Guided group reflection versus SNAPPS in enhancing clinical reasoning in final-year MBBS students at a public-sector teaching hospital in Pakistan: a quasi-experimental study

Abstract

Objectives

(1) To compare the effectiveness of guided group reflection (GGR) versus SNAPPS (Summarise history and physical examination, Narrow differentials, Analyse the differentials, Probe the preceptor, Plan of patient management, Self-directed learning), as a teaching intervention in improving clinical reasoning (CR) skills of final-year MBBS (Bachelor of Medicine, Bachelor of Surgery) students. (2) To determine which teaching intervention is more effective in improving specific components of CR (data acquisition, problem presentation, hypothesis generation, illness script comparison and diagnosis) among final-year MBBS students.

Design

Quasi-experimental study involving two parallel intervention groups with pretest and post-test research design.

Setting

The study was conducted from March to June 2025 at the medical department of Holy Family Hospital, affiliated with Rawalpindi Medical University.

Participants

A total of 128 final-year Bachelor of Medicine, Bachelor of Surgery students participated over three consecutive internal medicine clinical rotations, each rotation lasting four weeks and involving approximately 45 students.

Intervention

After the informed consent, the participants were randomly allocated into two groups using a lottery method: one receiving GGR intervention and the other receiving SNAPPS intervention. Four clinical cases: chronic dyspnoea, anaemia, chronic liver disease and chronic diarrhoea were addressed during the teaching sessions in both groups. To maintain ethical fairness, students were offered crossover exposure to the alternate teaching method after a 4-week gap.

Outcomes

A pretest comprised of 15 extended matching questions (EMQs) was used to assess baseline CR skills of the students. Post-tests using blueprint-aligned EMQs were administered after each session, with cumulative scoring at the end of each rotation. The primary outcome was overall improvement in CR, whereas the secondary outcome was improvement in performance of specific CR components.

Results

Both groups showed significant improvement in overall CR scores (p<0.001). The GGR group achieved higher postintervention scores compared with SNAPPS (32.25±6.2 vs 28.89±9.0; p=0.016), with a moderate effect size favouring GGR (Cohen’s d=0.43). Domain-level analysis showed greater gains with GGR in data acquisition, problem presentation and hypothesis generation, while no significant differences were observed for illness script comparison or final diagnosis.

Conclusion

Both GGR and SNAPPS teaching methods effectively enhance CR skills among undergraduate medical students. However, GGR demonstrated greater effectiveness in CR domains requiring deeper analytical thinking and collaborative reflection, especially when managing complex cases. Incorporating GGR within undergraduate clinical clerkships may strengthen diagnostic reasoning, particularly in settings with limited faculty–student teaching opportunities.

STRENGTH AND LIMITATIONS OF THIS STUDY.

• This study compared the two structured clinical teaching strategies, guided group reflection and Summarise history and physical examination, Narrow differentials, Analyse the differentials, Probe the preceptor, Plan of patient management, Self-directed learning, within an authentic internal medicine clerkship setting, making the findings relevant to real undergraduate clinical teaching.

• The study was conducted in a single public-sector teaching hospital, which may limit generalisability to institutions with different curriculums, clinical exposure or teaching environments. In addition, excluding students with <70% marks in the previous professional examination may have introduced selection bias and may limit the applicability of findings to lower-performing students.

• The 4-week intervention period only assessed the short-term effect of teaching interventions and did not assess long-term retention or transfer to real clinical performance.

• The Extended matching question (EMQ) tool, although valid and reliable, could not fully capture the complexity of real-time diagnostic reasoning.

• Partial overlap between pretest and post-test EMQ items introduced a minor practice effect; however, it was applied equally across both study groups.

Introduction

Clinical reasoning (CR) is the ability to gather, process and analyse patient information to arrive at a diagnosis.¹ As a critical skill for medical students, its underdevelopment increases the risk of diagnostic errors, directly impacting patient outcomes.² Teaching CR is challenging due to its complexity, but various strategies, such as structured reflection, self-explanation, differential diagnosis prompts, Summarise history and physical examination, Narrow differentials, Analyse the differentials, Probe the preceptor, Plan of patient management, Self-directed learning (SNAPPS) and illness scripts have been shown to facilitate its development.³

Learner-centred strategies like SNAPPS are effective in improving CR⁴ but often increase the duration of case presentations, which can strain already resource-limited clinical settings with faculty shortages and limited patient availability.⁵ These challenges hinder individualised feedback and necessitate teaching approaches that are both efficient and collaborative, enabling students to engage in CR development within constrained environments.

