Effectiveness of the flipped classroom in medical education: a systematic review and meta-analysis Authors: Yaqiu Wu, Meixiong Cheng Published: 2026 ABSTRACT AbstractPurposeThis systematic review and meta-analysis aimed to evaluate the effectiveness of the FC approach compared to conventional teaching methods in medical education.MethodsA comprehensive search of PubMed, Embase, and Web of Science databases was conducted. Eighteen eligible studies, involving diverse learner populations, were included in the analysis. The effectiveness of the FC model was assessed across five predefined domains: theoretical knowledge, procedural/technical skills, self-directed learning ability, case analysis/problem-solving ability, and learner satisfaction.ResultsPooled analyses demonstrated significant advantages of the FC model, with moderate to large effect sizes observed across most outcomes. Notable improvements were found in self-directed learning ability and procedural/technical skills. Subgroup analyses suggested that residents showed larger benefits, particularly in procedural/technical skills and learner satisfaction. Sensitivity analyses generally supported the direction of the findings, although statistical significance was not retained in all leave-one-out analyses for some outcomes.ConclusionsFlipped classrooms were associated with improved outcomes in medical education, with larger effects in residents. Given high heterogeneity and limited evidence in some domains, benefits appear context dependent. Implementation should be tailored, and future studies should define mechanisms and optimal conditions.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12909-026-09526-8.Keywords: Flipped Classroom, Medical Education, Systematic Review, Meta-Analysis, Educational Outcomes, Residents Introduction With rapid advances in biomedical science and evolving models of clinical diagnosis and care, expectations for the comprehensive competencies of medical professionals continue to rise. As a foundation of health-system development, medical education is under increasing pressure to adapt. For decades, lecture-based instruction has dominated medical classrooms [1]. While efficient for the systematic delivery of core knowledge, this teacher-centred approach often provides limited interaction and practice-oriented engagement [2], which may reduce learners’ motivation, hinder knowledge consolidation, and constrain translation of knowledge into clinical competence [3]. In parallel with the broader uptake of student-centred education [4], medical training has increasingly shifted from “knowledge transmission” towards competency-based learning [5, 6]. Identifying teaching approaches that are more interactive and competency-oriented is therefore of practical importance. The flipped classroom has gained traction in higher education and has been increasingly adopted in medical education. In this model, learners access preparatory materials (for example, videos or readings) before class to acquire foundational concepts, whereas class time is used for discussion, practice and application to deepen understanding and support skills development [7]. Compared with traditional formats, flipped teaching places greater emphasis on active participation, targeted facilitation by instructors and structured peer interaction. Prior studies suggest that flipped classrooms may improve clinical reasoning and procedural/technical skills in medical learners [8–10], with particular feasibility in practice-oriented courses such as procedural training and clinical skills instruction [11, 12]. The model also aligns with broader trends in digital and technology-enabled learning [13]. Despite growing interest, evidence regarding the effectiveness of flipped classrooms in medical education remains mixed [14]. Much of the literature consists of single-centre, small-sample observational studies, with substantial heterogeneity in curricular content, instructional tools and outcome assessment [15]. Moreover, studies often focus on selected educational domains rather than providing comprehensive, multidimensional evaluation [16], complicating interpretation of overall impact. Differences across training stages have also been insufficiently examined. In particular, residents, an essential population in medical training [17], face distinctive constraints related to workload and learning autonomy [18], and whether they benefit differentially from flipped teaching requires stronger evidence. A systematic review and meta-analysis was conducted to evaluate the effects of flipped classrooms in medical education across multiple outcomes, including theoretical knowledge, procedural/technical skills, self-directed learning ability, case analysis/problem-solving ability, and learner satisfaction. Potential differences in effects across learner groups at different training stages were also examined. By synthesizing available evidence, this study aims to provide quantitative support for instructional decision-making and curriculum optimization, including in