Case-based learning: how it works

There are medical textbooks that close each chapter with clinical cases to solve. The question is whether that case teaches something the chapter alone did not teach, and it is the same question the main systematic review on case-based learning in the health professions tried to answer (Thistlethwaite et al., 2012). You know the scene: the chapter seemed clear, and the case at the end shows that you knew the list of signs, but did not know where to start when they arrive out of order, mixed with data that do not matter.

NEXO is a daily medical diagnosis game: one clinical case a day, six progressive clues, and ICD-10 proximity feedback across five levels.

This text answers the textbook's question with what the studies allow us to say. The short answer is uncomfortable: students and teachers like learning with cases, but the evidence that the case teaches more than other formats is inconclusive (Thistlethwaite et al., 2012). No one can promise an exam grade based on it. The case has clear functions, such as organizing the diagnostic investigation, exposing the differential diagnosis and creating a safe place to make mistakes, and it has limits worth knowing before building your routine on it.

Main points

• Case-based learning is guided discussion of a case; problem-based learning (PBL) is open investigation, in which the group decides what to study. When the two were compared at the same school, most preferred the guided version.
• The largest review on the topic concludes that students like the method, but that the data on learning, compared with other activities, are inconclusive. In PBL, knowledge tests tied with traditional teaching; the gain appeared in clinical performance.
• Virtual patients win easily over studying nothing and tie, on average, with teaching without a computer. The serial clues format is the most used, and how much it teaches is an open question that the NEXO Researcher program invites researchers to study.
• Reflecting in a structured way on the case and explaining one's own reasoning improved diagnosis one week later, in experiments with students.
• Decision-making and discussion are trained with people: preceptorship, case sessions and simulation. A diagnosis game complements this practice with individual investigation training.

Summary

• Cases and problems: what is the difference
• Diagnostic investigation: the case as a sequence of questions
• Simulation of care and virtual patients
• Decision-making, discussion and error: what is trained with people
• What the evidence shows and where it stops
• How to study cases alone and in a group
• Frequently asked questions
• Checklist before the next case

Cases and problems: what is the difference

In medical school, "case" and "problem" often appear as synonyms, and the name of the method changes from one curriculum to another. In the literature, the difference is the degree of guidance. The clinical case is the starting point of both methods; what changes is who conducts the investigation and how much structure exists during the discussion.

Case-based learning

In case-based learning, the teacher presents a case and guides the group through questions, in a guided investigation method, with more structure during the session. This is the definition used in the study that followed the switch from PBL to cases at two University of California schools, UCLA and UC Davis (Srinivasan et al., 2007). A review of the world literature describes the method as a way of using cases to give relevance to content and link theory to practice, present on all continents and in various health professions, with Medicine in the lead (McLean, 2016).

Problem-based learning

In PBL, the group receives the problem with little guidance and decides what it needs to study, in an open investigation model that values self-directed study (Srinivasan et al., 2007). In many schools it is the axis of the entire curriculum, and not an isolated activity. This matters for reading the evidence: much of the research on PBL compares entire curricula, and not the effect of studying a case.

When the two California schools swapped PBL for guided cases, 89% of students and 84% of teachers who experienced both formats preferred the cases (Srinivasan et al., 2007). The reasons cited were fewer unfocused detours, less aimless busywork and more chance to apply clinical skills. The few who preferred PBL valued self-directed study and participation. The authors themselves warn that the study measured preference and could not say which method trains better doctors (Srinivasan et al., 2007).

Cases, problems and lecture side by side

The table sums up what the cited sources say about each format. Where the source says nothing, the cell has no statement.

	Case-based learning	Problem-based learning (PBL)	Lecture
How it works	Guided investigation, with more structure in the small group (Srinivasan, 2007)	Open investigation, with self-directed study (Srinivasan, 2007)	The teacher presents the content; it is the comparator in 20 of the 22 studies in a recent review (Varma, 2025)
What students report	They like it and think they learn more (Thistlethwaite, 2012); they preferred cases to PBL (Srinivasan, 2007)	Those who preferred it valued autonomy and participation (Srinivasan, 2007)	No data in the cited sources
Effect on learning	Inconclusive against other activities (Thistlethwaite, 2012); greater critical thinking than in the lecture, with a call for more rigorous trials (Varma, 2025)	Small effect on knowledge and diagnostic reasoning, in one school (Schmidt, 2009); no convincing evidence at the expected magnitude (Colliver, 2000)	Serves as a comparison; in critical thinking, it fell behind the cases (Varma, 2025)
After graduation	No data in the cited sources	Effects mainly on social and cognitive competencies, often self-assessed (Koh, 2008)	Control group in the PBL studies (Koh, 2008)
Main limitation	Many studies with a single class and assessment only after the intervention (Thistlethwaite, 2012)	High resource cost for the measured gain (Colliver, 2000)	No data in the cited sources