Although a variety of methods exist to enhance CR, limited research directly compares collaborative approaches like guided group reflection (GGR) with individual strategies such as SNAPPS.⁶ There is also a lack of evidence on their comparative effectiveness in improving specific components of CR (eg, data acquisition, hypothesis generation and diagnostic formulation) during the high-pressure, time-limited settings of medical clerkships.³

This study seeks to bridge the gap by comparing GGR and SNAPPS in developing CR skills among final-year Bachelor of Medicine, Bachelor of Surgery (MBBS) students during their medical clerkship. The findings may inform clinical educators about the potential role of structured reflective and learner-centred teaching strategies for supporting CR in resource-constrained clinical learning environments. This study further offers insight into the relative effectiveness of collaborative versus individual learning formats for enhancing CR competency.

Purpose of the study

The purpose of this study is to evaluate the effectiveness of GGR in comparison to SNAPPS in the enhancement of CR competency among final-year MBBS students during their internal medicine clerkship. The study aims to determine which strategy better supports the development of specific CR components, such as data acquisition, hypothesis generation, critical analysis of differential and diagnostic accuracy, while remaining feasible for integration into real-world undergraduate medical education settings.

Methods

Study design and setting

A quasi-experimental, parallel-group pre–post study was conducted among final-year MBBS students during their internal medicine clerkship at Holy Family Hospital (HFH), Rawalpindi. HFH is a tertiary-care teaching hospital affiliated with Rawalpindi Medical University (RMU). The study was carried out over three consecutive clerkship cycles between 15 March and 14 June 2025.

Participants

Eligible participants were final-year MBBS students coming for their mandatory 4-week clinical rotation in internal medicine at HFH. Students who were excluded from participating in the study were those who were posted at other allied hospitals of RMU. Furthermore, those students who had secured <70% marks in the fourth-year professional examination were also excluded to reduce baseline academic heterogeneity and reduce confounding related to prior academic performance.

Sample size and sampling

A total of 128 final-year MBBS students were included in the study, with equal allocation to the two study groups. The required sample size (n=128) was estimated using G*Power, assuming a medium effect size (0.5), α=0.05 and 80% power. Eligible students rotating through internal medicine at HFH during the three clerkship cycles were invited to participate using convenience sampling. This was followed by simple random allocation within each batch; students were assigned in a 1:1 ratio to either GGR or SNAPPS. Because students were first recruited through convenience sampling before allocation to intervention groups, the study was classified as quasi-experimental rather than a fully randomised controlled trial. To compensate for anticipated attrition, 138 students were initially enrolled; 10 subsequently withdrew, leaving 64 participants in each group.

Ethical considerations

Primary ethical approval for this study as MHPE (Master in Health Profession Education) thesis was obtained from the Institutional Review Committee of Riphah International University (Ref: Riphah/IIMC/25/1021). Site-specific approval for data collection was granted by the Institutional Review Board of RMU (Ref: 2088/IREF/RMU/25). All participants provided informed consent; participation was voluntary, and confidentiality was maintained. Both groups offered crossover exposure after data collection to ensure equal learning opportunity. A pilot study was conducted on 14 house officers to assess the feasibility of interventions and to assess the validity and reliability of the pretest and post-test extended matching questions (EMQs).

Patient and public involvement statement

The patients and public were not involved in design, conduct, reporting and dissemination of study results.

Teaching interventions

Guided group reflection (GGR)

GGR was delivered using the deliberate reflection framework described by Mamede and Schmidt.⁷ Each 60-min session followed six sequential steps:

Step 1: initial hypothesis generation

Step 2: identification of supportive findings

Step 3: identification of contradictory findings

Step 4: recognition of missing expected findings

Step 5: generation of alternative hypotheses (repeating steps 2 to 3 with each alternative hypothesis)

Step 6: prioritisation of differentials and facilitator feedback

Each case-based session lasted 1 hour and was facilitated by trained instructors.

Summarise history and physical examination, Narrow the differential diagnoses, Analyse the differentials, Probe the preceptor, Plan patient management and Select issues for self-directed learning (SNAPPS)

SNAPPS sessions followed the standard six-step framework. Students performed a focused history and examination (20 min), organised findings (10 min), and presented their reasoning to the facilitator (15 min). Total duration was 45–60 min. Facilitators adhered to a validated SNAPPS protocol.

