research article case

Have a language expert improve your writing

Run a free plagiarism check in 10 minutes, generate accurate citations for free.

• Knowledge Base

Methodology

• What Is a Case Study? | Definition, Examples & Methods

What Is a Case Study? | Definition, Examples & Methods

Published on May 8, 2019 by Shona McCombes . Revised on November 20, 2023.

A case study is a detailed study of a specific subject, such as a person, group, place, event, organization, or phenomenon. Case studies are commonly used in social, educational, clinical, and business research.

A case study research design usually involves qualitative methods , but quantitative methods are sometimes also used. Case studies are good for describing , comparing, evaluating and understanding different aspects of a research problem .

Table of contents

When to do a case study, step 1: select a case, step 2: build a theoretical framework, step 3: collect your data, step 4: describe and analyze the case, other interesting articles.

A case study is an appropriate research design when you want to gain concrete, contextual, in-depth knowledge about a specific real-world subject. It allows you to explore the key characteristics, meanings, and implications of the case.

Case studies are often a good choice in a thesis or dissertation . They keep your project focused and manageable when you don’t have the time or resources to do large-scale research.

You might use just one complex case study where you explore a single subject in depth, or conduct multiple case studies to compare and illuminate different aspects of your research problem.

Receive feedback on language, structure, and formatting

Professional editors proofread and edit your paper by focusing on:

• Academic style
• Vague sentences
• Style consistency

See an example

Once you have developed your problem statement and research questions , you should be ready to choose the specific case that you want to focus on. A good case study should have the potential to:

• Provide new or unexpected insights into the subject
• Challenge or complicate existing assumptions and theories
• Propose practical courses of action to resolve a problem
• Open up new directions for future research

TipIf your research is more practical in nature and aims to simultaneously investigate an issue as you solve it, consider conducting action research instead.

Unlike quantitative or experimental research , a strong case study does not require a random or representative sample. In fact, case studies often deliberately focus on unusual, neglected, or outlying cases which may shed new light on the research problem.

Example of an outlying case studyIn the 1960s the town of Roseto, Pennsylvania was discovered to have extremely low rates of heart disease compared to the US average. It became an important case study for understanding previously neglected causes of heart disease.

However, you can also choose a more common or representative case to exemplify a particular category, experience or phenomenon.

Example of a representative case studyIn the 1920s, two sociologists used Muncie, Indiana as a case study of a typical American city that supposedly exemplified the changing culture of the US at the time.

While case studies focus more on concrete details than general theories, they should usually have some connection with theory in the field. This way the case study is not just an isolated description, but is integrated into existing knowledge about the topic. It might aim to:

• Exemplify a theory by showing how it explains the case under investigation
• Expand on a theory by uncovering new concepts and ideas that need to be incorporated
• Challenge a theory by exploring an outlier case that doesn’t fit with established assumptions

To ensure that your analysis of the case has a solid academic grounding, you should conduct a literature review of sources related to the topic and develop a theoretical framework . This means identifying key concepts and theories to guide your analysis and interpretation.

There are many different research methods you can use to collect data on your subject. Case studies tend to focus on qualitative data using methods such as interviews , observations , and analysis of primary and secondary sources (e.g., newspaper articles, photographs, official records). Sometimes a case study will also collect quantitative data.

Example of a mixed methods case studyFor a case study of a wind farm development in a rural area, you could collect quantitative data on employment rates and business revenue, collect qualitative data on local people’s perceptions and experiences, and analyze local and national media coverage of the development.

The aim is to gain as thorough an understanding as possible of the case and its context.

Here's why students love Scribbr's proofreading services

Discover proofreading & editing

In writing up the case study, you need to bring together all the relevant aspects to give as complete a picture as possible of the subject.

How you report your findings depends on the type of research you are doing. Some case studies are structured like a standard scientific paper or thesis , with separate sections or chapters for the methods , results and discussion .

Others are written in a more narrative style, aiming to explore the case from various angles and analyze its meanings and implications (for example, by using textual analysis or discourse analysis ).

In all cases, though, make sure to give contextual details about the case, connect it back to the literature and theory, and discuss how it fits into wider patterns or debates.

If you want to know more about statistics , methodology , or research bias , make sure to check out some of our other articles with explanations and examples.

• Normal distribution
• Degrees of freedom
• Null hypothesis
• Discourse analysis
• Control groups
• Mixed methods research
• Non-probability sampling
• Quantitative research
• Ecological validity

Research bias

• Rosenthal effect
• Implicit bias
• Cognitive bias
• Selection bias
• Negativity bias
• Status quo bias

Cite this Scribbr article

If you want to cite this source, you can copy and paste the citation or click the “Cite this Scribbr article” button to automatically add the citation to our free Citation Generator.

McCombes, S. (2023, November 20). What Is a Case Study? | Definition, Examples & Methods. Scribbr. Retrieved September 13, 2024, from https://www.scribbr.com/methodology/case-study/

Is this article helpful?

Shona McCombes

Other students also liked, primary vs. secondary sources | difference & examples, what is a theoretical framework | guide to organizing, what is action research | definition & examples, get unlimited documents corrected.

✔ Free APA citation check included ✔ Unlimited document corrections ✔ Specialized in correcting academic texts

Organizing Your Social Sciences Research Assignments

• Annotated Bibliography
• Analyzing a Scholarly Journal Article
• Group Presentations
• Dealing with Nervousness
• Using Visual Aids
• Grading Someone Else's Paper
• Types of Structured Group Activities
• Group Project Survival Skills
• Leading a Class Discussion
• Multiple Book Review Essay
• Reviewing Collected Works
• Writing a Case Analysis Paper
• Writing a Case Study
• About Informed Consent
• Writing Field Notes
• Writing a Policy Memo
• Writing a Reflective Paper
• Writing a Research Proposal
• Generative AI and Writing
• Acknowledgments

A case study research paper examines a person, place, event, condition, phenomenon, or other type of subject of analysis in order to extrapolate  key themes and results that help predict future trends, illuminate previously hidden issues that can be applied to practice, and/or provide a means for understanding an important research problem with greater clarity. A case study research paper usually examines a single subject of analysis, but case study papers can also be designed as a comparative investigation that shows relationships between two or more subjects. The methods used to study a case can rest within a quantitative, qualitative, or mixed-method investigative paradigm.

Case Studies. Writing@CSU. Colorado State University; Mills, Albert J. , Gabrielle Durepos, and Eiden Wiebe, editors. Encyclopedia of Case Study Research . Thousand Oaks, CA: SAGE Publications, 2010 ; “What is a Case Study?” In Swanborn, Peter G. Case Study Research: What, Why and How? London: SAGE, 2010.

How to Approach Writing a Case Study Research Paper

General information about how to choose a topic to investigate can be found under the " Choosing a Research Problem " tab in the Organizing Your Social Sciences Research Paper writing guide. Review this page because it may help you identify a subject of analysis that can be investigated using a case study design.

However, identifying a case to investigate involves more than choosing the research problem . A case study encompasses a problem contextualized around the application of in-depth analysis, interpretation, and discussion, often resulting in specific recommendations for action or for improving existing conditions. As Seawright and Gerring note, practical considerations such as time and access to information can influence case selection, but these issues should not be the sole factors used in describing the methodological justification for identifying a particular case to study. Given this, selecting a case includes considering the following:

• The case represents an unusual or atypical example of a research problem that requires more in-depth analysis? Cases often represent a topic that rests on the fringes of prior investigations because the case may provide new ways of understanding the research problem. For example, if the research problem is to identify strategies to improve policies that support girl's access to secondary education in predominantly Muslim nations, you could consider using Azerbaijan as a case study rather than selecting a more obvious nation in the Middle East. Doing so may reveal important new insights into recommending how governments in other predominantly Muslim nations can formulate policies that support improved access to education for girls.
• The case provides important insight or illuminate a previously hidden problem? In-depth analysis of a case can be based on the hypothesis that the case study will reveal trends or issues that have not been exposed in prior research or will reveal new and important implications for practice. For example, anecdotal evidence may suggest drug use among homeless veterans is related to their patterns of travel throughout the day. Assuming prior studies have not looked at individual travel choices as a way to study access to illicit drug use, a case study that observes a homeless veteran could reveal how issues of personal mobility choices facilitate regular access to illicit drugs. Note that it is important to conduct a thorough literature review to ensure that your assumption about the need to reveal new insights or previously hidden problems is valid and evidence-based.
• The case challenges and offers a counter-point to prevailing assumptions? Over time, research on any given topic can fall into a trap of developing assumptions based on outdated studies that are still applied to new or changing conditions or the idea that something should simply be accepted as "common sense," even though the issue has not been thoroughly tested in current practice. A case study analysis may offer an opportunity to gather evidence that challenges prevailing assumptions about a research problem and provide a new set of recommendations applied to practice that have not been tested previously. For example, perhaps there has been a long practice among scholars to apply a particular theory in explaining the relationship between two subjects of analysis. Your case could challenge this assumption by applying an innovative theoretical framework [perhaps borrowed from another discipline] to explore whether this approach offers new ways of understanding the research problem. Taking a contrarian stance is one of the most important ways that new knowledge and understanding develops from existing literature.
• The case provides an opportunity to pursue action leading to the resolution of a problem? Another way to think about choosing a case to study is to consider how the results from investigating a particular case may result in findings that reveal ways in which to resolve an existing or emerging problem. For example, studying the case of an unforeseen incident, such as a fatal accident at a railroad crossing, can reveal hidden issues that could be applied to preventative measures that contribute to reducing the chance of accidents in the future. In this example, a case study investigating the accident could lead to a better understanding of where to strategically locate additional signals at other railroad crossings so as to better warn drivers of an approaching train, particularly when visibility is hindered by heavy rain, fog, or at night.
• The case offers a new direction in future research? A case study can be used as a tool for an exploratory investigation that highlights the need for further research about the problem. A case can be used when there are few studies that help predict an outcome or that establish a clear understanding about how best to proceed in addressing a problem. For example, after conducting a thorough literature review [very important!], you discover that little research exists showing the ways in which women contribute to promoting water conservation in rural communities of east central Africa. A case study of how women contribute to saving water in a rural village of Uganda can lay the foundation for understanding the need for more thorough research that documents how women in their roles as cooks and family caregivers think about water as a valuable resource within their community. This example of a case study could also point to the need for scholars to build new theoretical frameworks around the topic [e.g., applying feminist theories of work and family to the issue of water conservation].

Eisenhardt, Kathleen M. “Building Theories from Case Study Research.” Academy of Management Review 14 (October 1989): 532-550; Emmel, Nick. Sampling and Choosing Cases in Qualitative Research: A Realist Approach . Thousand Oaks, CA: SAGE Publications, 2013; Gerring, John. “What Is a Case Study and What Is It Good for?” American Political Science Review 98 (May 2004): 341-354; Mills, Albert J. , Gabrielle Durepos, and Eiden Wiebe, editors. Encyclopedia of Case Study Research . Thousand Oaks, CA: SAGE Publications, 2010; Seawright, Jason and John Gerring. "Case Selection Techniques in Case Study Research." Political Research Quarterly 61 (June 2008): 294-308.

Structure and Writing Style

The purpose of a paper in the social sciences designed around a case study is to thoroughly investigate a subject of analysis in order to reveal a new understanding about the research problem and, in so doing, contributing new knowledge to what is already known from previous studies. In applied social sciences disciplines [e.g., education, social work, public administration, etc.], case studies may also be used to reveal best practices, highlight key programs, or investigate interesting aspects of professional work.

In general, the structure of a case study research paper is not all that different from a standard college-level research paper. However, there are subtle differences you should be aware of. Here are the key elements to organizing and writing a case study research paper.

I.  Introduction

As with any research paper, your introduction should serve as a roadmap for your readers to ascertain the scope and purpose of your study . The introduction to a case study research paper, however, should not only describe the research problem and its significance, but you should also succinctly describe why the case is being used and how it relates to addressing the problem. The two elements should be linked. With this in mind, a good introduction answers these four questions:

• What is being studied? Describe the research problem and describe the subject of analysis [the case] you have chosen to address the problem. Explain how they are linked and what elements of the case will help to expand knowledge and understanding about the problem.
• Why is this topic important to investigate? Describe the significance of the research problem and state why a case study design and the subject of analysis that the paper is designed around is appropriate in addressing the problem.
• What did we know about this topic before I did this study? Provide background that helps lead the reader into the more in-depth literature review to follow. If applicable, summarize prior case study research applied to the research problem and why it fails to adequately address the problem. Describe why your case will be useful. If no prior case studies have been used to address the research problem, explain why you have selected this subject of analysis.
• How will this study advance new knowledge or new ways of understanding? Explain why your case study will be suitable in helping to expand knowledge and understanding about the research problem.

Each of these questions should be addressed in no more than a few paragraphs. Exceptions to this can be when you are addressing a complex research problem or subject of analysis that requires more in-depth background information.

II.  Literature Review

The literature review for a case study research paper is generally structured the same as it is for any college-level research paper. The difference, however, is that the literature review is focused on providing background information and  enabling historical interpretation of the subject of analysis in relation to the research problem the case is intended to address . This includes synthesizing studies that help to:

• Place relevant works in the context of their contribution to understanding the case study being investigated . This would involve summarizing studies that have used a similar subject of analysis to investigate the research problem. If there is literature using the same or a very similar case to study, you need to explain why duplicating past research is important [e.g., conditions have changed; prior studies were conducted long ago, etc.].
• Describe the relationship each work has to the others under consideration that informs the reader why this case is applicable . Your literature review should include a description of any works that support using the case to investigate the research problem and the underlying research questions.
• Identify new ways to interpret prior research using the case study . If applicable, review any research that has examined the research problem using a different research design. Explain how your use of a case study design may reveal new knowledge or a new perspective or that can redirect research in an important new direction.
• Resolve conflicts amongst seemingly contradictory previous studies . This refers to synthesizing any literature that points to unresolved issues of concern about the research problem and describing how the subject of analysis that forms the case study can help resolve these existing contradictions.
• Point the way in fulfilling a need for additional research . Your review should examine any literature that lays a foundation for understanding why your case study design and the subject of analysis around which you have designed your study may reveal a new way of approaching the research problem or offer a perspective that points to the need for additional research.
• Expose any gaps that exist in the literature that the case study could help to fill . Summarize any literature that not only shows how your subject of analysis contributes to understanding the research problem, but how your case contributes to a new way of understanding the problem that prior research has failed to do.
• Locate your own research within the context of existing literature [very important!] . Collectively, your literature review should always place your case study within the larger domain of prior research about the problem. The overarching purpose of reviewing pertinent literature in a case study paper is to demonstrate that you have thoroughly identified and synthesized prior studies in relation to explaining the relevance of the case in addressing the research problem.

III.  Method

In this section, you explain why you selected a particular case [i.e., subject of analysis] and the strategy you used to identify and ultimately decide that your case was appropriate in addressing the research problem. The way you describe the methods used varies depending on the type of subject of analysis that constitutes your case study.

If your subject of analysis is an incident or event . In the social and behavioral sciences, the event or incident that represents the case to be studied is usually bounded by time and place, with a clear beginning and end and with an identifiable location or position relative to its surroundings. The subject of analysis can be a rare or critical event or it can focus on a typical or regular event. The purpose of studying a rare event is to illuminate new ways of thinking about the broader research problem or to test a hypothesis. Critical incident case studies must describe the method by which you identified the event and explain the process by which you determined the validity of this case to inform broader perspectives about the research problem or to reveal new findings. However, the event does not have to be a rare or uniquely significant to support new thinking about the research problem or to challenge an existing hypothesis. For example, Walo, Bull, and Breen conducted a case study to identify and evaluate the direct and indirect economic benefits and costs of a local sports event in the City of Lismore, New South Wales, Australia. The purpose of their study was to provide new insights from measuring the impact of a typical local sports event that prior studies could not measure well because they focused on large "mega-events." Whether the event is rare or not, the methods section should include an explanation of the following characteristics of the event: a) when did it take place; b) what were the underlying circumstances leading to the event; and, c) what were the consequences of the event in relation to the research problem.

If your subject of analysis is a person. Explain why you selected this particular individual to be studied and describe what experiences they have had that provide an opportunity to advance new understandings about the research problem. Mention any background about this person which might help the reader understand the significance of their experiences that make them worthy of study. This includes describing the relationships this person has had with other people, institutions, and/or events that support using them as the subject for a case study research paper. It is particularly important to differentiate the person as the subject of analysis from others and to succinctly explain how the person relates to examining the research problem [e.g., why is one politician in a particular local election used to show an increase in voter turnout from any other candidate running in the election]. Note that these issues apply to a specific group of people used as a case study unit of analysis [e.g., a classroom of students].

If your subject of analysis is a place. In general, a case study that investigates a place suggests a subject of analysis that is unique or special in some way and that this uniqueness can be used to build new understanding or knowledge about the research problem. A case study of a place must not only describe its various attributes relevant to the research problem [e.g., physical, social, historical, cultural, economic, political], but you must state the method by which you determined that this place will illuminate new understandings about the research problem. It is also important to articulate why a particular place as the case for study is being used if similar places also exist [i.e., if you are studying patterns of homeless encampments of veterans in open spaces, explain why you are studying Echo Park in Los Angeles rather than Griffith Park?]. If applicable, describe what type of human activity involving this place makes it a good choice to study [e.g., prior research suggests Echo Park has more homeless veterans].

If your subject of analysis is a phenomenon. A phenomenon refers to a fact, occurrence, or circumstance that can be studied or observed but with the cause or explanation to be in question. In this sense, a phenomenon that forms your subject of analysis can encompass anything that can be observed or presumed to exist but is not fully understood. In the social and behavioral sciences, the case usually focuses on human interaction within a complex physical, social, economic, cultural, or political system. For example, the phenomenon could be the observation that many vehicles used by ISIS fighters are small trucks with English language advertisements on them. The research problem could be that ISIS fighters are difficult to combat because they are highly mobile. The research questions could be how and by what means are these vehicles used by ISIS being supplied to the militants and how might supply lines to these vehicles be cut off? How might knowing the suppliers of these trucks reveal larger networks of collaborators and financial support? A case study of a phenomenon most often encompasses an in-depth analysis of a cause and effect that is grounded in an interactive relationship between people and their environment in some way.

NOTE:   The choice of the case or set of cases to study cannot appear random. Evidence that supports the method by which you identified and chose your subject of analysis should clearly support investigation of the research problem and linked to key findings from your literature review. Be sure to cite any studies that helped you determine that the case you chose was appropriate for examining the problem.

IV.  Discussion

The main elements of your discussion section are generally the same as any research paper, but centered around interpreting and drawing conclusions about the key findings from your analysis of the case study. Note that a general social sciences research paper may contain a separate section to report findings. However, in a paper designed around a case study, it is common to combine a description of the results with the discussion about their implications. The objectives of your discussion section should include the following:

Reiterate the Research Problem/State the Major Findings Briefly reiterate the research problem you are investigating and explain why the subject of analysis around which you designed the case study were used. You should then describe the findings revealed from your study of the case using direct, declarative, and succinct proclamation of the study results. Highlight any findings that were unexpected or especially profound.

Explain the Meaning of the Findings and Why They are Important Systematically explain the meaning of your case study findings and why you believe they are important. Begin this part of the section by repeating what you consider to be your most important or surprising finding first, then systematically review each finding. Be sure to thoroughly extrapolate what your analysis of the case can tell the reader about situations or conditions beyond the actual case that was studied while, at the same time, being careful not to misconstrue or conflate a finding that undermines the external validity of your conclusions.

Relate the Findings to Similar Studies No study in the social sciences is so novel or possesses such a restricted focus that it has absolutely no relation to previously published research. The discussion section should relate your case study results to those found in other studies, particularly if questions raised from prior studies served as the motivation for choosing your subject of analysis. This is important because comparing and contrasting the findings of other studies helps support the overall importance of your results and it highlights how and in what ways your case study design and the subject of analysis differs from prior research about the topic.

Consider Alternative Explanations of the Findings Remember that the purpose of social science research is to discover and not to prove. When writing the discussion section, you should carefully consider all possible explanations revealed by the case study results, rather than just those that fit your hypothesis or prior assumptions and biases. Be alert to what the in-depth analysis of the case may reveal about the research problem, including offering a contrarian perspective to what scholars have stated in prior research if that is how the findings can be interpreted from your case.

Acknowledge the Study's Limitations You can state the study's limitations in the conclusion section of your paper but describing the limitations of your subject of analysis in the discussion section provides an opportunity to identify the limitations and explain why they are not significant. This part of the discussion section should also note any unanswered questions or issues your case study could not address. More detailed information about how to document any limitations to your research can be found here .

Suggest Areas for Further Research Although your case study may offer important insights about the research problem, there are likely additional questions related to the problem that remain unanswered or findings that unexpectedly revealed themselves as a result of your in-depth analysis of the case. Be sure that the recommendations for further research are linked to the research problem and that you explain why your recommendations are valid in other contexts and based on the original assumptions of your study.

V.  Conclusion

As with any research paper, you should summarize your conclusion in clear, simple language; emphasize how the findings from your case study differs from or supports prior research and why. Do not simply reiterate the discussion section. Provide a synthesis of key findings presented in the paper to show how these converge to address the research problem. If you haven't already done so in the discussion section, be sure to document the limitations of your case study and any need for further research.

The function of your paper's conclusion is to: 1) reiterate the main argument supported by the findings from your case study; 2) state clearly the context, background, and necessity of pursuing the research problem using a case study design in relation to an issue, controversy, or a gap found from reviewing the literature; and, 3) provide a place to persuasively and succinctly restate the significance of your research problem, given that the reader has now been presented with in-depth information about the topic.

Consider the following points to help ensure your conclusion is appropriate:

• If the argument or purpose of your paper is complex, you may need to summarize these points for your reader.
• If prior to your conclusion, you have not yet explained the significance of your findings or if you are proceeding inductively, use the conclusion of your paper to describe your main points and explain their significance.
• Move from a detailed to a general level of consideration of the case study's findings that returns the topic to the context provided by the introduction or within a new context that emerges from your case study findings.

Note that, depending on the discipline you are writing in or the preferences of your professor, the concluding paragraph may contain your final reflections on the evidence presented as it applies to practice or on the essay's central research problem. However, the nature of being introspective about the subject of analysis you have investigated will depend on whether you are explicitly asked to express your observations in this way.

Problems to Avoid

Overgeneralization One of the goals of a case study is to lay a foundation for understanding broader trends and issues applied to similar circumstances. However, be careful when drawing conclusions from your case study. They must be evidence-based and grounded in the results of the study; otherwise, it is merely speculation. Looking at a prior example, it would be incorrect to state that a factor in improving girls access to education in Azerbaijan and the policy implications this may have for improving access in other Muslim nations is due to girls access to social media if there is no documentary evidence from your case study to indicate this. There may be anecdotal evidence that retention rates were better for girls who were engaged with social media, but this observation would only point to the need for further research and would not be a definitive finding if this was not a part of your original research agenda.

Failure to Document Limitations No case is going to reveal all that needs to be understood about a research problem. Therefore, just as you have to clearly state the limitations of a general research study , you must describe the specific limitations inherent in the subject of analysis. For example, the case of studying how women conceptualize the need for water conservation in a village in Uganda could have limited application in other cultural contexts or in areas where fresh water from rivers or lakes is plentiful and, therefore, conservation is understood more in terms of managing access rather than preserving access to a scarce resource.

Failure to Extrapolate All Possible Implications Just as you don't want to over-generalize from your case study findings, you also have to be thorough in the consideration of all possible outcomes or recommendations derived from your findings. If you do not, your reader may question the validity of your analysis, particularly if you failed to document an obvious outcome from your case study research. For example, in the case of studying the accident at the railroad crossing to evaluate where and what types of warning signals should be located, you failed to take into consideration speed limit signage as well as warning signals. When designing your case study, be sure you have thoroughly addressed all aspects of the problem and do not leave gaps in your analysis that leave the reader questioning the results.

Case Studies. Writing@CSU. Colorado State University; Gerring, John. Case Study Research: Principles and Practices . New York: Cambridge University Press, 2007; Merriam, Sharan B. Qualitative Research and Case Study Applications in Education . Rev. ed. San Francisco, CA: Jossey-Bass, 1998; Miller, Lisa L. “The Use of Case Studies in Law and Social Science Research.” Annual Review of Law and Social Science 14 (2018): TBD; Mills, Albert J., Gabrielle Durepos, and Eiden Wiebe, editors. Encyclopedia of Case Study Research . Thousand Oaks, CA: SAGE Publications, 2010; Putney, LeAnn Grogan. "Case Study." In Encyclopedia of Research Design , Neil J. Salkind, editor. (Thousand Oaks, CA: SAGE Publications, 2010), pp. 116-120; Simons, Helen. Case Study Research in Practice . London: SAGE Publications, 2009;  Kratochwill,  Thomas R. and Joel R. Levin, editors. Single-Case Research Design and Analysis: New Development for Psychology and Education .  Hilldsale, NJ: Lawrence Erlbaum Associates, 1992; Swanborn, Peter G. Case Study Research: What, Why and How? London : SAGE, 2010; Yin, Robert K. Case Study Research: Design and Methods . 6th edition. Los Angeles, CA, SAGE Publications, 2014; Walo, Maree, Adrian Bull, and Helen Breen. “Achieving Economic Benefits at Local Events: A Case Study of a Local Sports Event.” Festival Management and Event Tourism 4 (1996): 95-106.

Writing Tip

At Least Five Misconceptions about Case Study Research

Social science case studies are often perceived as limited in their ability to create new knowledge because they are not randomly selected and findings cannot be generalized to larger populations. Flyvbjerg examines five misunderstandings about case study research and systematically "corrects" each one. To quote, these are:

Misunderstanding 1 :  General, theoretical [context-independent] knowledge is more valuable than concrete, practical [context-dependent] knowledge. Misunderstanding 2 :  One cannot generalize on the basis of an individual case; therefore, the case study cannot contribute to scientific development. Misunderstanding 3 :  The case study is most useful for generating hypotheses; that is, in the first stage of a total research process, whereas other methods are more suitable for hypotheses testing and theory building. Misunderstanding 4 :  The case study contains a bias toward verification, that is, a tendency to confirm the researcher’s preconceived notions. Misunderstanding 5 :  It is often difficult to summarize and develop general propositions and theories on the basis of specific case studies [p. 221].

While writing your paper, think introspectively about how you addressed these misconceptions because to do so can help you strengthen the validity and reliability of your research by clarifying issues of case selection, the testing and challenging of existing assumptions, the interpretation of key findings, and the summation of case outcomes. Think of a case study research paper as a complete, in-depth narrative about the specific properties and key characteristics of your subject of analysis applied to the research problem.

Flyvbjerg, Bent. “Five Misunderstandings About Case-Study Research.” Qualitative Inquiry 12 (April 2006): 219-245.

• << Previous: Writing a Case Analysis Paper
• Next: Writing a Field Report >>
• Last Updated: Jun 3, 2024 9:44 AM
• URL: https://libguides.usc.edu/writingguide/assignments

Case study research: opening up research opportunities

RAUSP Management Journal

ISSN : 2531-0488

Article publication date: 30 December 2019

Issue publication date: 3 March 2020

The case study approach has been widely used in management studies and the social sciences more generally. However, there are still doubts about when and how case studies should be used. This paper aims to discuss this approach, its various uses and applications, in light of epistemological principles, as well as the criteria for rigor and validity.

Design/methodology/approach

This paper discusses the various concepts of case and case studies in the methods literature and addresses the different uses of cases in relation to epistemological principles and criteria for rigor and validity.

The use of this research approach can be based on several epistemologies, provided the researcher attends to the internal coherence between method and epistemology, or what the authors call “alignment.”

Originality/value

This study offers a number of implications for the practice of management research, as it shows how the case study approach does not commit the researcher to particular data collection or interpretation methods. Furthermore, the use of cases can be justified according to multiple epistemological orientations.

• Epistemology

Takahashi, A.R.W. and Araujo, L. (2020), "Case study research: opening up research opportunities", RAUSP Management Journal , Vol. 55 No. 1, pp. 100-111. https://doi.org/10.1108/RAUSP-05-2019-0109

Emerald Publishing Limited

Copyright © 2019, Adriana Roseli Wünsch Takahashi and Luis Araujo.

Published in RAUSP Management Journal . Published by Emerald Publishing Limited. This article is published under the Creative Commons Attribution (CC BY 4.0) licence. Anyone may reproduce, distribute, translate and create derivative works of this article (for both commercial and non-commercial purposes), subject to full attribution to the original publication and authors. The full terms of this licence may be seen at http://creativecommons.org/licences/by/4.0/legalcode

1. Introduction

The case study as a research method or strategy brings us to question the very term “case”: after all, what is a case? A case-based approach places accords the case a central role in the research process ( Ragin, 1992 ). However, doubts still remain about the status of cases according to different epistemologies and types of research designs.

Despite these doubts, the case study is ever present in the management literature and represents the main method of management research in Brazil ( Coraiola, Sander, Maccali, & Bulgacov, 2013 ). Between 2001 and 2010, 2,407 articles (83.14 per cent of qualitative research) were published in conferences and management journals as case studies (Takahashi & Semprebom, 2013 ). A search on Spell.org.br for the term “case study” under title, abstract or keywords, for the period ranging from January 2010 to July 2019, yielded 3,040 articles published in the management field. Doing research using case studies, allows the researcher to immerse him/herself in the context and gain intensive knowledge of a phenomenon, which in turn demands suitable methodological principles ( Freitas et al. , 2017 ).

Our objective in this paper is to discuss notions of what constitutes a case and its various applications, considering epistemological positions as well as criteria for rigor and validity. The alignment between these dimensions is put forward as a principle advocating coherence among all phases of the research process.

This article makes two contributions. First, we suggest that there are several epistemological justifications for using case studies. Second, we show that the quality and rigor of academic research with case studies are directly related to the alignment between epistemology and research design rather than to choices of specific forms of data collection or analysis. The article is structured as follows: the following four sections discuss concepts of what is a case, its uses, epistemological grounding as well as rigor and quality criteria. The brief conclusions summarize the debate and invite the reader to delve into the literature on the case study method as a way of furthering our understanding of contemporary management phenomena.

2. What is a case study?

The debate over what constitutes a case in social science is a long-standing one. In 1988, Howard Becker and Charles Ragin organized a workshop to discuss the status of the case as a social science method. As the discussion was inconclusive, they posed the question “What is a case?” to a select group of eight social scientists in 1989, and later to participants in a symposium on the subject. Participants were unable to come up with a consensual answer. Since then, we have witnessed that further debates and different answers have emerged. The original question led to an even broader issue: “How do we, as social scientists, produce results and seem to know what we know?” ( Ragin, 1992 , p. 16).

An important step that may help us start a reflection on what is a case is to consider the phenomena we are looking at. To do that, we must know something about what we want to understand and how we might study it. The answer may be a causal explanation, a description of what was observed or a narrative of what has been experienced. In any case, there will always be a story to be told, as the choice of the case study method demands an answer to what the case is about.

A case may be defined ex ante , prior to the start of the research process, as in Yin’s (2015) classical definition. But, there is no compelling reason as to why cases must be defined ex ante . Ragin (1992 , p. 217) proposed the notion of “casing,” to indicate that what the case is emerges from the research process:

Rather than attempt to delineate the many different meanings of the term “case” in a formal taxonomy, in this essay I offer instead a view of cases that follows from the idea implicit in many of the contributions – that concocting cases is a varied but routine social scientific activity. […] The approach of this essay is that this activity, which I call “casing”, should be viewed in practical terms as a research tactic. It is selectively invoked at many different junctures in the research process, usually to resolve difficult issues in linking ideas and evidence.

In other words, “casing” is tied to the researcher’s practice, to the way he/she delimits or declares a case as a significant outcome of a process. In 2013, Ragin revisited the 1992 concept of “casing” and explored its multiple possibilities of use, paying particular attention to “negative cases.”

According to Ragin (1992) , a case can be centered on a phenomenon or a population. In the first scenario, cases are representative of a phenomenon, and are selected based on what can be empirically observed. The process highlights different aspects of cases and obscures others according to the research design, and allows for the complexity, specificity and context of the phenomenon to be explored. In the alternative, population-focused scenario, the selection of cases precedes the research. Both positive and negative cases are considered in exploring a phenomenon, with the definition of the set of cases dependent on theory and the central objective to build generalizations. As a passing note, it is worth mentioning here that a study of multiple cases requires a definition of the unit of analysis a priori . Otherwise, it will not be possible to make cross-case comparisons.

These two approaches entail differences that go beyond the mere opposition of quantitative and qualitative data, as a case often includes both types of data. Thus, the confusion about how to conceive cases is associated with Ragin’s (1992) notion of “small vs large N,” or McKeown’s (1999) “statistical worldview” – the notion that relevant findings are only those that can be made about a population based on the analysis of representative samples. In the same vein, Byrne (2013) argues that we cannot generate nomothetic laws that apply in all circumstances, periods and locations, and that no social science method can claim to generate invariant laws. According to the same author, case studies can help us understand that there is more than one ideographic variety and help make social science useful. Generalizations still matter, but they should be understood as part of defining the research scope, and that scope points to the limitations of knowledge produced and consumed in concrete time and space.

Thus, what defines the orientation and the use of cases is not the mere choice of type of data, whether quantitative or qualitative, but the orientation of the study. A statistical worldview sees cases as data units ( Byrne, 2013 ). Put differently, there is a clear distinction between statistical and qualitative worldviews; the use of quantitative data does not by itself means that the research is (quasi) statistical, or uses a deductive logic:

Case-based methods are useful, and represent, among other things, a way of moving beyond a useless and destructive tradition in the social sciences that have set quantitative and qualitative modes of exploration, interpretation, and explanation against each other ( Byrne, 2013 , p. 9).

Other authors advocate different understandings of what a case study is. To some, it is a research method, to others it is a research strategy ( Creswell, 1998 ). Sharan Merrian and Robert Yin, among others, began to write about case study research as a methodology in the 1980s (Merrian, 2009), while authors such as Eisenhardt (1989) called it a research strategy. Stake (2003) sees the case study not as a method, but as a choice of what to be studied, the unit of study. Regardless of their differences, these authors agree that case studies should be restricted to a particular context as they aim to provide an in-depth knowledge of a given phenomenon: “A case study is an in-depth description and analysis of a bounded system” (Merrian, 2009, p. 40). According to Merrian, a qualitative case study can be defined by the process through which the research is carried out, by the unit of analysis or the final product, as the choice ultimately depends on what the researcher wants to know. As a product of research, it involves the analysis of a given entity, phenomenon or social unit.

Thus, whether it is an organization, an individual, a context or a phenomenon, single or multiple, one must delimit it, and also choose between possible types and configurations (Merrian, 2009; Yin, 2015 ). A case study may be descriptive, exploratory, explanatory, single or multiple ( Yin, 2015 ); intrinsic, instrumental or collective ( Stake, 2003 ); and confirm or build theory ( Eisenhardt, 1989 ).

both went through the same process of implementing computer labs intended for the use of information and communication technologies in 2007;

both took part in the same regional program (Paraná Digital); and

they shared similar characteristics regarding location (operation in the same neighborhood of a city), number of students, number of teachers and technicians and laboratory sizes.

However, the two institutions differed in the number of hours of program use, with one of them displaying a significant number of hours/use while the other showed a modest number, according to secondary data for the period 2007-2013. Despite the context being similar and the procedures for implementing the technology being the same, the mechanisms of social integration – an idiosyncratic factor of each institution – were different in each case. This explained differences in their use of resource, processes of organizational learning and capacity to absorb new knowledge.

On the other hand, multiple case studies seek evidence in different contexts and do not necessarily require direct comparisons ( Stake, 2003 ). Rather, there is a search for patterns of convergence and divergence that permeate all the cases, as the same issues are explored in every case. Cases can be added progressively until theoretical saturation is achieved. An example is of a study that investigated how entrepreneurial opportunity and management skills were developed through entrepreneurial learning ( Zampier & Takahashi, 2014 ). The authors conducted nine case studies, based on primary and secondary data, with each one analyzed separately, so a search for patterns could be undertaken. The convergence aspects found were: the predominant way of transforming experience into knowledge was exploitation; managerial skills were developed through by taking advantages of opportunities; and career orientation encompassed more than one style. As for divergence patterns: the experience of success and failure influenced entrepreneurs differently; the prevailing rationality logic of influence was different; and the combination of styles in career orientation was diverse.

A full discussion of choice of case study design is outside the scope of this article. For the sake of illustration, we make a brief mention to other selection criteria such as the purpose of the research, the state of the art of the research theme, the time and resources involved and the preferred epistemological position of the researcher. In the next section, we look at the possibilities of carrying out case studies in line with various epistemological traditions, as the answers to the “what is a case?” question reveal varied methodological commitments as well as diverse epistemological and ontological positions ( Ragin, 2013 ).

