Clinical anatomy case discussions linked to improved national exam performance over 10 years

A decade-long study of medical students suggests that discussing anatomy through clinical cases can improve performance on national examinations, particularly when questions require students to connect structures to disease and diagnosis. The findings, reported by researchers at Guilin Medical University in China, offer a rare longitudinal look at whether case-based learning can deliver lasting benefits in one of medicine’s most traditionally memorization-heavy subjects. The analysis followed seven cohorts entering clinical medicine between 2015 and 2021. Students taught with conventional anatomy instruction from 2015 through 2017 were compared with students who experienced structured clinical case discussions from 2018 onward. Across the later cohorts, examination performance generally rose above national benchmarks, even as the university’s enrollment nearly doubled. Yet the results also revealed a warning: when the collection of teaching cases stopped evolving, some of the apparent gains began to fade.

Anatomy is often introduced as a catalogue of bones, muscles, nerves, vessels and organs, each with precise names and spatial relationships. That knowledge is essential, but memorizing isolated structures does not automatically teach students how anatomy functions in a living patient. A nerve injury, for example, becomes clinically meaningful only when a learner can connect the nerve’s course to a pattern of weakness, sensory loss or altered reflexes. Case-based learning, or CBL, is designed to build that connection. Instead of presenting anatomy solely as a sequence of facts, instructors frame it around a clinical problem: a fracture that damages a major artery, a tumor compressing a cranial nerve, or an abdominal pain syndrome whose diagnosis depends on understanding anatomical pathways. Students must retrieve structural knowledge, interpret symptoms and explain the mechanism linking one to the other. The cognitive task shifts from recognition to application, requiring integration across domains that are tested in many medical licensing systems.

The researchers conducted a retrospective cohort study using examination records and teaching information collected over 10 years. The intervention was introduced in 2018, when a major expansion in enrollment threatened to make small-group, clinically oriented teaching more difficult to sustain. The cohorts from 2015, 2016 and 2017 served as the comparison group and received traditional instruction. Those entering from 2018 through 2021 were taught with structured discussions built around clinical scenarios. The investigators assessed students using two national examinations. The Proficiency Test was examined separately for integrated questions, which combine anatomy with clinical reasoning, and non-integrated questions, which focus more directly on individual areas of knowledge. They also examined overall anatomy mastery in the National Medical Licensing Examination. Rather than reporting only raw scores, the study calculated the difference between the university’s performance and national averages, allowing the researchers to track whether students performed above or below the broader benchmark.

The strongest signal appeared in questions designed to test integration. In the Proficiency Test, the 2018 cohort achieved an anatomy mastery differential of 3.37 points above the national average on integrated questions. The gap was not uniform across years or question types, but non-integrated performance also rose, reaching a differential of 5.91 points for the 2020 cohort. On the National Medical Licensing Examination, the experimental group’s overall differential peaked at 5.69 points in 2019. These figures do not mean that every student improved by the same amount, nor do they establish that case discussions alone caused the gains. Examination performance can be influenced by faculty experience, student selection, curriculum changes, preparation strategies and changes in national testing. Still, the pattern is consistent with the central educational theory behind CBL: repeatedly retrieving anatomical facts in realistic contexts may help students organize those facts into usable mental models. Such models are more likely to support transfer, the ability to apply learning to a new patient or unfamiliar question.

The study’s enrollment data made the result particularly notable. The number of students grew from 408 to a peak of 795 during the observation period, an increase of nearly 95 percent. Large classes can create pressure on laboratories, instructors and discussion groups, potentially reducing opportunities for feedback and active participation. In theory, a teaching strategy that depends on interaction could become less effective as student numbers climb. Yet the researchers found that performance remained broadly stable across both national examinations despite the expansion. One possible explanation is that structured cases provide a common framework that helps standardize teaching across multiple groups. If each class works through defined scenarios with agreed learning objectives, instructors can preserve a degree of consistency even when the number of learners increases. The cases may also make it easier to focus limited teaching time on reasoning, rather than spending every session repeating lists of structures. However, the study did not directly measure classroom participation, instructor workload, group size or the quality of individual discussions, so the mechanisms behind the stability remain uncertain.

The most striking setback came with the 2021 cohort. Its performance on integrated Proficiency Test questions fell to 5.34 points below the national average, a sharp reversal from the gains seen in earlier case-based cohorts. The researchers noted that the decline occurred at the same time that the case repository had stopped changing: the clinical scenarios used for instruction had remained essentially unchanged since the 2020 cohort. That timing raises the possibility that students and instructors were relying on a limited and increasingly familiar set of examples. A static case collection can lose educational power in several ways. Students may learn the expected answers without developing flexible reasoning, while repeated scenarios may fail to represent the range of anatomical variations and clinical presentations encountered in examinations or practice. The content can also become misaligned with evolving curricula, diagnostic methods or patterns of disease. The authors therefore interpret the 2021 decline as a signal that case-based learning is not a set-and-forget intervention. Its effectiveness may depend on periodic renewal, careful review and deliberate expansion of the situations students must analyze.

Updating cases, however, is more complicated than simply adding new stories. A useful clinical scenario must be anatomically accurate, medically plausible and pitched at the appropriate level of difficulty. It should contain enough information to support reasoning without becoming a test of obscure trivia. Faculty reviewers must verify the anatomy, clinical terminology, diagnostic logic and learning objectives, while assessment specialists should examine whether the case measures understanding rather than reading speed or prior exposure. A dynamic repository could rotate cases across organ systems, patient ages, disease mechanisms and levels of complexity. It might also include variants of the same underlying problem, forcing students to reason through altered symptoms or imaging findings instead of memorizing a single solution. The paper points toward adaptive assessment and emerging artificial-intelligence tools as possible ways to support this process, but it does not test an AI-driven curriculum. Any automated system would still require expert oversight to prevent inaccurate anatomy, biased patient representations or cases whose apparent novelty masks repetitive reasoning.

The findings also illustrate the value and limitations of long-term educational research. A 10-year record can reveal trends that would be invisible in a single semester or one graduating class, including the possibility that an intervention works initially and then weakens as materials age. At the same time, this was a retrospective, non-randomized comparison of cohorts taught in different periods. The traditional-instruction students were not randomly assigned against case-discussion students, and other changes may have accompanied the 2018 curriculum shift. The study also focused on examination outcomes rather than direct measures of clinical competence, long-term retention or patient-care performance. National benchmark differentials are useful for comparison, but they do not show how students reasoned during teaching or which specific cases produced learning. The records were fully anonymized, and the study received approval from Guilin Medical University’s Medical Ethics Committee, with informed consent waived because the research involved minimal risk and no personal identifiers.

Even with those caveats, the study offers a compelling lesson for medical education at a time when universities are expanding class sizes and searching for ways to preserve quality. Anatomy may be memorized from atlases, models and lectures, but it is understood more deeply when students must use it to explain what is happening to a patient. The Guilin cohorts’ generally strong results suggest that structured clinical discussions can help bridge the gap between foundational science and the integrated questions that dominate modern medical assessment. The sudden decline associated with an aging case repository is equally important, because it shows why successful teaching methods can deteriorate when their content is not maintained. Case-based learning is not simply a matter of placing a diagnosis at the end of an anatomy lecture; it is a system of carefully designed prompts, discussion, feedback and reassessment. The next generation of studies will need to determine which kinds of cases produce durable gains, how frequently repositories should be renewed, and whether adaptive digital tools can improve learning without sacrificing scientific accuracy or the judgment of experienced educators.