clinical teaching contexts such as standardized residency training. Materials and methods Literature search and screening This study was conducted and reported in accordance with the Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) guidelines [19]. As this study is a systematic review and meta-analysis based on previously published data, no ethical approval or informed consent was required. PubMed, Embase, and Web of Science were searched from database inception to March 2025 to identify studies on the application and effects of flipped classrooms in medical education. Controlled vocabulary (e.g., MeSH) and free-text terms were combined. For PubMed, the search strategy was: (“flipped classroom”[Title/Abstract] OR “inverted classroom”[Title/Abstract] OR “flipped learning”[Title/Abstract]) AND (“education, medical”[MeSH Terms] OR “medical education”[Title/Abstract]). The same search logic was applied to the other databases with database-specific syntax adaptations (full strategies are provided in Supplementary Material 1). Search terms included “flipped classroom”, “inverted classroom”, and “blended learning”. Reference lists of included studies were screened. All records were imported into EndNote X7 and duplicates were removed. Two reviewers independently screened titles and abstracts using predefined eligibility criteria. Full texts of potentially eligible studies were retrieved and assessed independently by the same reviewers for final inclusion. Disagreements were resolved by a third reviewer. Inclusion and exclusion criteria Inclusion criteria were defined a priori as follows: (1) Population: learners in medical education, including undergraduate medical students, postgraduate students, or resident physicians, with no restriction on specialty; (2) Intervention: the experimental group received instruction using a flipped classroom model; (3) Comparator: the control group received a traditional, lecture-based classroom model. Control conditions were permitted to include minimal baseline teaching activities required to maintain comparability (for example, in-class question-and-answer sessions or group discussion), but studies were excluded if the control condition systematically integrated other structured educational innovations beyond conventional lecturing (for example, problem-based learning (PBL) or case-based learning (CBL)); and (4) Outcomes: outcomes included, but were not limited to, theoretical knowledge scores, procedural/technical skills scores, learner satisfaction, self-directed learning ability, and case analysis/problem-solving ability. Exclusion criteria were as follows: (1) studies with incomplete data that prevented extraction of relevant variables; and (2) publication types limited to conference abstracts, reviews, or case reports. Two reviewers (Reviewer A and Reviewer B) independently screened studies for eligibility and compared results. Disagreements were addressed by re-evaluating the title, abstract and, when needed, the full text against the predefined criteria, followed by discussion. If disagreement persisted, a third reviewer (Reviewer C) independently assessed the study and made the final decision. Data extraction Two reviewers (Reviewer A and Reviewer B) independently extracted data using a prespecified data-collection form. Extracted items included first author, year of publication, participant group, course type, sample size, teaching methods in the intervention and control groups, outcome measures and corresponding assessment instruments, and study duration. After independent extraction, the two data forms were cross-checked. Any discrepancies were resolved by returning to the original report for verification and discussion. If disagreement persisted, a third reviewer (Reviewer C) adjudicated until full consensus was reached. Outcome domain definitions Outcomes were categorized into five predefined domains: theoretical knowledge, procedural/technical skills, self-directed learning ability, case analysis/problem-solving ability, and learner satisfaction. Procedural/technical skills were defined as performance-based educational outcomes assessing learners’ ability to perform clinical or technical procedures, including operational skill scores, procedural performance scores, and skills examination scores reported in the included studies. Patient-level clinical outcomes were not included in this domain. Statistical analysis Statistical analyses were performed in R (version 4.3.1) using the meta package. Between-study heterogeneity was assessed using the I² statistic and Cochran’s Q test. A fixed-effects model was applied when I2 < 50% and p > 0.05; otherwise, a random-effects model was used. All outcomes—including theoretical knowledge scores, procedural/technical skills scores, learner satisfaction, self-directed learning ability, and case analysis/problem-solving ability—were treated as continuous variables and pooled as standardized mean differences (SMDs) with 95% confidence intervals. For random-effects models, 95% prediction intervals were