3. Epistemological positioning of case study research

Ontology and epistemology are like skin, not a garment to be occasionally worn ( Marsh & Furlong, 2002 ). According to these authors, ontology and epistemology guide the choice of theory and method because they cannot or should not be worn as a garment. Hence, one must practice philosophical “self-knowledge” to recognize one’s vision of what the world is and of how knowledge of that world is accessed and validated. Ontological and epistemological positions are relevant in that they involve the positioning of the researcher in social science and the phenomena he or she chooses to study. These positions do not tend to vary from one project to another although they can certainly change over time for a single researcher.

Ontology is the starting point from which the epistemological and methodological positions of the research arise ( Grix, 2002 ). Ontology expresses a view of the world, what constitutes reality, nature and the image one has of social reality; it is a theory of being ( Marsh & Furlong, 2002 ). The central question is the nature of the world out there regardless of our ability to access it. An essentialist or foundationalist ontology acknowledges that there are differences that persist over time and these differences are what underpin the construction of social life. An opposing, anti-foundationalist position presumes that the differences found are socially constructed and may vary – i.e. they are not essential but specific to a given culture at a given time ( Marsh & Furlong, 2002 ).

Epistemology is centered around a theory of knowledge, focusing on the process of acquiring and validating knowledge ( Grix, 2002 ). Positivists look at social phenomena as a world of causal relations where there is a single truth to be accessed and confirmed. In this tradition, case studies test hypotheses and rely on deductive approaches and quantitative data collection and analysis techniques. Scholars in the field of anthropology and observation-based qualitative studies proposed alternative epistemologies based on notions of the social world as a set of manifold and ever-changing processes. In management studies since the 1970s, the gradual acceptance of qualitative research has generated a diverse range of research methods and conceptions of the individual and society ( Godoy, 1995 ).

The interpretative tradition, in direct opposition to positivism, argues that there is no single objective truth to be discovered about the social world. The social world and our knowledge of it are the product of social constructions. Thus, the social world is constituted by interactions, and our knowledge is hermeneutic as the world does not exist independent of our knowledge ( Marsh & Furlong, 2002 ). The implication is that it is not possible to access social phenomena through objective, detached methods. Instead, the interaction mechanisms and relationships that make up social constructions have to be studied. Deductive approaches, hypothesis testing and quantitative methods are not relevant here. Hermeneutics, on the other hand, is highly relevant as it allows the analysis of the individual’s interpretation, of sayings, texts and actions, even though interpretation is always the “truth” of a subject. Methods such as ethnographic case studies, interviews and observations as data collection techniques should feed research designs according to interpretivism. It is worth pointing out that we are to a large extent, caricaturing polar opposites rather characterizing a range of epistemological alternatives, such as realism, conventionalism and symbolic interactionism.

If diverse ontologies and epistemologies serve as a guide to research approaches, including data collection and analysis methods, and if they should be regarded as skin rather than clothing, how does one make choices regarding case studies? What are case studies, what type of knowledge they provide and so on? The views of case study authors are not always explicit on this point, so we must delve into their texts to glean what their positions might be.

Two of the cited authors in case study research are Robert Yin and Kathleen Eisenhardt. Eisenhardt (1989) argues that a case study can serve to provide a description, test or generate a theory, the latter being the most relevant in contributing to the advancement of knowledge in a given area. She uses terms such as populations and samples, control variables, hypotheses and generalization of findings and even suggests an ideal number of case studies to allow for theory construction through replication. Although Eisenhardt includes observation and interview among her recommended data collection techniques, the approach is firmly anchored in a positivist epistemology:

Third, particularly in comparison with Strauss (1987) and Van Maanen (1988), the process described here adopts a positivist view of research. That is, the process is directed toward the development of testable hypotheses and theory which are generalizable across settings. In contrast, authors like Strauss and Van Maanen are more concerned that a rich, complex description of the specific cases under study evolve and they appear less concerned with development of generalizable theory ( Eisenhardt, 1989 , p. 546).

This position attracted a fair amount of criticism. Dyer & Wilkins (1991) in a critique of Eisenhardt’s (1989) article focused on the following aspects: there is no relevant justification for the number of cases recommended; it is the depth and not the number of cases that provides an actual contribution to theory; and the researcher’s purpose should be to get closer to the setting and interpret it. According to the same authors, discrepancies from prior expectations are also important as they lead researchers to reflect on existing theories. Eisenhardt & Graebner (2007 , p. 25) revisit the argument for the construction of a theory from multiple cases:

A major reason for the popularity and relevance of theory building from case studies is that it is one of the best (if not the best) of the bridges from rich qualitative evidence to mainstream deductive research.

Although they recognize the importance of single-case research to explore phenomena under unique or rare circumstances, they reaffirm the strength of multiple case designs as it is through them that better accuracy and generalization can be reached.

Likewise, Robert Yin emphasizes the importance of variables, triangulation in the search for “truth” and generalizable theoretical propositions. Yin (2015 , p. 18) suggests that the case study method may be appropriate for different epistemological orientations, although much of his work seems to invoke a realist epistemology. Authors such as Merrian (2009) and Stake (2003) suggest an interpretative version of case studies. Stake (2003) looks at cases as a qualitative option, where the most relevant criterion of case selection should be the opportunity to learn and understand a phenomenon. A case is not just a research method or strategy; it is a researcher’s choice about what will be studied:

Even if my definition of case study was agreed upon, and it is not, the term case and study defy full specification (Kemmis, 1980). A case study is both a process of inquiry about the case and the product of that inquiry ( Stake, 2003 , p. 136).

Later, Stake (2003 , p. 156) argues that:

[…] the purpose of a case report is not to represent the world, but to represent the case. […] The utility of case research to practitioners and policy makers is in its extension of experience.

Still according to Stake (2003 , pp. 140-141), to do justice to complex views of social phenomena, it is necessary to analyze the context and relate it to the case, to look for what is peculiar rather than common in cases to delimit their boundaries, to plan the data collection looking for what is common and unusual about facts, what could be valuable whether it is unique or common:

Reflecting upon the pertinent literature, I find case study methodology written largely by people who presume that the research should contribute to scientific generalization. The bulk of case study work, however, is done by individuals who have intrinsic interest in the case and little interest in the advance of science. Their designs aim the inquiry toward understanding of what is important about that case within its own world, which is seldom the same as the worlds of researchers and theorists. Those designs develop what is perceived to be the case’s own issues, contexts, and interpretations, its thick descriptions . In contrast, the methods of instrumental case study draw the researcher toward illustrating how the concerns of researchers and theorists are manifest in the case. Because the critical issues are more likely to be know in advance and following disciplinary expectations, such a design can take greater advantage of already developed instruments and preconceived coding schemes.

The aforementioned authors were listed to illustrate differences and sometimes opposing positions on case research. These differences are not restricted to a choice between positivism and interpretivism. It is worth noting that Ragin’s (2013 , p. 523) approach to “casing” is compatible with the realistic research perspective:

In essence, to posit cases is to engage in ontological speculation regarding what is obdurately real but only partially and indirectly accessible through social science. Bringing a realist perspective to the case question deepens and enriches the dialogue, clarifying some key issues while sweeping others aside.

cases are actual entities, reflecting their operations of real causal mechanism and process patterns;

case studies are interactive processes and are open to revisions and refinements; and

social phenomena are complex, contingent and context-specific.

Ragin (2013 , p. 532) concludes:

Lurking behind my discussion of negative case, populations, and possibility analysis is the implication that treating cases as members of given (and fixed) populations and seeking to infer the properties of populations may be a largely illusory exercise. While demographers have made good use of the concept of population, and continue to do so, it is not clear how much the utility of the concept extends beyond their domain. In case-oriented work, the notion of fixed populations of cases (observations) has much less analytic utility than simply “the set of relevant cases,” a grouping that must be specified or constructed by the researcher. The demarcation of this set, as the work of case-oriented researchers illustrates, is always tentative, fluid, and open to debate. It is only by casing social phenomena that social scientists perceive the homogeneity that allows analysis to proceed.

In summary, case studies are relevant and potentially compatible with a range of different epistemologies. Researchers’ ontological and epistemological positions will guide their choice of theory, methodologies and research techniques, as well as their research practices. The same applies to the choice of authors describing the research method and this choice should be coherent. We call this research alignment , an attribute that must be judged on the internal coherence of the author of a study, and not necessarily its evaluator. The following figure illustrates the interrelationship between the elements of a study necessary for an alignment ( Figure 1 ).

In addition to this broader aspect of the research as a whole, other factors should be part of the researcher’s concern, such as the rigor and quality of case studies. We will look into these in the next section taking into account their relevance to the different epistemologies.

4. Rigor and quality in case studies

Traditionally, at least in positivist studies, validity and reliability are the relevant quality criteria to judge research. Validity can be understood as external, internal and construct. External validity means identifying whether the findings of a study are generalizable to other studies using the logic of replication in multiple case studies. Internal validity may be established through the theoretical underpinning of existing relationships and it involves the use of protocols for the development and execution of case studies. Construct validity implies defining the operational measurement criteria to establish a chain of evidence, such as the use of multiple sources of evidence ( Eisenhardt, 1989 ; Yin, 2015 ). Reliability implies conducting other case studies, instead of just replicating results, to minimize the errors and bias of a study through case study protocols and the development of a case database ( Yin, 2015 ).

Several criticisms have been directed toward case studies, such as lack of rigor, lack of generalization potential, external validity and researcher bias. Case studies are often deemed to be unreliable because of a lack of rigor ( Seuring, 2008 ). Flyvbjerg (2006 , p. 219) addresses five misunderstandings about case-study research, and concludes that:

[…] a scientific discipline without a large number of thoroughly executed case studies is a discipline without systematic production of exemplars, and a discipline without exemplars is an ineffective one.

theoretical knowledge is more valuable than concrete, practical knowledge;

the case study cannot contribute to scientific development because it is not possible to generalize on the basis of an individual case;

the case study is more useful for generating rather than testing hypotheses;

the case study contains a tendency to confirm the researcher’s preconceived notions; and

it is difficult to summarize and develop general propositions and theories based on case studies.

These criticisms question the validity of the case study as a scientific method and should be corrected.

The critique of case studies is often framed from the standpoint of what Ragin (2000) labeled large-N research. The logic of small-N research, to which case studies belong, is different. Cases benefit from depth rather than breadth as they: provide theoretical and empirical knowledge; contribute to theory through propositions; serve not only to confirm knowledge, but also to challenge and overturn preconceived notions; and the difficulty in summarizing their conclusions is because of the complexity of the phenomena studies and not an intrinsic limitation of the method.

Thus, case studies do not seek large-scale generalizations as that is not their purpose. And yet, this is a limitation from a positivist perspective as there is an external reality to be “apprehended” and valid conclusions to be extracted for an entire population. If positivism is the epistemology of choice, the rigor of a case study can be demonstrated by detailing the criteria used for internal and external validity, construct validity and reliability ( Gibbert & Ruigrok, 2010 ; Gibbert, Ruigrok, & Wicki, 2008 ). An example can be seen in case studies in the area of information systems, where there is a predominant orientation of positivist approaches to this method ( Pozzebon & Freitas, 1998 ). In this area, rigor also involves the definition of a unit of analysis, type of research, number of cases, selection of sites, definition of data collection and analysis procedures, definition of the research protocol and writing a final report. Creswell (1998) presents a checklist for researchers to assess whether the study was well written, if it has reliability and validity and if it followed methodological protocols.

In case studies with a non-positivist orientation, rigor can be achieved through careful alignment (coherence among ontology, epistemology, theory and method). Moreover, the concepts of validity can be understood as concern and care in formulating research, research development and research results ( Ollaik & Ziller, 2012 ), and to achieve internal coherence ( Gibbert et al. , 2008 ). The consistency between data collection and interpretation, and the observed reality also help these studies meet coherence and rigor criteria. Siggelkow (2007) argues that a case study should be persuasive and that even a single case study may be a powerful example to contest a widely held view. To him, the value of a single case study or studies with few cases can be attained by their potential to provide conceptual insights and coherence to the internal logic of conceptual arguments: “[…] a paper should allow a reader to see the world, and not just the literature, in a new way” ( Siggelkow, 2007 , p. 23).

Interpretative studies should not be justified by criteria derived from positivism as they are based on a different ontology and epistemology ( Sandberg, 2005 ). The rejection of an interpretive epistemology leads to the rejection of an objective reality: “As Bengtsson points out, the life-world is the subjects’ experience of reality, at the same time as it is objective in the sense that it is an intersubjective world” ( Sandberg, 2005 , p. 47). In this event, how can one demonstrate what positivists call validity and reliability? What would be the criteria to justify knowledge as truth, produced by research in this epistemology? Sandberg (2005 , p. 62) suggests an answer based on phenomenology:

This was demonstrated first by explicating life-world and intentionality as the basic assumptions underlying the interpretative research tradition. Second, based on those assumptions, truth as intentional fulfillment, consisting of perceived fulfillment, fulfillment in practice, and indeterminate fulfillment, was proposed. Third, based on the proposed truth constellation, communicative, pragmatic, and transgressive validity and reliability as interpretative awareness were presented as the most appropriate criteria for justifying knowledge produced within interpretative approach. Finally, the phenomenological epoché was suggested as a strategy for achieving these criteria.

From this standpoint, the research site must be chosen according to its uniqueness so that one can obtain relevant insights that no other site could provide ( Siggelkow, 2007 ). Furthermore, the view of what is being studied is at the center of the researcher’s attention to understand its “truth,” inserted in a given context.

The case researcher is someone who can reduce the probability of misinterpretations by analyzing multiple perceptions, searches for data triangulation to check for the reliability of interpretations ( Stake, 2003 ). It is worth pointing out that this is not an option for studies that specifically seek the individual’s experience in relation to organizational phenomena.

In short, there are different ways of seeking rigor and quality in case studies, depending on the researcher’s worldview. These different forms pervade everything from the research design, the choice of research questions, the theory or theories to look at a phenomenon, research methods, the data collection and analysis techniques, to the type and style of research report produced. Validity can also take on different forms. While positivism is concerned with validity of the research question and results, interpretivism emphasizes research processes without neglecting the importance of the articulation of pertinent research questions and the sound interpretation of results ( Ollaik & Ziller, 2012 ). The means to achieve this can be diverse, such as triangulation (of multiple theories, multiple methods, multiple data sources or multiple investigators), pre-tests of data collection instrument, pilot case, study protocol, detailed description of procedures such as field diary in observations, researcher positioning (reflexivity), theoretical-empirical consistency, thick description and transferability.

5. Conclusions

The central objective of this article was to discuss concepts of case study research, their potential and various uses, taking into account different epistemologies as well as criteria of rigor and validity. Although the literature on methodology in general and on case studies in particular, is voluminous, it is not easy to relate this approach to epistemology. In addition, method manuals often focus on the details of various case study approaches which confuse things further.

Faced with this scenario, we have tried to address some central points in this debate and present various ways of using case studies according to the preferred epistemology of the researcher. We emphasize that this understanding depends on how a case is defined and the particular epistemological orientation that underpins that conceptualization. We have argued that whatever the epistemological orientation is, it is possible to meet appropriate criteria of research rigor and quality provided there is an alignment among the different elements of the research process. Furthermore, multiple data collection techniques can be used in in single or multiple case study designs. Data collection techniques or the type of data collected do not define the method or whether cases should be used for theory-building or theory-testing.

Finally, we encourage researchers to consider case study research as one way to foster immersion in phenomena and their contexts, stressing that the approach does not imply a commitment to a particular epistemology or type of research, such as qualitative or quantitative. Case study research allows for numerous possibilities, and should be celebrated for that diversity rather than pigeon-holed as a monolithic research method.

The interrelationship between the building blocks of research

Byrne , D. ( 2013 ). Case-based methods: Why We need them; what they are; how to do them . Byrne D. In D Byrne. and C.C Ragin (Eds.), The SAGE handbooks of Case-Based methods , pp. 1 – 10 . London : SAGE Publications Inc .

Creswell , J. W. ( 1998 ). Qualitative inquiry and research design: choosing among five traditions , London : Sage Publications .

Coraiola , D. M. , Sander , J. A. , Maccali , N. & Bulgacov , S. ( 2013 ). Estudo de caso . In A. R. W. Takahashi , (Ed.), Pesquisa qualitativa em administração: Fundamentos, métodos e usos no Brasil , pp. 307 – 341 . São Paulo : Atlas .

Dyer , W. G. , & Wilkins , A. L. ( 1991 ). Better stories, not better constructs, to generate better theory: a rejoinder to Eisenhardt . The Academy of Management Review , 16 , 613 – 627 .

Eisenhardt , K. ( 1989 ). Building theory from case study research . Academy of Management Review , 14 , 532 – 550 .

Eisenhardt , K. M. , & Graebner , M. E. ( 2007 ). Theory building from cases: Opportunities and challenges . Academy of Management Journal , 50 , 25 – 32 .

Flyvbjerg , B. ( 2006 ). Five misunderstandings about case-study research . Qualitative Inquiry , 12 , 219 – 245 .

Freitas , J. S. , Ferreira , J. C. A. , Campos , A. A. R. , Melo , J. C. F. , Cheng , L. C. , & Gonçalves , C. A. ( 2017 ). Methodological roadmapping: a study of centering resonance analysis . RAUSP Management Journal , 53 , 459 – 475 .

Gibbert , M. , Ruigrok , W. , & Wicki , B. ( 2008 ). What passes as a rigorous case study? . Strategic Management Journal , 29 , 1465 – 1474 .

Gibbert , M. , & Ruigrok , W. ( 2010 ). The “what” and “how” of case study rigor: Three strategies based on published work . Organizational Research Methods , 13 , 710 – 737 .

Godoy , A. S. ( 1995 ). Introdução à pesquisa qualitativa e suas possibilidades . Revista de Administração de Empresas , 35 , 57 – 63 .

Grix , J. ( 2002 ). Introducing students to the generic terminology of social research . Politics , 22 , 175 – 186 .

Marsh , D. , & Furlong , P. ( 2002 ). A skin, not a sweater: ontology and epistemology in political science . In D Marsh. , & G Stoker , (Eds.), Theory and Methods in Political Science , New York, NY : Palgrave McMillan , pp. 17 – 41 .

McKeown , T. J. ( 1999 ). Case studies and the statistical worldview: Review of King, Keohane, and Verba’s designing social inquiry: Scientific inference in qualitative research . International Organization , 53 , 161 – 190 .

Merriam , S. B. ( 2009 ). Qualitative research: a guide to design and implementation .

Ollaik , L. G. , & Ziller , H. ( 2012 ). Distintas concepções de validade em pesquisas qualitativas . Educação e Pesquisa , 38 , 229 – 241 .

Picoli , F. R. , & Takahashi , A. R. W. ( 2016 ). Capacidade de absorção, aprendizagem organizacional e mecanismos de integração social . Revista de Administração Contemporânea , 20 , 1 – 20 .

Pozzebon , M. , & Freitas , H. M. R. ( 1998 ). Pela aplicabilidade: com um maior rigor científico – dos estudos de caso em sistemas de informação . Revista de Administração Contemporânea , 2 , 143 – 170 .

Sandberg , J. ( 2005 ). How do we justify knowledge produced within interpretive approaches? . Organizational Research Methods , 8 , 41 – 68 .

Seuring , S. A. ( 2008 ). Assessing the rigor of case study research in supply chain management. Supply chain management . Supply Chain Management: An International Journal , 13 , 128 – 137 .

Siggelkow , N. ( 2007 ). Persuasion with case studies . Academy of Management Journal , 50 , 20 – 24 .

Stake , R. E. ( 2003 ). Case studies . In N. K. , Denzin , & Y. S. , Lincoln (Eds.). Strategies of Qualitative Inquiry , London : Sage Publications . pp. 134 – 164 .

Takahashi , A. R. W. , & Semprebom , E. ( 2013 ). Resultados gerais e desafios . In A. R. W. , Takahashi (Ed.), Pesquisa qualitativa em administração: Fundamentos, métodos e usos no brasil , pp. 343 – 354 . São Paulo : Atlas .

Ragin , C. C. ( 1992 ). Introduction: Cases of “what is a case? . In H. S. , Becker , & C. C. Ragin and (Eds). What is a case? Exploring the foundations of social inquiry , pp. 1 – 18 .

Ragin , C. C. ( 2013 ). Reflections on casing and Case-Oriented research . In D , Byrne. , & C. C. Ragin (Eds.), The SAGE handbooks of Case-Based methods , London : SAGE Publications , pp. 522 – 534 .

Yin , R. K. ( 2015 ). Estudo de caso: planejamento e métodos , Porto Alegre : Bookman .

Zampier , M. A. , & Takahashi , A. R. W. ( 2014 ). Aprendizagem e competências empreendedoras: Estudo de casos de micro e pequenas empresas do setor educacional . RGO Revista Gestão Organizacional , 6 , 1 – 18 .

Acknowledgements

Author contributions: Both authors contributed equally.

Corresponding author

Related articles, all feedback is valuable.

Please share your general feedback

Report an issue or find answers to frequently asked questions

Contact Customer Support

Research Writing and Analysis

• NVivo Group and Study Sessions
• SPSS This link opens in a new window
• Statistical Analysis Group sessions
• Using Qualtrics
• Dissertation and Data Analysis Group Sessions
• Defense Schedule - Commons Calendar This link opens in a new window
• Research Process Flow Chart
• Research Alignment Chapter 1 This link opens in a new window
• Step 1: Seek Out Evidence
• Step 2: Explain
• Step 3: The Big Picture
• Step 4: Own It
• Step 5: Illustrate
• Annotated Bibliography
• Seminal Authors
• Systematic Reviews & Meta-Analyses
• How to Synthesize and Analyze
• Synthesis and Analysis Practice
• Synthesis and Analysis Group Sessions
• Problem Statement
• Purpose Statement
• Conceptual Framework
• Theoretical Framework
• Locating Theoretical and Conceptual Frameworks This link opens in a new window
• Quantitative Research Questions
• Qualitative Research Questions
• Trustworthiness of Qualitative Data
• Analysis and Coding Example- Qualitative Data
• Thematic Data Analysis in Qualitative Design
• Dissertation to Journal Article This link opens in a new window
• International Journal of Online Graduate Education (IJOGE) This link opens in a new window
• Journal of Research in Innovative Teaching & Learning (JRIT&L) This link opens in a new window

Writing a Case Study

What is a case study?

A Case study is: 

• An in-depth research design that primarily uses a qualitative methodology but sometimes​​ includes quantitative methodology.
• Used to examine an identifiable problem confirmed through research.
• Used to investigate an individual, group of people, organization, or event.
• Used to mostly answer "how" and "why" questions.

What are the different types of case studies?

Note: These are the primary case studies. As you continue to research and learn

about case studies you will begin to find a robust list of different types. 

Who are your case study participants?

What is triangulation ? 

Validity and credibility are an essential part of the case study. Therefore, the researcher should include triangulation to ensure trustworthiness while accurately reflecting what the researcher seeks to investigate.

How to write a Case Study?

When developing a case study, there are different ways you could present the information, but remember to include the five parts for your case study.

Was this resource helpful?

• << Previous: Thematic Data Analysis in Qualitative Design
• Next: Journal Article Reporting Standards (JARS) >>
• Last Updated: Sep 7, 2024 9:42 AM
• URL: https://resources.nu.edu/researchtools

Educational resources and simple solutions for your research journey

What is a Case Study in Research? Definition, Methods, and Examples

Case study methodology offers researchers an exciting opportunity to explore intricate phenomena within specific contexts using a wide range of data sources and collection methods. It is highly pertinent in health and social sciences, environmental studies, social work, education, and business studies. Its diverse applications, such as advancing theory, program evaluation, and intervention development, make it an invaluable tool for driving meaningful research and fostering positive change.[ 1]  

Table of Contents

What is a Case Study?  

A case study method involves a detailed examination of a single subject, such as an individual, group, organization, event, or community, to explore and understand complex issues in real-life contexts. By focusing on one specific case, researchers can gain a deep understanding of the factors and dynamics at play, understanding their complex relationships, which might be missed in broader, more quantitative studies.  

When to do a Case Study?  

A case study design is useful when you want to explore a phenomenon in-depth and in its natural context. Here are some examples of when to use a case study :[ 2]  

• Exploratory Research: When you want to explore a new topic or phenomenon, a case study can help you understand the subject deeply. For example , a researcher studying a newly discovered plant species might use a case study to document its characteristics and behavior.  
• Descriptive Research: If you want to describe a complex phenomenon or process, a case study can provide a detailed and comprehensive description. For instance, a case study design   could describe the experiences of a group of individuals living with a rare disease.  
• Explanatory Research: When you want to understand why a particular phenomenon occurs, a case study can help you identify causal relationships. A case study design could investigate the reasons behind the success or failure of a particular business strategy.  
• Theory Building: Case studies can also be used to develop or refine theories. By systematically analyzing a series of cases, researchers can identify patterns and relationships that can contribute to developing new theories or refining existing ones.  
• Critical Instance: Sometimes, a single case can be used to study a rare or unusual phenomenon, but it is important for theoretical or practical reasons. For example , the case of Phineas Gage, a man who survived a severe brain injury, has been widely studied to understand the relationship between the brain and behavior.  
• Comparative Analysis: Case studies can also compare different cases or contexts. A case study example involves comparing the implementation of a particular policy in different countries to understand its effectiveness and identifying best practices.  

How to Create a Case Study – Step by Step  

Step 1: select a case  .

Careful case selection ensures relevance, insight, and meaningful contribution to existing knowledge in your field. Here’s how you can choose a case study design :[ 3]  

• Define Your Objectives: Clarify the purpose of your case study and what you hope to achieve. Do you want to provide new insights, challenge existing theories, propose solutions to a problem, or explore new research directions?  
• Consider Unusual or Outlying Cases: Focus on unusual, neglected, or outlying cases that can provide unique insights.  
• Choose a Representative Case: Alternatively, select a common or representative case to exemplify a particular category, experience, or phenomenon.   
• Avoid Bias: Ensure your selection process is unbiased using random or criteria-based selection.  
• Be Clear and Specific: Clearly define the boundaries of your study design , including the scope, timeframe, and key stakeholders.   
• Ethical Considerations: Consider ethical issues, such as confidentiality and informed consent.  

Step 2: Build a Theoretical Framework  

To ensure your case study has a solid academic foundation, it’s important to build a theoretical framework:   

• Conduct a Literature Review: Identify key concepts and theories relevant to your case study .  
• Establish Connections with Theory: Connect your case study with existing theories in the field.  
• Guide Your Analysis and Interpretation: Use your theoretical framework to guide your analysis, ensuring your findings are grounded in established theories and concepts.   

Step 3: Collect Your Data  

To conduct a comprehensive case study , you can use various research methods. These include interviews, observations, primary and secondary sources analysis, surveys, and a mixed methods approach. The aim is to gather rich and diverse data to enable a detailed analysis of your case study .  

Step 4: Describe and Analyze the Case  

How you report your findings will depend on the type of research you’re conducting. Here are two approaches:   

• Structured Approach: Follows a scientific paper format, making it easier for readers to follow your argument.  
• Narrative Approach: A more exploratory style aiming to analyze meanings and implications.  

Regardless of the approach you choose, it’s important to include the following elements in your case study :   

• Contextual Details: Provide background information about the case, including relevant historical, cultural, and social factors that may have influenced the outcome.  
• Literature and Theory: Connect your case study to existing literature and theory in the field. Discuss how your findings contribute to or challenge existing knowledge.  
• Wider Patterns or Debates: Consider how your case study fits into wider patterns or debates within the field. Discuss any implications your findings may have for future research or practice.  

What Are the Benefits of a Case Study   

Case studies offer a range of benefits , making them a powerful tool in research.  

1. In-Depth Analysis  

• Comprehensive Understanding: Case studies allow researchers to thoroughly explore a subject, understanding the complexities and nuances involved.  
• Rich Data: They offer rich qualitative and sometimes quantitative data, capturing the intricacies of real-life contexts.  

2. Contextual Insight  

• Real-World Application: Case studies provide insights into real-world applications, making the findings highly relevant and practical.  
• Context-Specific: They highlight how various factors interact within a specific context, offering a detailed picture of the situation.  

3. Flexibility  

• Methodological Diversity: Case studies can use various data collection methods, including interviews, observations, document analysis, and surveys.  
• Adaptability: Researchers can adapt the case study approach to fit the specific needs and circumstances of the research.  

4. Practical Solutions  

• Actionable Insights: The detailed findings from case studies can inform practical solutions and recommendations for practitioners and policymakers.  
• Problem-Solving: They help understand the root causes of problems and devise effective strategies to address them.  

5. Unique Cases  

• Rare Phenomena: Case studies are particularly valuable for studying rare or unique cases that other research methods may not capture.  
• Detailed Documentation: They document and preserve detailed information about specific instances that might otherwise be overlooked.  

What Are the Limitations of a Case Study   

While case studies offer valuable insights and a detailed understanding of complex issues, they have several limitations .  

1. Limited Generalizability  

• Specific Context: Case studies often focus on a single case or a small number of cases, which may limit the generalization of findings to broader populations or different contexts.  
• Unique Situations: The unique characteristics of the case may not be representative of other situations, reducing the applicability of the results.  

2. Subjectivity  

• Researcher Bias: The researcher’s perspectives and interpretations can influence the analysis and conclusions, potentially introducing bias.  
• Participant Bias: Participants’ responses and behaviors may be influenced by their awareness of being studied, known as the Hawthorne effect.  

3. Time-Consuming  

• Data Collection and Analysis: Gathering detailed, in-depth data requires significant time and effort, making case studies more time-consuming than other research methods.  
• Longitudinal Studies: If the case study observes changes over time, it can become even more prolonged.  

4. Resource Intensive  

• Financial and Human Resources: Conducting comprehensive case studies may require significant financial investment and human resources, including trained researchers and participant access.  
• Access to Data: Accessing relevant and reliable data sources can be challenging, particularly in sensitive or proprietary contexts.  

5. Replication Difficulties  

• Unique Contexts: A case study’s specific and detailed context makes it difficult to replicate the study exactly, limiting the ability to validate findings through repetition.  
• Variability: Differences in contexts, researchers, and methodologies can lead to variations in findings, complicating efforts to achieve consistent results.  

By acknowledging and addressing these limitations , researchers can enhance the rigor and reliability of their case study findings.  

Key Takeaways  

Case studies are valuable in research because they provide an in-depth, contextual analysis of a single subject, event, or organization. They allow researchers to explore complex issues in real-world settings, capturing detailed qualitative and quantitative data. This method is useful for generating insights, developing theories, and offering practical solutions to problems. They are versatile, applicable in diverse fields such as business, education, and health, and can complement other research methods by providing rich, contextual evidence. However, their findings may have limited generalizability due to the focus on a specific case.  

Frequently Asked Questions  

Q: What is a case study in research?  

A case study in research is an impactful tool for gaining a deep understanding of complex issues within their real-life context. It combines various data collection methods and provides rich, detailed insights that can inform theory development and practical applications.  

Q: What are the advantages of using case studies in research?  

Case studies are a powerful research method, offering advantages such as in-depth analysis, contextual insights, flexibility, rich data, and the ability to handle complex issues. They are particularly valuable for exploring new areas, generating hypotheses, and providing detailed, illustrative examples that can inform theory and practice.  

Q: Can case studies be used in quantitative research?  

While case studies are predominantly associated with qualitative research, they can effectively incorporate quantitative methods to provide a more comprehensive analysis. A mixed-methods approach leverages qualitative and quantitative research strengths, offering a powerful tool for exploring complex issues in a real-world context. For example , a new medical treatment case study can incorporate quantitative clinical outcomes (e.g., patient recovery rates and dosage levels) along with qualitative patient interviews.  

Q: What are the key components of a case study?  

A case study typically includes several key components:   

• Introductio n, which provides an overview and sets the context by presenting the problem statement and research objectives;  
• Literature review , which connects the study to existing theories and prior research;  
• Methodology , which details the case study design , data collection methods, and analysis techniques;   
• Findings , which present the data and results, including descriptions, patterns, and themes;   
• Discussion and conclusion , which interpret the findings, discuss their implications, and offer conclusions, practical applications, limitations, and suggestions for future research.  

Together, these components ensure a comprehensive, systematic, and insightful exploration of the case.  

References  

• de Vries, K. (2020). Case study methodology. In  Critical qualitative health research  (pp. 41-52). Routledge.  
• Fidel, R. (1984). The case study method: A case study.  Library and Information Science Research ,  6 (3), 273-288.  
• Thomas, G. (2021). How to do your case study.  How to do your case study , 1-320.  

Editage All Access is a subscription-based platform that unifies the best AI tools and services designed to speed up, simplify, and streamline every step of a researcher’s journey. The Editage All Access Pack is a one-of-a-kind subscription that unlocks full access to an AI writing assistant, literature recommender, journal finder, scientific illustration tool, and exclusive discounts on professional publication services from Editage.  

Based on 22+ years of experience in academia, Editage All Access empowers researchers to put their best research forward and move closer to success. Explore our top AI Tools pack, AI Tools + Publication Services pack, or Build Your Own Plan. Find everything a researcher needs to succeed, all in one place –  Get All Access now starting at just $14 a month !    

Related Posts

Join Us for Peer Review Week 2024

How Editage All Access is Boosting Productivity for Academics in India

• Browse All Articles
• Newsletter Sign-Up

• 26 Jul 2024
• Research & Ideas

Why Great Ideas Get Stuck in Universities

Academic institutions incubate novel ideas and are important innovation partners for companies, but getting products out of universities and onto the market can be challenging. A study by Maria Roche and Justine Boudou illustrates how researchers might be getting in their own way.

• 18 Jul 2024

New Hires Lose Psychological Safety After Year One. How to Fix It.

New hires begin their roles eager to offer ideas. But research by Amy Edmondson shows how they become more reluctant to share over time. She explains how psychological safety erodes on the job and provides advice for strengthening it.

• 27 Jun 2024

Gen AI Marketing: How Some 'Gibberish' Code Can Give Products an Edge

An increasing number of consumers are turning to generative AI for buying recommendations. But if companies can subtly manipulate the technology to favor their own products, some businesses may gain unfair advantage, says Himabindu Lakkaraju.

• 09 May 2024

Called Back to the Office? How You Benefit from Ideas You Didn't Know You Were Missing

As companies continue to weigh the benefits and drawbacks of remote work, a study of how knowledge flows among academic researchers by Karim Lakhani, Eamon Duede, and colleagues offers lessons for hybrid workplaces. Does in-person work provide more opportunities for innovation than people realize?

• 06 May 2024

The Critical Minutes After a Virtual Meeting That Can Build Up or Tear Down Teams

Weak communication and misunderstandings during virtual meetings can give way to resentment and rifts when the cameras turn off. Research by Leslie Perlow probes the nuances of digital communication. She offers advice for improving remote teamwork.

• 26 Mar 2024

How Humans Outshine AI in Adapting to Change

Could artificial intelligence systems eventually perform surgeries or fly planes? First, AI will have to learn to navigate shifting conditions as well as people do. Julian De Freitas and colleagues pit humans against machines in a video game to study AI's current limits and mine insights for the real world.

• 12 Mar 2024

Publish or Perish: What the Research Says About Productivity in Academia

Universities tend to evaluate professors based on their research output, but does that measure reflect the realities of higher ed? A study of 4,300 professors by Kyle Myers, Karim Lakhani, and colleagues probes the time demands, risk appetite, and compensation of faculty.

• 24 Jan 2024

Why Boeing’s Problems with the 737 MAX Began More Than 25 Years Ago

Aggressive cost cutting and rocky leadership changes have eroded the culture at Boeing, a company once admired for its engineering rigor, says Bill George. What will it take to repair the reputational damage wrought by years of crises involving its 737 MAX?

• 19 Sep 2023

What Chandrayaan-3 Says About India's Entrepreneurial Approach to Space

India reached an unexplored part of the moon despite its limited R&D funding compared with NASA and SpaceX. Tarun Khanna discusses the significance of the landing, and the country's advancements in data and digital technology.

• 28 Mar 2023

The FDA’s Speedy Drug Approvals Are Safe: A Win-Win for Patients and Pharma Innovation

Expediting so-called breakthrough therapies has saved millions of dollars in research time without compromising drug safety or efficacy, says research by Ariel Stern, Amitabh Chandra, and colleagues. Could policymakers harness the approach to bring life-saving treatments to the market faster?

• 16 Mar 2023

Why Business Travel Still Matters in a Zoom World

Meeting in person can make all the difference for colleagues from different time zones or cultural backgrounds. A study by Prithwiraj Choudhury traces flight patterns among 5,000 airports around the world to show how business travel propels innovation.

• 13 Apr 2021
• Working Paper Summaries

Population Interference in Panel Experiments

In panel experiments, units are exposed to different interventions over time. This article introduces a unifying framework for studying panel experiments with population interference, in which a treatment assigned to one experimental unit affects another experimental unit's outcome. Findings have implications for fields as diverse as education, economics, and public health.