Diagnostic investigation: the case as a sequence of questions

The book chapter organizes the disease from the name to the signs: definition, causes, clinical picture, tests, treatment. The case goes the opposite way, from the signs to the name. That is why it exposes gaps that reading hides. You may know how to describe a disease and still not think of it when the complaint arrives on its own.

Hypothesis early, review with every data point

The concepts used to teach clinical reasoning, such as generating hypotheses, recognizing patterns, framing the context, interpreting tests, building the differential diagnosis and verifying the diagnosis, give the language and the method of clinical problem solving (Kassirer, 2010). In the practice of studying with cases, this becomes a simple habit: after each new data point, ask what it does to your list. Does it raise one hypothesis, knock down another or change nothing? If nothing changes, why was that data point collected?

A schematic example helps. A complaint of chest pain opens a broad list, with cardiac, pulmonary, digestive and musculoskeletal causes. The history of onset, duration and aggravating factors reorders that list. The physical exam cuts branches. The laboratory or basic imaging test confirms or knocks down the hypotheses that remain. The order of these steps is the order of the diagnostic investigation, and it is what a well-built case forces you to go through. (The example is didactic and does not replace the reference text for the area.)

Differential diagnosis: the list that must survive

The differential diagnosis is the list of hypotheses that explain the picture, in order of probability and severity. In studying with cases, it plays a role that reading rarely plays: forcing the comparison between similar diseases. If you want a method to avoid forgetting categories of cause, the post on VINDICATE in differential diagnosis shows a way to organize the list.

Generating differentials is not, by itself, what teaches the most. In an experiment with 4th-year students, those who practiced with structured reflection, justifying each hypothesis with the findings for and against, did worse on the immediate test and better a week later than those who only gave the immediate diagnosis or listed differentials. The average score of the reflection group on the delayed test was 0.66, against 0.48 for the group that generated differentials and 0.52 for the immediate diagnosis (Mamede et al., 2012). The gain appeared later, not right away, which matches the feeling that reflecting takes more work and seems to pay off less.

Closing too early

One risk for those who study with cases is stopping at the first plausible hypothesis and failing to consider the alternatives. Stopping to reflect seems to help precisely in difficult cases: in internal medicine residents in Brazil, reflective reasoning did not change accuracy in simple cases, but improved the diagnosis of complex and uncommon cases (Mamede et al., 2008). Before closing a difficult case, ask what else would explain the same findings.

Simulation of care and virtual patients

Simulation of care puts you in a clinical situation without risk to anyone. It ranges from a mannequin in a skills lab to a virtual patient on the screen, and each format trains different things.

What a virtual patient is

A virtual patient is a digital and interactive simulation of a clinical scenario, made to teach health professionals (Kononowicz et al., 2019). In general, you collect the history, choose tests and decide on management, and the case responds to your choices. It is broader than a case in clues, in which information arrives in a fixed order.

What the meta-analyses show

Compared to not studying anything, virtual patients had large effects: 0.94 in knowledge and 0.80 in clinical reasoning. Compared to teaching without a computer, the effects were close to zero, with reasoning at -0.004 (Cook et al., 2010). In other words, the virtual patient teaches, but does not teach more than a good class or a good discussion with paper and pen. The same meta-analysis points to designs that seem to help: repeating until mastery is shown, giving more complete feedback and presenting cases that contrast explicitly.

In a later review, with randomized trials up to 2018, virtual patients tied with traditional teaching in knowledge (standardized difference of 0.11) and were better in skills (0.90), among them clinical reasoning. The quality of the evidence was, in general, low (Kononowicz et al., 2019).

The serial clues format

There are two ways to present a case to teach clinical reasoning: gradually, in serial clues, or whole at once. A narrative review found that the serial clues format is the most used, perhaps because it imitates what the doctor does when investigating, and that its effectiveness still needs to be measured. There was some evidence in favor of the whole case aimed at knowledge, and the approaches that try to teach "a way of thinking" were, in large part, ineffective (Schmidt and Mamede, 2015). It is also the format of NEXO, which presents the case in the order in which the investigation happens. To measure what this format teaches, NEXO maintains the NEXO Researcher program, open to researchers and institutions that want to propose studies on learning and diagnostic reasoning. It is worth combining it with the reading of complete cases, which had some evidence in favor when the goal is knowledge.