calculated to estimate the range within which the true effect of a future comparable study may be expected to fall. Statistical significance was defined as p < 0.05. Sensitivity analyses were conducted using a leave-one-out approach by repeating the meta-analysis after sequentially removing each included study to evaluate the influence of individual studies on the pooled estimates. Publication bias was explored using funnel plots, with findings interpreted cautiously when the number of studies was small. Results Literature search results A total of 1,951 records were retrieved using the prespecified search strategy (PubMed, n = 544; Embase, n = 791; Web of Science, n = 616). After removing 714 duplicates, 1,237 records underwent screening. Title/abstract screening first excluded 79 records with ineligible publication types (systematic reviews, n = 53; meta-analyses, n = 17; case reports, n = 9). A further 1,118 records were excluded at the subsequent title/abstract assessment because the topic was not relevant or eligibility criteria were not met, leaving 40 reports for full-text evaluation. At the full-text stage, 22 reports were excluded (full text unavailable, n = 17; insufficient data for extraction, n = 5), resulting in 18 studies included in the systematic review and meta-analysis (Fig. 1A). Because individual studies could report multiple outcomes, the total number of studies contributing to outcome-specific meta-analyses exceeded 18. Flowchart of Literature Screening Process. A Study selection flow diagram. B Summary of risk-of-bias assessment for included studies All included studies evaluated the application and educational effects of flipped classrooms in medical education, covering both didactic and practice-oriented courses. Thirteen studies directly compared a flipped classroom with traditional lecture-based teaching, whereas five evaluated a flipped-classroom–centred composite intervention that incorporated additional components within the flipped framework, including case-based teaching, group discussion, virtual video-based instruction, or the Peyton four-step approach. Control conditions used lecture-based teaching as the primary format and could include minimal baseline in-class activities required for comparability (for example, Q&A or group discussion), but studies were not included if the control arm systematically integrated other structured innovations (for example, PBL or CBL) as major components. Outcomes reported across studies included theoretical knowledge attainment (15 studies; Outcome 1), learner satisfaction (8 studies; Outcome 2), case analysis/problem-solving ability (6 studies; Outcome 3), self-directed learning ability (4 studies; Outcome 4), and procedural/technical skills (4 studies; Outcome 5). Several studies discussed potential effects on clinical decision-making, critical thinking, and knowledge retention; however, these were not reported in extractable, poolable quantitative formats and were not synthesized. Study characteristics are summarized in Table 1. Risk of bias was assessed using the Cochrane tool across domains including sequence generation, allocation concealment, blinding, completeness of outcome data, and selective reporting, with judgements categorized as low, high, or unclear risk and summarized graphically at the study and domain levels (Fig. 1B). Table 1Characteristics of the studies included in this systematic review and meta-analysisFirst AuthorYearType of ParticipantsCourse TypeAge (T/C)GradeGender(M/F)Sample Size (T/C)Experimental Group Teaching MethodControl Group Teaching MethodOutcome MeasuresHu2024Medical StudentsTheory21.15/20.85Fourth and Fifth Year55.4%/44.6%46/46Case Discussion + Group LearningTraditional Classroom Teaching①②③Khoshnoodi2019ResidentsTheory//51%/49%60/60Video Teaching + Group LearningTraditional Classroom Teaching①Lin2025Medical StudentsTheory20.1 ± 3.5/20.6 ± 1.1//380/380Multidisciplinary DiscussionTraditional Classroom Teaching + Group Discussion①③④Beom2018Medical StudentsPractice26.4/26.4Fourth Year69.8%/30.2%55/53Video Teaching + Group LearningTraditional Classroom Teaching②⑤Lu2023Medical StudentsTheory/Second Year/62/69Video Teaching + Group LearningTraditional Classroom Teaching①Rui2017Medical StudentsTheory20.84/20.90Third Year/90/91Video Teaching + Group LearningTraditional Classroom Teaching + Q&A①②Wang H2024ResidentsTheory24.9/25.2Fourth Year/52/51Video Teaching + Group LearningTraditional Classroom Teaching①②④⑤Wang J2024ResidentsPractice26.18/26.61First Year of Residency/33/33Video Teaching + Peyton Four-step Method + Case DiscussionTraditional Classroom Teaching②③⑤Wu Y2022Medical StudentsTheory///73/85Video Teaching + Group LearningTraditional Classroom Teaching①②③④Ruprai2024Medical StudentsTheory/First Year/100/100Pre-class Study + Group DiscussionTraditional Classroom Teaching①Seidi2024Medical StudentsTheory20.76/21.36//30/30Video Teaching + Group LearningTraditional Classroom Teaching + Practice①③Skowronski2020ResidentsTheory///13/21Case Discussion + Group LearningTraditional Classroom