• 22 Feb 2021

Private and Social Returns to R&D: Drug Development and Demographics

Research and development (R&D) by pharmaceutical firms focuses disproportionately on medical conditions afflicting the elderly. The proportion of R&D spending targeting older age groups is increasing over time. Even though these investments in R&D prolong life expectancy and improve quality of life, they have little effect on measured productivity and output growth.

• 15 Dec 2020

Designing, Not Checking, for Policy Robustness: An Example with Optimal Taxation

The approach used by most economists to check academic research results is flawed for policymaking and evaluation. The authors propose an alternative method for designing economic policy analyses that might be applied to a wide range of economic policies.

• 30 Nov 2020

Short-Termism, Shareholder Payouts, and Investment in the EU

Shareholder-driven “short-termism,” as evidenced by increasing payouts to shareholders, is said to impede long-term investment in EU public firms. But a deep dive into the data reveals a different story.

• 22 Oct 2020

Estimating Causal Effects in the Presence of Partial Interference Using Multivariate Bayesian Structural Time Series Models

A case study of an Italian supermarket introducing a new pricing policy—in which it reduced prices on some brands—offers managers a new approach to reduce uncertainty. The approach is flexible and can be applied to different business problems.

• 06 Oct 2020

Design and Analysis of Switchback Experiments

This paper presents a framework for managers to design and run switchback experiments.

• 28 Sep 2020

What Can Economics Say About Alzheimer's Disease?

This essay discusses the role of market frictions and "missing medicines" in drug innovation and highlights how frameworks and toolkits of economists can help our understanding of the determinants and effects of Alzheimer's disease on health.

• 24 Aug 2020

When Do Experts Listen to Other Experts? The Role of Negative Information in Expert Evaluations for Novel Projects

Evaluators of early-stage scientific proposals tend to systematically focus on the weaknesses of proposed work rather than its strengths, according to evidence from two field experiments.

• 10 Aug 2020

COVID's Surprising Toll on Careers of Women Scientists

Women scientists and those with young children are paying a steep career price in the pandemic, according to new research by Karim Lakhani, Kyle Myers, and colleagues. Open for comment; 0 Comments.

• Open access
• Published: 27 June 2011

The case study approach

• Sarah Crowe 1 ,
• Kathrin Cresswell 2 ,
• Ann Robertson 2 ,
• Guro Huby 3 ,
• Anthony Avery 1 &
• Aziz Sheikh 2  

BMC Medical Research Methodology volume  11 , Article number:  100 ( 2011 ) Cite this article

797k Accesses

1115 Citations

42 Altmetric

Metrics details

The case study approach allows in-depth, multi-faceted explorations of complex issues in their real-life settings. The value of the case study approach is well recognised in the fields of business, law and policy, but somewhat less so in health services research. Based on our experiences of conducting several health-related case studies, we reflect on the different types of case study design, the specific research questions this approach can help answer, the data sources that tend to be used, and the particular advantages and disadvantages of employing this methodological approach. The paper concludes with key pointers to aid those designing and appraising proposals for conducting case study research, and a checklist to help readers assess the quality of case study reports.

Peer Review reports

Introduction

The case study approach is particularly useful to employ when there is a need to obtain an in-depth appreciation of an issue, event or phenomenon of interest, in its natural real-life context. Our aim in writing this piece is to provide insights into when to consider employing this approach and an overview of key methodological considerations in relation to the design, planning, analysis, interpretation and reporting of case studies.

The illustrative 'grand round', 'case report' and 'case series' have a long tradition in clinical practice and research. Presenting detailed critiques, typically of one or more patients, aims to provide insights into aspects of the clinical case and, in doing so, illustrate broader lessons that may be learnt. In research, the conceptually-related case study approach can be used, for example, to describe in detail a patient's episode of care, explore professional attitudes to and experiences of a new policy initiative or service development or more generally to 'investigate contemporary phenomena within its real-life context' [ 1 ]. Based on our experiences of conducting a range of case studies, we reflect on when to consider using this approach, discuss the key steps involved and illustrate, with examples, some of the practical challenges of attaining an in-depth understanding of a 'case' as an integrated whole. In keeping with previously published work, we acknowledge the importance of theory to underpin the design, selection, conduct and interpretation of case studies[ 2 ]. In so doing, we make passing reference to the different epistemological approaches used in case study research by key theoreticians and methodologists in this field of enquiry.

This paper is structured around the following main questions: What is a case study? What are case studies used for? How are case studies conducted? What are the potential pitfalls and how can these be avoided? We draw in particular on four of our own recently published examples of case studies (see Tables 1 , 2 , 3 and 4 ) and those of others to illustrate our discussion[ 3 – 7 ].

What is a case study?

A case study is a research approach that is used to generate an in-depth, multi-faceted understanding of a complex issue in its real-life context. It is an established research design that is used extensively in a wide variety of disciplines, particularly in the social sciences. A case study can be defined in a variety of ways (Table 5 ), the central tenet being the need to explore an event or phenomenon in depth and in its natural context. It is for this reason sometimes referred to as a "naturalistic" design; this is in contrast to an "experimental" design (such as a randomised controlled trial) in which the investigator seeks to exert control over and manipulate the variable(s) of interest.

Stake's work has been particularly influential in defining the case study approach to scientific enquiry. He has helpfully characterised three main types of case study: intrinsic , instrumental and collective [ 8 ]. An intrinsic case study is typically undertaken to learn about a unique phenomenon. The researcher should define the uniqueness of the phenomenon, which distinguishes it from all others. In contrast, the instrumental case study uses a particular case (some of which may be better than others) to gain a broader appreciation of an issue or phenomenon. The collective case study involves studying multiple cases simultaneously or sequentially in an attempt to generate a still broader appreciation of a particular issue.

These are however not necessarily mutually exclusive categories. In the first of our examples (Table 1 ), we undertook an intrinsic case study to investigate the issue of recruitment of minority ethnic people into the specific context of asthma research studies, but it developed into a instrumental case study through seeking to understand the issue of recruitment of these marginalised populations more generally, generating a number of the findings that are potentially transferable to other disease contexts[ 3 ]. In contrast, the other three examples (see Tables 2 , 3 and 4 ) employed collective case study designs to study the introduction of workforce reconfiguration in primary care, the implementation of electronic health records into hospitals, and to understand the ways in which healthcare students learn about patient safety considerations[ 4 – 6 ]. Although our study focusing on the introduction of General Practitioners with Specialist Interests (Table 2 ) was explicitly collective in design (four contrasting primary care organisations were studied), is was also instrumental in that this particular professional group was studied as an exemplar of the more general phenomenon of workforce redesign[ 4 ].

What are case studies used for?

According to Yin, case studies can be used to explain, describe or explore events or phenomena in the everyday contexts in which they occur[ 1 ]. These can, for example, help to understand and explain causal links and pathways resulting from a new policy initiative or service development (see Tables 2 and 3 , for example)[ 1 ]. In contrast to experimental designs, which seek to test a specific hypothesis through deliberately manipulating the environment (like, for example, in a randomised controlled trial giving a new drug to randomly selected individuals and then comparing outcomes with controls),[ 9 ] the case study approach lends itself well to capturing information on more explanatory ' how ', 'what' and ' why ' questions, such as ' how is the intervention being implemented and received on the ground?'. The case study approach can offer additional insights into what gaps exist in its delivery or why one implementation strategy might be chosen over another. This in turn can help develop or refine theory, as shown in our study of the teaching of patient safety in undergraduate curricula (Table 4 )[ 6 , 10 ]. Key questions to consider when selecting the most appropriate study design are whether it is desirable or indeed possible to undertake a formal experimental investigation in which individuals and/or organisations are allocated to an intervention or control arm? Or whether the wish is to obtain a more naturalistic understanding of an issue? The former is ideally studied using a controlled experimental design, whereas the latter is more appropriately studied using a case study design.

Case studies may be approached in different ways depending on the epistemological standpoint of the researcher, that is, whether they take a critical (questioning one's own and others' assumptions), interpretivist (trying to understand individual and shared social meanings) or positivist approach (orientating towards the criteria of natural sciences, such as focusing on generalisability considerations) (Table 6 ). Whilst such a schema can be conceptually helpful, it may be appropriate to draw on more than one approach in any case study, particularly in the context of conducting health services research. Doolin has, for example, noted that in the context of undertaking interpretative case studies, researchers can usefully draw on a critical, reflective perspective which seeks to take into account the wider social and political environment that has shaped the case[ 11 ].

How are case studies conducted?

Here, we focus on the main stages of research activity when planning and undertaking a case study; the crucial stages are: defining the case; selecting the case(s); collecting and analysing the data; interpreting data; and reporting the findings.

Defining the case

Carefully formulated research question(s), informed by the existing literature and a prior appreciation of the theoretical issues and setting(s), are all important in appropriately and succinctly defining the case[ 8 , 12 ]. Crucially, each case should have a pre-defined boundary which clarifies the nature and time period covered by the case study (i.e. its scope, beginning and end), the relevant social group, organisation or geographical area of interest to the investigator, the types of evidence to be collected, and the priorities for data collection and analysis (see Table 7 )[ 1 ]. A theory driven approach to defining the case may help generate knowledge that is potentially transferable to a range of clinical contexts and behaviours; using theory is also likely to result in a more informed appreciation of, for example, how and why interventions have succeeded or failed[ 13 ].

For example, in our evaluation of the introduction of electronic health records in English hospitals (Table 3 ), we defined our cases as the NHS Trusts that were receiving the new technology[ 5 ]. Our focus was on how the technology was being implemented. However, if the primary research interest had been on the social and organisational dimensions of implementation, we might have defined our case differently as a grouping of healthcare professionals (e.g. doctors and/or nurses). The precise beginning and end of the case may however prove difficult to define. Pursuing this same example, when does the process of implementation and adoption of an electronic health record system really begin or end? Such judgements will inevitably be influenced by a range of factors, including the research question, theory of interest, the scope and richness of the gathered data and the resources available to the research team.

Selecting the case(s)

The decision on how to select the case(s) to study is a very important one that merits some reflection. In an intrinsic case study, the case is selected on its own merits[ 8 ]. The case is selected not because it is representative of other cases, but because of its uniqueness, which is of genuine interest to the researchers. This was, for example, the case in our study of the recruitment of minority ethnic participants into asthma research (Table 1 ) as our earlier work had demonstrated the marginalisation of minority ethnic people with asthma, despite evidence of disproportionate asthma morbidity[ 14 , 15 ]. In another example of an intrinsic case study, Hellstrom et al.[ 16 ] studied an elderly married couple living with dementia to explore how dementia had impacted on their understanding of home, their everyday life and their relationships.

For an instrumental case study, selecting a "typical" case can work well[ 8 ]. In contrast to the intrinsic case study, the particular case which is chosen is of less importance than selecting a case that allows the researcher to investigate an issue or phenomenon. For example, in order to gain an understanding of doctors' responses to health policy initiatives, Som undertook an instrumental case study interviewing clinicians who had a range of responsibilities for clinical governance in one NHS acute hospital trust[ 17 ]. Sampling a "deviant" or "atypical" case may however prove even more informative, potentially enabling the researcher to identify causal processes, generate hypotheses and develop theory.

In collective or multiple case studies, a number of cases are carefully selected. This offers the advantage of allowing comparisons to be made across several cases and/or replication. Choosing a "typical" case may enable the findings to be generalised to theory (i.e. analytical generalisation) or to test theory by replicating the findings in a second or even a third case (i.e. replication logic)[ 1 ]. Yin suggests two or three literal replications (i.e. predicting similar results) if the theory is straightforward and five or more if the theory is more subtle. However, critics might argue that selecting 'cases' in this way is insufficiently reflexive and ill-suited to the complexities of contemporary healthcare organisations.

The selected case study site(s) should allow the research team access to the group of individuals, the organisation, the processes or whatever else constitutes the chosen unit of analysis for the study. Access is therefore a central consideration; the researcher needs to come to know the case study site(s) well and to work cooperatively with them. Selected cases need to be not only interesting but also hospitable to the inquiry [ 8 ] if they are to be informative and answer the research question(s). Case study sites may also be pre-selected for the researcher, with decisions being influenced by key stakeholders. For example, our selection of case study sites in the evaluation of the implementation and adoption of electronic health record systems (see Table 3 ) was heavily influenced by NHS Connecting for Health, the government agency that was responsible for overseeing the National Programme for Information Technology (NPfIT)[ 5 ]. This prominent stakeholder had already selected the NHS sites (through a competitive bidding process) to be early adopters of the electronic health record systems and had negotiated contracts that detailed the deployment timelines.

It is also important to consider in advance the likely burden and risks associated with participation for those who (or the site(s) which) comprise the case study. Of particular importance is the obligation for the researcher to think through the ethical implications of the study (e.g. the risk of inadvertently breaching anonymity or confidentiality) and to ensure that potential participants/participating sites are provided with sufficient information to make an informed choice about joining the study. The outcome of providing this information might be that the emotive burden associated with participation, or the organisational disruption associated with supporting the fieldwork, is considered so high that the individuals or sites decide against participation.

In our example of evaluating implementations of electronic health record systems, given the restricted number of early adopter sites available to us, we sought purposively to select a diverse range of implementation cases among those that were available[ 5 ]. We chose a mixture of teaching, non-teaching and Foundation Trust hospitals, and examples of each of the three electronic health record systems procured centrally by the NPfIT. At one recruited site, it quickly became apparent that access was problematic because of competing demands on that organisation. Recognising the importance of full access and co-operative working for generating rich data, the research team decided not to pursue work at that site and instead to focus on other recruited sites.

Collecting the data

In order to develop a thorough understanding of the case, the case study approach usually involves the collection of multiple sources of evidence, using a range of quantitative (e.g. questionnaires, audits and analysis of routinely collected healthcare data) and more commonly qualitative techniques (e.g. interviews, focus groups and observations). The use of multiple sources of data (data triangulation) has been advocated as a way of increasing the internal validity of a study (i.e. the extent to which the method is appropriate to answer the research question)[ 8 , 18 – 21 ]. An underlying assumption is that data collected in different ways should lead to similar conclusions, and approaching the same issue from different angles can help develop a holistic picture of the phenomenon (Table 2 )[ 4 ].

Brazier and colleagues used a mixed-methods case study approach to investigate the impact of a cancer care programme[ 22 ]. Here, quantitative measures were collected with questionnaires before, and five months after, the start of the intervention which did not yield any statistically significant results. Qualitative interviews with patients however helped provide an insight into potentially beneficial process-related aspects of the programme, such as greater, perceived patient involvement in care. The authors reported how this case study approach provided a number of contextual factors likely to influence the effectiveness of the intervention and which were not likely to have been obtained from quantitative methods alone.

In collective or multiple case studies, data collection needs to be flexible enough to allow a detailed description of each individual case to be developed (e.g. the nature of different cancer care programmes), before considering the emerging similarities and differences in cross-case comparisons (e.g. to explore why one programme is more effective than another). It is important that data sources from different cases are, where possible, broadly comparable for this purpose even though they may vary in nature and depth.

Analysing, interpreting and reporting case studies

Making sense and offering a coherent interpretation of the typically disparate sources of data (whether qualitative alone or together with quantitative) is far from straightforward. Repeated reviewing and sorting of the voluminous and detail-rich data are integral to the process of analysis. In collective case studies, it is helpful to analyse data relating to the individual component cases first, before making comparisons across cases. Attention needs to be paid to variations within each case and, where relevant, the relationship between different causes, effects and outcomes[ 23 ]. Data will need to be organised and coded to allow the key issues, both derived from the literature and emerging from the dataset, to be easily retrieved at a later stage. An initial coding frame can help capture these issues and can be applied systematically to the whole dataset with the aid of a qualitative data analysis software package.

The Framework approach is a practical approach, comprising of five stages (familiarisation; identifying a thematic framework; indexing; charting; mapping and interpretation) , to managing and analysing large datasets particularly if time is limited, as was the case in our study of recruitment of South Asians into asthma research (Table 1 )[ 3 , 24 ]. Theoretical frameworks may also play an important role in integrating different sources of data and examining emerging themes. For example, we drew on a socio-technical framework to help explain the connections between different elements - technology; people; and the organisational settings within which they worked - in our study of the introduction of electronic health record systems (Table 3 )[ 5 ]. Our study of patient safety in undergraduate curricula drew on an evaluation-based approach to design and analysis, which emphasised the importance of the academic, organisational and practice contexts through which students learn (Table 4 )[ 6 ].

Case study findings can have implications both for theory development and theory testing. They may establish, strengthen or weaken historical explanations of a case and, in certain circumstances, allow theoretical (as opposed to statistical) generalisation beyond the particular cases studied[ 12 ]. These theoretical lenses should not, however, constitute a strait-jacket and the cases should not be "forced to fit" the particular theoretical framework that is being employed.

When reporting findings, it is important to provide the reader with enough contextual information to understand the processes that were followed and how the conclusions were reached. In a collective case study, researchers may choose to present the findings from individual cases separately before amalgamating across cases. Care must be taken to ensure the anonymity of both case sites and individual participants (if agreed in advance) by allocating appropriate codes or withholding descriptors. In the example given in Table 3 , we decided against providing detailed information on the NHS sites and individual participants in order to avoid the risk of inadvertent disclosure of identities[ 5 , 25 ].

What are the potential pitfalls and how can these be avoided?

The case study approach is, as with all research, not without its limitations. When investigating the formal and informal ways undergraduate students learn about patient safety (Table 4 ), for example, we rapidly accumulated a large quantity of data. The volume of data, together with the time restrictions in place, impacted on the depth of analysis that was possible within the available resources. This highlights a more general point of the importance of avoiding the temptation to collect as much data as possible; adequate time also needs to be set aside for data analysis and interpretation of what are often highly complex datasets.

Case study research has sometimes been criticised for lacking scientific rigour and providing little basis for generalisation (i.e. producing findings that may be transferable to other settings)[ 1 ]. There are several ways to address these concerns, including: the use of theoretical sampling (i.e. drawing on a particular conceptual framework); respondent validation (i.e. participants checking emerging findings and the researcher's interpretation, and providing an opinion as to whether they feel these are accurate); and transparency throughout the research process (see Table 8 )[ 8 , 18 – 21 , 23 , 26 ]. Transparency can be achieved by describing in detail the steps involved in case selection, data collection, the reasons for the particular methods chosen, and the researcher's background and level of involvement (i.e. being explicit about how the researcher has influenced data collection and interpretation). Seeking potential, alternative explanations, and being explicit about how interpretations and conclusions were reached, help readers to judge the trustworthiness of the case study report. Stake provides a critique checklist for a case study report (Table 9 )[ 8 ].

Conclusions

The case study approach allows, amongst other things, critical events, interventions, policy developments and programme-based service reforms to be studied in detail in a real-life context. It should therefore be considered when an experimental design is either inappropriate to answer the research questions posed or impossible to undertake. Considering the frequency with which implementations of innovations are now taking place in healthcare settings and how well the case study approach lends itself to in-depth, complex health service research, we believe this approach should be more widely considered by researchers. Though inherently challenging, the research case study can, if carefully conceptualised and thoughtfully undertaken and reported, yield powerful insights into many important aspects of health and healthcare delivery.

Yin RK: Case study research, design and method. 2009, London: Sage Publications Ltd., 4

Google Scholar  

Keen J, Packwood T: Qualitative research; case study evaluation. BMJ. 1995, 311: 444-446.

Article   CAS   PubMed   PubMed Central   Google Scholar  

Sheikh A, Halani L, Bhopal R, Netuveli G, Partridge M, Car J, et al: Facilitating the Recruitment of Minority Ethnic People into Research: Qualitative Case Study of South Asians and Asthma. PLoS Med. 2009, 6 (10): 1-11.

Article   Google Scholar  

Pinnock H, Huby G, Powell A, Kielmann T, Price D, Williams S, et al: The process of planning, development and implementation of a General Practitioner with a Special Interest service in Primary Care Organisations in England and Wales: a comparative prospective case study. Report for the National Co-ordinating Centre for NHS Service Delivery and Organisation R&D (NCCSDO). 2008, [ http://www.sdo.nihr.ac.uk/files/project/99-final-report.pdf ]

Robertson A, Cresswell K, Takian A, Petrakaki D, Crowe S, Cornford T, et al: Prospective evaluation of the implementation and adoption of NHS Connecting for Health's national electronic health record in secondary care in England: interim findings. BMJ. 2010, 41: c4564-

Pearson P, Steven A, Howe A, Sheikh A, Ashcroft D, Smith P, the Patient Safety Education Study Group: Learning about patient safety: organisational context and culture in the education of healthcare professionals. J Health Serv Res Policy. 2010, 15: 4-10. 10.1258/jhsrp.2009.009052.

Article   PubMed   Google Scholar  

van Harten WH, Casparie TF, Fisscher OA: The evaluation of the introduction of a quality management system: a process-oriented case study in a large rehabilitation hospital. Health Policy. 2002, 60 (1): 17-37. 10.1016/S0168-8510(01)00187-7.

Stake RE: The art of case study research. 1995, London: Sage Publications Ltd.

Sheikh A, Smeeth L, Ashcroft R: Randomised controlled trials in primary care: scope and application. Br J Gen Pract. 2002, 52 (482): 746-51.

PubMed   PubMed Central   Google Scholar  

King G, Keohane R, Verba S: Designing Social Inquiry. 1996, Princeton: Princeton University Press

Doolin B: Information technology as disciplinary technology: being critical in interpretative research on information systems. Journal of Information Technology. 1998, 13: 301-311. 10.1057/jit.1998.8.

George AL, Bennett A: Case studies and theory development in the social sciences. 2005, Cambridge, MA: MIT Press

Eccles M, the Improved Clinical Effectiveness through Behavioural Research Group (ICEBeRG): Designing theoretically-informed implementation interventions. Implementation Science. 2006, 1: 1-8. 10.1186/1748-5908-1-1.

Article   PubMed Central   Google Scholar  

Netuveli G, Hurwitz B, Levy M, Fletcher M, Barnes G, Durham SR, Sheikh A: Ethnic variations in UK asthma frequency, morbidity, and health-service use: a systematic review and meta-analysis. Lancet. 2005, 365 (9456): 312-7.

Sheikh A, Panesar SS, Lasserson T, Netuveli G: Recruitment of ethnic minorities to asthma studies. Thorax. 2004, 59 (7): 634-

CAS   PubMed   PubMed Central   Google Scholar  

Hellström I, Nolan M, Lundh U: 'We do things together': A case study of 'couplehood' in dementia. Dementia. 2005, 4: 7-22. 10.1177/1471301205049188.

Som CV: Nothing seems to have changed, nothing seems to be changing and perhaps nothing will change in the NHS: doctors' response to clinical governance. International Journal of Public Sector Management. 2005, 18: 463-477. 10.1108/09513550510608903.

Lincoln Y, Guba E: Naturalistic inquiry. 1985, Newbury Park: Sage Publications

Barbour RS: Checklists for improving rigour in qualitative research: a case of the tail wagging the dog?. BMJ. 2001, 322: 1115-1117. 10.1136/bmj.322.7294.1115.

Mays N, Pope C: Qualitative research in health care: Assessing quality in qualitative research. BMJ. 2000, 320: 50-52. 10.1136/bmj.320.7226.50.

Mason J: Qualitative researching. 2002, London: Sage

Brazier A, Cooke K, Moravan V: Using Mixed Methods for Evaluating an Integrative Approach to Cancer Care: A Case Study. Integr Cancer Ther. 2008, 7: 5-17. 10.1177/1534735407313395.

Miles MB, Huberman M: Qualitative data analysis: an expanded sourcebook. 1994, CA: Sage Publications Inc., 2

Pope C, Ziebland S, Mays N: Analysing qualitative data. Qualitative research in health care. BMJ. 2000, 320: 114-116. 10.1136/bmj.320.7227.114.

Cresswell KM, Worth A, Sheikh A: Actor-Network Theory and its role in understanding the implementation of information technology developments in healthcare. BMC Med Inform Decis Mak. 2010, 10 (1): 67-10.1186/1472-6947-10-67.

Article   PubMed   PubMed Central   Google Scholar  

Malterud K: Qualitative research: standards, challenges, and guidelines. Lancet. 2001, 358: 483-488. 10.1016/S0140-6736(01)05627-6.

Article   CAS   PubMed   Google Scholar  

Yin R: Case study research: design and methods. 1994, Thousand Oaks, CA: Sage Publishing, 2

Yin R: Enhancing the quality of case studies in health services research. Health Serv Res. 1999, 34: 1209-1224.

Green J, Thorogood N: Qualitative methods for health research. 2009, Los Angeles: Sage, 2

Howcroft D, Trauth E: Handbook of Critical Information Systems Research, Theory and Application. 2005, Cheltenham, UK: Northampton, MA, USA: Edward Elgar

Book   Google Scholar  

Blakie N: Approaches to Social Enquiry. 1993, Cambridge: Polity Press

Doolin B: Power and resistance in the implementation of a medical management information system. Info Systems J. 2004, 14: 343-362. 10.1111/j.1365-2575.2004.00176.x.

Bloomfield BP, Best A: Management consultants: systems development, power and the translation of problems. Sociological Review. 1992, 40: 533-560.

Shanks G, Parr A: Positivist, single case study research in information systems: A critical analysis. Proceedings of the European Conference on Information Systems. 2003, Naples

Pre-publication history

The pre-publication history for this paper can be accessed here: http://www.biomedcentral.com/1471-2288/11/100/prepub

Download references

Acknowledgements

We are grateful to the participants and colleagues who contributed to the individual case studies that we have drawn on. This work received no direct funding, but it has been informed by projects funded by Asthma UK, the NHS Service Delivery Organisation, NHS Connecting for Health Evaluation Programme, and Patient Safety Research Portfolio. We would also like to thank the expert reviewers for their insightful and constructive feedback. Our thanks are also due to Dr. Allison Worth who commented on an earlier draft of this manuscript.

Author information

Authors and affiliations.

Division of Primary Care, The University of Nottingham, Nottingham, UK

Sarah Crowe & Anthony Avery

Centre for Population Health Sciences, The University of Edinburgh, Edinburgh, UK

Kathrin Cresswell, Ann Robertson & Aziz Sheikh

School of Health in Social Science, The University of Edinburgh, Edinburgh, UK

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Sarah Crowe .

Additional information

Competing interests.

The authors declare that they have no competing interests.

Authors' contributions

AS conceived this article. SC, KC and AR wrote this paper with GH, AA and AS all commenting on various drafts. SC and AS are guarantors.

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/2.0 ), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Reprints and permissions

About this article

Cite this article.

Crowe, S., Cresswell, K., Robertson, A. et al. The case study approach. BMC Med Res Methodol 11 , 100 (2011). https://doi.org/10.1186/1471-2288-11-100

Download citation

Received : 29 November 2010

Accepted : 27 June 2011

Published : 27 June 2011

DOI : https://doi.org/10.1186/1471-2288-11-100

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

• Case Study Approach
• Electronic Health Record System
• Case Study Design
• Case Study Site
• Case Study Report

BMC Medical Research Methodology

ISSN: 1471-2288

• General enquiries: [email protected]

• Open access
• Published: 10 November 2020

Case study research for better evaluations of complex interventions: rationale and challenges

• Sara Paparini   ORCID: orcid.org/0000-0002-1909-2481 1 ,
• Judith Green 2 ,
• Chrysanthi Papoutsi 1 ,
• Jamie Murdoch 3 ,
• Mark Petticrew 4 ,
• Trish Greenhalgh 1 ,
• Benjamin Hanckel 5 &
• Sara Shaw 1  

BMC Medicine volume  18 , Article number:  301 ( 2020 ) Cite this article

19k Accesses

47 Citations

35 Altmetric

Metrics details

The need for better methods for evaluation in health research has been widely recognised. The ‘complexity turn’ has drawn attention to the limitations of relying on causal inference from randomised controlled trials alone for understanding whether, and under which conditions, interventions in complex systems improve health services or the public health, and what mechanisms might link interventions and outcomes. We argue that case study research—currently denigrated as poor evidence—is an under-utilised resource for not only providing evidence about context and transferability, but also for helping strengthen causal inferences when pathways between intervention and effects are likely to be non-linear.

Case study research, as an overall approach, is based on in-depth explorations of complex phenomena in their natural, or real-life, settings. Empirical case studies typically enable dynamic understanding of complex challenges and provide evidence about causal mechanisms and the necessary and sufficient conditions (contexts) for intervention implementation and effects. This is essential evidence not just for researchers concerned about internal and external validity, but also research users in policy and practice who need to know what the likely effects of complex programmes or interventions will be in their settings. The health sciences have much to learn from scholarship on case study methodology in the social sciences. However, there are multiple challenges in fully exploiting the potential learning from case study research. First are misconceptions that case study research can only provide exploratory or descriptive evidence. Second, there is little consensus about what a case study is, and considerable diversity in how empirical case studies are conducted and reported. Finally, as case study researchers typically (and appropriately) focus on thick description (that captures contextual detail), it can be challenging to identify the key messages related to intervention evaluation from case study reports.

Whilst the diversity of published case studies in health services and public health research is rich and productive, we recommend further clarity and specific methodological guidance for those reporting case study research for evaluation audiences.

Peer Review reports

The need for methodological development to address the most urgent challenges in health research has been well-documented. Many of the most pressing questions for public health research, where the focus is on system-level determinants [ 1 , 2 ], and for health services research, where provisions typically vary across sites and are provided through interlocking networks of services [ 3 ], require methodological approaches that can attend to complexity. The need for methodological advance has arisen, in part, as a result of the diminishing returns from randomised controlled trials (RCTs) where they have been used to answer questions about the effects of interventions in complex systems [ 4 , 5 , 6 ]. In conditions of complexity, there is limited value in maintaining the current orientation to experimental trial designs in the health sciences as providing ‘gold standard’ evidence of effect.

There are increasing calls for methodological pluralism [ 7 , 8 ], with the recognition that complex intervention and context are not easily or usefully separated (as is often the situation when using trial design), and that system interruptions may have effects that are not reducible to linear causal pathways between intervention and outcome. These calls are reflected in a shifting and contested discourse of trial design, seen with the emergence of realist [ 9 ], adaptive and hybrid (types 1, 2 and 3) [ 10 , 11 ] trials that blend studies of effectiveness with a close consideration of the contexts of implementation. Similarly, process evaluation has now become a core component of complex healthcare intervention trials, reflected in MRC guidance on how to explore implementation, causal mechanisms and context [ 12 ].

Evidence about the context of an intervention is crucial for questions of external validity. As Woolcock [ 4 ] notes, even if RCT designs are accepted as robust for maximising internal validity, questions of transferability (how well the intervention works in different contexts) and generalisability (how well the intervention can be scaled up) remain unanswered [ 5 , 13 ]. For research evidence to have impact on policy and systems organisation, and thus to improve population and patient health, there is an urgent need for better methods for strengthening external validity, including a better understanding of the relationship between intervention and context [ 14 ].

Policymakers, healthcare commissioners and other research users require credible evidence of relevance to their settings and populations [ 15 ], to perform what Rosengarten and Savransky [ 16 ] call ‘careful abstraction’ to the locales that matter for them. They also require robust evidence for understanding complex causal pathways. Case study research, currently under-utilised in public health and health services evaluation, can offer considerable potential for strengthening faith in both external and internal validity. For example, in an empirical case study of how the policy of free bus travel had specific health effects in London, UK, a quasi-experimental evaluation (led by JG) identified how important aspects of context (a good public transport system) and intervention (that it was universal) were necessary conditions for the observed effects, thus providing useful, actionable evidence for decision-makers in other contexts [ 17 ].

The overall approach of case study research is based on the in-depth exploration of complex phenomena in their natural, or ‘real-life’, settings. Empirical case studies typically enable dynamic understanding of complex challenges rather than restricting the focus on narrow problem delineations and simple fixes. Case study research is a diverse and somewhat contested field, with multiple definitions and perspectives grounded in different ways of viewing the world, and involving different combinations of methods. In this paper, we raise awareness of such plurality and highlight the contribution that case study research can make to the evaluation of complex system-level interventions. We review some of the challenges in exploiting the current evidence base from empirical case studies and conclude by recommending that further guidance and minimum reporting criteria for evaluation using case studies, appropriate for audiences in the health sciences, can enhance the take-up of evidence from case study research.

Case study research offers evidence about context, causal inference in complex systems and implementation

Well-conducted and described empirical case studies provide evidence on context, complexity and mechanisms for understanding how, where and why interventions have their observed effects. Recognition of the importance of context for understanding the relationships between interventions and outcomes is hardly new. In 1943, Canguilhem berated an over-reliance on experimental designs for determining universal physiological laws: ‘As if one could determine a phenomenon’s essence apart from its conditions! As if conditions were a mask or frame which changed neither the face nor the picture!’ ([ 18 ] p126). More recently, a concern with context has been expressed in health systems and public health research as part of what has been called the ‘complexity turn’ [ 1 ]: a recognition that many of the most enduring challenges for developing an evidence base require a consideration of system-level effects [ 1 ] and the conceptualisation of interventions as interruptions in systems [ 19 ].

The case study approach is widely recognised as offering an invaluable resource for understanding the dynamic and evolving influence of context on complex, system-level interventions [ 20 , 21 , 22 , 23 ]. Empirically, case studies can directly inform assessments of where, when, how and for whom interventions might be successfully implemented, by helping to specify the necessary and sufficient conditions under which interventions might have effects and to consolidate learning on how interdependencies, emergence and unpredictability can be managed to achieve and sustain desired effects. Case study research has the potential to address four objectives for improving research and reporting of context recently set out by guidance on taking account of context in population health research [ 24 ], that is to (1) improve the appropriateness of intervention development for specific contexts, (2) improve understanding of ‘how’ interventions work, (3) better understand how and why impacts vary across contexts and (4) ensure reports of intervention studies are most useful for decision-makers and researchers.

However, evaluations of complex healthcare interventions have arguably not exploited the full potential of case study research and can learn much from other disciplines. For evaluative research, exploratory case studies have had a traditional role of providing data on ‘process’, or initial ‘hypothesis-generating’ scoping, but might also have an increasing salience for explanatory aims. Across the social and political sciences, different kinds of case studies are undertaken to meet diverse aims (description, exploration or explanation) and across different scales (from small N qualitative studies that aim to elucidate processes, or provide thick description, to more systematic techniques designed for medium-to-large N cases).

Case studies with explanatory aims vary in terms of their positioning within mixed-methods projects, with designs including (but not restricted to) (1) single N of 1 studies of interventions in specific contexts, where the overall design is a case study that may incorporate one or more (randomised or not) comparisons over time and between variables within the case; (2) a series of cases conducted or synthesised to provide explanation from variations between cases; and (3) case studies of particular settings within RCT or quasi-experimental designs to explore variation in effects or implementation.

Detailed qualitative research (typically done as ‘case studies’ within process evaluations) provides evidence for the plausibility of mechanisms [ 25 ], offering theoretical generalisations for how interventions may function under different conditions. Although RCT designs reduce many threats to internal validity, the mechanisms of effect remain opaque, particularly when the causal pathways between ‘intervention’ and ‘effect’ are long and potentially non-linear: case study research has a more fundamental role here, in providing detailed observational evidence for causal claims [ 26 ] as well as producing a rich, nuanced picture of tensions and multiple perspectives [ 8 ].

Longitudinal or cross-case analysis may be best suited for evidence generation in system-level evaluative research. Turner [ 27 ], for instance, reflecting on the complex processes in major system change, has argued for the need for methods that integrate learning across cases, to develop theoretical knowledge that would enable inferences beyond the single case, and to develop generalisable theory about organisational and structural change in health systems. Qualitative Comparative Analysis (QCA) [ 28 ] is one such formal method for deriving causal claims, using set theory mathematics to integrate data from empirical case studies to answer questions about the configurations of causal pathways linking conditions to outcomes [ 29 , 30 ].

Nonetheless, the single N case study, too, provides opportunities for theoretical development [ 31 ], and theoretical generalisation or analytical refinement [ 32 ]. How ‘the case’ and ‘context’ are conceptualised is crucial here. Findings from the single case may seem to be confined to its intrinsic particularities in a specific and distinct context [ 33 ]. However, if such context is viewed as exemplifying wider social and political forces, the single case can be ‘telling’, rather than ‘typical’, and offer insight into a wider issue [ 34 ]. Internal comparisons within the case can offer rich possibilities for logical inferences about causation [ 17 ]. Further, case studies of any size can be used for theory testing through refutation [ 22 ]. The potential lies, then, in utilising the strengths and plurality of case study to support theory-driven research within different methodological paradigms.