Decision-making, discussion and error: what is trained with people

Diagnosing is half the case. The other half is deciding: which test to order, what to do now, what to expect, when to call for help. This part depends on context, on risk and on someone who challenges your choice. It is trained outside a diagnostic game: in preceptorship, in case sessions, in skills lab simulations and in discussions at the bedside.

Deciding requires someone who challenges

The theory behind teaching clinical reasoning proposes that memory of reasoning strategies improves when information, judgment and reasoning errors are pointed out and discussed right away, and that the participation of a preceptor increases the value of the experience (Kassirer, 2010).

Justify the hypothesis out loud

In the case session, the most useful question is "why?". Why does this hypothesis come before the other? Which finding supports it and which contradicts it? The structured reflection of the experiment with 4th-year students followed this logic and had an effect one week later (Mamede et al., 2012). In the group discussion, save the detailed justification for the difficult cases, where reflection seems to pay off more, as in the study with residents cited above (Mamede et al., 2008).

A practical recommendation to test in your session: each person writes their own answer before hearing the others. That way the discussion compares lines of reasoning, instead of following whoever spoke first.

Record the error and correct it

Laboratory studies indicate that making a mistake and receiving correction right after helps learning, and that errors made with great confidence are corrected more easily. Feedback that analyzes the reasoning that led to the error is decisive (Metcalfe, 2017). In practice, the error only teaches if you look at it again: write down the diagnosis you gave, the correct one, the clue that should have changed your list and the likely reason for the error (did not know, knew and did not remember, or closed too early). Studies show the value of making a mistake and correcting it; the error notebook is a practical way to apply this day to day.

What the evidence shows and where it stops

Back to the book's question: does the case teach something the chapter did not teach? The reviews allow a partial answer.

Liking is not learning

The BEME systematic review on case-based learning in the health professions included 104 articles, with a search up to 2010. The conclusion is direct: students like it and think they learn more, teachers like it too, but the empirical data, as a whole, are inconclusive as to the effect on learning compared with other activities. The review also could not separate the effect of the case from the effect of studying in a small group. In 75% of the articles, the assessment was done only after the intervention, with no measurement before it (Thistlethwaite et al., 2012).

The review of the world literature goes in the same direction: the most common assessments were a satisfaction questionnaire and a knowledge test, and outcomes in practice were less frequent (McLean, 2016). A more recent systematic review, with studies almost all against lecture-based teaching, found greater critical thinking in the group with cases, with a standardized difference of 0.75, and the authors themselves call for more rigorous randomized trials (Varma et al., 2025).

Full-curriculum PBL

For full-curriculum PBL, the reviews are older and more divided, and it is worth separating guided cases from open PBL, as in the California study (Srinivasan et al., 2007). A critical review found no convincing evidence that PBL improves the knowledge base and clinical performance to the magnitude expected from the resources it requires (Colliver, 2000). A meta-analysis of a single school with a PBL curriculum calculated effects in 270 comparisons: much better performance in interpersonal and practical skills, and positive but small differences in medical knowledge and diagnostic reasoning (Schmidt et al., 2009). A systematic review on the physician after graduation found effects of PBL mainly in social and cognitive competencies, such as dealing with uncertainty, communication and self-directed study, with part of the evidence based on self-assessment (Koh et al., 2008).

Exam, clinical performance and theoretical basis

The question "will this be on the test?" has an answer in PBL reviews. In a meta-analysis of studies from 1970 to 1992, PBL and traditional teaching did not differ on tests of factual knowledge or clinical knowledge (a difference of +0.08 in the latter). PBL was superior in clinical performance (+0.28) and in the students' evaluation of the course, while traditional teaching students were better in the first step of the United States national exam (NBME I), with large variation between schools (Vernon and Blake, 1993). In a systematic review of the preclinical cycle, 12 of 15 studies found no significant difference in test scores; the authors conclude that PBL does not change knowledge acquisition (Hartling et al., 2010).

There is also a possible cost. A 1993 review found PBL students who, in some cases, did worse on basic science exams and felt less prepared in them, with apparent gaps in their knowledge base (Albanese and Mitchell, 1993). Hence the rule of thumb in this text: the case complements theory, it does not replace it. That is what the book in the opening does, with the case after the chapter.

