kept the team’s focus on the methods and use cases of Epic for conducting research. Each domain was determined according to the stated purpose and method of each study. If the research fits in more than one domain, a primary domain is identified. The eligibility criteria are presented in Textbox 1. We also focused on capturing facilitators and barriers to using the Epic EHR for clinical or health research, which could include specific details about the software, ease or difficulty in extracting data, building new tools or fields, and achieving research objectives in general.

The search strategy for peer-reviewed literature was formulated and implemented by an information specialist (DL) on the team. We included 3 major health databases (MEDLINE, Embase, and CINAHL) and 2 interdisciplinary databases (Scopus and Web of Science). We did not place a time limit on the search, searching for all sources until April 25, 2023. To identify previously published reviews, we searched the Cochrane Database of Systematic Reviews and the first 250 results of Google Scholar and Google search engines. This allowed us to cast a wide net in our search for relevant literature. The databases, platforms, and their results are presented in the PRISMA diagram (Figure 1). A detailed search strategy is presented in Multimedia Appendix 1.

We developed a list of search keywords based on seminal sources [2,3,14,26-29] and the National Library of Medicine’s Medical Subject Headings trees to search databases [30,31]. This resulted in a single concept search for “Epic” combined with terms that included “electronic health record*” or “electronic medical record*” or “medical record*” or “health record*” or “EHR“ or “EMR” or “system*.” Keywords and word variants were searched as title and abstract words and subject headings (eg, MEDLINE Medical Subject Headings) as appropriate.

For gray literature, we searched conference proceedings for short or full papers. We excluded conference abstracts from the Embase search because it was not possible to glean the methods and results presented in abstracts alone. In addition, we explored other gray literature sources, including Google Scholar, Google search engines, and the Epic Research site [32]. We did not include articles from the Epic Research site as they varied in detail.

The Covidence (Level 10) web platform was used to import citations and facilitate the screening process [33]. Duplicates were removed using Covidence. Endnote (Clarivate) [34] was used to manage the references.

After thorough development of the inclusion and exclusion criteria, we commenced the title and abstract screening processes. Each abstract was reviewed in duplicate by 2 independent reviewers (NS and NW), and conflicts were resolved by a third reviewer (JC) in consultation with others. The full-text articles were screened in the same manner.

The data abstraction platform was developed in consultation with our multidisciplinary team to capture the breadth of findings from the identified sources. The reviewers piloted the form independently and abstracted 10 full-text articles each. There was excellent agreement between the reviewers on the abstraction form, with some clarification required in defining the research domains and study design. The major categories for abstracting the results included (1) study characteristics (year of publication, country, journal, or conference name); (2) research problem and setting (problem studied, study design, population, and settings); (3) research domains; (4) analytic methods; and (5) specific facilitators and barriers to using Epic for research. We followed the PRISMA-ScR checklist to report the review [25]. The checklist is presented in Multimedia Appendix 2.

During our preliminary literature search, we found varying uses of terms, such as the interchangeable use of EMR and EHR, and related concepts we aimed to cover in the review. Here, we detail the operationalized concepts for clarity.

Epic has various software applications, modules, and tools that can be leveraged for research. We curated information from various Epic websites to gather information on the modules mentioned in the literature and provide a high-level overview of their functions. Not every module we encountered had information available on the Epic website; therefore, additional sources were required to discern the use of some modules and applications.

Epic has a variety of specialty care applications, including ASAP for emergency and urgent care, Beacons for oncology care, Radiant for radiology reporting, tracking, and scheduling, and OpTime for surgery and anesthesia [42]. EpicCare is Epic’s ambulatory and inpatient care EHR.

Epic applications that appear to be particularly suited to research include Cosmos, SlicerDicer, Clarity, Chronicles, Reporting Workbench, and Signal. Cosmos is an aggregate data set of over 217 million patients in the United States that has been used in numerous research applications [43]. SlicerDicer is a data extraction tool within the Epic EHR that facilitates the search and visualization of data in large patient populations [44]. Reporting Workbench, a business intelligence tool, can be used to generate research reports [45]. These reports use Chronicles as their data source, which is a real-time database that runs most Epic software. Data from Chronicles are transferred to Clarity, a Microsoft SQL Server database that can be used for complex reporting [46,47]. The analytics tool Signal allows researchers to observe how providers spend time in the EHR, which can help improve workflow and productivity [48].