Evaluation research in health has much to learn from a range of social sciences where case study methodology has been used to develop various kinds of causal inference. For instance, Gerring [ 35 ] expands on the within-case variations utilised to make causal claims. For Gerring [ 35 ], case studies come into their own with regard to invariant or strong causal claims (such as X is a necessary and/or sufficient condition for Y) rather than for probabilistic causal claims. For the latter (where experimental methods might have an advantage in estimating effect sizes), case studies offer evidence on mechanisms: from observations of X affecting Y, from process tracing or from pattern matching. Case studies also support the study of emergent causation, that is, the multiple interacting properties that account for particular and unexpected outcomes in complex systems, such as in healthcare [ 8 ].

Finally, efficacy (or beliefs about efficacy) is not the only contributor to intervention uptake, with a range of organisational and policy contingencies affecting whether an intervention is likely to be rolled out in practice. Case study research is, therefore, invaluable for learning about contextual contingencies and identifying the conditions necessary for interventions to become normalised (i.e. implemented routinely) in practice [ 36 ].

The challenges in exploiting evidence from case study research

At present, there are significant challenges in exploiting the benefits of case study research in evaluative health research, which relate to status, definition and reporting. Case study research has been marginalised at the bottom of an evidence hierarchy, seen to offer little by way of explanatory power, if nonetheless useful for adding descriptive data on process or providing useful illustrations for policymakers [ 37 ]. This is an opportune moment to revisit this low status. As health researchers are increasingly charged with evaluating ‘natural experiments’—the use of face masks in the response to the COVID-19 pandemic being a recent example [ 38 ]—rather than interventions that take place in settings that can be controlled, research approaches using methods to strengthen causal inference that does not require randomisation become more relevant.

A second challenge for improving the use of case study evidence in evaluative health research is that, as we have seen, what is meant by ‘case study’ varies widely, not only across but also within disciplines. There is indeed little consensus amongst methodologists as to how to define ‘a case study’. Definitions focus, variously, on small sample size or lack of control over the intervention (e.g. [ 39 ] p194), on in-depth study and context [ 40 , 41 ], on the logic of inference used [ 35 ] or on distinct research strategies which incorporate a number of methods to address questions of ‘how’ and ‘why’ [ 42 ]. Moreover, definitions developed for specific disciplines do not capture the range of ways in which case study research is carried out across disciplines. Multiple definitions of case study reflect the richness and diversity of the approach. However, evidence suggests that a lack of consensus across methodologists results in some of the limitations of published reports of empirical case studies [ 43 , 44 ]. Hyett and colleagues [ 43 ], for instance, reviewing reports in qualitative journals, found little match between methodological definitions of case study research and how authors used the term.

This raises the third challenge we identify that case study reports are typically not written in ways that are accessible or useful for the evaluation research community and policymakers. Case studies may not appear in journals widely read by those in the health sciences, either because space constraints preclude the reporting of rich, thick descriptions, or because of the reported lack of willingness of some biomedical journals to publish research that uses qualitative methods [ 45 ], signalling the persistence of the aforementioned evidence hierarchy. Where they do, however, the term ‘case study’ is used to indicate, interchangeably, a qualitative study, an N of 1 sample, or a multi-method, in-depth analysis of one example from a population of phenomena. Definitions of what constitutes the ‘case’ are frequently lacking and appear to be used as a synonym for the settings in which the research is conducted. Despite offering insights for evaluation, the primary aims may not have been evaluative, so the implications may not be explicitly drawn out. Indeed, some case study reports might properly be aiming for thick description without necessarily seeking to inform about context or causality.

Acknowledging plurality and developing guidance

We recognise that definitional and methodological plurality is not only inevitable, but also a necessary and creative reflection of the very different epistemological and disciplinary origins of health researchers, and the aims they have in doing and reporting case study research. Indeed, to provide some clarity, Thomas [ 46 ] has suggested a typology of subject/purpose/approach/process for classifying aims (e.g. evaluative or exploratory), sample rationale and selection and methods for data generation of case studies. We also recognise that the diversity of methods used in case study research, and the necessary focus on narrative reporting, does not lend itself to straightforward development of formal quality or reporting criteria.

Existing checklists for reporting case study research from the social sciences—for example Lincoln and Guba’s [ 47 ] and Stake’s [ 33 ]—are primarily orientated to the quality of narrative produced, and the extent to which they encapsulate thick description, rather than the more pragmatic issues of implications for intervention effects. Those designed for clinical settings, such as the CARE (CAse REports) guidelines, provide specific reporting guidelines for medical case reports about single, or small groups of patients [ 48 ], not for case study research.

The Design of Case Study Research in Health Care (DESCARTE) model [ 44 ] suggests a series of questions to be asked of a case study researcher (including clarity about the philosophy underpinning their research), study design (with a focus on case definition) and analysis (to improve process). The model resembles toolkits for enhancing the quality and robustness of qualitative and mixed-methods research reporting, and it is usefully open-ended and non-prescriptive. However, even if it does include some reflections on context, the model does not fully address aspects of context, logic and causal inference that are perhaps most relevant for evaluative research in health.

Hence, for evaluative research where the aim is to report empirical findings in ways that are intended to be pragmatically useful for health policy and practice, this may be an opportune time to consider how to best navigate plurality around what is (minimally) important to report when publishing empirical case studies, especially with regards to the complex relationships between context and interventions, information that case study research is well placed to provide.

The conventional scientific quest for certainty, predictability and linear causality (maximised in RCT designs) has to be augmented by the study of uncertainty, unpredictability and emergent causality [ 8 ] in complex systems. This will require methodological pluralism, and openness to broadening the evidence base to better understand both causality in and the transferability of system change intervention [ 14 , 20 , 23 , 25 ]. Case study research evidence is essential, yet is currently under exploited in the health sciences. If evaluative health research is to move beyond the current impasse on methods for understanding interventions as interruptions in complex systems, we need to consider in more detail how researchers can conduct and report empirical case studies which do aim to elucidate the contextual factors which interact with interventions to produce particular effects. To this end, supported by the UK’s Medical Research Council, we are embracing the challenge to develop guidance for case study researchers studying complex interventions. Following a meta-narrative review of the literature, we are planning a Delphi study to inform guidance that will, at minimum, cover the value of case study research for evaluating the interrelationship between context and complex system-level interventions; for situating and defining ‘the case’, and generalising from case studies; as well as provide specific guidance on conducting, analysing and reporting case study research. Our hope is that such guidance can support researchers evaluating interventions in complex systems to better exploit the diversity and richness of case study research.

Availability of data and materials

Not applicable (article based on existing available academic publications)

Abbreviations

Qualitative comparative analysis

Quasi-experimental design

Randomised controlled trial

Diez Roux AV. Complex systems thinking and current impasses in health disparities research. Am J Public Health. 2011;101(9):1627–34.

Article   Google Scholar  

Ogilvie D, Mitchell R, Mutrie N, M P, Platt S. Evaluating health effects of transport interventions: methodologic case study. Am J Prev Med 2006;31:118–126.

Walshe C. The evaluation of complex interventions in palliative care: an exploration of the potential of case study research strategies. Palliat Med. 2011;25(8):774–81.

Woolcock M. Using case studies to explore the external validity of ‘complex’ development interventions. Evaluation. 2013;19:229–48.

Cartwright N. Are RCTs the gold standard? BioSocieties. 2007;2(1):11–20.

Deaton A, Cartwright N. Understanding and misunderstanding randomized controlled trials. Soc Sci Med. 2018;210:2–21.

Salway S, Green J. Towards a critical complex systems approach to public health. Crit Public Health. 2017;27(5):523–4.

Greenhalgh T, Papoutsi C. Studying complexity in health services research: desperately seeking an overdue paradigm shift. BMC Med. 2018;16(1):95.

Bonell C, Warren E, Fletcher A. Realist trials and the testing of context-mechanism-outcome configurations: a response to Van Belle et al. Trials. 2016;17:478.

Pallmann P, Bedding AW, Choodari-Oskooei B. Adaptive designs in clinical trials: why use them, and how to run and report them. BMC Med. 2018;16:29.

Curran G, Bauer M, Mittman B, Pyne J, Stetler C. Effectiveness-implementation hybrid designs: combining elements of clinical effectiveness and implementation research to enhance public health impact. Med Care. 2012;50(3):217–26. https://doi.org/10.1097/MLR.0b013e3182408812 .

Moore GF, Audrey S, Barker M, Bond L, Bonell C, Hardeman W, et al. Process evaluation of complex interventions: Medical Research Council guidance. BMJ. 2015 [cited 2020 Jun 27];350. Available from: https://www.bmj.com/content/350/bmj.h1258 .

Evans RE, Craig P, Hoddinott P, Littlecott H, Moore L, Murphy S, et al. When and how do ‘effective’ interventions need to be adapted and/or re-evaluated in new contexts? The need for guidance. J Epidemiol Community Health. 2019;73(6):481–2.

Shoveller J. A critical examination of representations of context within research on population health interventions. Crit Public Health. 2016;26(5):487–500.

Treweek S, Zwarenstein M. Making trials matter: pragmatic and explanatory trials and the problem of applicability. Trials. 2009;10(1):37.

Rosengarten M, Savransky M. A careful biomedicine? Generalization and abstraction in RCTs. Crit Public Health. 2019;29(2):181–91.

Green J, Roberts H, Petticrew M, Steinbach R, Goodman A, Jones A, et al. Integrating quasi-experimental and inductive designs in evaluation: a case study of the impact of free bus travel on public health. Evaluation. 2015;21(4):391–406.

Canguilhem G. The normal and the pathological. New York: Zone Books; 1991. (1949).

Google Scholar  

Hawe P, Shiell A, Riley T. Theorising interventions as events in systems. Am J Community Psychol. 2009;43:267–76.

King G, Keohane RO, Verba S. Designing social inquiry: scientific inference in qualitative research: Princeton University Press; 1994.

Greenhalgh T, Robert G, Macfarlane F, Bate P, Kyriakidou O. Diffusion of innovations in service organizations: systematic review and recommendations. Milbank Q. 2004;82(4):581–629.

Yin R. Enhancing the quality of case studies in health services research. Health Serv Res. 1999;34(5 Pt 2):1209.

CAS   PubMed   PubMed Central   Google Scholar  

Raine R, Fitzpatrick R, Barratt H, Bevan G, Black N, Boaden R, et al. Challenges, solutions and future directions in the evaluation of service innovations in health care and public health. Health Serv Deliv Res. 2016 [cited 2020 Jun 30];4(16). Available from: https://www.journalslibrary.nihr.ac.uk/hsdr/hsdr04160#/abstract .

Craig P, Di Ruggiero E, Frohlich KL, E M, White M, Group CCGA. Taking account of context in population health intervention research: guidance for producers, users and funders of research. NIHR Evaluation, Trials and Studies Coordinating Centre; 2018.

Grant RL, Hood R. Complex systems, explanation and policy: implications of the crisis of replication for public health research. Crit Public Health. 2017;27(5):525–32.

Mahoney J. Strategies of causal inference in small-N analysis. Sociol Methods Res. 2000;4:387–424.

Turner S. Major system change: a management and organisational research perspective. In: Rosalind Raine, Ray Fitzpatrick, Helen Barratt, Gywn Bevan, Nick Black, Ruth Boaden, et al. Challenges, solutions and future directions in the evaluation of service innovations in health care and public health. Health Serv Deliv Res. 2016;4(16) 2016. https://doi.org/10.3310/hsdr04160.

Ragin CC. Using qualitative comparative analysis to study causal complexity. Health Serv Res. 1999;34(5 Pt 2):1225.

Hanckel B, Petticrew M, Thomas J, Green J. Protocol for a systematic review of the use of qualitative comparative analysis for evaluative questions in public health research. Syst Rev. 2019;8(1):252.

Schneider CQ, Wagemann C. Set-theoretic methods for the social sciences: a guide to qualitative comparative analysis: Cambridge University Press; 2012. 369 p.

Flyvbjerg B. Five misunderstandings about case-study research. Qual Inq. 2006;12:219–45.

Tsoukas H. Craving for generality and small-N studies: a Wittgensteinian approach towards the epistemology of the particular in organization and management studies. Sage Handb Organ Res Methods. 2009:285–301.

Stake RE. The art of case study research. London: Sage Publications Ltd; 1995.

Mitchell JC. Typicality and the case study. Ethnographic research: A guide to general conduct. Vol. 238241. 1984.

Gerring J. What is a case study and what is it good for? Am Polit Sci Rev. 2004;98(2):341–54.

May C, Mort M, Williams T, F M, Gask L. Health technology assessment in its local contexts: studies of telehealthcare. Soc Sci Med 2003;57:697–710.

McGill E. Trading quality for relevance: non-health decision-makers’ use of evidence on the social determinants of health. BMJ Open. 2015;5(4):007053.

Greenhalgh T. We can’t be 100% sure face masks work – but that shouldn’t stop us wearing them | Trish Greenhalgh. The Guardian. 2020 [cited 2020 Jun 27]; Available from: https://www.theguardian.com/commentisfree/2020/jun/05/face-masks-coronavirus .

Hammersley M. So, what are case studies? In: What’s wrong with ethnography? New York: Routledge; 1992.

Crowe S, Cresswell K, Robertson A, Huby G, Avery A, Sheikh A. The case study approach. BMC Med Res Methodol. 2011;11(1):100.

Luck L, Jackson D, Usher K. Case study: a bridge across the paradigms. Nurs Inq. 2006;13(2):103–9.

Yin RK. Case study research and applications: design and methods: Sage; 2017.

Hyett N, A K, Dickson-Swift V. Methodology or method? A critical review of qualitative case study reports. Int J Qual Stud Health Well-Being. 2014;9:23606.

Carolan CM, Forbat L, Smith A. Developing the DESCARTE model: the design of case study research in health care. Qual Health Res. 2016;26(5):626–39.

Greenhalgh T, Annandale E, Ashcroft R, Barlow J, Black N, Bleakley A, et al. An open letter to the BMJ editors on qualitative research. Bmj. 2016;352.

Thomas G. A typology for the case study in social science following a review of definition, discourse, and structure. Qual Inq. 2011;17(6):511–21.

Lincoln YS, Guba EG. Judging the quality of case study reports. Int J Qual Stud Educ. 1990;3(1):53–9.

Riley DS, Barber MS, Kienle GS, Aronson JK, Schoen-Angerer T, Tugwell P, et al. CARE guidelines for case reports: explanation and elaboration document. J Clin Epidemiol. 2017;89:218–35.

Download references

Acknowledgements

Not applicable

This work was funded by the Medical Research Council - MRC Award MR/S014632/1 HCS: Case study, Context and Complex interventions (TRIPLE C). SP was additionally funded by the University of Oxford's Higher Education Innovation Fund (HEIF).

Author information

Authors and affiliations.

Nuffield Department of Primary Care Health Sciences, University of Oxford, Oxford, UK

Sara Paparini, Chrysanthi Papoutsi, Trish Greenhalgh & Sara Shaw

Wellcome Centre for Cultures & Environments of Health, University of Exeter, Exeter, UK

Judith Green

School of Health Sciences, University of East Anglia, Norwich, UK

Jamie Murdoch

Public Health, Environments and Society, London School of Hygiene & Tropical Medicin, London, UK

Mark Petticrew

Institute for Culture and Society, Western Sydney University, Penrith, Australia

Benjamin Hanckel

You can also search for this author in PubMed   Google Scholar

Contributions

JG, MP, SP, JM, TG, CP and SS drafted the initial paper; all authors contributed to the drafting of the final version, and read and approved the final manuscript.

Corresponding author

Correspondence to Sara Paparini .

Ethics declarations

Ethics approval and consent to participate, consent for publication, competing interests.

The authors declare that they have no competing interests.

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.

Paparini, S., Green, J., Papoutsi, C. et al. Case study research for better evaluations of complex interventions: rationale and challenges. BMC Med 18 , 301 (2020). https://doi.org/10.1186/s12916-020-01777-6

Download citation

Received : 03 July 2020

Accepted : 07 September 2020

Published : 10 November 2020

DOI : https://doi.org/10.1186/s12916-020-01777-6

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

• Qualitative
• Case studies
• Mixed-method
• Public health
• Health services research
• Interventions

BMC Medicine

ISSN: 1741-7015

• General enquiries: [email protected]

Log in using your username and password

• Search More Search for this keyword Advanced search
• Latest content
• Current issue
• Write for Us
• BMJ Journals

You are here

• Volume 21, Issue 1
• What is a case study?
• Article Text
• Article info
• Citation Tools
• Rapid Responses
• Article metrics

• Roberta Heale 1 ,
• Alison Twycross 2
• 1 School of Nursing , Laurentian University , Sudbury , Ontario , Canada
• 2 School of Health and Social Care , London South Bank University , London , UK
• Correspondence to Dr Roberta Heale, School of Nursing, Laurentian University, Sudbury, ON P3E2C6, Canada; rheale{at}laurentian.ca

https://doi.org/10.1136/eb-2017-102845

Statistics from Altmetric.com

Request permissions.

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

What is it?

Case study is a research methodology, typically seen in social and life sciences. There is no one definition of case study research. 1 However, very simply… ‘a case study can be defined as an intensive study about a person, a group of people or a unit, which is aimed to generalize over several units’. 1 A case study has also been described as an intensive, systematic investigation of a single individual, group, community or some other unit in which the researcher examines in-depth data relating to several variables. 2

Often there are several similar cases to consider such as educational or social service programmes that are delivered from a number of locations. Although similar, they are complex and have unique features. In these circumstances, the evaluation of several, similar cases will provide a better answer to a research question than if only one case is examined, hence the multiple-case study. Stake asserts that the cases are grouped and viewed as one entity, called the quintain . 6  ‘We study what is similar and different about the cases to understand the quintain better’. 6

The steps when using case study methodology are the same as for other types of research. 6 The first step is defining the single case or identifying a group of similar cases that can then be incorporated into a multiple-case study. A search to determine what is known about the case(s) is typically conducted. This may include a review of the literature, grey literature, media, reports and more, which serves to establish a basic understanding of the cases and informs the development of research questions. Data in case studies are often, but not exclusively, qualitative in nature. In multiple-case studies, analysis within cases and across cases is conducted. Themes arise from the analyses and assertions about the cases as a whole, or the quintain, emerge. 6

Benefits and limitations of case studies

If a researcher wants to study a specific phenomenon arising from a particular entity, then a single-case study is warranted and will allow for a in-depth understanding of the single phenomenon and, as discussed above, would involve collecting several different types of data. This is illustrated in example 1 below.

Using a multiple-case research study allows for a more in-depth understanding of the cases as a unit, through comparison of similarities and differences of the individual cases embedded within the quintain. Evidence arising from multiple-case studies is often stronger and more reliable than from single-case research. Multiple-case studies allow for more comprehensive exploration of research questions and theory development. 6

Despite the advantages of case studies, there are limitations. The sheer volume of data is difficult to organise and data analysis and integration strategies need to be carefully thought through. There is also sometimes a temptation to veer away from the research focus. 2 Reporting of findings from multiple-case research studies is also challenging at times, 1 particularly in relation to the word limits for some journal papers.

Examples of case studies

Example 1: nurses’ paediatric pain management practices.

One of the authors of this paper (AT) has used a case study approach to explore nurses’ paediatric pain management practices. This involved collecting several datasets:

Observational data to gain a picture about actual pain management practices.

Questionnaire data about nurses’ knowledge about paediatric pain management practices and how well they felt they managed pain in children.

Questionnaire data about how critical nurses perceived pain management tasks to be.

These datasets were analysed separately and then compared 7–9 and demonstrated that nurses’ level of theoretical did not impact on the quality of their pain management practices. 7 Nor did individual nurse’s perceptions of how critical a task was effect the likelihood of them carrying out this task in practice. 8 There was also a difference in self-reported and observed practices 9 ; actual (observed) practices did not confirm to best practice guidelines, whereas self-reported practices tended to.

Example 2: quality of care for complex patients at Nurse Practitioner-Led Clinics (NPLCs)

The other author of this paper (RH) has conducted a multiple-case study to determine the quality of care for patients with complex clinical presentations in NPLCs in Ontario, Canada. 10 Five NPLCs served as individual cases that, together, represented the quatrain. Three types of data were collected including:

Review of documentation related to the NPLC model (media, annual reports, research articles, grey literature and regulatory legislation).

Interviews with nurse practitioners (NPs) practising at the five NPLCs to determine their perceptions of the impact of the NPLC model on the quality of care provided to patients with multimorbidity.

Chart audits conducted at the five NPLCs to determine the extent to which evidence-based guidelines were followed for patients with diabetes and at least one other chronic condition.

The three sources of data collected from the five NPLCs were analysed and themes arose related to the quality of care for complex patients at NPLCs. The multiple-case study confirmed that nurse practitioners are the primary care providers at the NPLCs, and this positively impacts the quality of care for patients with multimorbidity. Healthcare policy, such as lack of an increase in salary for NPs for 10 years, has resulted in issues in recruitment and retention of NPs at NPLCs. This, along with insufficient resources in the communities where NPLCs are located and high patient vulnerability at NPLCs, have a negative impact on the quality of care. 10

These examples illustrate how collecting data about a single case or multiple cases helps us to better understand the phenomenon in question. Case study methodology serves to provide a framework for evaluation and analysis of complex issues. It shines a light on the holistic nature of nursing practice and offers a perspective that informs improved patient care.

• Gustafsson J
• Calanzaro M
• Sandelowski M

Competing interests None declared.

Provenance and peer review Commissioned; internally peer reviewed.

Read the full text or download the PDF:

• SpringerLink shop

Types of journal articles

It is helpful to familiarise yourself with the different types of articles published by journals. Although it may appear there are a large number of types of articles published due to the wide variety of names they are published under, most articles published are one of the following types; Original Research, Review Articles, Short reports or Letters, Case Studies, Methodologies.

Original Research:

This is the most common type of journal manuscript used to publish full reports of data from research. It may be called an  Original Article, Research Article, Research, or just  Article, depending on the journal. The Original Research format is suitable for many different fields and different types of studies. It includes full Introduction, Methods, Results, and Discussion sections.

Short reports or Letters:

These papers communicate brief reports of data from original research that editors believe will be interesting to many researchers, and that will likely stimulate further research in the field. As they are relatively short the format is useful for scientists with results that are time sensitive (for example, those in highly competitive or quickly-changing disciplines). This format often has strict length limits, so some experimental details may not be published until the authors write a full Original Research manuscript. These papers are also sometimes called Brief communications .

Review Articles:

Review Articles provide a comprehensive summary of research on a certain topic, and a perspective on the state of the field and where it is heading. They are often written by leaders in a particular discipline after invitation from the editors of a journal. Reviews are often widely read (for example, by researchers looking for a full introduction to a field) and highly cited. Reviews commonly cite approximately 100 primary research articles.

TIP: If you would like to write a Review but have not been invited by a journal, be sure to check the journal website as some journals to not consider unsolicited Reviews. If the website does not mention whether Reviews are commissioned it is wise to send a pre-submission enquiry letter to the journal editor to propose your Review manuscript before you spend time writing it.  

Case Studies:

These articles report specific instances of interesting phenomena. A goal of Case Studies is to make other researchers aware of the possibility that a specific phenomenon might occur. This type of study is often used in medicine to report the occurrence of previously unknown or emerging pathologies.

Methodologies or Methods

These articles present a new experimental method, test or procedure. The method described may either be completely new, or may offer a better version of an existing method. The article should describe a demonstrable advance on what is currently available.

Back │ Next

Free APA Journals ™ Articles

Recently published articles from subdisciplines of psychology covered by more than 90 APA Journals™ publications.

For additional free resources (such as article summaries, podcasts, and more), please visit the Highlights in Psychological Research page.

• Basic / Experimental Psychology
• Clinical Psychology
• Core of Psychology
• Developmental Psychology
• Educational Psychology, School Psychology, and Training
• Forensic Psychology
• Health Psychology & Medicine
• Industrial / Organizational Psychology & Management
• Neuroscience & Cognition
• Social Psychology & Social Processes
• Moving While Black: Intergroup Attitudes Influence Judgments of Speed (PDF, 71KB) Journal of Experimental Psychology: General February 2016 by Andreana C. Kenrick, Stacey Sinclair, Jennifer Richeson, Sara C. Verosky, and Janetta Lun
• Recognition Without Awareness: Encoding and Retrieval Factors (PDF, 116KB) Journal of Experimental Psychology: Learning, Memory, and Cognition September 2015 by Fergus I. M. Craik, Nathan S. Rose, and Nigel Gopie
• The Tip-of-the-Tongue Heuristic: How Tip-of-the-Tongue States Confer Perceptibility on Inaccessible Words (PDF, 91KB) Journal of Experimental Psychology: Learning, Memory, and Cognition September 2015 by Anne M. Cleary and Alexander B. Claxton
• Cognitive Processes in the Breakfast Task: Planning and Monitoring (PDF, 146KB) Canadian Journal of Experimental Psychology / Revue canadienne de psychologie expérimentale September 2015 by Nathan S. Rose, Lin Luo, Ellen Bialystok, Alexandra Hering, Karen Lau, and Fergus I. M. Craik
• Searching for Explanations: How the Internet Inflates Estimates of Internal Knowledge (PDF, 138KB) Journal of Experimental Psychology: General June 2015 by Matthew Fisher, Mariel K. Goddu, and Frank C. Keil
• Client Perceptions of Corrective Experiences in Cognitive Behavioral Therapy and Motivational Interviewing for Generalized Anxiety Disorder: An Exploratory Pilot Study (PDF, 62KB) Journal of Psychotherapy Integration March 2017 by Jasmine Khattra, Lynne Angus, Henny Westra, Christianne Macaulay, Kathrin Moertl, and Michael Constantino
• Attention-Deficit/Hyperactivity Disorder Developmental Trajectories Related to Parental Expressed Emotion (PDF, 160KB) Journal of Abnormal Psychology February 2016 by Erica D. Musser, Sarah L. Karalunas, Nathan Dieckmann, Tara S. Peris, and Joel T. Nigg
• The Integrated Scientist-Practitioner: A New Model for Combining Research and Clinical Practice in Fee-For-Service Settings (PDF, 58KB) Professional Psychology: Research and Practice December 2015 by Jenna T. LeJeune and Jason B. Luoma
• Psychotherapists as Gatekeepers: An Evidence-Based Case Study Highlighting the Role and Process of Letter Writing for Transgender Clients (PDF, 76KB) Psychotherapy September 2015 by Stephanie L. Budge
• Perspectives of Family and Veterans on Family Programs to Support Reintegration of Returning Veterans With Posttraumatic Stress Disorder (PDF, 70KB) Psychological Services August 2015 by Ellen P. Fischer, Michelle D. Sherman, Jean C. McSweeney, Jeffrey M. Pyne, Richard R. Owen, and Lisa B. Dixon
• "So What Are You?": Inappropriate Interview Questions for Psychology Doctoral and Internship Applicants (PDF, 79KB) Training and Education in Professional Psychology May 2015 by Mike C. Parent, Dana A. Weiser, and Andrea McCourt
• Cultural Competence as a Core Emphasis of Psychoanalytic Psychotherapy (PDF, 81KB) Psychoanalytic Psychology April 2015 by Pratyusha Tummala-Narra
• The Role of Gratitude in Spiritual Well-Being in Asymptomatic Heart Failure Patients (PDF, 123KB) Spirituality in Clinical Practice March 2015 by Paul J. Mills, Laura Redwine, Kathleen Wilson, Meredith A. Pung, Kelly Chinh, Barry H. Greenberg, Ottar Lunde, Alan Maisel, Ajit Raisinghani, Alex Wood, and Deepak Chopra
• Nepali Bhutanese Refugees Reap Support Through Community Gardening (PDF, 104KB) International Perspectives in Psychology: Research, Practice, Consultation January 2017 by Monica M. Gerber, Jennifer L. Callahan, Danielle N. Moyer, Melissa L. Connally, Pamela M. Holtz, and Beth M. Janis
• Does Monitoring Goal Progress Promote Goal Attainment? A Meta-Analysis of the Experimental Evidence (PDF, 384KB) Psychological Bulletin February 2016 by Benjamin Harkin, Thomas L. Webb, Betty P. I. Chang, Andrew Prestwich, Mark Conner, Ian Kellar, Yael Benn, and Paschal Sheeran
• Youth Violence: What We Know and What We Need to Know (PDF, 388KB) American Psychologist January 2016 by Brad J. Bushman, Katherine Newman, Sandra L. Calvert, Geraldine Downey, Mark Dredze, Michael Gottfredson, Nina G. Jablonski, Ann S. Masten, Calvin Morrill, Daniel B. Neill, Daniel Romer, and Daniel W. Webster
• Supervenience and Psychiatry: Are Mental Disorders Brain Disorders? (PDF, 113KB) Journal of Theoretical and Philosophical Psychology November 2015 by Charles M. Olbert and Gary J. Gala
• Constructing Psychological Objects: The Rhetoric of Constructs (PDF, 108KB) Journal of Theoretical and Philosophical Psychology November 2015 by Kathleen L. Slaney and Donald A. Garcia
• Expanding Opportunities for Diversity in Positive Psychology: An Examination of Gender, Race, and Ethnicity (PDF, 119KB) Canadian Psychology / Psychologie canadienne August 2015 by Meghana A. Rao and Stewart I. Donaldson
• Racial Microaggression Experiences and Coping Strategies of Black Women in Corporate Leadership (PDF, 132KB) Qualitative Psychology August 2015 by Aisha M. B. Holder, Margo A. Jackson, and Joseph G. Ponterotto
• An Appraisal Theory of Empathy and Other Vicarious Emotional Experiences (PDF, 151KB) Psychological Review July 2015 by Joshua D. Wondra and Phoebe C. Ellsworth
• An Attachment Theoretical Framework for Personality Disorders (PDF, 100KB) Canadian Psychology / Psychologie canadienne May 2015 by Kenneth N. Levy, Benjamin N. Johnson, Tracy L. Clouthier, J. Wesley Scala, and Christina M. Temes
• Emerging Approaches to the Conceptualization and Treatment of Personality Disorder (PDF, 111KB) Canadian Psychology / Psychologie canadienne May 2015 by John F. Clarkin, Kevin B. Meehan, and Mark F. Lenzenweger
• A Complementary Processes Account of the Development of Childhood Amnesia and a Personal Past (PDF, 585KB) Psychological Review April 2015 by Patricia J. Bauer
• Terminal Decline in Well-Being: The Role of Social Orientation (PDF, 238KB) Psychology and Aging March 2016 by Denis Gerstorf, Christiane A. Hoppmann, Corinna E. Löckenhoff, Frank J. Infurna, Jürgen Schupp, Gert G. Wagner, and Nilam Ram
• Student Threat Assessment as a Standard School Safety Practice: Results From a Statewide Implementation Study (PDF, 97KB) School Psychology Quarterly June 2018 by Dewey Cornell, Jennifer L. Maeng, Anna Grace Burnette, Yuane Jia, Francis Huang, Timothy Konold, Pooja Datta, Marisa Malone, and Patrick Meyer
• Can a Learner-Centered Syllabus Change Students’ Perceptions of Student–Professor Rapport and Master Teacher Behaviors? (PDF, 90KB) Scholarship of Teaching and Learning in Psychology September 2016 by Aaron S. Richmond, Jeanne M. Slattery, Nathanael Mitchell, Robin K. Morgan, and Jared Becknell
• Adolescents' Homework Performance in Mathematics and Science: Personal Factors and Teaching Practices (PDF, 170KB) Journal of Educational Psychology November 2015 by Rubén Fernández-Alonso, Javier Suárez-Álvarez, and José Muñiz
• Teacher-Ready Research Review: Clickers (PDF, 55KB) Scholarship of Teaching and Learning in Psychology September 2015 by R. Eric Landrum
• Enhancing Attention and Memory During Video-Recorded Lectures (PDF, 83KB) Scholarship of Teaching and Learning in Psychology March 2015 by Daniel L. Schacter and Karl K. Szpunar
• The Alleged "Ferguson Effect" and Police Willingness to Engage in Community Partnership (PDF, 70KB) Law and Human Behavior February 2016 by Scott E. Wolfe and Justin Nix
• Randomized Controlled Trial of an Internet Cognitive Behavioral Skills-Based Program for Auditory Hallucinations in Persons With Psychosis (PDF, 92KB) Psychiatric Rehabilitation Journal September 2017 by Jennifer D. Gottlieb, Vasudha Gidugu, Mihoko Maru, Miriam C. Tepper, Matthew J. Davis, Jennifer Greenwold, Ruth A. Barron, Brian P. Chiko, and Kim T. Mueser
• Preventing Unemployment and Disability Benefit Receipt Among People With Mental Illness: Evidence Review and Policy Significance (PDF, 134KB) Psychiatric Rehabilitation Journal June 2017 by Bonnie O'Day, Rebecca Kleinman, Benjamin Fischer, Eric Morris, and Crystal Blyler
• Sending Your Grandparents to University Increases Cognitive Reserve: The Tasmanian Healthy Brain Project (PDF, 88KB) Neuropsychology July 2016 by Megan E. Lenehan, Mathew J. Summers, Nichole L. Saunders, Jeffery J. Summers, David D. Ward, Karen Ritchie, and James C. Vickers
• The Foundational Principles as Psychological Lodestars: Theoretical Inspiration and Empirical Direction in Rehabilitation Psychology (PDF, 68KB) Rehabilitation Psychology February 2016 by Dana S. Dunn, Dawn M. Ehde, and Stephen T. Wegener
• Feeling Older and Risk of Hospitalization: Evidence From Three Longitudinal Cohorts (PDF, 55KB) Health Psychology Online First Publication — February 11, 2016 by Yannick Stephan, Angelina R. Sutin, and Antonio Terracciano
• Anger Intensification With Combat-Related PTSD and Depression Comorbidity (PDF, 81KB) Psychological Trauma: Theory, Research, Practice, and Policy January 2016 by Oscar I. Gonzalez, Raymond W. Novaco, Mark A. Reger, and Gregory A. Gahm
• Special Issue on eHealth and mHealth: Challenges and Future Directions for Assessment, Treatment, and Dissemination (PDF, 32KB) Health Psychology December 2015 by Belinda Borrelli and Lee M. Ritterband
• Posttraumatic Growth Among Combat Veterans: A Proposed Developmental Pathway (PDF, 110KB) Psychological Trauma: Theory, Research, Practice, and Policy July 2015 by Sylvia Marotta-Walters, Jaehwa Choi, and Megan Doughty Shaine
• Racial and Sexual Minority Women's Receipt of Medical Assistance to Become Pregnant (PDF, 111KB) Health Psychology June 2015 by Bernadette V. Blanchfield and Charlotte J. Patterson
• An Examination of Generational Stereotypes as a Path Towards Reverse Ageism (PDF, 205KB) The Psychologist-Manager Journal August 2017 By Michelle Raymer, Marissa Reed, Melissa Spiegel, and Radostina K. Purvanova
• Sexual Harassment: Have We Made Any Progress? (PDF, 121KB) Journal of Occupational Health Psychology July 2017 By James Campbell Quick and M. Ann McFadyen
• Multidimensional Suicide Inventory-28 (MSI-28) Within a Sample of Military Basic Trainees: An Examination of Psychometric Properties (PDF, 79KB) Military Psychology November 2015 By Serena Bezdjian, Danielle Burchett, Kristin G. Schneider, Monty T. Baker, and Howard N. Garb
• Cross-Cultural Competence: The Role of Emotion Regulation Ability and Optimism (PDF, 100KB) Military Psychology September 2015 By Bianca C. Trejo, Erin M. Richard, Marinus van Driel, and Daniel P. McDonald
• The Effects of Stress on Prospective Memory: A Systematic Review (PDF, 149KB) Psychology & Neuroscience September 2017 by Martina Piefke and Katharina Glienke
• Don't Aim Too High for Your Kids: Parental Overaspiration Undermines Students' Learning in Mathematics (PDF, 164KB) Journal of Personality and Social Psychology November 2016 by Kou Murayama, Reinhard Pekrun, Masayuki Suzuki, Herbert W. Marsh, and Stephanie Lichtenfeld
• Sex Differences in Sports Interest and Motivation: An Evolutionary Perspective (PDF, 155KB) Evolutionary Behavioral Sciences April 2016 by Robert O. Deaner, Shea M. Balish, and Michael P. Lombardo
• Asian Indian International Students' Trajectories of Depression, Acculturation, and Enculturation (PDF, 210KB) Asian American Journal of Psychology March 2016 By Dhara T. Meghani and Elizabeth A. Harvey
• Cynical Beliefs About Human Nature and Income: Longitudinal and Cross-Cultural Analyses (PDF, 163KB) January 2016 Journal of Personality and Social Psychology by Olga Stavrova and Daniel Ehlebracht
• Annual Review of Asian American Psychology, 2014 (PDF, 384KB) Asian American Journal of Psychology December 2015 By Su Yeong Kim, Yishan Shen, Yang Hou, Kelsey E. Tilton, Linda Juang, and Yijie Wang
• Resilience in the Study of Minority Stress and Health of Sexual and Gender Minorities (PDF, 40KB) Psychology of Sexual Orientation and Gender Diversity September 2015 by Ilan H. Meyer
• Self-Reported Psychopathy and Its Association With Criminal Cognition and Antisocial Behavior in a Sample of University Undergraduates (PDF, 91KB) Canadian Journal of Behavioural Science / Revue canadienne des sciences du comportement July 2015 by Samantha J. Riopka, Richard B. A. Coupland, and Mark E. Olver

Journals Publishing Resource Center

Find resources for writing, reviewing, and editing articles for publishing with APA Journals™.