What can be concluded

Three things hold up. First, the case is not a shortcut to a test score: the gains in knowledge, when they appear, are small. Second, what seems to make a difference is what you do with the case: reflect in a structured way, explain your own reasoning and receive correction. Third, the effect depends on the comparison: against not studying, almost everything works; against a good class, the difference shrinks (Cook et al., 2010).

How to study cases alone and in a group

The case pays off more when it comes after the content, as in the book that closes the chapter with cases. It complements theory; it does not replace it.

Alone: explain before checking

In an experiment with third-year students in clinical clerkship, those who generated self-explanations while solving cases, telling themselves why each finding mattered, had better diagnostic performance one week later. The effect appeared in cases on less familiar topics, and that was without any feedback (Chamberland et al., 2011). A possible routine:

1. Read the complaint and write three hypotheses before continuing.
2. With each new piece of data, reorder the list and note in one line why it changed.
3. Before seeing the answer, write the diagnosis and the finding that weighed most.
4. When checking, compare your list with the explanation and mark the clue where you should have changed your mind.
5. For the cases you got wrong, do the structured reflection: findings in favor, findings against, what you would expect to find if it were the other disease (Mamede et al., 2012).

Prefer cases on topics you know less well. It was in these that self-explanation showed an effect in the experiment cited.

In a group: each person answers before discussing

A well-run case session has simple rules. As a practical recommendation, each person writes their own hypothesis before hearing the others. Someone plays the role of the challenger and asks "what if it is not this?". At the end, the group names the most common error of the session and the clue that undid it. If there is a preceptor, ask them to point out the error right away, which is what teaching with a preceptor proposes (Kassirer, 2010).

The group is also the place for the part that a diagnostic case does not cover: management, priority, communication. Bring to the session the case you got wrong alone and discuss what you would do next, with someone who knows more than you. If you produce medical content and use NEXO for studying, learn about the NEXO Creators program.

Frequently asked questions

What is the difference between case-based learning and problem-based learning?

Both start from a clinical case. In case-based learning, the investigation is guided by the teacher, with more structure; in PBL, it is open, and the group decides what to study. When the two were compared by the same students and teachers, most preferred the guided version, although the study did not measure which form produces better doctors (Srinivasan et al., 2007).

Does case-based learning improve test scores?

The evidence does not allow promising that. The main review of the method concludes that the data on learning, compared with other activities, are inconclusive (Thistlethwaite et al., 2012). In PBL, test scores generally did not change (Hartling et al., 2010), and gains in knowledge were small (Schmidt et al., 2009). What has experimental support is the way of studying the case: structured reflection and self-explanation improved diagnosis one week later (Mamede et al., 2012; Chamberland et al., 2011).

How do you study cases in a group?

Each person writes their own hypothesis before hearing the others; then the group discusses and someone plays the role of the one who challenges. If there is a preceptor, ask them to point out the error on the spot (Kassirer, 2010).

How does a case in clues relate to a virtual patient?

Both simulate a consultation. In the virtual patient, in general, you choose what to ask, which tests to order and what to do, and the case responds to your choices (Kononowicz et al., 2019). In the case in clues, the information arrives in a fixed order, which reproduces the sequence of the diagnostic investigation. The clues format is the most used in teaching clinical reasoning, and the comparison of effectiveness between the formats is an open question, which NEXO Researcher receives as a study proposal (Schmidt and Mamede, 2015).

How can NEXO help in the discussion of college cases?

NEXO trains the individual part of the diagnostic investigation: one case per day, up to six clues, feedback by proximity in ICD-10 and clinical explanation. The cases you played become material for the case session: bring your list of hypotheses, the clue at which you changed your mind and the clinical explanation, and discuss management and priorities with the group and the preceptor. If you are a doctor and want to review the cases in your specialty, learn about the NEXO Curator program. To bring NEXO to your medical school, academic league or residency, learn about the NEXO Closer program.

Checklist before the next case

1. Study the content beforehand, and use the case to test what you read.
2. Write three hypotheses before moving past the complaint.
3. Reorder the list with each piece of data and note why it changed.
4. Explain to yourself why each finding matters, especially in topics you are less comfortable with.
5. Do the structured reflection on the cases you got wrong: findings for, against and what you would expect if it were the other disease.
6. Record the error and the clue that undid it, and return to it in the Archive or in your review cards.
7. Take the decision to someone who can discuss it: preceptor, case session or simulation lab.

Want to practice diagnostic investigation with one case per day? Play today's case on NEXO.