Clinical cases

The participants of both study groups (GGR and SNAPPS) were exposed to similar clinical cases (chronic dyspnoea, anaemia, chronic liver disease and chronic diarrhoea) with predefined learning objectives as per the clerkship curriculum. The students completed 4 weekly sessions, one per clinical case.

Faculty training and standardisation

Facilitators attended a 60-min orientation session for each intervention before data collection. The sessions focused on standardising instructional delivery and familiarising facilitators with the GGR and SNAPPS frameworks. All training sessions were conducted by the same instructor to ensure consistency. After training, facilitators were randomly assigned to the GGR and SNAPPS groups. To minimise contamination, GGR and SNAPPS sessions were conducted in separate medical units. Furthermore, students and facilitators were also blinded to the study hypothesis.

Data collection instrument

CR was assessed using a 15-item EMQ test developed for this study. A blueprint was prepared to align items with the four teaching topics: chronic dyspnoea, anaemia, chronic liver disease and chronic diarrhoea and five CR components: data acquisition, problem presentation, hypothesis generation, comparing/contrasting illness scripts and diagnosis (tables1 2).

Table 1. EMQ distribution across clinical topics.

Clinical topic	No. of EMQs	Cognitive level	Total marks
Chronic dyspnoea	4	C4–C5	12
Anaemia	3	C4–C5	9
Chronic liver disease	4	C4–C5	12
Chronic diarrhoea	4	C4–C5	12

EMQ, extended matching question.

Table 2. EMQ distribution across clinical reasoning components.

Clinical reasoning component	No. of EMQs	Cognitive level	Total marks
Data acquisition	3	C4–C5	9
Problem presentation	3	C4–C5	9
Hypothesis generation	3	C4–C5	9
Comparing illness scripts	3	C4–C5	9
Diagnosis	3	C4–C5	9

EMQ, extended matching question.

The principal investigator initially developed 18 EMQs. Each item included a theme: topic, a subtheme: CR component, eight response options, a lead-in statement and three clinical vignettes. All items were written at Bloom’s taxonomy levels C4 (analysis) and C5 (evaluation), reflecting the higher-order cognitive processes required for CR.

Six internal medicine experts, each with at least 5 years of undergraduate teaching and EMQ-writing experience, reviewed the items for content relevance, clarity, face validity, cognitive level and blueprint alignment. All 18 items achieved Content Validity Index (CVI) and Face Validity Index (FVI) values >0.8.

15 validated EMQs were selected for the pretest. The post-test also contained 15 EMQs, including 12 repeated items and three matched new items to maintain construct alignment while reducing the practice effect. Post-tests were administered after each weekly session and aggregated into a cumulative postintervention score. Both assessments followed the same topic and reasoning-domain blueprint (tables 1 and 2). Reliability coefficients for both tests were >0.8.

Pilot testing was conducted among 14 house officers to assess feasibility, clarity, item difficulty and internal consistency. Cronbach’s alpha exceeded 0.9 during pilot testing and exceeded 0.8 in the main study, demonstrating good reliability. Test–retest reliability was not performed because the pilot phase primarily focused on feasibility and internal consistency assessment.

Data collection procedures

Each batch completed a pretest on the first day of rotation. Students then attended orientation sessions, followed by weekly GGR or SNAPPS sessions linked to the four clinical cases. After each session, participants completed a case-specific EMQ post-test. All post-test scores across 4 weeks were aggregated into a cumulative postintervention score. This process was repeated for all three batches following identical procedures, as illustrated in the study flow diagram (figure 1).

Figure 1. Flowchart diagram showing the data collection process. CR, Clinical Reasoning; EMQs, extended matching questions; GGR, guided group reflection; MBBS, Bachelor of Medicine, Bachelor of Surgery; SNAPPS, Summarise history and physical examination, Narrow the differential diagnoses, Analyse the differentials, Probe the preceptor, Plan patient management and Select issues for self-directed learning.

Data analysis

Data were analysed using SPSS V.25. Descriptive statistics were calculated for all variables. Within-group change was assessed using paired t-tests; between-group comparisons used independent t-tests. Assumptions of normality were verified using Shapiro–Wilk testing and visual plots. Statistical significance was set at p <0.05. Sub-scores for the five reasoning components were analysed to identify specific areas of improvement.