Visit the resource center

Journals information

• Frequently Asked Questions
• Journal statistics and operations data
• Special Issues
• Email alerts
• Copyright and permissions

Contact APA Publications

A case controlled study of risk factors for metastatic squamous cell carcinoma in organ transplant recipients: single academic medical center

• ORIGINAL PAPER
• Published: 11 September 2024
• Volume 316 , article number  612 , ( 2024 )

Cite this article

• Cynthia F. Griffith   ORCID: orcid.org/0000-0002-0566-8611 1 ,
• Anthony Solhjoo 1 ,
• Luke Mahan 2 &
• Rajiv I. Nijhawan 1  

Solid organ transplant recipients (SOTRs) are at high risk of cutaneous squamous cell carcinoma (cSCC) metastasis. Despite prior studies identifying risk factors, mortality remains high. Understanding additional risk factors may aid in reducing mortality in this population. This study aimed to investigate risk factors and predictive variables for metastatic cSCC in SOTRs. The primary goal was to accurately identify transplant patients at increased risk of metastatic cSCC. A retrospective case–control study in a single institution of 3576 cases of organ transplants were identified from January 1991 to July 2022. A cohort of metastatic cancer patients and two randomly generated age and organ matched control cohorts were identified. 16 SOTR patients developed metastatic cSCC. The majority were male, with high-risk tumor sites. Tumor depth varied and half exhibited perineural invasion. Cylex® (p = 0.05) and white blood cell counts (p = 0.04) were significantly lower in these patients compared to control. Lung transplants were at highest risk relative to other solid organ transplants. Voriconazole exposure was also associated with increased metastatic risk (p = 0.04). Small sample size at a single institution. Close monitoring of SOTR, especially those with lung transplants given their increased risk, reducing immunosuppression, and limiting exposure to voriconazole can improve outcomes in SOTRs with metastatic cSCC.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Subscribe and save.

• Get 10 units per month
• Download Article/Chapter or eBook
• 1 Unit = 1 Article or 1 Chapter
• Cancel anytime

Price includes VAT (Russian Federation)

Instant access to the full article PDF.

Rent this article via DeepDyve

Institutional subscriptions

Similar content being viewed by others

Update on Staging, Definition, and Chemoprevention of “High-Risk Squamous Cell Carcinoma” in Organ Transplant Recipients

Oncological Outcome After Lymph Node Dissection for Cutaneous Squamous Cell Carcinoma

Management of Cutaneous Squamous Cell Carcinoma in Organ Transplant Recipients

Data availability.

De-identified data will be made available upon reasonable request.

Lanz J, Bouwes Bavinck JN, Westhuis M et al (2019) Aggressive squamous cell carcinoma in organ transplant recipients. JAMA Dermatol 155(1):66–71. https://doi.org/10.1001/jamadermatol.2018.4406

Article   PubMed   Google Scholar  

Goyal RK (2015) Voriconazole-associated phototoxic dermatoses and skin cancer. Expert Rev Anti Infect Ther 13(12):1537–1546. https://doi.org/10.1586/14787210.2015.1102053

Article   CAS   PubMed   Google Scholar  

Garret GL, Lowenstein SE, Singer JP, He SY, Arron ST (2016) Trends of skin cancer mortality after transplant in the United States: 1987 to 2013. J Am Acad Dermatol 75(1):106–112

Article   Google Scholar  

Brougham ND, Dennett ER, Cameron R, Tan ST (2012) The incidence of metastasis from cutaneous squamous cell carcinoma and the impact of its risk factors. J Surg Oncol 106(7):811–815

Genders RE, Osinga JAJ, Tromp EE, O’Rourke P, Bouwes Bavinck JN, Plasmeijer EI (2018) Metastasis risk of cutaneous squamous cell carcinoma in organ transplant recipients and immunocompetent patients. Acta Derm Venereol 98(6):551–555

Kumar V, Shinagare AB, Rennke HG et al (2020) The safety and efficacy of checkpoint inhibitors in transplant recipients: a case series and systematic review of literature. Oncologist 25(6):505–514

Article   CAS   PubMed   PubMed Central   Google Scholar  

Zeevi A, Lunz J (2013) Cylex immuknow cell function assay. Methods Mol Biol Transpl Immunol 1034:343–351

Article   CAS   Google Scholar  

McLaughlin EJ, Miller L, Shin TM et al (2017) Rate of regional nodal metastases of cutaneous squamous cell carcinoma in the immunosuppressed patient. Am J Otolaryngol 38(3):325–328

D’Arcy ME, Pfeiffer RM, Rivera DR et al (2020) Voriconazole and the risk of keratinocyte carcinomas among lung transplant recipients in the United States. JAMA Dermatol 156(7):772–779

Jiyad Z, Olsen CM, Burke MT, Isbel NM, Green AC (2016) Azathioprine and risk of skin cancer in organ transplant recipients: systematic review and meta-analysis. Am J Transplant 16(12):3490–3503

Wang L, Motter J, Bae S et al (2020) Induction immunosuppression and the risk of incident malignancies among older and younger kidney transplant recipients: a prospective cohort study. Clin Transplant 34(12):e14121

Haisma MS, Greven N, Logendran M et al (2023) Chronic use of hydrochlorothiazide and risk of skin cancer in caucasian adults: a pharmlines initiative inception cohort study. Acta Derm Venereol 4:103

Google Scholar  

Bossart S, Daneluzzi C, Moor MB et al (2023) HPV vaccination in immunosuppressed patients with established skin warts and non-melanoma skin cancer: a single-institutional cohort study. Vaccines (Basel) 1511(9):1490

Matsuzaki J, Tsuji T, Luescher IF et al (2015) Direct tumor recognition by a human CD4(+) T-cell subset potently mediates tumor growth inhibition and orchestrates anti-tumor immune responses. Sci Rep 5:14896

Tay RE, Richardson EK, Toh HC (2021) Revisiting the role of CD4+ T cells in cancer immunotherapy—new insights into old paradigms. Cancer Gene Ther 28:5–17

Download references

Acknowledgements

Thank you to Dr. Ahmed Shalaby for statistical analysis. Research reported in this publication was supported by the National Center for Advancing Translational Sciences of the National Institutes of Health under award Number UL1 TR003163. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH (National Institutes of Health). Thank you to Anusha Mithani, PA-C, Yadaris Bonilla, PA-C and Rafael Basa PA-C for data collection.

No funding sources for this study.

Author information

Authors and affiliations.

Department of Dermatology, University of Texas Southwestern Medical Center, Harry Hines, Dallas, TX, 75390, USA

Cynthia F. Griffith, Anthony Solhjoo & Rajiv I. Nijhawan

Department of Lung Transplant, University of Texas Southwestern Medical Center, Dallas, TX, 75390, USA

You can also search for this author in PubMed   Google Scholar

Contributions

C.G., A.S., and L.M. wrote the main manuscript text, prepared figures/tables, and made substantial contributions to the conception of the work, acquisition, analysis, and interpretation of data. R.I.N. revised the manuscript text critically for important intellectual content. All authors reviewed the manuscript. C.G. approved the version to be published.

Corresponding author

Correspondence to Cynthia F. Griffith .

Ethics declarations

Conflict of interest.

The authors declare no competing interests.

Patient consent

Not applicable.

Additional information

Publisher's note.

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

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary file1 (DOCX 85 KB)

Rights and permissions.

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Griffith, C.F., Solhjoo, A., Mahan, L. et al. A case controlled study of risk factors for metastatic squamous cell carcinoma in organ transplant recipients: single academic medical center. Arch Dermatol Res 316 , 612 (2024). https://doi.org/10.1007/s00403-024-03284-7

Download citation

Received : 10 July 2024

Revised : 10 July 2024

Accepted : 05 August 2024

Published : 11 September 2024

DOI : https://doi.org/10.1007/s00403-024-03284-7

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

• Squamous cell carcinoma
• Organ transplant recipient
• Immunosuppression
• Solid organ transplant
• Metastatic squamous carcinoma
• Metastatic cutaneous squamous carcinoma
• Find a journal
• Publish with us
• Track your research

Our systems are now restored following recent technical disruption, and we’re working hard to catch up on publishing. We apologise for the inconvenience caused. Find out more: https://www.cambridge.org/universitypress/about-us/news-and-blogs/cambridge-university-press-publishing-update-following-technical-disruption

We use cookies to distinguish you from other users and to provide you with a better experience on our websites. Close this message to accept cookies or find out how to manage your cookie settings .

Login Alert

• > Journals
• > PS: Political Science & Politics
• > Volume 46 Issue 3
• > Implicit and Explicit Measurement Approaches to Research...

Article contents

Implicit and explicit measurement approaches to research on policy implementation: the case of race-based disparities in criminal justice.

Published online by Cambridge University Press:  21 June 2013

In 2011, more than 3% of all black men in the United States were imprisoned, compared to only 0.5% of all white men. Among prisoners ages 18 to 24, black men were imprisoned at a rate more than seven times that of white men (Carson and Sabol 2012). It is becoming increasingly urgent for researchers to understand what accounts for these race-based disparities. While a broad constellation of social problems exist that likely contribute to these disparities in concert, different fields of social science tend to focus on different types of explanations. Political scientists and sociologists have tended to emphasize the role of institutional factors, including criminal justice policies and practices, in maintaining race-based disparities. Social psychologists, in contrast, have tended to emphasize individual factors, including punitive responses to crime by jurors, judges, and criminal justice professionals.

Access options

This article has been cited by the following publications. This list is generated based on data provided by Crossref .

• Google Scholar

View all Google Scholar citations for this article.

Save article to Kindle

To save this article to your Kindle, first ensure [email protected] is added to your Approved Personal Document E-mail List under your Personal Document Settings on the Manage Your Content and Devices page of your Amazon account. Then enter the ‘name’ part of your Kindle email address below. Find out more about saving to your Kindle .

Note you can select to save to either the @free.kindle.com or @kindle.com variations. ‘@free.kindle.com’ emails are free but can only be saved to your device when it is connected to wi-fi. ‘@kindle.com’ emails can be delivered even when you are not connected to wi-fi, but note that service fees apply.

Find out more about the Kindle Personal Document Service.

• Volume 46, Issue 3
• Eugene Borgida (a1) and Andrea Miller (a2)
• DOI: https://doi.org/10.1017/S1049096513000668

Save article to Dropbox

To save this article to your Dropbox account, please select one or more formats and confirm that you agree to abide by our usage policies. If this is the first time you used this feature, you will be asked to authorise Cambridge Core to connect with your Dropbox account. Find out more about saving content to Dropbox .

Save article to Google Drive

To save this article to your Google Drive account, please select one or more formats and confirm that you agree to abide by our usage policies. If this is the first time you used this feature, you will be asked to authorise Cambridge Core to connect with your Google Drive account. Find out more about saving content to Google Drive .

Reply to: Submit a response

- No HTML tags allowed - Web page URLs will display as text only - Lines and paragraphs break automatically - Attachments, images or tables are not permitted

Your details

Your email address will be used in order to notify you when your comment has been reviewed by the moderator and in case the author(s) of the article or the moderator need to contact you directly.

You have entered the maximum number of contributors

Conflicting interests.

Please list any fees and grants from, employment by, consultancy for, shared ownership in or any close relationship with, at any time over the preceding 36 months, any organisation whose interests may be affected by the publication of the response. Please also list any non-financial associations or interests (personal, professional, political, institutional, religious or other) that a reasonable reader would want to know about in relation to the submitted work. This pertains to all the authors of the piece, their spouses or partners.

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
• Open access
• Published: 12 September 2024

An open-source framework for end-to-end analysis of electronic health record data

• Lukas Heumos 1 , 2 , 3 ,
• Philipp Ehmele 1 ,
• Tim Treis 1 , 3 ,
• Julius Upmeier zu Belzen   ORCID: orcid.org/0000-0002-0966-4458 4 ,
• Eljas Roellin 1 , 5 ,
• Lilly May 1 , 5 ,
• Altana Namsaraeva 1 , 6 ,
• Nastassya Horlava 1 , 3 ,
• Vladimir A. Shitov   ORCID: orcid.org/0000-0002-1960-8812 1 , 3 ,
• Xinyue Zhang   ORCID: orcid.org/0000-0003-4806-4049 1 ,
• Luke Zappia   ORCID: orcid.org/0000-0001-7744-8565 1 , 5 ,
• Rainer Knoll 7 ,
• Niklas J. Lang 2 ,
• Leon Hetzel 1 , 5 ,
• Isaac Virshup 1 ,
• Lisa Sikkema   ORCID: orcid.org/0000-0001-9686-6295 1 , 3 ,
• Fabiola Curion 1 , 5 ,
• Roland Eils 4 , 8 ,
• Herbert B. Schiller 2 , 9 ,
• Anne Hilgendorff 2 , 10 &
• Fabian J. Theis   ORCID: orcid.org/0000-0002-2419-1943 1 , 3 , 5  

Nature Medicine ( 2024 ) Cite this article

86 Altmetric

Metrics details

• Epidemiology
• Translational research

With progressive digitalization of healthcare systems worldwide, large-scale collection of electronic health records (EHRs) has become commonplace. However, an extensible framework for comprehensive exploratory analysis that accounts for data heterogeneity is missing. Here we introduce ehrapy, a modular open-source Python framework designed for exploratory analysis of heterogeneous epidemiology and EHR data. ehrapy incorporates a series of analytical steps, from data extraction and quality control to the generation of low-dimensional representations. Complemented by rich statistical modules, ehrapy facilitates associating patients with disease states, differential comparison between patient clusters, survival analysis, trajectory inference, causal inference and more. Leveraging ontologies, ehrapy further enables data sharing and training EHR deep learning models, paving the way for foundational models in biomedical research. We demonstrate ehrapy’s features in six distinct examples. We applied ehrapy to stratify patients affected by unspecified pneumonia into finer-grained phenotypes. Furthermore, we reveal biomarkers for significant differences in survival among these groups. Additionally, we quantify medication-class effects of pneumonia medications on length of stay. We further leveraged ehrapy to analyze cardiovascular risks across different data modalities. We reconstructed disease state trajectories in patients with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) based on imaging data. Finally, we conducted a case study to demonstrate how ehrapy can detect and mitigate biases in EHR data. ehrapy, thus, provides a framework that we envision will standardize analysis pipelines on EHR data and serve as a cornerstone for the community.

Similar content being viewed by others

Data-driven identification of heart failure disease states and progression pathways using electronic health records

EHR foundation models improve robustness in the presence of temporal distribution shift

Harnessing EHR data for health research

Electronic health records (EHRs) are becoming increasingly common due to standardized data collection 1 and digitalization in healthcare institutions. EHRs collected at medical care sites serve as efficient storage and sharing units of health information 2 , enabling the informed treatment of individuals using the patient’s complete history 3 . Routinely collected EHR data are approaching genomic-scale size and complexity 4 , posing challenges in extracting information without quantitative analysis methods. The application of such approaches to EHR databases 1 , 5 , 6 , 7 , 8 , 9 has enabled the prediction and classification of diseases 10 , 11 , study of population health 12 , determination of optimal treatment policies 13 , 14 , simulation of clinical trials 15 and stratification of patients 16 .

However, current EHR datasets suffer from serious limitations, such as data collection issues, inconsistencies and lack of data diversity. EHR data collection and sharing problems often arise due to non-standardized formats, with disparate systems using exchange protocols, such as Health Level Seven International (HL7) and Fast Healthcare Interoperability Resources (FHIR) 17 . In addition, EHR data are stored in various on-disk formats, including, but not limited to, relational databases and CSV, XML and JSON formats. These variations pose challenges with respect to data retrieval, scalability, interoperability and data sharing.

Beyond format variability, inherent biases of the collected data can compromise the validity of findings. Selection bias stemming from non-representative sample composition can lead to skewed inferences about disease prevalence or treatment efficacy 18 , 19 . Filtering bias arises through inconsistent criteria for data inclusion, obscuring true variable relationships 20 . Surveillance bias exaggerates associations between exposure and outcomes due to differential monitoring frequencies 21 . EHR data are further prone to missing data 22 , 23 , which can be broadly classified into three categories: missing completely at random (MCAR), where missingness is unrelated to the data; missing at random (MAR), where missingness depends on observed data; and missing not at random (MNAR), where missingness depends on unobserved data 22 , 23 . Information and coding biases, related to inaccuracies in data recording or coding inconsistencies, respectively, can lead to misclassification and unreliable research conclusions 24 , 25 . Data may even contradict itself, such as when measurements were reported for deceased patients 26 , 27 . Technical variation and differing data collection standards lead to distribution differences and inconsistencies in representation and semantics across EHR datasets 28 , 29 . Attrition and confounding biases, resulting from differential patient dropout rates or unaccounted external variable effects, can significantly skew study outcomes 30 , 31 , 32 . The diversity of EHR data that comprise demographics, laboratory results, vital signs, diagnoses, medications, x-rays, written notes and even omics measurements amplifies all the aforementioned issues.

Addressing these challenges requires rigorous study design, careful data pre-processing and continuous bias evaluation through exploratory data analysis. Several EHR data pre-processing and analysis workflows were previously developed 4 , 33 , 34 , 35 , 36 , 37 , but none of them enables the analysis of heterogeneous data, provides in-depth documentation, is available as a software package or allows for exploratory visual analysis. Current EHR analysis pipelines, therefore, differ considerably in their approaches and are often commercial, vendor-specific solutions 38 . This is in contrast to strategies using community standards for the analysis of omics data, such as Bioconductor 39 or scverse 40 . As a result, EHR data frequently remain underexplored and are commonly investigated only for a particular research question 41 . Even in such cases, EHR data are then frequently input into machine learning models with serious data quality issues that greatly impact prediction performance and generalizability 42 .

To address this lack of analysis tooling, we developed the EHR Analysis in Python framework, ehrapy, which enables exploratory analysis of diverse EHR datasets. The ehrapy package is purpose-built to organize, analyze, visualize and statistically compare complex EHR data. ehrapy can be applied to datasets of different data types, sizes, diseases and origins. To demonstrate this versatility, we applied ehrapy to datasets obtained from EHR and population-based studies. Using the Pediatric Intensive Care (PIC) EHR database 43 , we stratified patients diagnosed with ‘unspecified pneumonia’ into distinct clinically relevant groups, extracted clinical indicators of pneumonia through statistical analysis and quantified medication-class effects on length of stay (LOS) with causal inference. Using the UK Biobank 44 (UKB), a population-scale cohort comprising over 500,000 participants from the United Kingdom, we employed ehrapy to explore cardiovascular risk factors using clinical predictors, metabolomics, genomics and retinal imaging-derived features. Additionally, we performed image analysis to project disease progression through fate mapping in patients affected by coronavirus disease 2019 (COVID-19) using chest x-rays. Finally, we demonstrate how exploratory analysis with ehrapy unveils and mitigates biases in over 100,000 visits by patients with diabetes across 130 US hospitals. We provide online links to additional use cases that demonstrate ehrapy’s usage with further datasets, including MIMIC-II (ref. 45 ), and for various medical conditions, such as patients subject to indwelling arterial catheter usage. ehrapy is compatible with any EHR dataset that can be transformed into vectors and is accessible as a user-friendly open-source software package hosted at https://github.com/theislab/ehrapy and installable from PyPI. It comes with comprehensive documentation, tutorials and further examples, all available at https://ehrapy.readthedocs.io .

ehrapy: a framework for exploratory EHR data analysis

The foundation of ehrapy is a robust and scalable data storage backend that is combined with a series of pre-processing and analysis modules. In ehrapy, EHR data are organized as a data matrix where observations are individual patient visits (or patients, in the absence of follow-up visits), and variables represent all measured quantities ( Methods ). These data matrices are stored together with metadata of observations and variables. By leveraging the AnnData (annotated data) data structure that implements this design, ehrapy builds upon established standards and is compatible with analysis and visualization functions provided by the omics scverse 40 ecosystem. Readers are also available in R, Julia and Javascript 46 . We additionally provide a dataset module with more than 20 public loadable EHR datasets in AnnData format to kickstart analysis and development with ehrapy.

For standardized analysis of EHR data, it is crucial that these data are encoded and stored in consistent, reusable formats. Thus, ehrapy requires that input data are organized in structured vectors. Readers for common formats, such as CSV, OMOP 47 or SQL databases, are available in ehrapy. Data loaded into AnnData objects can be mapped against several hierarchical ontologies 48 , 49 , 50 , 51 ( Methods ). Clinical keywords of free text notes can be automatically extracted ( Methods ).

Powered by scanpy, which scales to millions of observations 52 ( Methods and Supplementary Table 1 ) and the machine learning library scikit-learn 53 , ehrapy provides more than 100 composable analysis functions organized in modules from which custom analysis pipelines can be built. Each function directly interacts with the AnnData object and adds all intermediate results for simple access and reuse of information to it. To facilitate setting up these pipelines, ehrapy guides analysts through a general analysis pipeline (Fig. 1 ). At any step of an analysis pipeline, community software packages can be integrated without any vendor lock-in. Because ehrapy is built on open standards, it can be purposefully extended to solve new challenges, such as the development of foundational models ( Methods ).

a , Heterogeneous health data are first loaded into memory as an AnnData object with patient visits as observational rows and variables as columns. Next, the data can be mapped against ontologies, and key terms are extracted from free text notes. b , The EHR data are subject to quality control where low-quality or spurious measurements are removed or imputed. Subsequently, numerical data are normalized, and categorical data are encoded. Data from different sources with data distribution shifts are integrated, embedded, clustered and annotated in a patient landscape. c , Further downstream analyses depend on the question of interest and can include the inference of causal effects and trajectories, survival analysis or patient stratification.

In the ehrapy analysis pipeline, EHR data are initially inspected for quality issues by analyzing feature distributions that may skew results and by detecting visits and features with high missing rates that ehrapy can then impute ( Methods ). ehrapy tracks all filtering steps while keeping track of population dynamics to highlight potential selection and filtering biases ( Methods ). Subsequently, ehrapy’s normalization and encoding functions ( Methods ) are applied to achieve a uniform numerical representation that facilitates data integration and corrects for dataset shift effects ( Methods ). Calculated lower-dimensional representations can subsequently be visualized, clustered and annotated to obtain a patient landscape ( Methods ). Such annotated groups of patients can be used for statistical comparisons to find differences in features among them to ultimately learn markers of patient states.

As analysis goals can differ between users and datasets, the ehrapy analysis pipeline is customizable during the final knowledge inference step. ehrapy provides statistical methods for group comparison and extensive support for survival analysis ( Methods ), enabling the discovery of biomarkers. Furthermore, ehrapy offers functions for causal inference to go from statistically determined associations to causal relations ( Methods ). Moreover, patient visits in aggregated EHR data can be regarded as snapshots where individual measurements taken at specific timepoints might not adequately reflect the underlying progression of disease and result from unrelated variation due to, for example, day-to-day differences 54 , 55 , 56 . Therefore, disease progression models should rely on analysis of the underlying clinical data, as disease progression in an individual patient may not be monotonous in time. ehrapy allows for the use of advanced trajectory inference methods to overcome sparse measurements 57 , 58 , 59 . We show that this approach can order snapshots to calculate a pseudotime that can adequately reflect the progression of the underlying clinical process. Given a sufficient number of snapshots, ehrapy increases the potential to understand disease progression, which is likely not robustly captured within a single EHR but, rather, across several.

ehrapy enables patient stratification in pneumonia cases

To demonstrate ehrapy’s capability to analyze heterogeneous datasets from a broad patient set across multiple care units, we applied our exploratory strategy to the PIC 43 database. The PIC database is a single-center database hosting information on children admitted to critical care units at the Children’s Hospital of Zhejiang University School of Medicine in China. It contains 13,499 distinct hospital admissions of 12,881 individual pediatric patients admitted between 2010 and 2018 for whom demographics, diagnoses, doctors’ notes, vital signs, laboratory and microbiology tests, medications, fluid balances and more were collected (Extended Data Figs. 1 and 2a and Methods ). After missing data imputation and subsequent pre-processing (Extended Data Figs. 2b,c and 3 and Methods ), we generated a uniform manifold approximation and projection (UMAP) embedding to visualize variation across all patients using ehrapy (Fig. 2a ). This visualization of the low-dimensional patient manifold shows the heterogeneity of the collected data in the PIC database, with malformations, perinatal and respiratory being the most abundant International Classification of Diseases (ICD) chapters (Fig. 2b ). The most common respiratory disease categories (Fig. 2c ) were labeled pneumonia and influenza ( n  = 984). We focused on pneumonia to apply ehrapy to a challenging, broad-spectrum disease that affects all age groups. Pneumonia is a prevalent respiratory infection that poses a substantial burden on public health 60 and is characterized by inflammation of the alveoli and distal airways 60 . Individuals with pre-existing chronic conditions are particularly vulnerable, as are children under the age of 5 (ref. 61 ). Pneumonia can be caused by a range of microorganisms, encompassing bacteria, respiratory viruses and fungi.

a , UMAP of all patient visits in the ICU with primary discharge diagnosis grouped by ICD chapter. b , The prevalence of respiratory diseases prompted us to investigate them further. c , Respiratory categories show the abundance of influenza and pneumonia diagnoses that we investigated more closely. d , We observed the ‘unspecified pneumonia’ subgroup, which led us to investigate and annotate it in more detail. e , The previously ‘unspecified pneumonia’-labeled patients were annotated using several clinical features (Extended Data Fig. 5 ), of which the most important ones are shown in the heatmap ( f ). g , Example disease progression of an individual child with pneumonia illustrating pharmacotherapy over time until positive A. baumannii swab.

We selected the age group ‘youths’ (13 months to 18 years of age) for further analysis, addressing a total of 265 patients who dominated the pneumonia cases and were diagnosed with ‘unspecified pneumonia’ (Fig. 2d and Extended Data Fig. 4 ). Neonates (0–28 d old) and infants (29 d to 12 months old) were excluded from the analysis as the disease context is significantly different in these age groups due to distinct anatomical and physical conditions. Patients were 61% male, had a total of 277 admissions, had a mean age at admission of 54 months (median, 38 months) and had an average LOS of 15 d (median, 7 d). Of these, 152 patients were admitted to the pediatric intensive care unit (PICU), 118 to the general ICU (GICU), four to the surgical ICU (SICU) and three to the cardiac ICU (CICU). Laboratory measurements typically had 12–14% missing data, except for serum procalcitonin (PCT), a marker for bacterial infections, with 24.5% missing, and C-reactive protein (CRP), a marker of inflammation, with 16.8% missing. Measurements assigned as ‘vital signs’ contained between 44% and 54% missing values. Stratifying patients with unspecified pneumonia further enables a more nuanced understanding of the disease, potentially facilitating tailored approaches to treatment.

To deepen clinical phenotyping for the disease group ‘unspecified pneumonia’, we calculated a k -nearest neighbor graph to cluster patients into groups and visualize these in UMAP space ( Methods ). Leiden clustering 62 identified four patient groupings with distinct clinical features that we annotated (Fig. 2e ). To identify the laboratory values, medications and pathogens that were most characteristic for these four groups (Fig. 2f ), we applied t -tests for numerical data and g -tests for categorical data between the identified groups using ehrapy (Extended Data Fig. 5 and Methods ). Based on this analysis, we identified patient groups with ‘sepsis-like, ‘severe pneumonia with co-infection’, ‘viral pneumonia’ and ‘mild pneumonia’ phenotypes. The ‘sepsis-like’ group of patients ( n  = 28) was characterized by rapid disease progression as exemplified by an increased number of deaths (adjusted P  ≤ 5.04 × 10 −3 , 43% ( n  = 28), 95% confidence interval (CI): 23%, 62%); indication of multiple organ failure, such as elevated creatinine (adjusted P  ≤ 0.01, 52.74 ± 23.71 μmol L −1 ) or reduced albumin levels (adjusted P  ≤ 2.89 × 10 −4 , 33.40 ± 6.78 g L −1 ); and increased expression levels and peaks of inflammation markers, including PCT (adjusted P  ≤ 3.01 × 10 −2 , 1.42 ± 2.03 ng ml −1 ), whole blood cell count, neutrophils, lymphocytes, monocytes and lower platelet counts (adjusted P  ≤ 6.3 × 10 −2 , 159.30 ± 142.00 × 10 9 per liter) and changes in electrolyte levels—that is, lower potassium levels (adjusted P  ≤ 0.09 × 10 −2 , 3.14 ± 0.54 mmol L −1 ). Patients whom we associated with the term ‘severe pneumonia with co-infection’ ( n  = 74) were characterized by prolonged ICU stays (adjusted P  ≤ 3.59 × 10 −4 , 15.01 ± 29.24 d); organ affection, such as higher levels of creatinine (adjusted P  ≤ 1.10 × 10 −4 , 52.74 ± 23.71 μmol L −1 ) and lower platelet count (adjusted P  ≤ 5.40 × 10 −23 , 159.30 ± 142.00 × 10 9 per liter); increased inflammation markers, such as peaks of PCT (adjusted P  ≤ 5.06 × 10 −5 , 1.42 ± 2.03 ng ml −1 ), CRP (adjusted P  ≤ 1.40 × 10 −6 , 50.60 ± 37.58 mg L −1 ) and neutrophils (adjusted P  ≤ 8.51 × 10 −6 , 13.01 ± 6.98 × 10 9 per liter); detection of bacteria in combination with additional pathogen fungals in sputum samples (adjusted P  ≤ 1.67 × 10 −2 , 26% ( n  = 74), 95% CI: 16%, 36%); and increased application of medication, including antifungals (adjusted P  ≤ 1.30 × 10 −4 , 15% ( n  = 74), 95% CI: 7%, 23%) and catecholamines (adjusted P  ≤ 2.0 × 10 −2 , 45% ( n  = 74), 95% CI: 33%, 56%). Patients in the ‘mild pneumonia’ group were characterized by positive sputum cultures in the presence of relatively lower inflammation markers, such as PCT (adjusted P  ≤ 1.63 × 10 −3 , 1.42 ± 2.03 ng ml −1 ) and CRP (adjusted P  ≤ 0.03 × 10 −1 , 50.60 ± 37.58 mg L −1 ), while receiving antibiotics more frequently (adjusted P  ≤ 1.00 × 10 −5 , 80% ( n  = 78), 95% CI: 70%, 89%) and additional medications (electrolytes, blood thinners and circulation-supporting medications) (adjusted P  ≤ 1.00 × 10 −5 , 82% ( n  = 78), 95% CI: 73%, 91%). Finally, patients in the ‘viral pneumonia’ group were characterized by shorter LOSs (adjusted P  ≤ 8.00 × 10 −6 , 15.01 ± 29.24 d), a lack of non-viral pathogen detection in combination with higher lymphocyte counts (adjusted P  ≤ 0.01, 4.11 ± 2.49 × 10 9 per liter), lower levels of PCT (adjusted P  ≤ 0.03 × 10 −2 , 1.42 ± 2.03 ng ml −1 ) and reduced application of catecholamines (adjusted P  ≤ 5.96 × 10 −7 , 15% (n = 97), 95% CI: 8%, 23%), antibiotics (adjusted P  ≤ 8.53 × 10 −6 , 41% ( n  = 97), 95% CI: 31%, 51%) and antifungals (adjusted P  ≤ 5.96 × 10 −7 , 0% ( n  = 97), 95% CI: 0%, 0%).

To demonstrate the ability of ehrapy to examine EHR data from different levels of resolution, we additionally reconstructed a case from the ‘severe pneumonia with co-infection’ group (Fig. 2g ). In this case, the analysis revealed that CRP levels remained elevated despite broad-spectrum antibiotic treatment until a positive Acinetobacter baumannii result led to a change in medication and a subsequent decrease in CRP and monocyte levels.

ehrapy facilitates extraction of pneumonia indicators

ehrapy’s survival analysis module allowed us to identify clinical indicators of disease stages that could be used as biomarkers through Kaplan–Meier analysis. We found strong variance in overall aspartate aminotransferase (AST), alanine aminotransferase (ALT), gamma-glutamyl transferase (GGT) and bilirubin levels (Fig. 3a ), including changes over time (Extended Data Fig. 6a,b ), in all four ‘unspecified pneumonia’ groups. Routinely used to assess liver function, studies provide evidence that AST, ALT and GGT levels are elevated during respiratory infections 63 , including severe pneumonia 64 , and can guide diagnosis and management of pneumonia in children 63 . We confirmed reduced survival in more severely affected children (‘sepsis-like pneumonia’ and ‘severe pneumonia with co-infection’) using Kaplan–Meier curves and a multivariate log-rank test (Fig. 3b ; P  ≤ 1.09 × 10 −18 ) through ehrapy. To verify the association of this trajectory with altered AST, ALT and GGT expression levels, we further grouped all patients based on liver enzyme reference ranges ( Methods and Supplementary Table 2 ). By Kaplan–Meier survival analysis, cases with peaks of GGT ( P  ≤ 1.4 × 10 −2 , 58.01 ± 2.03 U L −1 ), ALT ( P  ≤ 2.9 × 10 −2 , 43.59 ± 38.02 U L −1 ) and AST ( P  ≤ 4.8 × 10 −4 , 78.69 ± 60.03 U L −1 ) in ‘outside the norm’ were found to correlate with lower survival in all groups (Fig. 3c and Extended Data Fig. 6 ), in line with previous studies 63 , 65 . Bilirubin was not found to significantly affect survival ( P  ≤ 2.1 × 10 −1 , 12.57 ± 21.22 mg dl −1 ).

a , Line plots of major hepatic system laboratory measurements per group show variance in the measurements per pneumonia group. b , Kaplan–Meier survival curves demonstrate lower survival for ‘sepsis-like’ and ‘severe pneumonia with co-infection’ groups. c , Kaplan–Meier survival curves for children with GGT measurements outside the norm range display lower survival.

ehrapy quantifies medication class effect on LOS

Pneumonia requires case-specific medications due to its diverse causes. To demonstrate the potential of ehrapy’s causal inference module, we quantified the effect of medication on ICU LOS to evaluate case-specific administration of medication. In contrast to causal discovery that attempts to find a causal graph reflecting the causal relationships, causal inference is a statistical process used to investigate possible effects when altering a provided system, as represented by a causal graph and observational data (Fig. 4a ) 66 . This approach allows identifying and quantifying the impact of specific interventions or treatments on outcome measures, thereby providing insight for evidence-based decision-making in healthcare. Causal inference relies on datasets incorporating interventions to accurately quantify effects.

a , ehrapy’s causal module is based on the strategy of the tool ‘dowhy’. Here, EHR data containing treatment, outcome and measurements and a causal graph serve as input for causal effect quantification. The process includes the identification of the target estimand based on the causal graph, the estimation of causal effects using various models and, finally, refutation where sensitivity analyses and refutation tests are performed to assess the robustness of the results and assumptions. b , Curated causal graph using age, liver damage and inflammation markers as disease progression proxies together with medications as interventions to assess the causal effect on length of ICU stay. c , Determined causal effect strength on LOS in days of administered medication categories.