References

1. Srinivasan M, Wilkes M, Stevenson F, et al. Comparing problem-based learning with case-based learning: effects of a major curricular shift at two institutions. Acad Med. 2007. https://pubmed.ncbi.nlm.nih.gov/17198294/
2. McLean SF. Case-Based Learning and its Application in Medical and Health-Care Fields: A Review of Worldwide Literature. J Med Educ Curric Dev. 2016. https://pubmed.ncbi.nlm.nih.gov/29349306/
3. Varma B, Karuveettil V, Fernandez R, et al. Effectiveness of case-based learning in comparison to alternate learning methods on learning competencies and student satisfaction among healthcare professional students: A systematic review. J Educ Health Promot. 2025. https://pubmed.ncbi.nlm.nih.gov/40144176/
4. Thistlethwaite JE, Davies D, Ekeocha S, et al. The effectiveness of case-based learning in health professional education. A BEME systematic review: BEME Guide No. 23. Med Teach. 2012. https://pubmed.ncbi.nlm.nih.gov/22578051/
5. Schmidt HG, van der Molen HT, te Winkel WWR, Wijnen WHFW. Constructivist, Problem-Based Learning Does Work: A Meta-Analysis of Curricular Comparisons Involving a Single Medical School. Educational Psychologist. 2009. https://doi.org/10.1080/00461520903213592
6. Colliver JA. Effectiveness of problem-based learning curricula: research and theory. Acad Med. 2000. https://pubmed.ncbi.nlm.nih.gov/10724315/
7. Koh GC, Khoo HE, Wong ML, Koh D. The effects of problem-based learning during medical school on physician competency: a systematic review. CMAJ. 2008. https://pubmed.ncbi.nlm.nih.gov/18166729/
8. Kassirer JP. Teaching clinical reasoning: case-based and coached. Acad Med. 2010. https://pubmed.ncbi.nlm.nih.gov/20603909/
9. Mamede S, van Gog T, Moura AS, et al. [Translated title: Reflection as a strategy to promote the acquisition of diagnostic competence by medical students]. Med Educ. 2012. https://pubmed.ncbi.nlm.nih.gov/22515754/
10. Kononowicz AA, Woodham LA, Edelbring S, et al. Virtual Patient Simulations in Health Professions Education: Systematic Review and Meta-Analysis by the Digital Health Education Collaboration. J Med Internet Res. 2019. https://pubmed.ncbi.nlm.nih.gov/31267981/
11. Cook DA, Erwin PJ, Triola MM. Computerized virtual patients in health professions education: a systematic review and meta-analysis. Acad Med. 2010. https://pubmed.ncbi.nlm.nih.gov/20703150/
12. Schmidt HG, Mamede S. How to improve the teaching of clinical reasoning: a narrative review and a proposal. Med Educ. 2015. https://pubmed.ncbi.nlm.nih.gov/26383068/
13. Mamede S, Schmidt HG, Penaforte JC. Effects of reflective practice on the accuracy of medical diagnoses. Med Educ. 2008. https://pubmed.ncbi.nlm.nih.gov/18412886/
14. Metcalfe J. Learning from Errors. Annu Rev Psychol. 2017. https://pubmed.ncbi.nlm.nih.gov/27648988/
15. Chamberland M, St-Onge C, Setrakian J, et al. The influence of medical students' self-explanations on diagnostic performance. Med Educ. 2011. https://pubmed.ncbi.nlm.nih.gov/21649701/
16. Vernon DT, Blake RL. Does problem-based learning work? A meta-analysis of evaluative research. Acad Med. 1993. https://pubmed.ncbi.nlm.nih.gov/8323649/
17. Hartling L, Spooner C, Tjosvold L, Oswald A. Problem-based learning in pre-clinical medical education: 22 years of outcome research. Med Teach. 2010. https://pubmed.ncbi.nlm.nih.gov/20095771/
18. Albanese MA, Mitchell S. Problem-based learning: a review of literature on its outcomes and implementation issues. Acad Med. 1993. https://pubmed.ncbi.nlm.nih.gov/8447896/
19. Lee CY, Lee CH, Lai HY, et al. Do games work? A meta-analytic synthesis of gamified learning for clinical reasoning in medical and allied health education. Med Educ Online. 2026. https://pubmed.ncbi.nlm.nih.gov/41532745/
20. Schmidt M. Outcomes of a Foundational Sciences Daily Engagement Game in Medical Education. Biochem Mol Biol Educ. 2026. https://pubmed.ncbi.nlm.nih.gov/42204775/