Teaching①Sourg2023Medical StudentsTheory/Second Year40%/60%30/33Video Teaching + Group LearningTraditional Classroom Teaching①Tsao2022Medical StudentsTheory23.2/23.8Fifth Year62.2%/37.8%45/45Video Teaching + Case DiscussionTraditional Classroom Teaching①②③Tusa2018Medical StudentsTheory/Fourth YearT:46.2%/53.8%C: 50%/50%40/39Video Teaching + Group LearningTraditional Classroom Teaching①Wu T2024Medical StudentsPractice///40/40Video Teaching + Group LearningTraditional Classroom Teaching②Zhang2024ResidentsPractice26.4/26.4First Year of Residency51%/49%45/45Video Teaching + Case DiscussionTraditional Classroom Teaching①④⑤Yang2020Medical StudentsTheory19.7First Year11%/89%45/44Video Teaching + Group LearningTraditional Classroom Teaching①①Theoretical Knowledge ②learner satisfaction ③Case Analysis Problem-Solving Ability ④Self-directed Learning Ability ⑤Skill Operation Score Characteristics of the studies included in this systematic review and meta-analysis 20.1 ± 3.5/ 20.6 ± 1.1 T:46.2%/53.8% C: 50%/50% ①Theoretical Knowledge ②learner satisfaction ③Case Analysis Problem-Solving Ability ④Self-directed Learning Ability ⑤Skill Operation Score Meta-analysis results Overall effect of the flipped classroom teaching model on theoretical knowledge scores A total of 15 studies were included, involving 1,111 participants in the intervention group and 1,139 in the control group. The results of the meta-analysis indicated that the flipped classroom teaching model had a significant advantage in improving theoretical knowledge among medical learners (SMD = 1.26, 95% CI: 0.80–1.71), with the difference being statistically significant. Substantial heterogeneity was observed across studies (I² = 93.1%, p < 0.0001) (Fig. 2A), suggesting a high degree of variability among the included studies. Meta-Analysis of the Effect of the Flipped Classroom on Theoretical Knowledge Scores. A Forest plot of the overall effect on theoretical knowledge scores; B Forest plot of subgroup analysis by integrated teaching method; C Forest plot of subgroup analysis by learner type; D Forest plot of subgroup analysis by course type. SMD = standardized mean difference; 95% CI = 95% confidence interval To explore the sources of heterogeneity, subgroup analyses were conducted based on instructional strategies, learner types, and course types (Fig. 2B-D). The results showed no statistically significant differences between subgroups (instructional strategy: p = 0.2533; learner type: p = 0.1394; course type: p = 0.9271), and the heterogeneity within each subgroup remained high. These findings suggest that the knowledge-related benefits of the flipped classroom were consistent across different teaching approaches, learner groups, and course types. Leave-one-out sensitivity analyses showed that the pooled effect remained statistically significant after exclusion of any single study (SMD range, 1.12–1.33), indicating that the estimate was not driven by an individual study (Figure S1A). Funnel-plot inspection suggested approximate symmetry (Figure S2A), although the ability to detect publication bias is limited. Overall, the effect of flipped classrooms on knowledge attainment should be interpreted cautiously. Despite a pooled benefit, heterogeneity was very high (I2 = 93.1%), and prespecified subgroup analyses did not explain this variability; the prediction interval indicated wide dispersion of effects. Thus, flipped teaching was associated with improved knowledge outcomes, but the magnitude of benefit appears context dependent, potentially influenced by learner characteristics, course design and assessment methods not captured in the included studies. Overall effect of the flipped classroom model on procedural/technical skills Four studies were included, comprising 185 participants in the intervention group and 182 in the control group. The meta-analysis results indicated that the flipped classroom teaching model had a significant advantage in enhancing procedural/technical skills among learners in medical education, with a pooled SMD of 1.80 (95% CI: 0.19–3.41), and the difference was statistically significant. However, substantial heterogeneity was detected among the studies (I² = 97.0%, p < 0.0001), indicating considerable variability in the results (Fig. 3A). Meta-Analysis of the Effect of the Flipped Classroom on Procedural/Technical Skills. A Forest plot of the overall effect on procedural/technical skills scores; B Forest plot of subgroup analysis by integrated teaching method; C Forest plot of subgroup analysis by course type; D Forest plot of subgroup analysis by learner type. SMD = standardized mean difference; 95% CI = 95% confidence interval To further investigate potential sources of heterogeneity, subgroup analyses were conducted based on instructional methods, learner types, and course types. No significant difference was found between subgroups combining case-based teaching and simulation-based teaching (p = 0.7562), although the heterogeneity for the simulation-based subgroup was I² = 0%, suggesting that the instructional method