We manually constructed a minimal causal graph with ehrapy (Fig. 4b ) on records of treatment with corticosteroids, carbapenems, penicillins, cephalosporins and antifungal and antiviral medications as interventions (Extended Data Fig. 7 and Methods ). We assumed that the medications affect disease progression proxies, such as inflammation markers and markers of organ function. The selection of ‘interventions’ is consistent with current treatment standards for bacterial pneumonia and respiratory distress 67 , 68 . Based on the approach of the tool ‘dowhy’ 69 (Fig. 4a ), ehrapy’s causal module identified the application of corticosteroids, antivirals and carbapenems to be associated with shorter LOSs, in line with current evidence 61 , 70 , 71 , 72 . In contrast, penicillins and cephalosporins were associated with longer LOSs, whereas antifungal medication did not strongly influence LOS (Fig. 4c ).

ehrapy enables deriving population-scale risk factors

To illustrate the advantages of using a unified data management and quality control framework, such as ehrapy, we modeled myocardial infarction risk using Cox proportional hazards models on UKB 44 data. Large population cohort studies, such as the UKB, enable the investigation of common diseases across a wide range of modalities, including genomics, metabolomics, proteomics, imaging data and common clinical variables (Fig. 5a,b ). From these, we used a publicly available polygenic risk score for coronary heart disease 73 comprising 6.6 million variants, 80 nuclear magnetic resonance (NMR) spectroscopy-based metabolomics 74 features, 81 features derived from retinal optical coherence tomography 75 , 76 and the Framingham Risk Score 77 feature set, which includes known clinical predictors, such as age, sex, body mass index, blood pressure, smoking behavior and cholesterol levels. We excluded features with more than 10% missingness and imputed the remaining missing values ( Methods ). Furthermore, individuals with events up to 1 year after the sampling time were excluded from the analyses, ultimately selecting 29,216 individuals for whom all mentioned data types were available (Extended Data Figs. 8 and 9 and Methods ). Myocardial infarction, as defined by our mapping to the phecode nomenclature 51 , was defined as the endpoint (Fig. 5c ). We modeled the risk for myocardial infarction 1 year after either the metabolomic sample was obtained or imaging was performed.

a , The UKB includes 502,359 participants from 22 assessment centers. Most participants have genetic data (97%) and physical measurement data (93%), but fewer have data for complex measures, such as metabolomics, retinal imaging or proteomics. b , We found a distinct cluster of individuals (bottom right) from the Birmingham assessment center in the retinal imaging data, which is an artifact of the image acquisition process and was, thus, excluded. c , Myocardial infarctions are recorded for 15% of the male and 7% of the female study population. Kaplan–Meier estimators with 95% CIs are shown. d , For every modality combination, a linear Cox proportional hazards model was fit to determine the prognostic potential of these for myocardial infarction. Cardiovascular risk factors show expected positive log hazard ratios (log (HRs)) for increased blood pressure or total cholesterol and negative ones for sampling age and systolic blood pressure (BP). log (HRs) with 95% CIs are shown. e , Combining all features yields a C-index of 0.81. c – e , Error bars indicate 95% CIs ( n  = 29,216).

Predictive performance for each modality was assessed by fitting Cox proportional hazards (Fig. 5c ) models on each of the feature sets using ehrapy (Fig. 5d ). The age of the first occurrence served as the time to event; alternatively, date of death or date of the last record in the EHR served as censoring times. Models were evaluated using the concordance index (C-index) ( Methods ). The combination of multiple modalities successfully improved the predictive performance for coronary heart disease by increasing the C-index from 0.63 (genetic) to 0.76 (genetics, age and sex) and to 0.77 (clinical predictors) with 0.81 (imaging and clinical predictors) for combinations of feature sets (Fig. 5e ). Our finding is in line with previous observations of complementary effects between different modalities, where a broader ‘major adverse cardiac event’ phenotype was modeled in the UKB achieving a C-index of 0.72 (ref. 78 ). Adding genetic data improves predictive potential, as it is independent of sampling age and has limited prediction of other modalities 79 . The addition of metabolomic data did not improve predictive power (Fig. 5e ).

Imaging-based disease severity projection via fate mapping

To demonstrate ehrapy’s ability to handle diverse image data and recover disease stages, we embedded pulmonary imaging data obtained from patients with COVID-19 into a lower-dimensional space and computationally inferred disease progression trajectories using pseudotemporal ordering. This describes a continuous trajectory or ordering of individual points based on feature similarity 80 . Continuous trajectories enable mapping the fate of new patients onto precise states to potentially predict their future condition.

In COVID-19, a highly contagious respiratory illness caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), symptoms range from mild flu-like symptoms to severe respiratory distress. Chest x-rays typically show opacities (bilateral patchy, ground glass) associated with disease severity 81 .

We used COVID-19 chest x-ray images from the BrixIA 82 dataset consisting of 192 images (Fig. 6a ) with expert annotations of disease severity. We used the BrixIA database scores, which are based on six regions annotated by radiologists, to classify disease severity ( Methods ). We embedded raw image features using a pre-trained DenseNet model ( Methods ) and further processed this embedding into a nearest-neighbors-based UMAP space using ehrapy (Fig. 6b and Methods ). Fate mapping based on imaging information ( Methods ) determined a severity ordering from mild to critical cases (Fig. 6b–d ). Images labeled as ‘normal’ are projected to stay within the healthy group, illustrating the robustness of our approach. Images of diseased patients were ordered by disease severity, highlighting clear trajectories from ‘normal’ to ‘critical’ states despite the heterogeneity of the x-ray images stemming from, for example, different zoom levels (Fig. 6a ).

a , Randomly selected chest x-ray images from the BrixIA dataset demonstrate its variance. b , UMAP visualization of the BrixIA dataset embedding shows a separation of disease severity classes. c , Calculated pseudotime for all images increases with distance to the ‘normal’ images. d , Stream projection of fate mapping in UMAP space showcases disease severity trajectory of the COVID-19 chest x-ray images.

Detecting and mitigating biases in EHR data with ehrapy

To showcase how exploratory analysis using ehrapy can reveal and mitigate biases, we analyzed the Fairlearn 83 version of the Diabetes 130-US Hospitals 84 dataset. The dataset covers 10 years (1999–2008) of clinical records from 130 US hospitals, detailing 47 features of diabetes diagnoses, laboratory tests, medications and additional data from up to 14 d of inpatient care of 101,766 diagnosed patient visits ( Methods ). It was originally collected to explore the link between the measurement of hemoglobin A1c (HbA1c) and early readmission.

The cohort primarily consists of White and African American individuals, with only a minority of cases from Asian or Hispanic backgrounds (Extended Data Fig. 10a ). ehrapy’s cohort tracker unveiled selection and surveillance biases when filtering for Medicare recipients for further analysis, resulting in a shift of age distribution toward an age of over 60 years in addition to an increasing ratio of White participants. Using ehrapy’s visualization modules, our analysis showed that HbA1c was measured in only 18.4% of inpatients, with a higher frequency in emergency admissions compared to referral cases (Extended Data Fig. 10b ). Normalization biases can skew data relationships when standardization techniques ignore subgroup variability or assume incorrect distributions. The choice of normalization strategy must be carefully considered to avoid obscuring important factors. When normalizing the number of applied medications individually, differences in distributions between age groups remained. However, when normalizing both distributions jointly with age group as an additional group variable, differences between age groups were masked (Extended Data Fig. 10c ). To investigate missing data and imputation biases, we introduced missingness for the number of applied medications according to an MCAR mechanism, which we verified using ehrapy’s Little’s test ( P  ≤ 0.01 × 10 −2 ), and an MAR mechanism ( Methods ). Whereas imputing the mean in the MCAR case did not affect the overall location of the distribution, it led to an underestimation of the variance, with the standard deviation dropping from 8.1 in the original data to 6.8 in the imputed data (Extended Data Fig. 10d ). Mean imputation in the MAR case skewed both location and variance of the mean from 16.02 to 14.66, with a standard deviation of only 5.72 (Extended Data Fig. 10d ). Using ehrapy’s multiple imputation based MissForest 85 imputation on the MAR data resulted in a mean of 16.04 and a standard deviation of 6.45. To predict patient readmission in fewer than 30 d, we merged the three smallest race groups, ‘Asian’, ‘Hispanic’ and ‘Other’. Furthermore, we dropped the gender group ‘Unknown/Invalid’ owing to the small sample size making meaningful assessment impossible, and we performed balanced random undersampling, resulting in 5,677 cases from each condition. We observed an overall balanced accuracy of 0.59 using a logistic regression model. However, the false-negative rate was highest for the races ‘Other’ and ‘Unknown’, whereas their selection rate was lowest, and this model was, therefore, biased (Extended Data Fig. 10e ). Using ehrapy’s compatibility with existing machine learning packages, we used Fairlearn’s ThresholdOptimizer ( Methods ), which improved the selection rates for ‘Other’ from 0.32 to 0.38 and for ‘Unknown’ from 0.23 to 0.42 and the false-negative rates for ‘Other’ from 0.48 to 0.42 and for ‘Unknown’ from 0.61 to 0.45 (Extended Data Fig. 10e ).

Clustering offers a hypothesis-free alternative to supervised classification when clear hypotheses or labels are missing. It has enabled the identification of heart failure subtypes 86 and progression pathways 87 and COVID-19 severity states 88 . This concept, which is central to ehrapy, further allowed us to identify fine-grained groups of ‘unspecified pneumonia’ cases in the PIC dataset while discovering biomarkers and quantifying effects of medications on LOS. Such retroactive characterization showcases ehrapy’s ability to put complex evidence into context. This approach supports feedback loops to improve diagnostic and therapeutic strategies, leading to more efficiently allocated resources in healthcare.

ehrapy’s flexible data structures enabled us to integrate the heterogeneous UKB data for predictive performance in myocardial infarction. The different data types and distributions posed a challenge for predictive models that were overcome with ehrapy’s pre-processing modules. Our analysis underscores the potential of combining phenotypic and health data at population scale through ehrapy to enhance risk prediction.

By adapting pseudotime approaches that are commonly used in other omics domains, we successfully recovered disease trajectories from raw imaging data with ehrapy. The determined pseudotime, however, only orders data but does not necessarily provide a future projection per patient. Understanding the driver features for fate mapping in image-based datasets is challenging. The incorporation of image segmentation approaches could mitigate this issue and provide a deeper insight into the spatial and temporal dynamics of disease-related processes.

Limitations of our analyses include the lack of control for informative missingness where the absence of information represents information in itself 89 . Translation from Chinese to English in the PIC database can cause information loss and inaccuracies because the Chinese ICD-10 codes are seven characters long compared to the five-character English codes. Incompleteness of databases, such as the lack of radiology images in the PIC database, low sample sizes, underrepresentation of non-White ancestries and participant self-selection, cannot be accounted for and limit generalizability. This restricts deeper phenotyping of, for example, all ‘unspecified pneumonia’ cases with respect to their survival, which could be overcome by the use of multiple databases. Our causal inference use case is limited by unrecorded variables, such as Sequential Organ Failure Assessment (SOFA) scores, and pneumonia-related pathogens that are missing in the causal graph due to dataset constraints, such as high sparsity and substantial missing data, which risk overfitting and can lead to overinterpretation. We counterbalanced this by employing several refutation methods that statistically reject the causal hypothesis, such as a placebo treatment, a random common cause or an unobserved common cause. The longer hospital stays associated with penicillins and cephalosporins may be dataset specific and stem from higher antibiotic resistance, their use as first-line treatments, more severe initial cases, comorbidities and hospital-specific protocols.

Most analysis steps can introduce algorithmic biases where results are misleading or unfavorably affect specific groups. This is particularly relevant in the context of missing data 22 where determining the type of missing data is necessary to handle it correctly. ehrapy includes an implementation of Little’s test 90 , which tests whether data are distributed MCAR to discern missing data types. For MCAR data single-imputation approaches, such as mean, median or mode, imputation can suffice, but these methods are known to reduce variability 91 , 92 . Multiple imputation strategies, such as Multiple Imputation by Chained Equations (MICE) 93 and MissForest 85 , as implemented in ehrapy, are effective for both MCAR and MAR data 22 , 94 , 95 . MNAR data require pattern-mixture or shared-parameter models that explicitly incorporate the mechanism by which data are missing 96 . Because MNAR involves unobserved data, the assumptions about the missingness mechanism cannot be directly verified, making sensitivity analysis crucial 21 . ehrapy’s wide range of normalization functions and grouping functionality enables to account for intrinsic variability within subgroups, and its compatibility with Fairlearn 83 can potentially mitigate predictor biases. Generally, we recommend to assess all pre-processing in an iterative manner with respect to downstream applications, such as patient stratification. Moreover, sensitivity analysis can help verify the robustness of all inferred knowledge 97 .

These diverse use cases illustrate ehrapy’s potential to sufficiently address the need for a computationally efficient, extendable, reproducible and easy-to-use framework. ehrapy is compatible with major standards, such as Observational Medical Outcomes Partnership (OMOP), Common Data Model (CDM) 47 , HL7, FHIR or openEHR, with flexible support for common tabular data formats. Once loaded into an AnnData object, subsequent sharing of analysis results is made easy because AnnData objects can be stored and read platform independently. ehrapy’s rich documentation of the application programming interface (API) and extensive hands-on tutorials make EHR analysis accessible to both novices and experienced analysts.

As ehrapy remains under active development, users can expect ehrapy to continuously evolve. We are improving support for the joint analysis of EHR, genetics and molecular data where ehrapy serves as a bridge between the EHR and the omics communities. We further anticipate the generation of EHR-specific reference datasets, so-called atlases 98 , to enable query-to-reference mapping where new datasets get contextualized by transferring annotations from the reference to the new dataset. To promote the sharing and collective analysis of EHR data, we envision adapted versions of interactive single-cell data explorers, such as CELLxGENE 99 or the UCSC Cell Browser 100 , for EHR data. Such web interfaces would also include disparity dashboards 20 to unveil trends of preferential outcomes for distinct patient groups. Additional modules specifically for high-frequency time-series data, natural language processing and other data types are currently under development. With the widespread availability of code-generating large language models, frameworks such as ehrapy are becoming accessible to medical professionals without coding expertise who can leverage its analytical power directly. Therefore, ehrapy, together with a lively ecosystem of packages, has the potential to enhance the scientific discovery pipeline to shape the era of EHR analysis.

All datasets that were used during the development of ehrapy and the use cases were used according to their terms of use as indicated by each provider.

Design and implementation of ehrapy

A unified pipeline as provided by our ehrapy framework streamlines the analysis of EHR data by providing an efficient, standardized approach, which reduces the complexity and variability in data pre-processing and analysis. This consistency ensures reproducibility of results and facilitates collaboration and sharing within the research community. Additionally, the modular structure allows for easy extension and customization, enabling researchers to adapt the pipeline to their specific needs while building on a solid foundational framework.

ehrapy was designed from the ground up as an open-source effort with community support. The package, as well as all associated tutorials and dataset preparation scripts, are open source. Development takes place publicly on GitHub where the developers discuss feature requests and issues directly with users. This tight interaction between both groups ensures that we implement the most pressing needs to cater the most important use cases and can guide users when difficulties arise. The open-source nature, extensive documentation and modular structure of ehrapy are designed for other developers to build upon and extend ehrapy’s functionality where necessary. This allows us to focus ehrapy on the most important features to keep the number of dependencies to a minimum.

ehrapy was implemented in the Python programming language and builds upon numerous existing numerical and scientific open-source libraries, specifically matplotlib 101 , seaborn 102 , NumPy 103 , numba 104 , Scipy 105 , scikit-learn 53 and Pandas 106 . Although taking considerable advantage of all packages implemented, ehrapy also shares the limitations of these libraries, such as a lack of GPU support or small performance losses due to the translation layer cost for operations between the Python interpreter and the lower-level C language for matrix operations. However, by building on very widely used open-source software, we ensure seamless integration and compatibility with a broad range of tools and platforms to promote community contributions. Additionally, by doing so, we enhance security by allowing a larger pool of developers to identify and address vulnerabilities 107 . All functions are grouped into task-specific modules whose implementation is complemented with additional dependencies.

Data preparation

Dataloaders.

ehrapy is compatible with any type of vectorized data, where vectorized refers to the data being stored in structured tables in either on-disk or database form. The input and output module of ehrapy provides readers for common formats, such as OMOP, CSV tables or SQL databases through Pandas. When reading in such datasets, the data are stored in the appropriate slots in a new AnnData 46 object. ehrapy’s data module provides access to more than 20 public EHR datasets that feature diseases, including, but not limited to, Parkinson’s disease, breast cancer, chronic kidney disease and more. All dataloaders return AnnData objects to allow for immediate analysis.

AnnData for EHR data

Our framework required a versatile data structure capable of handling various matrix formats, including Numpy 103 for general use cases and interoperability, Scipy 105 sparse matrices for efficient storage, Dask 108 matrices for larger-than-memory analysis and Awkward array 109 for irregular time-series data. We needed a single data structure that not only stores data but also includes comprehensive annotations for thorough contextual analysis. It was essential for this structure to be widely used and supported, which ensures robustness and continual updates. Interoperability with other analytical packages was a key criterion to facilitate seamless integration within existing tools and workflows. Finally, the data structure had to support both in-memory operations and on-disk storage using formats such as HDF5 (ref. 110 ) and Zarr 111 , ensuring efficient handling and accessibility of large datasets and the ability to easily share them with collaborators.

All of these requirements are fulfilled by the AnnData format, which is a popular data structure in single-cell genomics. At its core, an AnnData object encapsulates diverse components, providing a holistic representation of data and metadata that are always aligned in dimensions and easily accessible. A data matrix (commonly referred to as ‘ X ’) stands as the foundational element, embodying the measured data. This matrix can be dense (as Numpy array), sparse (as Scipy sparse matrix) or ragged (as Awkward array) where dimensions do not align within the data matrix. The AnnData object can feature several such data matrices stored in ‘layers’. Examples of such layers can be unnormalized or unencoded data. These data matrices are complemented by an observations (commonly referred to as ‘obs’) segment where annotations on the level of patients or visits are stored. Patients’ age or sex, for instance, are often used as such annotations. The variables (commonly referred to as ‘var’) section complements the observations, offering supplementary details about the features in the dataset, such as missing data rates. The observation-specific matrices (commonly referred to as ‘obsm’) section extends the capabilities of the AnnData structure by allowing the incorporation of observation-specific matrices. These matrices can represent various types of information at the individual cell level, such as principal component analysis (PCA) results, t-distributed stochastic neighbor embedding (t-SNE) coordinates or other dimensionality reduction outputs. Analogously, AnnData features a variables-specific variables (commonly referred to as ‘varm’) component. The observation-specific pairwise relationships (commonly referred to as ‘obsp’) segment complements the ‘obsm’ section by accommodating observation-specific pairwise relationships. This can include connectivity matrices, indicating relationships between patients. The inclusion of an unstructured annotations (commonly referred to as ‘uns’) component further enhances flexibility. This segment accommodates unstructured annotations or arbitrary data that might not conform to the structured observations or variables categories. Any AnnData object can be stored on disk in h5ad or Zarr format to facilitate data exchange.

ehrapy natively interfaces with the scientific Python ecosystem via Pandas 112 and Numpy 103 . The development of deep learning models for EHR data 113 is further accelerated through compatibility with pathml 114 , a unified framework for whole-slide image analysis in pathology, and scvi-tools 115 , which provides data loaders for loading tensors from AnnData objects into PyTorch 116 or Jax arrays 117 to facilitate the development of generalizing foundational models for medical artificial intelligence 118 .

Feature annotation

After AnnData creation, any metadata can be mapped against ontologies using Bionty ( https://github.com/laminlabs/bionty-base ). Bionty provides access to the Human Phenotype, Phecodes, Phenotype and Trait, Drug, Mondo and Human Disease ontologies.

Key medical terms stored in an AnnData object in free text can be extracted using the Medical Concept Annotation Toolkit (MedCAT) 119 .

Data processing

Cohort tracking.

ehrapy provides a CohortTracker tool that traces all filtering steps applied to an associated AnnData object. To calculate cohort summary statistics, the implementation makes use of tableone 120 and can subsequently be plotted as bar charts together with flow diagrams 121 that visualize the order and reasoning of filtering operations.

Basic pre-processing and quality control

ehrapy encompasses a suite of functionalities for fundamental data processing that are adopted from scanpy 52 but adapted to EHR data:

Regress out: To address unwanted sources of variation, a regression procedure is integrated, enhancing the dataset’s robustness.

Subsample: Selects a specified fraction of observations.

Balanced sample: Balances groups in the dataset by random oversampling or undersampling.

Highly variable features: The identification and annotation of highly variable features following the ‘highly variable genes’ function of scanpy is seamlessly incorporated, providing users with insights into pivotal elements influencing the dataset.

To identify and minimize quality issues, ehrapy provides several quality control functions:

Basic quality control: Determines the relative and absolute number of missing values per feature and per patient.

Winsorization: For data refinement, ehrapy implements a winsorization process, creating a version of the input array less susceptible to extreme values.

Feature clipping: Imposes limits on features to enhance dataset reliability.

Detect biases: Computes pairwise correlations between features, standardized mean differences for numeric features between groups of sensitive features, categorical feature value count differences between groups of sensitive features and feature importances when predicting a target variable.

Little’s MCAR test: Applies Little’s MCAR test whose null hypothesis is that data are MCAR. Rejecting the null hypothesis may not always mean that data are not MCAR, nor is accepting the null hypothesis a guarantee that data are MCAR. For more details, see Schouten et al. 122 .

Summarize features: Calculates statistical indicators per feature, including minimum, maximum and average values. This can be especially useful to reduce complex data with multiple measurements per feature per patient into sets of columns with single values.

Imputation is crucial in data analysis to address missing values, ensuring the completeness of datasets that can be required for specific algorithms. The ‘ehrapy’ pre-processing module offers a range of imputation techniques:

Explicit Impute: Replaces missing values, in either all columns or a user-specified subset, with a designated replacement value.

Simple Impute: Imputes missing values in numerical data using mean, median or the most frequent value, contributing to a more complete dataset.

KNN Impute: Uses k -nearest neighbor imputation to fill in missing values in the input AnnData object, preserving local data patterns.

MissForest Impute: Implements the MissForest strategy for imputing missing data, providing a robust approach for handling complex datasets.

MICE Impute: Applies the MICE algorithm for imputing data. This implementation is based on the miceforest ( https://github.com/AnotherSamWilson/miceforest ) package.

Data encoding can be required if categoricals are a part of the dataset to obtain numerical values only. Most algorithms in ehrapy are compatible only with numerical values. ehrapy offers two encoding algorithms based on scikit-learn 53 :

One-Hot Encoding: Transforms categorical variables into binary vectors, creating a binary feature for each category and capturing the presence or absence of each category in a concise representation.

Label Encoding: Assigns a unique numerical label to each category, facilitating the representation of categorical data as ordinal values and supporting algorithms that require numerical input.

To ensure that the distributions of the heterogeneous data are aligned, ehrapy offers several normalization procedures:

Log Normalization: Applies the natural logarithm function to the data, useful for handling skewed distributions and reducing the impact of outliers.

Max-Abs Normalization: Scales each feature by its maximum absolute value, ensuring that the maximum absolute value for each feature is 1.

Min-Max Normalization: Transforms the data to a specific range (commonly (0, 1)) by scaling each feature based on its minimum and maximum values.

Power Transformation Normalization: Applies a power transformation to make the data more Gaussian like, often useful for stabilizing variance and improving the performance of models sensitive to distributional assumptions.

Quantile Normalization: Aligns the distributions of multiple variables, ensuring that their quantiles match, which can be beneficial for comparing datasets or removing batch effects.

Robust Scaling Normalization: Scales data using the interquartile range, making it robust to outliers and suitable for datasets with extreme values.

Scaling Normalization: Standardizes data by subtracting the mean and dividing by the standard deviation, creating a distribution with a mean of 0 and a standard deviation of 1.

Offset to Positive Values: Shifts all values by a constant offset to make all values non-negative, with the lowest negative value becoming 0.

Dataset shifts can be corrected using the scanpy implementation of the ComBat 123 algorithm, which employs a parametric and non-parametric empirical Bayes framework for adjusting data for batch effects that is robust to outliers.

Finally, a neighbors graph can be efficiently computed using scanpy’s implementation.

To obtain meaningful lower-dimensional embeddings that can subsequently be visualized and reused for downstream algorithms, ehrapy provides the following algorithms based on scanpy’s implementation:

t-SNE: Uses a probabilistic approach to embed high-dimensional data into a lower-dimensional space, emphasizing the preservation of local similarities and revealing clusters in the data.

UMAP: Embeds data points by modeling their local neighborhood relationships, offering an efficient and scalable technique that captures both global and local structures in high-dimensional data.

Force-Directed Graph Drawing: Uses a physical simulation to position nodes in a graph, with edges representing pairwise relationships, creating a visually meaningful representation that emphasizes connectedness and clustering in the data.

Diffusion Maps: Applies spectral methods to capture the intrinsic geometry of high-dimensional data by modeling diffusion processes, providing a way to uncover underlying structures and patterns.

Density Calculation in Embedding: Quantifies the density of observations within an embedding, considering conditions or groups, offering insights into the concentration of data points in different regions and aiding in the identification of densely populated areas.

ehrapy further provides algorithms for clustering and trajectory inference based on scanpy:

Leiden Clustering: Uses the Leiden algorithm to cluster observations into groups, revealing distinct communities within the dataset with an emphasis on intra-cluster cohesion.

Hierarchical Clustering Dendrogram: Constructs a dendrogram through hierarchical clustering based on specified group by categories, illustrating the hierarchical relationships among observations and facilitating the exploration of structured patterns.

Feature ranking

ehrapy provides two ways of ranking feature contributions to clusters and target variables:

Statistical tests: To compare any obtained clusters to obtain marker features that are significantly different between the groups, ehrapy extends scanpy’s ‘rank genes groups’. The original implementation, which features a t -test for numerical data, is complemented by a g -test for categorical data.

Feature importance: Calculates feature rankings for a target variable using linear regression, support vector machine or random forest models from scikit-learn. ehrapy evaluates the relative importance of each predictor by fitting the model and extracting model-specific metrics, such as coefficients or feature importances.

Dataset integration

Based on scanpy’s ‘ingest’ function, ehrapy facilitates the integration of labels and embeddings from a well-annotated reference dataset into a new dataset, enabling the mapping of cluster annotations and spatial relationships for consistent comparative analysis. This process ensures harmonized clinical interpretations across datasets, especially useful when dealing with multiple experimental diseases or batches.

Knowledge inference

Survival analysis.

ehrapy’s implementation of survival analysis algorithms is based on lifelines 124 :

Ordinary Least Squares (OLS) Model: Creates a linear regression model using OLS from a specified formula and an AnnData object, allowing for the analysis of relationships between variables and observations.

Generalized Linear Model (GLM): Constructs a GLM from a given formula, distribution and AnnData, providing a versatile framework for modeling relationships with nonlinear data structures.

Kaplan–Meier: Fits the Kaplan–Meier curve to generate survival curves, offering a visual representation of the probability of survival over time in a dataset.

Cox Hazard Model: Constructs a Cox proportional hazards model using a specified formula and an AnnData object, enabling the analysis of survival data by modeling the hazard rates and their relationship to predictor variables.

Log-Rank Test: Calculates the P value for the log-rank test, comparing the survival functions of two groups, providing statistical significance for differences in survival distributions.

GLM Comparison: Given two fit GLMs, where the larger encompasses the parameter space of the smaller, this function returns the P value, indicating the significance of the larger model and adding explanatory power beyond the smaller model.

Trajectory inference

Trajectory inference is a computational approach that reconstructs and models the developmental paths and transitions within heterogeneous clinical data, providing insights into the temporal progression underlying complex systems. ehrapy offers several inbuilt algorithms for trajectory inference based on scanpy:

Diffusion Pseudotime: Infers the progression of observations by measuring geodesic distance along the graph, providing a pseudotime metric that represents the developmental trajectory within the dataset.

Partition-based Graph Abstraction (PAGA): Maps out the coarse-grained connectivity structures of complex manifolds using a partition-based approach, offering a comprehensive visualization of relationships in high-dimensional data and aiding in the identification of macroscopic connectivity patterns.

Because ehrapy is compatible with scverse, further trajectory inference-based algorithms, such as CellRank, can be seamlessly applied.

Causal inference

ehrapy’s causal inference module is based on ‘dowhy’ 69 . It is based on four key steps that are all implemented in ehrapy:

Graphical Model Specification: Define a causal graphical model representing relationships between variables and potential causal effects.

Causal Effect Identification: Automatically identify whether a causal effect can be inferred from the given data, addressing confounding and selection bias.

Causal Effect Estimation: Employ automated tools to estimate causal effects, using methods such as matching, instrumental variables or regression.

Sensitivity Analysis and Testing: Perform sensitivity analysis to assess the robustness of causal inferences and conduct statistical testing to determine the significance of the estimated causal effects.

Patient stratification

ehrapy’s complete pipeline from pre-processing to the generation of lower-dimensional embeddings, clustering, statistical comparison between determined groups and more facilitates the stratification of patients.

Visualization

ehrapy features an extensive visualization pipeline that is customizable and yet offers reasonable defaults. Almost every analysis function is matched with at least one visualization function that often shares the name but is available through the plotting module. For example, after importing ehrapy as ‘ep’, ‘ep.tl.umap(adata)’ runs the UMAP algorithm on an AnnData object, and ‘ep.pl.umap(adata)’ would then plot a scatter plot of the UMAP embedding.

ehrapy further offers a suite of more generally usable and modifiable plots:

Scatter Plot: Visualizes data points along observation or variable axes, offering insights into the distribution and relationships between individual data points.

Heatmap: Represents feature values in a grid, providing a comprehensive overview of the data’s structure and patterns.

Dot Plot: Displays count values of specified variables as dots, offering a clear depiction of the distribution of counts for each variable.

Filled Line Plot: Illustrates trends in data with filled lines, emphasizing variations in values over a specified axis.

Violin Plot: Presents the distribution of data through mirrored density plots, offering a concise view of the data’s spread.

Stacked Violin Plot: Combines multiple violin plots, stacked to allow for visual comparison of distributions across categories.

Group Mean Heatmap: Creates a heatmap displaying the mean count per group for each specified variable, providing insights into group-wise trends.

Hierarchically Clustered Heatmap: Uses hierarchical clustering to arrange data in a heatmap, revealing relationships and patterns among variables and observations.

Rankings Plot: Visualizes rankings within the data, offering a clear representation of the order and magnitude of values.

Dendrogram Plot: Plots a dendrogram of categories defined in a group by operation, illustrating hierarchical relationships within the dataset.

Benchmarking ehrapy

We generated a subset of the UKB data selecting 261 features and 488,170 patient visits. We removed all features with missingness rates greater than 70%. To demonstrate speed and memory consumption for various scenarios, we subsampled the data to 20%, 30% and 50%. We ran a minimal ehrapy analysis pipeline on each of those subsets and the full data, including the calculation of quality control metrics, filtering of variables by a missingness threshold, nearest neighbor imputation, normalization, dimensionality reduction and clustering (Supplementary Table 1 ). We conducted our benchmark on a single CPU with eight threads and 60 GB of maximum memory.

ehrapy further provides out-of-core implementations using Dask 108 for many algorithms in ehrapy, such as our normalization functions or our PCA implementation. Out-of-core computation refers to techniques that process data that do not fit entirely in memory, using disk storage to manage data overflow. This approach is crucial for handling large datasets without being constrained by system memory limits. Because the principal components get reused for other computationally expensive algorithms, such as the neighbors graph calculation, it effectively enables the analysis of very large datasets. We are currently working on supporting out-of-core computation for all computationally expensive algorithms in ehrapy.

We demonstrate the memory benefits in a hosted tutorial where the in-memory pipeline for 50,000 patients with 1,000 features required about 2 GB of memory, and the corresponding out-of-core implementation required less than 200 MB of memory.

The code for benchmarking is available at https://github.com/theislab/ehrapy-reproducibility . The implementation of ehrapy is accessible at https://github.com/theislab/ehrapy together with extensive API documentation and tutorials at https://ehrapy.readthedocs.io .

PIC database analysis

Study design.

We collected clinical data from the PIC 43 version 1.1.0 database. PIC is a single-center, bilingual (English and Chinese) database hosting information of children admitted to critical care units at the Children’s Hospital of Zhejiang University School of Medicine in China. The requirement for individual patient consent was waived because the study did not impact clinical care, and all protected health information was de-identified. The database contains 13,499 distinct hospital admissions of 12,881 distinct pediatric patients. These patients were admitted to five ICU units with 119 total critical care beds—GICU, PICU, SICU, CICU and NICU—between 2010 and 2018. The mean age of the patients was 2.5 years, of whom 42.5% were female. The in-hospital mortality was 7.1%; the mean hospital stay was 17.6 d; the mean ICU stay was 9.3 d; and 468 (3.6%) patients were admitted multiple times. Demographics, diagnoses, doctors’ notes, laboratory and microbiology tests, prescriptions, fluid balances, vital signs and radiographics reports were collected from all patients. For more details, see the original publication of Zeng et al. 43 .

Study participants

Individuals older than 18 years were excluded from the study. We grouped the data into three distinct groups: ‘neonates’ (0–28 d of age; 2,968 patients), ‘infants’ (1–12 months of age; 4,876 patients) and ‘youths’ (13 months to 18 years of age; 6,097 patients). We primarily analyzed the ‘youths’ group with the discharge diagnosis ‘unspecified pneumonia’ (277 patients).

Data collection

The collected clinical data included demographics, laboratory and vital sign measurements, diagnoses, microbiology and medication information and mortality outcomes. The five-character English ICD-10 codes were used, whose values are based on the seven-character Chinese ICD-10 codes.

Dataset extraction and analysis

We downloaded the PIC database of version 1.1.0 from Physionet 1 to obtain 17 CSV tables. Using Pandas, we selected all information with more than 50% coverage rate, including demographics and laboratory and vital sign measurements (Fig. 2 ). To reduce the amount of noise, we calculated and added only the minimum, maximum and average of all measurements that had multiple values per patient. Examination reports were removed because they describe only diagnostics and not detailed findings. All further diagnoses and microbiology and medication information were included into the observations slot to ensure that the data were not used for the calculation of embeddings but were still available for the analysis. This ensured that any calculated embedding would not be divided into treated and untreated groups but, rather, solely based on phenotypic features. We imputed all missing data through k -nearest neighbors imputation ( k  = 20) using the knn_impute function of ehrapy. Next, we log normalized the data with ehrapy using the log_norm function. Afterwards, we winsorized the data using ehrapy’s winsorize function to obtain 277 ICU visits ( n  = 265 patients) with 572 features. Of those 572 features, 254 were stored in the matrix X and the remaining 318 in the ‘obs’ slot in the AnnData object. For clustering and visualization purposes, we calculated 50 principal components using ehrapy’s pca function. The obtained principal component representation was then used to calculate a nearest neighbors graph using the neighbors function of ehrapy. The nearest neighbors graph then served as the basis for a UMAP embedding calculation using ehrapy’s umap function.

We applied the community detection algorithm Leiden with resolution 0.6 on the nearest neighbor graph using ehrapy’s leiden function. The four obtained clusters served as input for two-sided t -tests for all numerical values and two-sided g -tests for all categorical values for all four clusters against the union of all three other clusters, respectively. This was conducted using ehrapy’s rank_feature_groups function, which also corrects P values for multiple testing with the Benjamini–Hochberg method 125 . We presented the four groups and the statistically significantly different features between the groups to two pediatricians who annotated the groups with labels.

Our determined groups can be confidently labeled owing to their distinct clinical profiles. Nevertheless, we could only take into account clinical features that were measured. Insightful features, such as lung function tests, are missing. Moreover, the feature representation of the time-series data is simplified, which can hide some nuances between the groups. Generally, deciding on a clustering resolution is difficult. However, more fine-grained clusters obtained via higher clustering resolutions may become too specific and not generalize well enough.

Kaplan–Meier survival analysis

We selected patients with up to 360 h of total stay for Kaplan–Meier survival analysis to ensure a sufficiently high number of participants. We proceeded with the AnnData object prepared as described in the ‘Patient stratification’ subsection to conduct Kaplan–Meier analysis among all four determined pneumonia groups using ehrapy’s kmf function. Significance was tested through ehrapy’s test_kmf_logrank function, which tests whether two Kaplan–Meier series are statistically significant, employing a chi-squared test statistic under the null hypothesis. Let h i (t) be the hazard ratio of group i at time t and c a constant that represents a proportional change in the hazard ratio between the two groups, then:

This implicitly uses the log-rank weights. An additional Kaplan–Meier analysis was conducted for all children jointly concerning the liver markers AST, ALT and GGT. To determine whether measurements were inside or outside the norm range, we used reference ranges (Supplementary Table 2 ). P values less than 0.05 were labeled significant.

Our Kaplan–Meier curve analysis depends on the groups being well defined and shares the same limitations as the patient stratification. Additionally, the analysis is sensitive to the reference table where we selected limits that generalize well for the age ranges, but, due to children of different ages being examined, they may not necessarily be perfectly accurate for all children.

Causal effect of mechanism of action on LOS

Although the dataset was not initially intended for investigating causal effects of interventions, we adapted it for this purpose by focusing on the LOS in the ICU, measured in months, as the outcome variable. This choice aligns with the clinical aim of stabilizing patients sufficiently for ICU discharge. We constructed a causal graph to explore how different drug administrations could potentially reduce the LOS. Based on consultations with clinicians, we included several biomarkers of liver damage (AST, ALT and GGT) and inflammation (CRP and PCT) in our model. Patient age was also considered a relevant variable.

Because several different medications act by the same mechanisms, we grouped specific medications by their drug classes This grouping was achieved by cross-referencing the drugs listed in the dataset with DrugBank release 5.1 (ref. 126 ), using Levenshtein distances for partial string matching. After manual verification, we extracted the corresponding DrugBank categories, counted the number of features per category and compiled a list of commonly prescribed medications, as advised by clinicians. This approach facilitated the modeling of the causal graph depicted in Fig. 4 , where an intervention is defined as the administration of at least one drug from a specified category.