Some additional Epic modules and tools include MyChart, Healthy Planet, Care Everywhere, and SmartTools. MyChart is Epic’s web-based patient portal that enables patients to access their health information [49]. Healthy Planet, a population health module, allows providers to monitor patients’ health and care over time, as well as reduce barriers to health by addressing the social determinants of health [50]. Care Everywhere, an interoperability platform, assists with the coordination of care across different providers and health care organizations [51].

Ethics committee approval was not required for this scoping review.

A total of 4669 articles were identified through a combined database and gray literature search. A total of 3469 duplicates were removed by Covidence matching of the titles, abstracts, and article identifiers. Table 1 presents the sources, platforms, and search results.

We screened the titles and abstracts of 1230 articles, of which 626 (50.8%) were excluded. Of the 604 articles included in the full-text review, 397 (65.6%) were excluded according to our criteria. Ultimately, 206 articles were selected for this study. The PRISMA flow diagram is shown in Figure 1.

Of the 206 articles, 174 (83.5%) were peer-reviewed journal articles and 32 (15.5%) were gray literature articles. All gray literature articles were conference proceedings with clear objectives, methods, results, and discussion sections. We could not identify or report on other forms of gray literature because of the lack of detail necessary to qualify as research. We present the results under 5 main headings.

The first articles involving Epic being used for research surfaced in 2009. These trends are shown in Figure 2. Since then, the number of peer-reviewed articles has exponentially increased from 2015 to 2023. The research studies were conducted in the United States (199/206, 96.6%; Figure 3), the Netherlands (3/206, 1.5%), and Saudi Arabia, the United Arab Emirates, Canada, and France (1/206, 0.5% each)

Study settings where the Epic EHR system was used for research included single (133/206, 64.6%) and multiple sites (73/206, 35.4%). As Epic EHR can link across multiple sites, it was difficult to distinguish the number of people and units involved. Single sites included a single department (such as the department of urology or medicine), whereas multiple sites consisted of different practices, departments, and hospitals. Many studies have been conducted in university health systems (44/206, 21.5%), which comprise hospitals of various sizes. Research was also mentioned specifically using Epic EHR data in primary care settings (16/206, 0.1%) and emergency departments and outpatient departments (11/206, 0.1%, both cases).

The domains of research using the Epic EHR in the literature included clinical research (51/206, 24.8%), HSR (51/206, 24.8%), CDS (45/206, 21.8%), QI (27/206, 13.1%), population health (22/206, 10.7%), registry (8/206, 3.9%), biobanks (1/206, 0.5%), and economic analysis (1/206, 0.5%) emerging as distinct new domains. The research domains, according to the population studied, are presented in Table 2.

Clinical research includes examinations of the prevalence and frequency of clinical conditions and associated factors, as well as the efficacy and safety of clinical interventions. Examples of HSR include investigating the impact of EHR implementation, identifying factors predictive of the use of health services, analyzing EHR use and time spent by clinicians, and measuring health care costs. CDS methods have been used to improve clinical practice, implement screening programs, and increase referrals. QI domain studies assessed the effect of interventions on the quality of care, how to improve access to clinical information, and evaluated quality metrics. Studies in the population health domain have determined factors affecting public health programs and investigated the associations between socioeconomic factors and health. Registries were created and evaluated for use in future studies. The study in the biobank domain discussed the implementation of a biobank using residual clinical samples, whereas the economic analysis focused on behavioral economics to promote appropriate management of a chronic condition.

Longitudinal study designs, including retrospective cohort (85/206, 41.3%) and prospective observational (30/206, 14.6%), were the most common study designs across both peer-reviewed and gray literature. Other designs included pre-post intervention studies (32/206, 15.5%), cross-sectional studies (20/206, 9.7%), and randomized controlled trials (12/206, 5.8%). A variety of other study designs were encountered in the research using the Epic EHR, as presented in Table 3.

The populations studied using the Epic EHR were primarily adults (104/206, 50.5%) or combined adult and pediatric populations (53/206, 25.7%). There were 14 studies (6.8%) researching physician users, which examined their use of the Epic EHR and its various functions and tools, including CDS tools. The specialties of physicians studied varied and included primary care (3/14, 21%), internal medicine (4/14, 29%), and multispecialty physician research (3/14, 21%). The remaining studies focused on pediatric patients (23/206, 11.2%), adolescents (3/206, 1.5%), adolescents and pediatrics (1/206, 0.5%), and other health care staff and patients (4/206, 1.9%; Table 3).

The analytic methods included descriptive statistics (70/206, 34%), multivariable regression (61/206, 29.6%), development and testing of an algorithm (6/206, 2.9%), and machine learning methods (2/206, 1%). One qualitative study used ethnographic methods and another used economic analysis. Further details of the analytical methods used are presented in Multimedia Appendix 3.