may be a potential source of heterogeneity (Fig. 3B). In the course-type subgroup analysis, the single study involving a theoretical course reported a very large effect (SMD = 4.11, 95% CI 3.42–4.80), whereas the pooled effect across three practice-oriented course studies was SMD = 1.03 (95% CI 0.14–1.93). The difference between subgroups was statistically significant (p < 0.0001), suggesting that the flipped classroom yielded different effects across course types, with a notably more substantial impact on procedural/technical skills in theoretical courses (Fig. 3C). Regarding learner types, a statistically significant difference was observed between residents and undergraduates (p = 0.0131), indicating that the flipped classroom model produced significantly different effects across learner groups, with a more pronounced improvement in procedural/technical skills among residents (Fig. 3D). Leave-one-out sensitivity analyses showed a consistently positive pooled effect after removing any single study (SMD range, 1.03–2.35), with heterogeneity essentially unchanged; however, inference remained limited by few studies and very high heterogeneity (Figure S1B). With only four studies (k = 4), funnel plots were underpowered for assessing publication bias and are shown for completeness only (Figure S2B). The pooled estimate for procedural/technical skills therefore requires cautious interpretation because heterogeneity was very high (I2 = 97.0%) and the prediction interval was wide, indicating that the true effect may vary substantially across educational contexts. Although subgroup analyses suggested larger effects among residents and in theoretical-course settings, several subgroups were based on single studies; therefore, these findings should be regarded as exploratory and may not be universally applicable. Overall effect of the flipped classroom teaching model on self-directed learning ability Four studies were included (intervention, n = 550; control, n = 561). Pooled analysis showed a statistically significant improvement in self-directed learning ability with the flipped classroom compared with traditional lecture-based teaching (SMD = 2.54, 95% CI 1.04–4.04), with very high heterogeneity (I² = 98.8%, p < 0.0001) (Fig. 4A). Meta-Analysis of the Effect of the Flipped Classroom on Self-Directed Learning Ability. A Forest plot of the overall effect on self-directed learning ability; B Forest plot of subgroup analysis by integrated teaching method; C Forest plot of subgroup analysis by learner type; D Forest plot of subgroup analysis by course type. SMD = standardized mean difference; 95% CI = 95% confidence interval To explore heterogeneity, subgroup analyses were conducted by co-intervention/teaching format, learner group, and course type. No evidence of between-subgroup differences was observed by teaching format (p = 0.8683), learner group (p = 0.7025), or course type (p = 0.8683), while heterogeneity remained high within subgroups (Fig. 4B–D). Leave-one-out sensitivity analyses yielded statistically significant pooled effects across all iterations (SMD range, 2.08–3.25); however, inference was constrained by the small number of studies and extreme heterogeneity (Figure S1C). With only four studies, funnel-plot assessment was underpowered and was therefore not used to draw inferences about publication bias (Figure S2C). Given the extreme heterogeneity (I² = 98.8%), the large pooled effect for self-directed learning ability (SMD = 2.54) should not be interpreted as a uniform or universally applicable effect. The wide prediction interval indicates that effects may vary substantially across educational contexts, and prespecified subgroup analyses did not account for this variability, suggesting potential influence of unmeasured factors such as implementation fidelity, learner motivation, and differences in assessment instruments. Overall effect of the flipped classroom model on case analysis/problem-solving ability Six studies were included (intervention, n = 607; control, n = 619). Pooled analysis showed a statistically significant improvement in case analysis/problem-solving ability with the flipped classroom compared with traditional lecture-based teaching (SMD = 1.50, 95% CI 0.14–2.85), with substantial heterogeneity (I² = 97.3%, p < 0.0001) (Fig. 5A). Meta-Analysis of the Effect of the Flipped Classroom on Case Analysis and Problem-Solving Skills. A Forest plot of the overall effect on case analysis and problem-solving skills; B Forest plot of subgroup analysis by integrated teaching method; C Forest plot of subgroup analysis by course type; D Forest plot of subgroup analysis by learner type. SMD = standardized mean difference; 95% CI = 95% confidence interval To explore heterogeneity, subgroup analyses were conducted by teaching format (co-intervention), learner group, and course type. No evidence of between-subgroup differences was observed by teaching format (p = 0.3077), course type (p = 0.6437), or learner group (p = 0.6437), and heterogeneity remained within subgroups (Fig. 