Causal inference was then conducted with ehrapy’s ‘dowhy’ 69 -based causal inference module using the expert-curated causal graph. Medication groups were designated as causal interventions, and the LOS was the outcome of interest. Linear regression served as the estimation method for analyzing these causal effects. We excluded four patients from the analysis owing to their notably long hospital stays exceeding 90 d, which were deemed outliers. To validate the robustness of our causal estimates, we incorporated several refutation methods:

Placebo Treatment Refuter: This method involved replacing the treatment assignment with a placebo to test the effect of the treatment variable being null.

Random Common Cause: A randomly generated variable was added to the data to assess the sensitivity of the causal estimate to the inclusion of potential unmeasured confounders.

Data Subset Refuter: The stability of the causal estimate was tested across various random subsets of the data to ensure that the observed effects were not dependent on a specific subset.

Add Unobserved Common Cause: This approach tested the effect of an omitted variable by adding a theoretically relevant unobserved confounder to the model, evaluating how much an unmeasured variable could influence the causal relationship.

Dummy Outcome: Replaces the true outcome variable with a random variable. If the causal effect nullifies, it supports the validity of the original causal relationship, indicating that the outcome is not driven by random factors.

Bootstrap Validation: Employs bootstrapping to generate multiple samples from the dataset, testing the consistency of the causal effect across these samples.

The selection of these refuters addresses a broad spectrum of potential biases and model sensitivities, including unobserved confounders and data dependencies. This comprehensive approach ensures robust verification of the causal analysis. Each refuter provides an orthogonal perspective, targeting specific vulnerabilities in causal analysis, which strengthens the overall credibility of the findings.

UKB analysis

Study population.

We used information from the UKB cohort, which includes 502,164 study participants from the general UK population without enrichment for specific diseases. The study involved the enrollment of individuals between 2006 and 2010 across 22 different assessment centers throughout the United Kingdom. The tracking of participants is still ongoing. Within the UKB dataset, metabolomics, proteomics and retinal optical coherence tomography data are available for a subset of individuals without any enrichment for specific diseases. Additionally, EHRs, questionnaire responses and other physical measures are available for almost everyone in the study. Furthermore, a variety of genotype information is available for nearly the entire cohort, including whole-genome sequencing, whole-exome sequencing, genotyping array data as well as imputed genotypes from the genotyping array 44 . Because only the latter two are available for download, and are sufficient for polygenic risk score calculation as performed here, we used the imputed genotypes in the present study. Participants visited the assessment center up to four times for additional and repeat measurements and completed additional online follow-up questionnaires.

In the present study, we restricted the analyses to data obtained from the initial assessment, including the blood draw, for obtaining the metabolomics data and the retinal imaging as well as physical measures. This restricts the study population to 33,521 individuals for whom all of these modalities are available. We have a clear study start point for each individual with the date of their initial assessment center visit. The study population has a mean age of 57 years, is 54% female and is censored at age 69 years on average; 4.7% experienced an incident myocardial infarction; and 8.1% have prevalent type 2 diabetes. The study population comes from six of the 22 assessment centers due to the retinal imaging being performed only at those.

For the myocardial infarction endpoint definition, we relied on the first occurrence data available in the UKB, which compiles the first date that each diagnosis was recorded for a participant in a hospital in ICD-10 nomenclature. Subsequently, we mapped these data to phecodes and focused on phecode 404.1 for myocardial infarction.

The Framingham Risk Score was developed on data from 8,491 participants in the Framingham Heart Study to assess general cardiovascular risk 77 . It includes easily obtainable predictors and is, therefore, easily applicable in clinical practice, although newer and more specific risk scores exist and might be used more frequently. It includes age, sex, smoking behavior, blood pressure, total and low-density lipoprotein cholesterol as well as information on insulin, antihypertensive and cholesterol-lowering medications, all of which are routinely collected in the UKB and used in this study as the Framingham feature set.

The metabolomics data used in this study were obtained using proton NMR spectroscopy, a low-cost method with relatively low batch effects. It covers established clinical predictors, such as albumin and cholesterol, as well as a range of lipids, amino acids and carbohydrate-related metabolites.

The retinal optical coherence tomography–derived features were returned by researchers to the UKB 75 , 76 . They used the available scans and determined the macular volume, macular thickness, retinal pigment epithelium thickness, disc diameter, cup-to-disk ratio across different regions as well as the thickness between the inner nuclear layer and external limiting membrane, inner and outer photoreceptor segments and the retinal pigment epithelium across different regions. Furthermore, they determined a wide range of quality metrics for each scan, including the image quality score, minimum motion correlation and inner limiting membrane (ILM) indicator.

Data analysis

After exporting the data from the UKB, all timepoints were transformed into participant age entries. Only participants without prevalent myocardial infarction (relative to the first assessment center visit at which all data were collected) were included.

The data were pre-processed for retinal imaging and metabolomics subsets separately, to enable a clear analysis of missing data and allow for the k -nearest neighbors–based imputation ( k  = 20) of missing values when less than 10% were missing for a given participant. Otherwise, participants were dropped from the analyses. The imputed genotypes and Framingham analyses were available for almost every participant and, therefore, not imputed. Individuals without them were, instead, dropped from the analyses. Because genetic risk modeling poses entirely different methodological and computational challenges, we applied a published polygenic risk score for coronary heart disease using 6.6 million variants 73 . This was computed using the plink2 score option on the imputed genotypes available in the UKB.

UMAP embeddings were computed using default parameters on the full feature sets with ehrapy’s umap function. For all analyses, the same time-to-event and event-indicator columns were used. The event indicator is a Boolean variable indicating whether a myocardial infarction was observed for a study participant. The time to event is defined as the timespan between the start of the study, in this case the date of the first assessment center visit. Otherwise, it is the timespan from the start of the study to the start of censoring; in this case, this is set to the last date for which EHRs were available, unless a participant died, in which case the date of death is the start of censoring. Kaplan–Meier curves and Cox proportional hazards models were fit using ehrapy’s survival analysis module and the lifelines 124 package’s Cox-PHFitter function with default parameters. For Cox proportional hazards models with multiple feature sets, individually imputed and quality-controlled feature sets were concatenated, and the model was fit on the resulting matrix. Models were evaluated using the C-index 127 as a metric. It can be seen as an extension of the common area under the receiver operator characteristic score to time-to-event datasets, in which events are not observed for every sample and which ranges from 0.0 (entirely false) over 0.5 (random) to 1.0 (entirely correct). CIs for the C-index were computed based on bootstrapping by sampling 1,000 times with replacement from all computed partial hazards and computing the C-index over each of these samples. The percentiles at 2.5% and 97.5% then give the upper and lower confidence bound for the 95% CIs.

In all UKB analyses, the unit of study for a statistical test or predictive model is always an individual study participant.

The generalizability of the analysis is limited as the UK Biobank cohort may not represent the general population, with potential selection biases and underrepresentation of the different demographic groups. Additionally, by restricting analysis to initial assessment data and censoring based on the last available EHR or date of death, our analysis does not account for longitudinal changes and can introduce follow-up bias, especially if participants lost to follow-up have different risk profiles.

In-depth quality control of retina-derived features

A UMAP plot of the retina-derived features indicating the assessment centers shows a cluster of samples that lie somewhat outside the general population and mostly attended the Birmingham assessment center (Fig. 5b ). To further investigate this, we performed Leiden clustering of resolution 0.3 (Extended Data Fig. 9a ) and isolated this group in cluster 5. When comparing cluster 5 to the rest of the population in the retina-derived feature space, we noticed that many individuals in cluster 5 showed overall retinal pigment epithelium (RPE) thickness measures substantially elevated over the rest of the population in both eyes (Extended Data Fig. 9b ), which is mostly a feature of this cluster (Extended Data Fig. 9c ). To investigate potential confounding, we computed ratios between cluster 5 and the rest of the population over the ‘obs’ DataFrame containing the Framingham features, diabetes-related phecodes and genetic principal components. Out of the top and bottom five highest ratios observed, six are in genetic principal components, which are commonly used to represent genetic ancestry in a continuous space (Extended Data Fig. 9d ). Additionally, diagnoses for type 1 and type 2 diabetes and antihypertensive use are enriched in cluster 5. Further investigating the ancestry, we computed log ratios for self-reported ancestries and absolute counts, which showed no robust enrichment and depletion effects.

A closer look at three quality control measures of the imaging pipeline revealed that cluster 5 was an outlier in terms of either image quality (Extended Data Fig. 9e ) or minimum motion correlation (Extended Data Fig. 9f ) and the ILM indicator (Extended Data Fig. 9g ), all of which can be indicative of artifacts in image acquisition and downstream processing 128 . Subsequently, we excluded 301 individuals from cluster 5 from all analyses.

COVID-19 chest-x-ray fate determination

Dataset overview.

We used the public BrixIA COVID-19 dataset, which contains 192 chest x-ray images annotated with BrixIA scores 82 . Hereby, six regions were annotated by a senior radiologist with more than 20 years of experience and a junior radiologist with a disease severity score ranging from 0 to 3. A global score was determined as the sum of all of these regions and, therefore, ranges from 0 to 18 (S-Global). S-Global scores of 0 were classified as normal. Images that only had severity values up to 1 in all six regions were classified as mild. Images with severity values greater than or equal to 2, but a S-Global score of less than 7, were classified as moderate. All images that contained at least one 3 in any of the six regions with a S-Global score between 7 and 10 were classified as severe, and all remaining images with S-Global scores greater than 10 with at least one 3 were labeled critical. The dataset and instructions to download the images can be found at https://github.com/ieee8023/covid-chestxray-dataset .

We first resized all images to 224 × 224. Afterwards, the images underwent a random affine transformation that involved rotation, translation and scaling. The rotation angle was randomly selected from a range of −45° to 45°. The images were also subject to horizontal and vertical translation, with the maximum translation being 15% of the image size in either direction. Additionally, the images were scaled by a factor ranging from 0.85 to 1.15. The purpose of applying these transformations was to enhance the dataset and introduce variations, ultimately improving the robustness and generalization of the model.

To generate embeddings, we used a pre-trained DenseNet model with weights densenet121-res224-all of TorchXRayVision 129 . A DenseNet is a convolutional neural network that makes use of dense connections between layers (Dense Blocks) where all layers (with matching feature map sizes) directly connect with each other. To maintain a feed-forward nature, every layer in the DenseNet architecture receives supplementary inputs from all preceding layers and transmits its own feature maps to all subsequent layers. The model was trained on the nih-pc- chex-mimic_ch-google-openi-rsna dataset 130 .

Next, we calculated 50 principal components on the feature representation of the DenseNet model of all images using ehrapy’s pca function. The principal component representation served as input for a nearest neighbors graph calculation using ehrapy’s neighbors function. This graph served as the basis for the calculation of a UMAP embedding with three components that was finally visualized using ehrapy.

We randomly picked a root in the group of images that was labeled ‘Normal’. First, we calculated so-called pseudotime by fitting a trajectory through the calculated UMAP space using diffusion maps as implemented in ehrapy’s dpt function 57 . Each image’s pseudotime value represents its estimated position along this trajectory, serving as a proxy for its severity stage relative to others in the dataset. To determine fates, we employed CellRank 58 , 59 with the PseudotimeKernel . This kernel computes transition probabilities for patient visits based on the connectivity of the k -nearest neighbors graph and the pseudotime values of patient visits, which resembles their progression through a process. Directionality is infused in the nearest neighbors graph in this process where the kernel either removes or downweights edges in the graph that contradict the directional flow of increasing pseudotime, thereby refining the graph to better reflect the developmental trajectory. We computed the transition matrix with a soft threshold scheme (Parameter of the PseudotimeKernel ), which downweights edges that point against the direction of increasing pseudotime. Finally, we calculated a projection on top of the UMAP embedding with CellRank using the plot_projection function of the PseudotimeKernel that we subsequently plotted.

This analysis is limited by the small dataset of 192 chest x-ray images, which may affect the model’s generalizability and robustness. Annotation subjectivity from radiologists can further introduce variability in severity scores. Additionally, the random selection of a root from ‘Normal’ images can introduce bias in pseudotime calculations and subsequent analyses.

Diabetes 130-US hospitals analysis

We used data from the Diabetes 130-US hospitals dataset that were collected between 1999 and 2008. It contains clinical care information at 130 hospitals and integrated delivery networks. The extracted database information pertains to hospital admissions specifically for patients diagnosed with diabetes. These encounters required a hospital stay ranging from 1 d to 14 d, during which both laboratory tests and medications were administered. The selection criteria focused exclusively on inpatient encounters with these defined characteristics. More specifically, we used a version that was curated by the Fairlearn team where the target variable ‘readmitted’ was binarized and a few features renamed or binned ( https://fairlearn.org/main/user_guide/datasets/diabetes_hospital_data.html ). The dataset contains 101,877 patient visits and 25 features. The dataset predominantly consists of White patients (74.8%), followed by African Americans (18.9%), with other racial groups, such as Hispanic, Asian and Unknown categories, comprising smaller percentages. Females make up a slight majority in the data at 53.8%, with males accounting for 46.2% and a negligible number of entries listed as unknown or invalid. A substantial majority of the patients are over 60 years of age (67.4%), whereas those aged 30–60 years represent 30.2%, and those 30 years or younger constitute just 2.5%.

All of the following descriptions start by loading the Fairlearn version of the Diabetes 130-US hospitals dataset using ehrapy’s dataloader as an AnnData object.

Selection and filtering bias

An overview of sensitive variables was generated using tableone. Subsequently, ehrapy’s CohortTracker was used to track the age, gender and race variables. The cohort was filtered for all Medicare recipients and subsequently plotted.

Surveillance bias

We plotted the HbA1c measurement ratios using ehrapy’s catplot .

Missing data and imputation bias

MCAR-type missing data for the number of medications variable (‘num_medications‘) were introduced by randomly setting 30% of the variables to be missing using Numpy’s choice function. We tested that the data are MCAR by applying ehrapy’s implementation of Little’s MCAR test, which returned a non-significant P value of 0.71. MAR data for the number of medications variable (‘num_medications‘) were introduced by scaling the ‘time_in_hospital’ variable to have a mean of 0 and a standard deviation of 1, adjusting these values by multiplying by 1.2 and subtracting 0.6 to influence overall missingness rate, and then using these values to generate MAR data in the ‘num_medications’ variable via a logistic transformation and binomial sampling. We verified that the newly introduced missing values are not MCAR with respect to the ‘time_in_hospital’ variable by applying ehrapy’s implementation of Little’s test, which was significant (0.01 × 10 −2 ). The missing data were imputed using ehrapy’s mean imputation and MissForest implementation.

Algorithmic bias

Variables ‘race’, ‘gender’, ‘age’, ‘readmitted’, ‘readmit_binary’ and ‘discharge_disposition_id’ were moved to the ‘obs’ slot of the AnnData object to ensure that they were not used for model training. We built a binary label ‘readmit_30_days’ indicating whether a patient had been readmitted in fewer than 30 d. Next, we combined the ‘Asian’ and ‘Hispanic’ categories into a single ‘Other’ category within the ‘race’ column of our AnnData object and then filtered out and discarded any samples labeled as ‘Unknown/Invalid’ under the ‘gender‘ column and subsequently moved the ‘gender’ data to the variable matrix X of the AnnData object. All categorical variables got encoded. The data were split into train and test groups with a test size of 50%. The data were scaled, and a logistic regression model was trained using scikit-learn, which was also used to determine the balanced accuracy score. Fairlearn’s MetricFrame function was used to inspect the target model performance against the sensitive variable ‘race’. We subsequently fit Fairlearn’s ThresholdOptimizer using the logistic regression estimator with balanced_accuracy_score as the target object. The algorithmic demonstration of Fairlearn’s abilities on this dataset is shown here: https://github.com/fairlearn/talks/tree/main/2021_scipy_tutorial .

Normalization bias

We one-hot encoded all categorical variables with ehrapy using the encode function. We applied ehrapy’s implementation of scaling normalization with and without the ‘Age group’ variable as group key to scale the data jointly and separately using ehrapy’s scale_norm function.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

Physionet provides access to the PIC database 43 at https://physionet.org/content/picdb/1.1.0 for credentialed users. The BrixIA images 82 are available at https://github.com/BrixIA/Brixia-score-COVID-19 . The data used in this study were obtained from the UK Biobank 44 ( https://www.ukbiobank.ac.uk/ ). Access to the UK Biobank resource was granted under application number 49966. The data are available to researchers upon application to the UK Biobank in accordance with their data access policies and procedures. The Diabetes 130-US Hospitals dataset is available at https://archive.ics.uci.edu/dataset/296/diabetes+130-us+hospitals+for+years+1999-2008 .

Code availability

The ehrapy source code is available at https://github.com/theislab/ehrapy under an Apache 2.0 license. Further documentation, tutorials and examples are available at https://ehrapy.readthedocs.io . We are actively developing the software and invite contributions from the community.

Jupyter notebooks to reproduce our analysis and figures, including Conda environments that specify all versions, are available at https://github.com/theislab/ehrapy-reproducibility .

Goldberger, A. L. et al. PhysioBank, PhysioToolkit, and PhysioNet: components of a new research resource for complex physiologic signals. Circulation 101 , E215–E220 (2000).

Article   CAS   PubMed   Google Scholar  

Atasoy, H., Greenwood, B. N. & McCullough, J. S. The digitization of patient care: a review of the effects of electronic health records on health care quality and utilization. Annu. Rev. Public Health 40 , 487–500 (2019).

Article   PubMed   Google Scholar  

Jamoom, E. W., Patel, V., Furukawa, M. F. & King, J. EHR adopters vs. non-adopters: impacts of, barriers to, and federal initiatives for EHR adoption. Health (Amst.) 2 , 33–39 (2014).

Google Scholar  

Rajkomar, A. et al. Scalable and accurate deep learning with electronic health records. NPJ Digit. Med. 1 , 18 (2018).

Article   PubMed   PubMed Central   Google Scholar  

Wolf, A. et al. Data resource profile: Clinical Practice Research Datalink (CPRD) Aurum. Int. J. Epidemiol. 48 , 1740–1740g (2019).

Sudlow, C. et al. UK biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age. PLoS Med. 12 , e1001779 (2015).

Pollard, T. J. et al. The eICU Collaborative Research Database, a freely available multi-center database for critical care research. Sci. Data 5 , 180178 (2018).

Johnson, A. E. W. et al. MIMIC-III, a freely accessible critical care database. Sci. Data 3 , 160035 (2016).

Article   CAS   PubMed   PubMed Central   Google Scholar  

Hyland, S. L. et al. Early prediction of circulatory failure in the intensive care unit using machine learning. Nat. Med. 26 , 364–373 (2020).

Rasmy, L. et al. Recurrent neural network models (CovRNN) for predicting outcomes of patients with COVID-19 on admission to hospital: model development and validation using electronic health record data. Lancet Digit. Health 4 , e415–e425 (2022).

Marcus, J. L. et al. Use of electronic health record data and machine learning to identify candidates for HIV pre-exposure prophylaxis: a modelling study. Lancet HIV 6 , e688–e695 (2019).

Kruse, C. S., Stein, A., Thomas, H. & Kaur, H. The use of electronic health records to support population health: a systematic review of the literature. J. Med. Syst. 42 , 214 (2018).

Sheikh, A., Jha, A., Cresswell, K., Greaves, F. & Bates, D. W. Adoption of electronic health records in UK hospitals: lessons from the USA. Lancet 384 , 8–9 (2014).

Sheikh, A. et al. Health information technology and digital innovation for national learning health and care systems. Lancet Digit. Health 3 , e383–e396 (2021).

Cord, K. A. M., Mc Cord, K. A. & Hemkens, L. G. Using electronic health records for clinical trials: where do we stand and where can we go? Can. Med. Assoc. J. 191 , E128–E133 (2019).

Article   Google Scholar  

Landi, I. et al. Deep representation learning of electronic health records to unlock patient stratification at scale. NPJ Digit. Med. 3 , 96 (2020).

Ayaz, M., Pasha, M. F., Alzahrani, M. Y., Budiarto, R. & Stiawan, D. The Fast Health Interoperability Resources (FHIR) standard: systematic literature review of implementations, applications, challenges and opportunities. JMIR Med. Inform. 9 , e21929 (2021).

Peskoe, S. B. et al. Adjusting for selection bias due to missing data in electronic health records-based research. Stat. Methods Med. Res. 30 , 2221–2238 (2021).

Haneuse, S. & Daniels, M. A general framework for considering selection bias in EHR-based studies: what data are observed and why? EGEMS (Wash. DC) 4 , 1203 (2016).

PubMed   Google Scholar  

Gallifant, J. et al. Disparity dashboards: an evaluation of the literature and framework for health equity improvement. Lancet Digit. Health 5 , e831–e839 (2023).

Sauer, C. M. et al. Leveraging electronic health records for data science: common pitfalls and how to avoid them. Lancet Digit. Health 4 , e893–e898 (2022).

Li, J. et al. Imputation of missing values for electronic health record laboratory data. NPJ Digit. Med. 4 , 147 (2021).

Rubin, D. B. Inference and missing data. Biometrika 63 , 581 (1976).

Scheid, L. M., Brown, L. S., Clark, C. & Rosenfeld, C. R. Data electronically extracted from the electronic health record require validation. J. Perinatol. 39 , 468–474 (2019).

Phelan, M., Bhavsar, N. A. & Goldstein, B. A. Illustrating informed presence bias in electronic health records data: how patient interactions with a health system can impact inference. EGEMS (Wash. DC). 5 , 22 (2017).

PubMed   PubMed Central   Google Scholar  

Secondary Analysis of Electronic Health Records (ed MIT Critical Data) (Springer, 2016).

Jetley, G. & Zhang, H. Electronic health records in IS research: quality issues, essential thresholds and remedial actions. Decis. Support Syst. 126 , 113137 (2019).

McCormack, J. P. & Holmes, D. T. Your results may vary: the imprecision of medical measurements. BMJ 368 , m149 (2020).

Hobbs, F. D. et al. Is the international normalised ratio (INR) reliable? A trial of comparative measurements in hospital laboratory and primary care settings. J. Clin. Pathol. 52 , 494–497 (1999).

Huguet, N. et al. Using electronic health records in longitudinal studies: estimating patient attrition. Med. Care 58 Suppl 6 Suppl 1 , S46–S52 (2020).

Zeng, J., Gensheimer, M. F., Rubin, D. L., Athey, S. & Shachter, R. D. Uncovering interpretable potential confounders in electronic medical records. Nat. Commun. 13 , 1014 (2022).

Getzen, E., Ungar, L., Mowery, D., Jiang, X. & Long, Q. Mining for equitable health: assessing the impact of missing data in electronic health records. J. Biomed. Inform. 139 , 104269 (2023).

Tang, S. et al. Democratizing EHR analyses with FIDDLE: a flexible data-driven preprocessing pipeline for structured clinical data. J. Am. Med. Inform. Assoc. 27 , 1921–1934 (2020).

Dagliati, A. et al. A process mining pipeline to characterize COVID-19 patients’ trajectories and identify relevant temporal phenotypes from EHR data. Front. Public Health 10 , 815674 (2022).

Sun, Y. & Zhou, Y.-H. A machine learning pipeline for mortality prediction in the ICU. Int. J. Digit. Health 2 , 3 (2022).

Article   CAS   Google Scholar  

Mandyam, A., Yoo, E. C., Soules, J., Laudanski, K. & Engelhardt, B. E. COP-E-CAT: cleaning and organization pipeline for EHR computational and analytic tasks. In Proc. of the 12th ACM Conference on Bioinformatics, Computational Biology, and Health Informatics. https://doi.org/10.1145/3459930.3469536 (Association for Computing Machinery, 2021).

Gao, C. A. et al. A machine learning approach identifies unresolving secondary pneumonia as a contributor to mortality in patients with severe pneumonia, including COVID-19. J. Clin. Invest. 133 , e170682 (2023).

Makam, A. N. et al. The good, the bad and the early adopters: providers’ attitudes about a common, commercial EHR. J. Eval. Clin. Pract. 20 , 36–42 (2014).

Amezquita, R. A. et al. Orchestrating single-cell analysis with Bioconductor. Nat. Methods 17 , 137–145 (2020).

Virshup, I. et al. The scverse project provides a computational ecosystem for single-cell omics data analysis. Nat. Biotechnol. 41 , 604–606 (2023).

Zou, Q. et al. Predicting diabetes mellitus with machine learning techniques. Front. Genet. 9 , 515 (2018).

Cios, K. J. & William Moore, G. Uniqueness of medical data mining. Artif. Intell. Med. 26 , 1–24 (2002).

Zeng, X. et al. PIC, a paediatric-specific intensive care database. Sci. Data 7 , 14 (2020).

Bycroft, C. et al. The UK Biobank resource with deep phenotyping and genomic data. Nature 562 , 203–209 (2018).

Lee, J. et al. Open-access MIMIC-II database for intensive care research. Annu. Int. Conf. Proc. IEEE Eng. Med. Biol. Soc. 2011 , 8315–8318 (2011).

Virshup, I., Rybakov, S., Theis, F. J., Angerer, P. & Alexander Wolf, F. anndata: annotated data. Preprint at bioRxiv https://doi.org/10.1101/2021.12.16.473007 (2021).

Voss, E. A. et al. Feasibility and utility of applications of the common data model to multiple, disparate observational health databases. J. Am. Med. Inform. Assoc. 22 , 553–564 (2015).

Vasilevsky, N. A. et al. Mondo: unifying diseases for the world, by the world. Preprint at medRxiv https://doi.org/10.1101/2022.04.13.22273750 (2022).

Harrison, J. E., Weber, S., Jakob, R. & Chute, C. G. ICD-11: an international classification of diseases for the twenty-first century. BMC Med. Inform. Decis. Mak. 21 , 206 (2021).

Köhler, S. et al. Expansion of the Human Phenotype Ontology (HPO) knowledge base and resources. Nucleic Acids Res. 47 , D1018–D1027 (2019).

Wu, P. et al. Mapping ICD-10 and ICD-10-CM codes to phecodes: workflow development and initial evaluation. JMIR Med. Inform. 7 , e14325 (2019).

Wolf, F. A., Angerer, P. & Theis, F. J. SCANPY: large-scale single-cell gene expression data analysis. Genome Biol. 19 , 15 (2018).

Pedregosa, F. et al. Scikit-learn: machine learning in Python. J. Mach. Learn. Res . 12 , 2825–2830 (2011).

de Haan-Rietdijk, S., de Haan-Rietdijk, S., Kuppens, P. & Hamaker, E. L. What’s in a day? A guide to decomposing the variance in intensive longitudinal data. Front. Psychol. 7 , 891 (2016).

Pedersen, E. S. L., Danquah, I. H., Petersen, C. B. & Tolstrup, J. S. Intra-individual variability in day-to-day and month-to-month measurements of physical activity and sedentary behaviour at work and in leisure-time among Danish adults. BMC Public Health 16 , 1222 (2016).

Roffey, D. M., Byrne, N. M. & Hills, A. P. Day-to-day variance in measurement of resting metabolic rate using ventilated-hood and mouthpiece & nose-clip indirect calorimetry systems. JPEN J. Parenter. Enter. Nutr. 30 , 426–432 (2006).

Haghverdi, L., Büttner, M., Wolf, F. A., Buettner, F. & Theis, F. J. Diffusion pseudotime robustly reconstructs lineage branching. Nat. Methods 13 , 845–848 (2016).

Lange, M. et al. CellRank for directed single-cell fate mapping. Nat. Methods 19 , 159–170 (2022).

Weiler, P., Lange, M., Klein, M., Pe'er, D. & Theis, F. CellRank 2: unified fate mapping in multiview single-cell data. Nat. Methods 21 , 1196–1205 (2024).

Zhang, S. et al. Cost of management of severe pneumonia in young children: systematic analysis. J. Glob. Health 6 , 010408 (2016).

Torres, A. et al. Pneumonia. Nat. Rev. Dis. Prim. 7 , 25 (2021).

Traag, V. A., Waltman, L. & van Eck, N. J. From Louvain to Leiden: guaranteeing well-connected communities. Sci. Rep. 9 , 5233 (2019).

Kamin, W. et al. Liver involvement in acute respiratory infections in children and adolescents—results of a non-interventional study. Front. Pediatr. 10 , 840008 (2022).

Shi, T. et al. Risk factors for mortality from severe community-acquired pneumonia in hospitalized children transferred to the pediatric intensive care unit. Pediatr. Neonatol. 61 , 577–583 (2020).

Dudnyk, V. & Pasik, V. Liver dysfunction in children with community-acquired pneumonia: the role of infectious and inflammatory markers. J. Educ. Health Sport 11 , 169–181 (2021).

Charpignon, M.-L. et al. Causal inference in medical records and complementary systems pharmacology for metformin drug repurposing towards dementia. Nat. Commun. 13 , 7652 (2022).

Grief, S. N. & Loza, J. K. Guidelines for the evaluation and treatment of pneumonia. Prim. Care 45 , 485–503 (2018).

Paul, M. Corticosteroids for pneumonia. Cochrane Database Syst. Rev. 12 , CD007720 (2017).

Sharma, A. & Kiciman, E. DoWhy: an end-to-end library for causal inference. Preprint at arXiv https://doi.org/10.48550/ARXIV.2011.04216 (2020).

Khilnani, G. C. et al. Guidelines for antibiotic prescription in intensive care unit. Indian J. Crit. Care Med. 23 , S1–S63 (2019).

Harris, L. K. & Crannage, A. J. Corticosteroids in community-acquired pneumonia: a review of current literature. J. Pharm. Technol. 37 , 152–160 (2021).

Dou, L. et al. Decreased hospital length of stay with early administration of oseltamivir in patients hospitalized with influenza. Mayo Clin. Proc. Innov. Qual. Outcomes 4 , 176–182 (2020).

Khera, A. V. et al. Genome-wide polygenic scores for common diseases identify individuals with risk equivalent to monogenic mutations. Nat. Genet. 50 , 1219–1224 (2018).

Julkunen, H. et al. Atlas of plasma NMR biomarkers for health and disease in 118,461 individuals from the UK Biobank. Nat. Commun. 14 , 604 (2023).

Ko, F. et al. Associations with retinal pigment epithelium thickness measures in a large cohort: results from the UK Biobank. Ophthalmology 124 , 105–117 (2017).

Patel, P. J. et al. Spectral-domain optical coherence tomography imaging in 67 321 adults: associations with macular thickness in the UK Biobank study. Ophthalmology 123 , 829–840 (2016).

D’Agostino Sr, R. B. et al. General cardiovascular risk profile for use in primary care: the Framingham Heart Study. Circulation 117 , 743–753 (2008).

Buergel, T. et al. Metabolomic profiles predict individual multidisease outcomes. Nat. Med. 28 , 2309–2320 (2022).

Xu, Y. et al. An atlas of genetic scores to predict multi-omic traits. Nature 616 , 123–131 (2023).

Saelens, W., Cannoodt, R., Todorov, H. & Saeys, Y. A comparison of single-cell trajectory inference methods. Nat. Biotechnol. 37 , 547–554 (2019).

Rousan, L. A., Elobeid, E., Karrar, M. & Khader, Y. Chest x-ray findings and temporal lung changes in patients with COVID-19 pneumonia. BMC Pulm. Med. 20 , 245 (2020).

Signoroni, A. et al. BS-Net: learning COVID-19 pneumonia severity on a large chest X-ray dataset. Med. Image Anal. 71 , 102046 (2021).

Bird, S. et al. Fairlearn: a toolkit for assessing and improving fairness in AI. https://www.microsoft.com/en-us/research/publication/fairlearn-a-toolkit-for-assessing-and-improving-fairness-in-ai/ (2020).

Strack, B. et al. Impact of HbA1c measurement on hospital readmission rates: analysis of 70,000 clinical database patient records. BioMed. Res. Int. 2014 , 781670 (2014).

Stekhoven, D. J. & Bühlmann, P. MissForest—non-parametric missing value imputation for mixed-type data. Bioinformatics 28 , 112–118 (2012).

Banerjee, A. et al. Identifying subtypes of heart failure from three electronic health record sources with machine learning: an external, prognostic, and genetic validation study. Lancet Digit. Health 5 , e370–e379 (2023).

Nagamine, T. et al. Data-driven identification of heart failure disease states and progression pathways using electronic health records. Sci. Rep. 12 , 17871 (2022).

Da Silva Filho, J. et al. Disease trajectories in hospitalized COVID-19 patients are predicted by clinical and peripheral blood signatures representing distinct lung pathologies. Preprint at bioRxiv https://doi.org/10.1101/2023.09.08.23295024 (2023).

Haneuse, S., Arterburn, D. & Daniels, M. J. Assessing missing data assumptions in EHR-based studies: a complex and underappreciated task. JAMA Netw. Open 4 , e210184 (2021).

Little, R. J. A. A test of missing completely at random for multivariate data with missing values. J. Am. Stat. Assoc. 83 , 1198–1202 (1988).

Jakobsen, J. C., Gluud, C., Wetterslev, J. & Winkel, P. When and how should multiple imputation be used for handling missing data in randomised clinical trials—a practical guide with flowcharts. BMC Med. Res. Methodol. 17 , 162 (2017).

Dziura, J. D., Post, L. A., Zhao, Q., Fu, Z. & Peduzzi, P. Strategies for dealing with missing data in clinical trials: from design to analysis. Yale J. Biol. Med. 86 , 343–358 (2013).

White, I. R., Royston, P. & Wood, A. M. Multiple imputation using chained equations: issues and guidance for practice. Stat. Med. 30 , 377–399 (2011).

Jäger, S., Allhorn, A. & Bießmann, F. A benchmark for data imputation methods. Front. Big Data 4 , 693674 (2021).

Waljee, A. K. et al. Comparison of imputation methods for missing laboratory data in medicine. BMJ Open 3 , e002847 (2013).

Ibrahim, J. G. & Molenberghs, G. Missing data methods in longitudinal studies: a review. Test (Madr.) 18 , 1–43 (2009).

Li, C., Alsheikh, A. M., Robinson, K. A. & Lehmann, H. P. Use of recommended real-world methods for electronic health record data analysis has not improved over 10 years. Preprint at bioRxiv https://doi.org/10.1101/2023.06.21.23291706 (2023).

Regev, A. et al. The Human Cell Atlas. eLife 6 , e27041 (2017).

Megill, C. et al. cellxgene: a performant, scalable exploration platform for high dimensional sparse matrices. Preprint at bioRxiv https://doi.org/10.1101/2021.04.05.438318 (2021).

Speir, M. L. et al. UCSC Cell Browser: visualize your single-cell data. Bioinformatics 37 , 4578–4580 (2021).

Hunter, J. D. Matplotlib: a 2D graphics environment. Comput. Sci. Eng. 9 , 90–95 (2007).

Waskom, M. seaborn: statistical data visualization. J. Open Source Softw. 6 , 3021 (2021).

Harris, C. R. et al. Array programming with NumPy. Nature 585 , 357–362 (2020).

Lam, S. K., Pitrou, A. & Seibert, S. Numba: a LLVM-based Python JIT compiler. In Proc. of the Second Workshop on the LLVM Compiler Infrastructure in HPC. https://doi.org/10.1145/2833157.2833162 (Association for Computing Machinery, 2015).

Virtanen, P. et al. SciPy 1.0: fundamental algorithms for scientific computing in Python. Nat. Methods 17 , 261–272 (2020).

McKinney, W. Data structures for statistical computing in Python. In Proc. of the 9th Python in Science Conference (eds van der Walt, S. & Millman, J.). https://doi.org/10.25080/majora-92bf1922-00a (SciPy, 2010).

Boulanger, A. Open-source versus proprietary software: is one more reliable and secure than the other? IBM Syst. J. 44 , 239–248 (2005).

Rocklin, M. Dask: parallel computation with blocked algorithms and task scheduling. In Proc. of the 14th Python in Science Conference. https://doi.org/10.25080/majora-7b98e3ed-013 (SciPy, 2015).

Pivarski, J. et al. Awkward Array. https://doi.org/10.5281/ZENODO.4341376

Collette, A. Python and HDF5: Unlocking Scientific Data (‘O’Reilly Media, Inc., 2013).

Miles, A. et al. zarr-developers/zarr-python: v2.13.6. https://doi.org/10.5281/zenodo.7541518 (2023).

The pandas development team. pandas-dev/pandas: Pandas. https://doi.org/10.5281/ZENODO.3509134 (2024).

Weberpals, J. et al. Deep learning-based propensity scores for confounding control in comparative effectiveness research: a large-scale, real-world data study. Epidemiology 32 , 378–388 (2021).

Rosenthal, J. et al. Building tools for machine learning and artificial intelligence in cancer research: best practices and a case study with the PathML toolkit for computational pathology. Mol. Cancer Res. 20 , 202–206 (2022).