Most studies did not use or did not mention using a specific module or application within the Epic EHR for research purposes (161/206, 78.2%). Among the studies that indicated the Epic modules used (45/206, 21.8%), tools included MyChart (10/45, 22%), Clarity (10/45, 22%), SlicerDicer (6/45, 13%), and SmartTools (6/45, 13%). Other modules used were EpicCare for ambulatory and inpatient care (3/45, 7%), ASAP for the emergency room (2/45, 4%), Optime for surgery and anesthesia (2/45, 4%), Beacon for medical oncology (2/45, 4%), and Radiant for radiology (1/45, 2%). Additional Epic modules and applications reported included Healthy Planet (1/45, 2%), Care Everywhere (3/45, 7%), Cosmos (1/45, 2%), Chronicles (1/45, 2%), Reporting Workbench (1/45, 2%), and Signal (2/45, 4%). Two studies reported the use of multiple Epic modules (2/45, 4%).

Related to its use in research, we examined the literature for facilitators and barriers that could be helpful for researchers considering the use of Epic EHR data or module features. Facilitators for using the Epic EHR included developing guideline-based clinical order sets, pooling survey resources into dashboards, using the SlicerDicer visualization tool used for presenting data, identifying cohorts, physician prompts, audit logs for user interaction, and use of the EHR for patient education, and access to data by various audiences, including patients.

Studies listing barriers to the use of the Epic EHR for research discussed inaccurate coding and missing data, missing specific terminology for the detection of particular conditions, underreported social determinants of health and patient-reported outcomes, lack of documentation of mental health issues, dependence on diagnostic codes within Epic to flag conditions and reports, and variation in the use of the EHR by staff, physicians, and across teams.

As scoping reviews capture the breadth of literature [21], this review presents the current state of science on the use of the Epic EHR for research in medicine and health sciences. Currently, reviews related to Epic implementation [13], transition [8], and integration to the Epic EHR for personalized care [14] are available. In addition, our search for systematic and narrative reviews on the same topic related to other major EHRs in use, such as Cerner, could not identify any peer-reviewed published work, at least in the last 2 decades. Hence, this scoping review attempts to fill a critical gap in the literature, showcasing trends, opportunities, and issues in the use of Epic EHR for healthcare research. We discuss and contrast our findings with considerations and recommendations that would be helpful for researchers and practitioners intending to use EHRs as data sources or research tools.

We found that Epic EHR has been related to a plethora of research and analytic tools that can be used in a wide variety of ways, including for different study designs and clinical populations, with an increase in the use of studies from 2015 onwards. This increase could be due to EHR enhancements, better understanding of EHR data among researchers, and improved data access. We also found that most studies were from the United States, which could be due to the high uptake of Epic in the country, and capital and recurring costs related to the implementation and maintenance of EHRs.

Epic’s architecture allows the integration of different data sources, including large medical and health databases, population-based registries, and patient-reported data [53,250,256]. This is an important capability that allows linkage and access to administrative and clinical data [53,186,257-259]. Linkage enhances the capacity to discover patient care patterns within and between hospitals and community care providers [249]. In addition, deidentified data can be exported for analysis [185]. Epic can also be used to scale interventions at multiple sites, thereby expanding research capabilities across geographic boundaries [155]. Furthermore, studies point to the advantage afforded by EHRs as a cheaper clinical data source that can be used for research [18].

Among the Epic EHR tools, MyChart, SlicerDicer, and Clarity were the most frequently used. For the research domains of their use, half of MyChart belonged to HSR, whereas others included clinical research, CDS, and QI. Of the studies that reported the use of SmartTools, half belonged to CDS, with the remainder in the QI and clinical research domains. Clarity and SlicerDicer have been used in various research domains. The most reported modules were EpicCare for ambulatory and inpatient care, ASAP for the emergency room, Optime for surgery and anesthesia, and Beacon for medical oncology.

An earlier narrative review of EHRs identified multivariate modeling as the most common method for targeting confounding and selection bias [17]. Epic’s built-in applications and features, CDS tools, and predictive analytic algorithms can guide clinical management and provide a wide variety of opportunities for data analysis and reporting [114,155,159,260,261]. In this review, the most used analytical methods were descriptive statistics, closely followed by multivariable regression modeling, developing algorithms, and other machine learning methods. The least reported methods include ethnographic and economic analysis.