5B–D). Leave-one-out sensitivity analyses produced pooled point estimates ranging from SMD = 0.81 to 1.72. The point estimates remained positive, but statistical significance was not retained in all iterations, particularly after omitting Wang J or Tsao (Figure S1D). This indicates that the direction of effect was generally stable, whereas the statistical significance of this outcome was less robust. Funnel-plot inspection showed some asymmetry (Figure S2D); however, with only six studies (k = 6), funnel plots have limited power for assessing publication bias. Overall, the pooled estimate indicated an association between flipped teaching and improved case analysis/problem-solving outcomes. Interpretation was limited by heterogeneity (I² = 97.3%) and uncertainty around the summary effect. Prespecified subgroup analyses did not identify effect modifiers, and variability may reflect methodological differences across studies (for example, case complexity, instructional facilitation strategies, or scoring criteria). Given the high heterogeneity and wide prediction interval, both the consistency and magnitude of this association require confirmation in studies with more comparable instructional designs and outcome assessments. These findings should therefore be interpreted as context-dependent rather than universally applicable. Overall effect of the flipped classroom model on learner satisfaction Eight studies were included (intervention, n = 434; control, n = 444). Pooled analysis showed a statistically significant improvement in learner satisfaction with the flipped classroom compared with traditional lecture-based teaching (SMD = 0.69, 95% CI 0.10–1.28), with heterogeneity across studies (I² = 93.6%, p < 0.0001) (Fig. 6A). Meta-Analysis of the Effect of the Flipped Classroom on Student Satisfaction. A Forest plot of the overall effect on learner satisfaction; B Forest plot of subgroup analysis by learner type; C Forest plot of subgroup analysis by integrated teaching method; D Forest plot of subgroup analysis by course type. SMD = standardized mean difference; 95% CI = 95% confidence interval To explore heterogeneity, subgroup analyses were conducted by learner group, teaching format (co-intervention), and course type. By learner group, the pooled effect for undergraduates was SMD = 0.37 (95% CI − 0.15 to 0.90), whereas the pooled effect for residents was SMD = 1.69 (95% CI 0.82 to 2.56); the between-subgroup difference was statistically significant (p = 0.0107) (Fig. 6B). By teaching format, no between-subgroup difference was observed between studies incorporating case-based teaching and those incorporating simulation-based teaching (p = 0.6147) (Fig. 6C). By course type, no between-subgroup difference was observed (p = 0.2837) (Fig. 6D). Heterogeneity persisted within subgroups. Leave-one-out sensitivity analyses showed pooled effects ranging from SMD = 0.49 to 0.87. The direction of effect remained positive across iterations, but statistical significance was not retained in all leave-one-out analyses, indicating that the satisfaction outcome was directionally stable but statistically sensitive to the exclusion of several individual studies (Figure S1E). Funnel-plot inspection suggested approximate symmetry, but with only eight studies, funnel plots were underpowered for assessing publication bias and were not used for inference (Figure S2E). Overall, satisfaction outcomes differed by learner group, and interpretation of the pooled estimate was limited by heterogeneity and uncertainty around the summary effect. The resident subgroup showed a larger improvement, and the prediction interval indicated that effects may vary widely across settings. Variability may reflect differences in how satisfaction was defined and measured across studies; accordingly, the association between flipped teaching and satisfaction may be context dependent. Discussion This systematic review and meta-analysis evaluated the effects of flipped classrooms across multiple outcomes in medical education, including theoretical knowledge, procedural/technical skills, self-directed learning ability, case analysis/problem-solving ability, and learner satisfaction. Overall pooled estimates indicated statistically significant advantages of flipped teaching over predominantly lecture-based instruction across these outcomes; however, heterogeneity was present for most endpoints, and the evidence base for several outcomes/subgroups was limited, suggesting that effect sizes may vary with instructional context and measurement approaches. Subgroup analyses further showed larger effect estimates for procedural/technical skills and learner satisfaction among residents, indicating that training stage and implementation setting may be associated with differences in observed effects. Prior evidence also links learning achievement with learner engagement and active learning [20, 21], consistent with the overall direction observed in the present synthesis. From a mechanistic perspective, observed effects may relate to the instructional structure