Gayoso, A. et al. A Python library for probabilistic analysis of single-cell omics data. Nat. Biotechnol. 40 , 163–166 (2022).

Paszke, A. et al. PyTorch: an imperative style, high-performance deep learning library. In Advances in Neural Information Processing Systems 32 (eds Wallach, H. et al.). 8024–8035 (Curran Associates, 2019).

Frostig, R., Johnson, M. & Leary, C. Compiling machine learning programs via high-level tracing. https://cs.stanford.edu/~rfrostig/pubs/jax-mlsys2018.pdf (2018).

Moor, M. et al. Foundation models for generalist medical artificial intelligence. Nature 616 , 259–265 (2023).

Kraljevic, Z. et al. Multi-domain clinical natural language processing with MedCAT: the Medical Concept Annotation Toolkit. Artif. Intell. Med. 117 , 102083 (2021).

Pollard, T. J., Johnson, A. E. W., Raffa, J. D. & Mark, R. G. An open source Python package for producing summary statistics for research papers. JAMIA Open 1 , 26–31 (2018).

Ellen, J. G. et al. Participant flow diagrams for health equity in AI. J. Biomed. Inform. 152 , 104631 (2024).

Schouten, R. M. & Vink, G. The dance of the mechanisms: how observed information influences the validity of missingness assumptions. Sociol. Methods Res. 50 , 1243–1258 (2021).

Johnson, W. E., Li, C. & Rabinovic, A. Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics 8 , 118–127 (2007).

Davidson-Pilon, C. lifelines: survival analysis in Python. J. Open Source Softw. 4 , 1317 (2019).

Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Stat. Soc. Ser. B Stat. Methodol. 57 , 289–300 (1995).

Wishart, D. S. et al. DrugBank: a comprehensive resource for in silico drug discovery and exploration. Nucleic Acids Res. 34 , D668–D672 (2006).

Harrell, F. E. Jr, Califf, R. M., Pryor, D. B., Lee, K. L. & Rosati, R. A. Evaluating the yield of medical tests. JAMA 247 , 2543–2546 (1982).

Currant, H. et al. Genetic variation affects morphological retinal phenotypes extracted from UK Biobank optical coherence tomography images. PLoS Genet. 17 , e1009497 (2021).

Cohen, J. P. et al. TorchXRayVision: a library of chest X-ray datasets and models. In Proc. of the 5th International Conference on Medical Imaging with Deep Learning (eds Konukoglu, E. et al.). 172 , 231–249 (PMLR, 2022).

Cohen, J.P., Hashir, M., Brooks, R. & Bertrand, H. On the limits of cross-domain generalization in automated X-ray prediction. In Proceedings of Machine Learning Research , Vol. 121 (eds Arbel, T. et al.) 136–155 (PMLR, 2020).

Download references

Acknowledgements

We thank M. Ansari who designed the ehrapy logo. The authors thank F. A. Wolf, M. Lücken, J. Steinfeldt, B. Wild, G. Rätsch and D. Shung for feedback on the project. We further thank L. Halle, Y. Ji, M. Lücken and R. K. Rubens for constructive comments on the paper. We thank F. Hashemi for her help in implementing the survival analysis module. This research was conducted using data from the UK Biobank, a major biomedical database ( https://www.ukbiobank.ac.uk ), under application number 49966. This work was supported by the German Center for Lung Research (DZL), the Helmholtz Association and the CRC/TRR 359 Perinatal Development of Immune Cell Topology (PILOT). N.H. and F.J.T. acknowledge support from the German Federal Ministry of Education and Research (BMBF) (LODE, 031L0210A), co-funded by the European Union (ERC, DeepCell, 101054957). A.N. is supported by the Konrad Zuse School of Excellence in Learning and Intelligent Systems (ELIZA) through the DAAD program Konrad Zuse Schools of Excellence in Artificial Intelligence, sponsored by the Federal Ministry of Education and Research. This work was also supported by the Chan Zuckerberg Initiative (CZIF2022-007488; Human Cell Atlas Data Ecosystem).

Open access funding provided by Helmholtz Zentrum München - Deutsches Forschungszentrum für Gesundheit und Umwelt (GmbH).

Author information

Authors and affiliations.

Institute of Computational Biology, Helmholtz Munich, Munich, Germany

Lukas Heumos, Philipp Ehmele, Tim Treis, Eljas Roellin, Lilly May, Altana Namsaraeva, Nastassya Horlava, Vladimir A. Shitov, Xinyue Zhang, Luke Zappia, Leon Hetzel, Isaac Virshup, Lisa Sikkema, Fabiola Curion & Fabian J. Theis

Institute of Lung Health and Immunity and Comprehensive Pneumology Center with the CPC-M bioArchive; Helmholtz Zentrum Munich; member of the German Center for Lung Research (DZL), Munich, Germany

Lukas Heumos, Niklas J. Lang, Herbert B. Schiller & Anne Hilgendorff

TUM School of Life Sciences Weihenstephan, Technical University of Munich, Munich, Germany

Lukas Heumos, Tim Treis, Nastassya Horlava, Vladimir A. Shitov, Lisa Sikkema & Fabian J. Theis

Health Data Science Unit, Heidelberg University and BioQuant, Heidelberg, Germany

Julius Upmeier zu Belzen & Roland Eils

Department of Mathematics, School of Computation, Information and Technology, Technical University of Munich, Munich, Germany

Eljas Roellin, Lilly May, Luke Zappia, Leon Hetzel, Fabiola Curion & Fabian J. Theis

Konrad Zuse School of Excellence in Learning and Intelligent Systems (ELIZA), Darmstadt, Germany

Altana Namsaraeva

Systems Medicine, Deutsches Zentrum für Neurodegenerative Erkrankungen (DZNE), Bonn, Germany

Rainer Knoll

Center for Digital Health, Berlin Institute of Health (BIH) at Charité – Universitätsmedizin Berlin, Berlin, Germany

Roland Eils

Research Unit, Precision Regenerative Medicine (PRM), Helmholtz Munich, Munich, Germany

Herbert B. Schiller

Center for Comprehensive Developmental Care (CDeCLMU) at the Social Pediatric Center, Dr. von Hauner Children’s Hospital, LMU Hospital, Ludwig Maximilian University, Munich, Germany

Anne Hilgendorff

You can also search for this author in PubMed   Google Scholar

Contributions

L. Heumos and F.J.T. conceived the study. L. Heumos, P.E., X.Z., E.R., L.M., A.N., L.Z., V.S., T.T., L. Hetzel, N.H., R.K. and I.V. implemented ehrapy. L. Heumos, P.E., N.L., L.S., T.T. and A.H. analyzed the PIC database. J.U.z.B. and L. Heumos analyzed the UK Biobank database. X.Z. and L. Heumos analyzed the COVID-19 chest x-ray dataset. L. Heumos, P.E. and J.U.z.B. wrote the paper. F.J.T., A.H., H.B.S. and R.E. supervised the work. All authors read, corrected and approved the final paper.

Corresponding author

Correspondence to Fabian J. Theis .

Ethics declarations

Competing interests.

L. Heumos is an employee of LaminLabs. F.J.T. consults for Immunai Inc., Singularity Bio B.V., CytoReason Ltd. and Omniscope Ltd. and has ownership interest in Dermagnostix GmbH and Cellarity. The remaining authors declare no competing interests.

Peer review

Peer review information.

Nature Medicine thanks Leo Anthony Celi and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary handling editor: Lorenzo Righetto, in collaboration with the Nature Medicine team.

Additional information

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

Extended data

Extended data fig. 1 overview of the paediatric intensive care database (pic)..

The database consists of several tables corresponding to several data modalities and measurement types. All tables colored in green were selected for analysis and all tables in blue were discarded based on coverage rate. Despite the high coverage rate, we discarded the ‘OR_EXAM_REPORTS’ table because of the lack of detail in the exam reports.

Extended Data Fig. 2 Preprocessing of the Paediatric Intensive Care (PIC) dataset with ehrapy.

( a ) Heterogeneous data of the PIC database was stored in ‘data’ (matrix that is used for computations) and ‘observations’ (metadata per patient visit). During quality control, further annotations are added to the ‘variables’ (metadata per feature) slot. ( b ) Preprocessing steps of the PIC dataset. ( c ) Example of the function calls in the data analysis pipeline that resembles the preprocessing steps in (B) using ehrapy.

Extended Data Fig. 3 Missing data distribution for the ‘youths’ group of the PIC dataset.

The x-axis represents the percentage of missing values in each feature. The y-axis reflects the number of features in each bin with text labels representing the names of the individual features.

Extended Data Fig. 4 Patient selection during analysis of the PIC dataset.

Filtering for the pneumonia cohort of the youths filters out care units except for the general intensive care unit and the pediatric intensive care unit.

Extended Data Fig. 5 Feature rankings of stratified patient groups.

Scores reflect the z-score underlying the p-value per measurement for each group. Higher scores (above 0) reflect overrepresentation of the measurement compared to all other groups and vice versa. ( a ) By clinical chemistry. ( b ) By liver markers. ( c ) By medication type. ( d ) By infection markers.

Extended Data Fig. 6 Liver marker value progression for the ‘youths’ group and Kaplan-Meier curves.

( a ) Viral and severe pneumonia with co-infection groups display enriched gamma-glutamyl transferase levels in blood serum. ( b ) Aspartate transferase (AST) and Alanine transaminase (ALT) levels are enriched for severe pneumonia with co-infection during early ICU stay. ( c ) and ( d ) Kaplan-Meier curves for ALT and AST demonstrate lower survivability for children with measurements outside the norm.

Extended Data Fig. 7 Overview of medication categories used for causal inference.

( a ) Feature engineering process to group administered medications into medication categories using drugbank. ( b ) Number of medications per medication category. ( c ) Number of patients that received (dark blue) and did not receive specific medication categories (light blue).

Extended Data Fig. 8 UK-Biobank data overview and quality control across modalities.

( a ) UMAP plot of the metabolomics data demonstrating a clear gradient with respect to age at sampling, and ( b ) type 2 diabetes prevalence. ( c ) Analogously, the features derived from retinal imaging show a less pronounced age gradient, and ( d ) type 2 diabetes prevalence gradient. ( e ) Stratifying myocardial infarction risk by the type 2 diabetes comorbidity confirms vastly increased risk with a prior type 2 (T2D) diabetes diagnosis. Kaplan-Meier estimators with 95 % confidence intervals are shown. ( f ) Similarly, the polygenic risk score for coronary heart disease used in this work substantially enriches myocardial infarction risk in its top 5% percentile. Kaplan-Meier estimators with 95 % confidence intervals are shown. ( g ) UMAP visualization of the metabolomics features colored by the assessment center shows no discernable biases. (A-G) n = 29,216.

Extended Data Fig. 9 UK-Biobank retina derived feature quality control.

( a ) Leiden Clustering of retina derived feature space. ( b ) Comparison of ‘overall retinal pigment epithelium (RPE) thickness’ values between cluster 5 (n = 301) and the rest of the population (n = 28,915). ( c ) RPE thickness in the right eye outliers on the UMAP largely corresponds to cluster 5. ( d ) Log ratio of top and bottom 5 fields in obs dataframe between cluster 5 and the rest of the population. ( e ) Image Quality of the optical coherence tomography scan as reported in the UKB. ( f ) Minimum motion correlation quality control indicator. ( g ) Inner limiting membrane (ILM) quality control indicator. (D-G) Data are shown for the right eye only, comparable results for the left eye are omitted. (A-G) n = 29,216.

Extended Data Fig. 10 Bias detection and mitigation study on the Diabetes 130-US hospitals dataset (n = 101,766 hospital visits, one patient can have multiple visits).

( a ) Filtering to the visits of Medicare recipients results in an increase of Caucasians. ( b ) Proportion of visits where Hb1Ac measurements are recorded, stratified by admission type. Adjusted P values were calculated with Chi squared tests and Bonferroni correction (Adjusted P values: Emergency vs Referral 3.3E-131, Emergency vs Other 1.4E-101, Referral vs Other 1.6E-4.) ( c ) Normalizing feature distributions jointly vs. separately can mask distribution differences. ( d ) Imputing the number of medications for visits. Onto the complete data (blue), MCAR (30% missing data) and MAR (38% missing data) were introduced (orange), with the MAR mechanism depending on the time in hospital. Mean imputation (green) can reduce the variance of the distribution under MCAR and MAR mechanisms, and bias the center of the distribution under an MAR mechanism. Multiple imputation, such as MissForest imputation can impute meaningfully even in MAR cases, when having access to variables involved in the MAR mechanism. Each boxplot represents the IQR of the data, with the horizontal line inside the box indicating the median value. The left and right bounds of the box represent the first and third quartiles, respectively. The ‘whiskers’ extend to the minimum and maximum values within 1.5 times the IQR from the lower and upper quartiles, respectively. ( e ) Predicting the early readmission within 30 days after release on a per-stay level. Balanced accuracy can mask differences in selection and false negative rate between sensitive groups.

Supplementary information

Supplementary tables 1 and 2, reporting summary, 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

Cite this article.

Heumos, L., Ehmele, P., Treis, T. et al. An open-source framework for end-to-end analysis of electronic health record data. Nat Med (2024). https://doi.org/10.1038/s41591-024-03214-0

Download citation

Received : 11 December 2023

Accepted : 25 July 2024

Published : 12 September 2024

DOI : https://doi.org/10.1038/s41591-024-03214-0

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

Quick links

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

Sign up for the Nature Briefing: Translational Research newsletter — top stories in biotechnology, drug discovery and pharma.

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
• BMJ Open Qual
• v.11(2); 2022

Defining case management success: a qualitative study of case manager perspectives from a large-scale health and social needs support program

Margae knox.

1 School of Public Health, University of California, Berkeley, Berkeley, California, USA

Emily E Esteban

2 Contra Costa Health Services, Martinez, California, USA

Elizabeth A Hernandez

Mark d fleming, nadia safaeinilli, amanda l brewster, associated data.

No data are available. Data are not publicly available to protect potentially sensitive information. For data inquiries, please contact the corresponding author.

Health systems are expanding efforts to address health and social risks, although the heterogeneity of early evidence indicates need for more nuanced exploration of how such programs work and how to holistically assess program success. This qualitative study aims to identify characteristics of success in a large-scale, health and social needs case management program from the perspective of interdisciplinary case managers.

Case management program for high-risk, complex patients run by an integrated, county-based public health system.

Participants

30 out of 70 case managers, purposively sampled to represent their interdisciplinary health and social work backgrounds. Interviews took place in March–November 2019.

Primary and secondary outcome measures

The analysis intended to identify characteristics of success working with patients.

Case managers described three characteristics of success working with patients: (1) establishing trust; (2) observing change in patients’ mindset or initiative and (3) promoting stability and independence. Cross-cutting these characteristics, case managers emphasised the importance of patients defining their own success, often demonstrated through individualised, incremental progress. Thus, moments of success commonly contrasted with external perceptions and operational or productivity metrics.

Conclusions

Themes emphasise the importance of compassion for complexity in patients’ lives, and success as a step-by-step process that is built over longitudinal relationships.

What is already known on this topic?

• Case management programs to support health and social needs have demonstrated promising yet mixed results. Underlying mechanisms and shared definitions of successful case management are underdeveloped.

What this study adds?

• Case managers emphasised building trust over time and individual, patient-defined objectives as key markers of success, a contrast to commonly used quantitative evaluation metrics.

How this study might affect research, practice or policy?

• Results suggest that lighter touch case management interventions face limitations without an established patient relationship. Results also support a need for alternative definitions of case management success including patient-centered measures such as trust in one’s case manager.

Introduction

Health system efforts to address both health and social needs are expanding. In the USA, some state Medicaid programmes are testing payments for non-medical services to address transportation, housing instability and food insecurity. Medicaid provides healthcare coverage for lower income individuals and families, jointly funded by federal and state governments. Similarly, social prescribing, or the linking of patients with social needs to community resources, is supported by the UK’s National Health Service and has also been piloted by Canada’s Alliance for Healthier Communities. 1

A growing evidence base suggests promising outcomes from healthcare interventions addressing social needs. In some contexts, case managers or navigators providing social needs assistance can improve health 2 and reduce costly hospital use. 3–5 Yet systematic reviews also report mixed results for measures of health and well-being, hospitalisation and emergency department use, and overall healthcare costs. 6–9 Notably, a randomised trial of the Camden Care Coalition programme for patients with frequent hospitalisations due to medically and socially complex needs 10 found no difference in 180-day readmission between patients assigned to a care transitions programme compared with usual hospital postdischarge care. In the care transition programme, patients received follow-up from a multidisciplinary team of nurses, social workers and community health workers. The team conducted home visits, scheduled and accompanied patients to follow-up outpatient visits, helped with managing medications, coached patients on self-care and connected patients with social services and behavioural healthcare. The usual care group received usual postdischarge care with limited follow-up. 11 This heterogeneity of early evidence indicates a need for more nuanced explorations of how social needs assistance programmes work, and how to holistically assess whether programmes are successful. 12 13

Social needs case management may lead to health and well-being improvements through multiple pathways involving both material and social support. 14 15 Improvements are often a long-term, non-linear process. 16 17 At the same time, quality measures specific to social needs assistance programmes currently remain largely undefined. Studies often analyse utilisation and cost outcomes but lack granularity on interim processes and markers of success.

In order to translate a complex and context-dependent intervention like social needs case management from one setting to another, these interim processes and outcomes need greater recognition. 18–20 Early efforts to refine complex care measures are underway and call out a need for person-centred and goal-concordant measures. 21 Further research on how frontline social needs case managers themselves define successes in their work could help leaders improve programme design and management and could also inform broader quality measure development efforts.

Our in-depth, qualitative study sought to understand how case managers defined success in their work with high-risk patients. Case managers were employed by CommunityConnect, a large-scale health and social needs care management programme that serves a mixed-age adult population with varying physical health, mental health and social needs. Each case manager’s workflow includes an individualised, regularly updated dashboard of operational metrics. It is unclear, however, whether or how these operational factors relate to patient success in a complex care programme. Thus, the case managers’ perspectives on defining success are critical for capturing how programmes work and identifying essential principles.

Study design and setting

In 2017, the Contra Costa County Health Services Department in California launched CommunityConnect, a case management programme to coordinate health, behavioural health and social services for County Medicaid patients with complex health and social conditions. The County Health Services Department serves approximately 15% (180 000) of Contra Costa’s nearly 1.2 million residents. CommunityConnect enrollees were selected based on a predictive model, which leveraged data from multiple county systems to identify individuals most likely to use hospital or emergency room services for preventable reasons. Enrollees are predominantly women (59%) and under age 40 (49%). Seventy-seven per cent of enrollees have more than one chronic condition, particularly hypertension (42%), mood disorders (40%) and chronic pain (35%). 22 Programme goals include improving beneficiary health and well-being through more efficient and effective use of resources.

Each case manager interviewed in this study worked full time with approximately 90 patients at a time. Case managers met patients in-person, ideally at least once a month for 1 year, although patients sometimes continue to receive ongoing support at the case manager’s discretion in cases of continued need. Overall, up to 6000 individuals at a time receive in-person case management services through CommunityConnect, with approximately 200–300 added and 200–300 graduated per month. At the time of the study, CommunityConnect employed approximately 70 case managers trained in various public health and social work disciplines (see table 1 , Interview Sample). Case managers and patients are matched based on an algorithm that prioritises mental health history, primary language and county region.

Interview sample

Although case managers bring unique experience from their respective discipline, all are expected to conduct similar case management services. Services included discussing any unmet social needs with patients, coordinating applicable resources and partnering with the patient and patient’s care team to improve physical and emotional health. The programme tracks hospital and emergency department utilisation as well as patient benefits such as food stamps, housing or transportation vouchers and continuous Medicaid coverage on an overall basis. Each case manager has access to an individualised dashboard that includes operational metrics such as new patients to contact, and frequency of patient contacts, timeliness for calling patients recently discharged from the hospital, whether patients have continuous Medicaid coverage, and completion of social risk screenings.

Study recruitment

Semistructured interviews were conducted with 30 field-based case managers as part of the programme’s evaluation and quality improvement process. Participants included four mental health clinical specialists, five substance abuse counsellors, six social workers, nine public health nurses, four housing support specialists and two community health worker specialists. Case managers were recruited by email and selected based on purposive sampling to reflect membership across disciplines and experience working with CommunityConnect for at least 1 year. Three case managers declined to participate. Interviews ended when data saturation was achieved. 23

Interview procedures

Interviews were conducted by five CommunityConnect evaluation staff members (including EEE), who received training and supervision from the evaluation director (EH), who also conducted interviews. The evaluation staff were bachelor and masters-level trained. The evaluation director was masters-level trained and held prior experience in healthcare quality and programme planning.

The evaluation team drafted the interview guide to ask about a variety of work processes and experiences with the goal of improving programme operations including staff and patient experiences. Specific questions analysed for this study were (1) how case managers define success with a patient and (2) examples where case managers considered work with patients a success.

Interviews took place in-person in private meeting rooms at case managers’ workplace from March 2019 – November 2019. Interviews lasted 60–90 min and only the interviewer and case manager were present. All interviewers were familiar with CommunityConnect yet did not have a prior relationship with case managers. Case managers did not receive compensation beyond their regular salary for participating in the study and were allowed to opt out of recruitment or end the interview early for any reason. All interviews were audio recorded, transcribed and entered into Nvivo V.12 for analysis.

Patient and public involvement

This project focused on case manager’s perspectives and thus did not directly involve patients. Rather, patients were involved through case manager recollections of experiences working with patients.

Data analysis

We used an integrated approach to develop an initial set of qualitative codes including deductive coding of programme processes and concepts, followed by inductive coding of how case managers defined success. All interviews were coded by two researchers experienced in qualitative research (EEE and MK). Themes were determined based on recurrence across interviews and illustrative examples and being described by more than one case manager type. The two researchers identified preliminary themes independently, then consulted with one another to achieve consensus on final themes. Themes and supporting quotes were then presented to the full author team to ensure collective agreement that key perspectives had been included. Preliminary results were also shared at a staff meeting attended by case managers and other staff as an opportunity for feedback on study findings. This manuscript addresses the Standards for Reporting Qualitative Research, 24 and the Consolidated Criteria for Reporting Qualitative Research checklist is provided as an appendix. 25

All case manager participants provided informed consent. Research procedures were approved by the Contra Costa Regional Medical Center and Health Centers Institutional Review Committee (Protocol 12-17-2018).

Case managers frequently and across multiple roles mentioned three characteristics of success when working with patients: (1) establishing trust; (2) fostering change in patients’ mindset or initiative and (3) promoting stability and independence. Across these characteristics, case managers expressed that success is patient-defined, with individualised and often incremental progress—a contrast with external perceptions of success and common operational or productivity metrics (see figure 1 ).

Illustration of key themes.

Success is establishing trust

Trusting relationships were the most widely noted characteristic of success. Trust was described as both a product of case managers’ consistent follow-up and helpfulness over time and a foundational step to enable progress on patient-centred goals. To build trust, case managers explained, patients must feel seen and heard, and understand the case managers’ desire to help: ‘Success is to know that she knows me very well…I look for her on the streets, and I’m waiting for her to call me back. Hopefully she knows that when she’s ready I will be there at least to provide that resource for her and so it’s that personal relationship that you build’ (Case manager 11, social worker). Case managers also highlighted the longitudinal relationship required to establish trust, distinguishing success as more than one-time information delivery or navigating bureaucratic processes to procure services.

Case managers also identified trust as foundational to provide better support for patients: ‘So they’re as honest with me as they can be. That way I have a clear understanding about realistically what I can do to help them coordinate their care or link them to services.’ (Case manager 2, mental health clinician specialist). Establishing trust was essential to improve communication with patients and produced an amplifying effect. That is, a case manager’s initial help and follow-up builds trust so that patients can be more open, and open communication helps the case manager know what specific services can be most useful. This positive feedback loop further cements trust and builds momentum for a longitudinal relationship.

Permission to have a home visit was mentioned as a valuable indicator of early success in building trust: ‘(Your home is) your sanctuary’, expressed one case manager (Case manager 29, public health nurse), acknowledging the vulnerability of opening one’s home to an outsider. For another case manager, regular home visits in the context of a trusting relationship made the case manager aware of and able to address a difficult situation: ‘Every time I was going to her home, I was noticing more and more gnats flying around… She said it’s because of the garbage…’ After establishing trust, the patient allowed the case manager access to the bedroom where the case manager uncovered numerous soiled diapers. The case manager arranged professional cleaning and sanitation through CommunityConnect, after which, ‘there was room for a dance floor in her bedroom. There was so much room, and the look on her face, it was almost as if her chest got proud, just in that day. She didn’t seem so burdened…So that’s a success’ (Case manager 4, substance abuse counsellor). Across multiple examples, case managers expressed trust as a critical element for effective patient partnerships.

However, the pathways to building trust are less clear cut. Quick wins through tangible support such as a transportation voucher to a medical appointment could help engage a patient initially. Yet case managers more frequently emphasised strategies based on relationships over time. Strategies included expressing empathy (putting yourself in the patient’s shoes), demonstrating respect (especially when the patient has experienced disrespect in other health system encounters), keeping appointments, following through on what you say you will do, calling to check in and ‘being there’. Overall, case managers expressed that trust lets patients know they are not alone and sets the stage for future success.

Success is fostering a change in patients’ mindset or initiative

Case managers described a change in patients’ mindset or initiative as evidence of further success. One case manager explained, ‘Really (success) could be a switch in mind state… If I can get someone to consider addressing an issue. Or just acknowledging an issue. That’s progress’ (Case manager 24, substance abuse counsellor). Another case manager spoke to the importance of mindset by stating, ‘what I try to do is not just change the surface of life’. This case manager elaborated, ‘You help (a patient) get their housing and they’re gonna lose it again, unless they change; something changes in their mindset, and then they see things differently.’ (Case manager 6, mental health clinician specialist). Some case managers suggested that the supportive resources they provide are only band-aid solutions if unaccompanied by a changed mindset to address root causes.

Case managers reported that shared goals and plans are essential, in contrast to solutions identified by case managers without patient involvement. ‘I can’t do everything for them’, expressed one case manager (Case manager 21, public health nurse), while others similarly acknowledged that imposing self-improvement goals or providing resources for which a patient may not be ready may be counterproductive. Rather, one case manager emphasised, ‘I think it’s really important to celebrate people’s ideas, their beliefs, their own goals and values’. (Case manager 4, substance abuse counsellor). As an example, the case manager applauded a patient’s ideas of getting a driver’s license and completing an education certificate. In summary, case managers viewed success as a two-way street where patient’s own ideas and motivation were essential for long-term impact.

Success is promoting stability and independence

Case managers also identified patients’ stability and independence as a characteristic of success. One case manager stated, ‘I define success as having them be more independent in their just manoeuvring the system…how they problem solve’ (Case manager 30, public health nurse). Relative to the other characteristics of success, stability and independence more closely built on resources and services coordinated or procured by the case manager. For example, CommunityConnect provides cell phones free-of-charge to patients who do not currently have a phone or continuous service, which has helped patients build a network beyond the case manager: ‘Once we get them that cell phone then they’re able to make a lot of connections … linking to services on their own. They actually become a lot more confident in themselves is what I’ve seen’. (Case manager 23, substance abuse counsellor). In another example, a case manager helped a patient experiencing complex health issues to reconcile and understand various medications. For this patient stability means, ‘when he does go into the emergency room, it’s needed. … even though he’s taking his medication like he’s supposed to… it’s just his health gets bad. So, yea I would say that one (is a success)’ (Case manager 8, social worker). Thus, stability represents maintained, improved well-being, supported by care coordination and resources, even while challenges may still be present.

As a step further, ‘Absolute success’, according to one case manager, ‘(is when a patient) drops off my caseload and I don’t hear from them, not because they’re not doing well but because they are doing well, because they are independent’ (Case manager 12, social worker). Patients may still need periodic help knowing who to contact but can follow through on their own. This independence may arise because patients have found personal support networks and other resources that allow them to rely less and less on the case manager. While not all patients reach this step of sustained independence and stability, it is an accomplishment programmatically and for case managers personally.

Success is patient-defined, built on individualised and incremental progress

Case managers widely recognised that success comes in different shapes and sizes, dependent on their patient’s situation. Irrespective of the primary concern, many identified the patient’s own judgement as the benchmark for success. One case manager explained, ‘I define success with my patients by they are telling me it was a success. It’s by their expression, it’s just not a success until they say it’s a success for them’ (Case manager 7, social worker). In a more specific example, a case manager highlighted checking in with a patient instead of assuming a change is successful: ‘It’s not just getting someone housed or getting someone income. Like the male who we’re working towards reconciliation with his parents… that’s a huge step but if he doesn’t feel good about it… then that’s not a success.’ The same case manager elaborated, ‘it’s really engaging with the knowing where the patient him or herself is at mentally, for me. Yeah. That’s a success’ (Case manager 18, homeless services specialist). This comment challenges the current paradigm where, for example, if a patient has a housing need and is matched to housing, then the case is a success. Rather, case managers viewed success as more than meeting a need but also reciprocal satisfaction from the patient.

Often, case managers valued individualised, even if seemingly small, achievements as successes: ‘Every person’s different you know. A success could be just getting up and brushing their teeth. Sometimes success is actually getting them out of the house or getting the care they need’ (Case manager 28, social worker). Another case manager echoed, ‘(Success) depends on where they’re at … it runs the gamut, you know, but they’re all successes’ (Case manager 10, public health nurse). CommunityConnect’s interdisciplinary focus was identified as an important facilitator for tailoring support to individualised client needs. In contrast with condition-specific case management settings, for example, a case manager with substance abuse training noted, ‘whether someone wants to address their substance use or not, they still have these other needs, and (with CommunityConnect) I can still provide assistance’ (Case manager 24).

However, the individualised and incremental successes are not well captured by common case management metrics. One case manager highlighted a tension between operational productivity metrics and patient success, noting, ‘I get it, that there has to be accountability. We’re out in the field, I mean people could really be doing just a whole lot of nothing… (Yet), for me I don’t find the success in the numbers. I don’t think people are a number. Oh, look I got a pamphlet for you, I’m dropping it off… I don’t think that that is what’s really going to make this programme successful’ (Case manager 8, social worker). One case manager mentioned change in healthcare utilisation as a marker of success, but more often, case managers offered stories of patient success that diverge from common programme measures. For example, one case manager observed, ‘The clear (successes) are nice: when you apply for Social Security and they get it that’s like a hurrah. And then there’s other times it’s just getting them to the dentist’ (Case manager 28, social worker). Another case manager elaborated, ‘It’s not always the big number—the how many people did I house this year. It’s the little stuff like the fact that this 58-year-old woman who believes she’s pregnant and has been living outside for years and years, a victim of domestic violence, has considered going inside. Like that is gigantic’ (Case manager 18, homeless services specialist). Overwhelmingly, case managers defined success through the interpersonal relationship with their patients within patients’ complex, daily life circumstances.

Case managers’ definitions of success focused on establishing trust, fostering patients change in mindset or initiative, and, for some patients, achieving independence and stability. Examples of success were commonly incremental and specific to an individual’s circumstances, contrasting with programmatic measures such as reduction in hospital or emergency department utilisation, benefits and other resources secured, or productivity expectations. Study themes heavily emphasise the interpersonal relationship that case managers have with patients and underscore the importance of patient-centred and patient-defined definitions of success over other outcome measures.

Our results complement prior work on clinic-based programmes for complex patients. For example, interdisciplinary staff in a qualitative study of an ambulatory intensive care centre also identified warm relationships between patients and staff as a marker of success. 26 In another study interviewing clinicians and leaders across 12 intensive outpatient programmes, three key facilitators of patient engagement emerged: (1) financial assistance and other resources to help meet basic needs, (2) working as a multi-disciplinary care team and (3) adequate time and resources to develop close relationships focused on patient goals. 27 Our results concur on the importance of a multi-disciplinary approach, establishing trusting relationships, and pursuing patient-centred goals. Our results diverge on the role of resources to meet basic needs. Case managers in our study indicated that while connections to social services benefits and other resources help initiate the case manager-patient relationship, lasting success involved longer-term relationships in which they supported patients in developing patients’ own goal setting skills and motivation.

An important takeaway from case managers’ definitions of success is the ‘how’ they go about their work, in contrast to the ‘what’ of particular care coordination activities. For example, case managers emphasise interpersonal approaches such as empathy and respect over specific processes and resource availability. Primary care clinicians, too, have expressed how standard HEDIS or CAHPS quality metrics fail to capture, and in some cases disincentivise, the intuitions in their work that are important for high quality care. 28 29 Complex care management programmes must also wrestle with this challenge of identifying standards without extinguishing underlying quality constructs.

Strengths and limitations

This study brings several strengths, including bringing to light the unique, unexplored perspective of case managers working on both health and social needs with patients facing diverse circumstances that contribute to high-risk of future hospital or emergency department utilisation. The fact that our study explores perspectives across an array of case manager disciplines is also a strength, however a limitation is that we are unable to distinguish how success differed by discipline based on smaller numbers of each discipline in this study sample. Other study limitations include generalisability to other settings, given that all case managers worked for a single large-scale social needs case management programme. Comments around productivity concerns or interdisciplinary perspectives on ways to support patients may be unique to the infrastructure or management of this organisation. In addition, at the time of the study, all case managers were able to meet with patients in-person; future studies may explore whether definitions of success change when interactions become virtual or telephonic as occurred amidst COVID-19 concerns.

This study is the first to our knowledge to inquire about holistic patient success from the perspective of case managers in the context of a social needs case management programme. The findings offer important implications for researchers as well as policy makers and managers who are designing complex case management programmes.

Our results identify patient-directed goals, stability and satisfaction, as aspects of social needs case management which are difficult to measure but nonetheless critical to fostering health and well-being. Case managers indicated these aspects are most likely to emerge through a longer-term connection with their patients. Thus, while resource-referral solutions may play an important role in addressing basic needs, 30 our findings suggest that weak patient–referrer rapport may be a limitation for such lighter touch interventions. The need for sustained rapport building is also one explanation why longer time horizons may be necessary to show outcome improvements in rigorous studies. 16

Relatedly, results point to trusting relationships as an under-recognised and understudied feature of social needs case management. Existing research finds that patients’ trust in their primary care physician is associated with greater self-reported medication adherence 31 along with health behaviours such as exercise and smoking cessation. 32 Similar quantitative results have not yet been illuminated in social needs case management contexts, yet the prominence of trusting relationships in this study as well as other sources 26 27 33 34 suggests that measures of trust should be used to complement currently emphasised outcomes such as inpatient and outpatient utilisation. Future research and programme evaluation will need to develop new trust measurement or modify existing trust measures for the social needs case management context. 31 35

In summary, study themes provide waypoints of how to conceptualise programme design, new staff training and potential measurement development for complex case management programmes like CommunityConnect. Despite the broad swath of social needs addressed, case managers coalesced on establishing a trusting relationship as a necessary foundation to appropriately identify needs and facilitate connections. Second, fostering patients’ own ideas, including a change their mindset or initiative, was important to fully make use of programme resources. Third, supporting new-found independence or stability was a gratifying, but not universally achieved marker of success. Commonly, case managers highlighted moments of success with mindfulness toward small victories, illuminating that success is non-linear with no certain path nor single end point. Themes emphasise the importance of bringing compassion for the complexity in patients’ lives and developing collaborative relationships one interaction at a time.

Acknowledgments

The authors would like to thank the CommunityConnect evaluation team for their support conducting and transcribing interviews and applying preliminary coding, especially Gabriella Quintana, Alison Stribling, Julia Surges and Camella Taylor.

Contributors: MK coded and analysed qualitative data, identified key themes and related discussion areas, and drafted and critically revised the manuscript. EEE conducted interviews, coded and analysed qualitative data, and drafted and critically revised the manuscript. EH developed the study instrument, conducted interviews, supervised data collection, contributed to the data interpretation and critically revised the manuscript. MDF contributed to the interpretation and critically revised the manuscript. NS contributed to the interpretation and critically revised the manuscript. ALB contributed to the design and interpretation and critically revised the manuscript. All authors approve of the final version to be published.

Funding: MK was supported by the Agency for Healthcare Research and Quality (AHRQ) under the Ruth L. Kirschstein National Research Service Award T32 (T32HS022241). MDF was supported by the Agency for Healthcare Research and Quality, grant # K01HS027648.

Disclaimer: Its contents are solely the responsibility of the authors and do not necessarily represent the official views of AHRQ. Funding had no role in the study’s design, conduct or reporting.

Competing interests: None declared.

Patient and public involvement: Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

Provenance and peer review: Not commissioned; externally peer reviewed.

Data availability statement

Ethics statements, patient consent for publication.

Not applicable.

Ethics approval

This study involves human participants and was approved by Contra Costa Regional Medical Center and Health Centers Institutional Review Committee (Protocol 12-17-2018). Participants gave informed consent to participate in the study before taking part.