Results

A total of 128 final-year medical students completed the study, with equal allocation to the two intervention groups: GGR and SNAPPS. Baseline demographic characteristics and academic performance were similar. The participants of both study groups did not differ in their pre-intervention EMQ scores, ensuring equivalent baseline CR skills at the start of the research (table 3).

Table 3. Continuous demographic characteristics.

Demographics	Mean±SD
Age	23.20±0.50 (23–35)
Fourth-year MBBS final professional exam (%)	78.5 (70–90)
Last academic year marks: GGR group (%)	78.3±5.1
Last academic year marks: SNAPPS group (%)	78.7±5.0

GGR, guided group reflection; MBBS, Bachelor of Medicine, Bachelor of Surgery; SNAPPS, Summarise history and physical examination, Narrow the differential diagnoses, Analyse the differentials, Probe the preceptor, Plan patient management and Select issues for self-directed learning.

Both study groups demonstrated significant within-group improvement following the intervention. Between-group comparison showed that the GGR group achieved significantly higher postintervention scores compared with SNAPPS (32.25±6.2 vs 28.89±9.0; p=0.016), indicating a moderately better overall effect of GGR on CR (table 4). Domain-wise analysis revealed improvement across all five components in both groups. However, GGR showed significantly higher postintervention scores in data acquisition (p=0.030), problem presentation (p=0.010) and hypothesis generation (p=0.027). However, no statistically significant differences were observed between groups for comparing illness scripts (p=0.249) or diagnosis (p=0.281), although the GGR group showed numerically higher means (table 4).

Table 4. Baseline and postintervention EMQ scores between groups.

Outcome	GGR mean±SD	SNAPPS mean±SD	P value	Effect size (d)
Pretest (overall)	25.88±6.8	24.18±7.5	0.181	–
Post-test (overall)	32.25±6.2	28.89±9.0	0.016	0.43
Data acquisition	6.94±1.6	5.94±2.1	0.030	–
Problem presentation	6.78±1.5	5.97±1.9	0.010	–
Hypothesis generation	6.17±1.9	5.73±2.2	0.027	–
C&C illness script	6.14±1.5	5.78±2.1	0.249	–
Diagnosis	6.23±1.6	5.48±2.1	0.281	–

C&C, Compare & Contrast; EMQ, extended matching question; GGR, guided group reflection; SNAPPS, Summarise history and physical examination, Narrow the differential diagnoses, Analyse the differentials, Probe the preceptor, Plan patient management and Select issues for self-directed learning.

In addition, the participants showed statistically significant improvement in scores within groups for all clinical cases (all p<0.001). Between-group comparisons showed that GGR produced significantly higher postintervention scores for chronic liver disease (p=0.005), chronic dyspnoea (p=0.028) and chronic diarrhoea (p=0.035). No significant difference was found in the anaemia topic (p=0.274), suggesting similar performance of both strategies for simple pattern recognition cases (table 5).

Table 5. Postintervention topic-wise EMQ scores between groups.

Topic	GGR mean±SD	SNAPPS mean±SD	P value
Anaemia	7.05±1.3	6.72±1.9	0.274
Chronic liver disease	8.06±1.9	6.94±2.4	0.005
Chronic dyspnoea	9.11±2.1	8.08±3.0	0.028
Chronic diarrhoea	8.06±2.2	7.11±2.8	0.035

EMQ, extended matching question; GGR, guided group reflection; SNAPPS, Summarise history and physical examination, Narrow the differential diagnoses, Analyse the differentials, Probe the preceptor, Plan patient management and Select issues for self-directed learning.

Lastly, the Cronbach’s alpha was 0.83 for the pretest and 0.85 for the post-test, demonstrating good internal consistency of the EMQ instrument.

Discussion

A total of 128 final-year MBBS students participated in the study during their internal medicine clinical clerkship at HFH. The data collection was conducted over three consecutive 4-week rotations. Each rotation involved one batch of approximately 45 students. Each batch was divided randomly into two study groups. Each study group was exposed to four clinical cases (anaemia, chronic dyspnoea, chronic diarrhoea and chronic liver disease) over 4 weeks. EMQ-based pretest and post-test were administered to evaluate changes in CR. To strengthen the validity of the observed differences between GGR and SNAPPS, the reliability of the 15-item EMQ was assessed at both measurement points. The results demonstrated acceptable internal consistency (α=0.83 pretest, α=0.85 post-test), confirming that the EMQ functioned as a stable and reliable assessment tool across the study cohort. These coefficient values were consistent with published evidence supporting the use of EMQs for evaluating CR in undergraduate medical education.⁸ Both instructional strategies improved students’ CR scores. The GGR group had a higher mean post-test score (32.25±6.2) than the SNAPPS group (28.89±9.0), with a moderate effect size (Cohen’s d=0.43). However, this difference should be interpreted cautiously, as the advantage was not consistent across all CR components. GGR showed greater improvement in data acquisition, problem presentation and hypothesis generation, while performance in comparing illness scripts and diagnosis did not differ significantly between groups. Thus, the findings suggested that GGR may offer additional benefit for selected analytic components of CR rather than demonstrating overall superiority over SNAPPS.