of the flipped classroom, which shifts initial content acquisition to pre-class activities and allocates class time to application-oriented tasks: learners review core concepts before class via videos or readings, whereas in-class time is used more often for problem solving, case discussion, and skills practice, thereby increasing opportunities for engagement and knowledge use [22]. Evidence that active learning is associated with improved student performance provides external consistency for this pathway [23]. In this setting, instructors function more as facilitators, providing real-time feedback and learning support (scaffolding) based on learners’ preparation and in-class performance, which may aid conceptual clarification and competence development [24]. For knowledge- and reasoning-related outcomes, the 15 included studies suggested that flipped classrooms were associated with higher knowledge assessment scores (SMD = 1.26), consistent with the direction reported in prior reviews [25, 26]. Flipped teaching was also associated with improvements in case analysis/problem-solving ability (SMD = 1.50), and in-class discussion and contextualized practice may support the translation of conceptual knowledge into clinical reasoning processes. Notably, some studies reported non-significant effects, indicating that knowledge and reasoning outcomes may be sensitive to course difficulty, adherence to pre-class preparation, instructor implementation experience, and variability in outcome measurement. In addition, publication-bias assessment for case analysis/problem-solving ability was limited by the small number of studies; therefore, funnel-plot findings should be interpreted cautiously and cannot reliably exclude publication bias. Further confirmation using more harmonized evaluation frameworks is warranted. Procedural/technical skills training is a core component of medical education [27]. The present analysis indicated an association between flipped teaching and improved procedural/technical skills (SMD = 1.80), with larger estimates among residents; this pattern may relate to residents’ higher practice demands and the organization of in-class demonstration, practice, and feedback within flipped formats. Prior work has also reported that flipped approaches may support understanding of standardized procedures and improve error identification and correction [28]. For self-directed learning ability, the pooled effect was relatively large (SMD = 2.54), which is consistent with the emphasis of flipped instruction on pre-class preparation and self-regulation. Given that self-directed learning is linked to later professional development and lifelong learning [29] and that medical education emphasizes early development of self-regulated learning [30], future studies should assess the durability of these gains across training stages and identify key implementation conditions (for example, task design, process support, and assessment strategies). Learner satisfaction was also higher with flipped teaching (SMD = 0.69). Although the effect estimate was smaller, it remained statistically significant. In relation to learner satisfaction and interaction, Zhang et al. reported that team-based online medical consultation may facilitate users’ adoption of medical advice by enhancing collaborative input and trust in professional recommendations [31]. Although this evidence comes from patient-facing online consultation rather than classroom-based education, it provides indirect support for the potential value of structured, team-based online interaction in flipped-classroom settings. Some studies reported increased classroom participation and interaction, whereas others noted an early adaptation burden, suggesting that implementation may benefit from clear course structure and appropriate learning and technical support to reduce transition costs and improve adherence. In addition, integration of educational technologies with flipped teaching may influence learning processes across phases. Artificial intelligence/adaptive learning systems can support individualized pre-class preparation and diagnostic testing [32], VR/AR can provide repeatable simulation environments and has been linked to improved skills performance in evidence syntheses [33], and learning analytics can inform formative feedback decisions using learning-process data [34]. Evidence from pharmacology education also suggests that new media teaching strategies may enhance learning outcomes across students with diverse educational backgrounds, supporting the value of media-rich instructional design in medical education [35]. Therefore, technology-enabled flipped approaches should be further evaluated under varying implementation intensities, learner backgrounds, and outcome measurement tools [36]. A major finding of this review was the consistently high heterogeneity across outcomes (I² > 90%). Although prespecified subgroup analyses by teaching format, learner group, and course type were conducted, most did not explain the observed