From 2015 onwards, we observed an increasing trend in the use of Epic EHR in population-based studies. An earlier narrative review focusing on population health applications using EHRs also identified diverse research, including cross-sectional and longitudinal studies [18]. A greater number of population-based studies could be due to a better understanding of EHR data in general among epidemiological researchers. For example, one seminal study reported observational data related to the Flint, Michigan, drinking water crisis, which resulted in both local and national population health policy changes [219]. Such studies, although rare, are excellent examples of not only generating and using research data but also translating evidence toward policy change.

With the rapid growth in the use of EHRs, a rise in population health-related research from EHR data can be expected in the near future. Disease surveillance is also a growing area, with examples from preconception screening [85], immunization [92], hypertension, and obesity [262].

Epic allows for the development, implementation, and validation of CDS tools to improve quality of care [55,66,260,263]. In addition, in-system evaluation is possible within the EHR [59]. For clinical research, tools offer support for trials, cohort selection, recruitment, randomization, and data collection [87,114]. QI initiatives using the Epic EHR focused on best practice advisories and workflows [222,245,264,265], reducing prescription and medication dosing errors [266,267], and linking data from health apps to the Epic system [256], which shows the potential for integration of various mobile and eHealth apps. Another important capability of the Epic EHR is reporting patient-reported outcomes, which is enhanced by the ability to integrate outcome assessment tools within the personal health portal [268]. Additional functionality for improving education and communication tools for patients and health care professionals has also been noted [246,269].

From 2021 onwards, we identified 15 COVID-19–related studies [88,101,112,125,128,137,181,189,195,202,209,213,215,216,244]. These ranged from descriptive studies on disparities due to social determinants of health, the association of the virus with other conditions, severity in specific groups, provision of virtual care, treatment options, and changes in physician behavior.

Other varied but little-used research functions using Epic EHR included developing registries, biobanks, and an economic analysis. For example, Stanford University’s breast cancer registry was created by linking 2 independent institutions to identify interventions including surgery, chemotherapy, and diagnostics [250]. We found only one study using the EHR for economic analysis [149] and one for creating a biobank [52]. This points to the opportunity afforded by such a diverse system, as economic analyses are traditionally conducted using administrative data.

On the basis of the opportunities and challenges outlined above, Table 4 summarizes the key considerations and recommendations that can be useful for future research applications. We highlight key aspects and recommendations according to the 5 research domains.

Although our review focuses on the use of the Epic EHR as a widely adopted system, the major lessons, facilitators, and barriers can be appreciated, compared, and generalized to other EHR systems. Researchers and practitioners can be better aware of the known and potential opportunities and challenges related to the use of an EHR system for research. We contend that our study is not only timely, but also highly relevant for health care institutions, as 8 out of 10 instances of EHR implementation involve an Epic EHR offering. As we observed, this body of literature is expanding rapidly as Epic sites become better established with the provision of collaborative research support, education, and governance processes.

Future research can inform many directions for the use of EHR data, not only for Epic EHR, but also for other major EHRs. There is a critical need for updated information on the use of EHRs and data and newer methodologies, such as the use of machine learning methods, particularly natural language processing [281], and visual analytic methods for knowledge generation and decision support [282,283]. Although EHR clinical notes are gold mines for research, data quality, completeness, and accuracy remain major challenges for which emerging methodologies for validation and reducing variability are important areas of research.

As a scoping review, this paper presents an overview of a large and diverse body of literature pertaining to areas of research using Epic EHR. There are several limitations to consider, especially pertaining to the feasibility of any scoping review. Although we conducted a broad search of the literature using databases and gray literature sources, we could not include articles from languages other than English. To improve the completeness of the literature, we diligently searched for and abstracted full-text articles that were previously published as conference proceedings. Thus, we were able to include updated information with key lessons for researchers and practitioners. As scoping reviews mainly aims to capture the breadth of literature, assessing the quality of literature is optional. For this review, we did not assess the quality of the literature; however, given our wide search of the databases and inclusion of mostly peer-reviewed literature, we are reasonably sure that the literature included in the review is of high quality.

As strengths, we employed established scoping review methods, along with constituting a multidisciplinary team, including an information specialist, to guide the research question, search strategy, and operationalize the scoping review. We also spent considerable time and effort in conceptualizing and reaching a consensus on the research domains and presented our results in a manner that would illustrate the breadth of literature, while making it more accessible.

EHRs are a new source of health data with an ever-expanding repertoire of functions. This makes them a unique tool for primary and secondary research. Epic is a globally known, well-established, and growing EHR. This review attempts to fill a critical gap in the literature regarding the use of EHRs for research, outlining the current state of science in this field. In addition, we offer opportunities, key considerations, and recommendations for the future use of Epic and other EHRs.