In this study, as both GGR and SNAPPS were associated with statistically significant improvement in post-test CR scores compared with baseline, suggesting that structured teaching strategies can support the development of CR among undergraduate medical students. These findings align well with Delavari et al, who similarly concluded that structured teaching interventions outperform traditional methods in fostering CR skills at the undergraduate level.⁹ Meanwhile, Lambe et al previously found that guided reflection did not significantly improve diagnostic accuracy among undergraduate medical students.¹⁰ They attributed this outcome to students’ existing reliance on analytical (type 2) thinking, due to their limited practical clinical exposure.¹⁰ However, our study showed contrasting results, as the GGR intervention, incorporating deliberate reflection, promoted deeper analytical thinking, collaborative discussion and idea sharing among participants. These structured group interactions enabled students to critically evaluate their reasoning, ultimately enhancing their diagnostic abilities.

Our study also evaluated which of the two interventions was more effective in improving specific components of CR. Five key domains were assessed: data acquisition, problem presentation, hypothesis generation, comparison and contrast of illness scripts and final diagnosis.⁶ Each of these components plays an important role in the development of CR and contributes directly to enhancing diagnostic accuracy in medical students. In our study, domain-specific performance score analysis revealed statistically significant dominance of GGR in data acquisition (p=0.03), problem presentation (p=0.010) and hypothesis generation (p=0.027). However, no statistically significant differences were noted between groups in comparing and contrasting illness scripts or diagnostic accuracy domains.

The better performance of the GGR group in the domains of data acquisition and problem presentation in our study could be explained by the structured nature of deliberate reflection. As supported by Mamede and Schmidt, in the GGR deliberate framework, learners were encouraged to repeatedly analyse the clinical case by reviewing the available information to confirm or revise their initial diagnostic hypothesis.⁷ This reflective process encouraged students to develop a habit of revisiting the clinical history and physical findings, allowing students to identify missing or inconsistent data. This cognitive intervention helped the students not to develop the diagnostic tunnel vision during their encounters with patients in the clinical setting.¹¹ This intervention also facilitated the students in having a better understanding of how different symptoms collectively present across various clinical cases. On the other hand, in our study, the improvement of the post-test scores was not statistically significant in the SNAPPS group. This was because in SNAPPS, students had to collect data from the patient and analyse the differentials individually without the help from peers as in GGR collaborative learning sessions. Hence, SNAPPS is considered more suitable for postgraduate students.⁵

The higher post-test scores of students in the GGR group compared with the SNAPPS group in our study for the hypothesis generation domain further demonstrated the benefits of its structured reflective process. Students were encouraged to refine their hypotheses by repeatedly evaluating the clinical clues that supported or opposed each differential diagnosis. This process likely strengthened their analytical thinking and CR. Similar findings were documented by Mamede et al, who observed that structured reflection enhanced learners’ ability to generate clinical hypotheses through deliberate cognitive engagement.⁷

In this study, both instructional strategies GGR and SNAPPS led to improvement in students’ post-test scores in the domain of comparing and contrasting illness scripts. However, the difference between the two groups was not statistically significant. This could be explained by the fact that both interventions provided a structured framework to students to consider multiple diagnostic possibilities, identify conflicting clinical cues and determine which additional findings could support or refute an alternative diagnosis. In GGR, this was achieved through the guided steps of deliberate reflection, while in SNAPPS, similar cognitive engagement occurred during the ‘analyze the differential’ step.^{7 12} Prior studies, including Grunewald et al, had reported mixed results on the effectiveness of structured interventions in improving illness script comparison, suggesting that outcomes may vary depending on context, facilitator expertise and learner maturity.¹³ These results again emphasised the importance of aligning instructional methods with learner needs and experience rather than assuming uniform superiority of one approach over another.

For the diagnosis domain, the GGR group showed a higher mean score than the SNAPPS group; however, the difference was not statistically significant. Although both interventions required learners to provide a working diagnosis and justify their reasoning, the format of GGR provided students with an additional benefit through group-based discussion. Justifying and defending a diagnosis in front of peers encouraged metacognitive monitoring, allowing students to reflect on their thought processes in real time.¹⁴ This reflective process was supported by immediate peer and facilitator feedback. This could explain the slightly better performance seen in the GGR group. In contrast, the students of the SNAPPS group relied on a one-on-one interaction with a preceptor. This hierarchical relation between medical students and preceptor could be a hurdle for students in challenging ideas or openly exploring uncertainty, which could limit deeper diagnostic reasoning during the session.¹⁵

In this study, four clinical cases of anaemia, chronic liver disease, chronic dyspnoea and chronic diarrhoea were used to evaluate the impact of both instructional strategies. Both the GGR and SNAPPS groups showed statistically significant improvement in post-test scores across all topics, indicating that structured teaching interventions are broadly effective in enhancing CR and medical knowledge. However, the GGR group demonstrated better performance in the cases of chronic liver disease, chronic dyspnoea and chronic diarrhoea. This result of our study highlighted the fact that GGR could be more effective in enhancing diagnostic reasoning for complex clinical problems in medical students. The absence of intergroup difference of scores for anaemia in our study suggested that SNAPPS was effective when students were dealing with a comparatively easy clinical condition. These findings were consistent with existing literature, which reported that SNAPPS produce better outcomes among postgraduate learners, who typically have more developed illness scripts and greater diagnostic experience.¹² In contrast, undergraduate students could be benefitted more from the structured, reflective process of GGR, particularly when encountered with diagnostically challenging cases in the clinical settings. This was supported by previous research suggesting that deliberate reflection is especially beneficial in promoting analytical reasoning in complex clinical scenarios.¹⁶

This study showed that both GGR and SNAPPS enhanced CR skills among final-year MBBS students during their internal medicine clerkship at Rawalpindi Medical University. However, GGR demonstrated better results, likely due to its structured format, peer interaction and reflective prompts that encouraged deeper analysis and hypothesis revision. The findings of this study aligned with existing literature, which also supported the benefit of deliberate reflection in improving diagnostic reasoning.⁷ Previous studies, like by Schmidt et al, emphasised the role of structured reflection in facilitating deeper learning by encouraging learners to question their initial impressions.² On the contrary, the SNAPPS technique, although helpful in promoting learner autonomy and clinical case presentation skills, provided less cognitive scaffolding in analysing clinical scenarios to reach the diagnosis.¹³ This interpretation was consistent with our observations that SNAPPS was effective in improving CR and its domains for undergraduate medical students but slightly less than GGR in clinical cases demanding thorough cognitive analysis.

The results of our study also held relevant within the clinical learning environment, where time constraints, high patient loads and limited patient interaction opportunities often restrict students’ ability to develop CR skills through traditional workplace-based learning.¹⁷ The group-based format of GGR, in particular, promoted collaborative learning, encouraged hypothesis revision and activated metacognitive monitoring, all of which were essential in the process of development of CR in an otherwise constrained clinical setting.

Recommendations for future research

Future studies should explore the long-term effects of these interventions on CR skill retention and real-world clinical application by the medical students beyond immediate post-test evaluations. Moreover, multicentric trials involving diverse institutional settings should be arranged to gather stronger evidence of generalisability.

Conclusion

The findings of this study conclude that both GGR and SNAPPS are effective in enhancing CR among final-year MBBS students during their internal medicine clerkship. However, GGR demonstrated superior outcomes, particularly in the domains of data acquisition, problem presentation and hypothesis generation and in critically analysing complex clinical cases. The better performance of GGR in this study can be explained by its incorporation of a deliberate reflection framework and collaborative nature. While SNAPPS offers value in promoting learner autonomy and focused case analysis, GGR appears more suitable for undergraduate medical students who are at an early CR stage. Therefore, this study recommends integrating structured reflection sessions like GGR into undergraduate clinical teaching, especially in settings where bedside exposure is limited, to support the development of diagnostic reasoning skills critical to safe and effective medical practice.

Data availability statement

Data are available upon reasonable request. All data relevant to the study are included in the article or uploaded as supplementary information.