variability, suggesting that the effects of flipped classrooms may not be stable across contexts and may depend on specific instructional conditions. Therefore, the large pooled effect sizes, particularly for self-directed learning ability and procedural/technical skills, should be interpreted cautiously and should not be taken to indicate uniform benefits across all flipped-classroom implementations. The prediction intervals from the random-effects models were wide, further indicating that the magnitude and practical relevance of the effects may vary substantially across educational settings. Factors such as the granularity of instructional design, instructor facilitation skills, implementation fidelity, the validity and reliability of assessment tools, and learners’ prior study habits may act as effect modifiers. Several limitations should be noted: (1) methodological reporting in some included studies was incomplete (for example, randomization and allocation concealment), introducing potential bias; (2) key outcomes such as procedural/technical skills and self-directed learning ability were informed by relatively few studies (k ≤ 4), limiting statistical power and stability; (3) blinding is difficult to implement in educational interventions; and (4) for outcomes with few studies, publication-bias assessment via funnel plots is underpowered and cannot reliably exclude bias, despite sensitivity analyses indicating stable directions of effect. Despite these limitations, this synthesis provides a multi-outcome evidence summary for flipped classrooms in medical education. Future research should prioritize larger, methodologically rigorous prospective studies (including improved reporting of randomization and blinding-related procedures where applicable), particularly in under-studied domains such as procedural/technical skills and self-directed learning ability. Standardized and validated outcome instruments are needed to improve comparability across studies, and longer-term endpoints (for example, knowledge retention, clinical decision-making, and professional development) should be incorporated. Theory-informed heterogeneity analyses, such as meta-regression, may further clarify for whom and under what conditions flipped teaching is most effective. Future studies should also examine whether AI-driven teaching assistants or educational LLM agents can support flipped-classroom implementation by helping instructors develop personalized preparatory materials and feedback resources, as emerging evidence highlights their potential role in teachers’ content creation [37]. Overall, current evidence supports the potential utility of flipped classrooms across multiple domains in medical education, while indicating that effect sizes are context dependent; implementation should therefore be tailored to learner stage (for example, residency training), course objectives, such as procedural/technical skills training and self-directed learning ability, and available resources rather than assuming a uniform effect across settings. To better illustrate these findings, Fig. 7 summarizes the conceptual shift from traditional to flipped classrooms and highlights the corresponding improvements across the five evaluated educational outcomes. Mechanism of How Traditional vs. Flipped Classroom Teaching Models Influence Educational Outcomes in Medical Education. This diagram illustrates the key differences in instructional flow between traditional and flipped classroom teaching models and their pathways of impact on learning outcomes. Traditional classrooms emphasize teacher-led instruction and post-class learning, whereas the flipped classroom emphasizes self-directed learning before class and interactive engagement during class. The meta-analysis showed that the flipped classroom was associated with improvements in theoretical knowledge, procedural/technical skills, self-directed learning ability, case analysis/problem-solving ability, and learner satisfaction. Created in BioRender Conclusion This study systematically synthesized 18 empirical studies on the application of the flipped classroom teaching model in medical education. The multidimensional meta-analysis showed that, compared with traditional lecture-based teaching, flipped classroom teaching was associated with improvements in theoretical knowledge, procedural/technical skills, self-directed learning ability, case analysis/problem-solving ability, and learner satisfaction among medical learners. Larger effects were observed among residents, especially for procedural/technical skills and learner satisfaction. Although substantial heterogeneity was present and publication-bias assessment was limited by the small number of studies for several outcomes, sensitivity analyses generally supported the direction of the findings. These findings support the potential utility of the flipped classroom model in medical education, while indicating that implementation should be tailored to learner stage, course objectives, and available resources. Supplementary Information Supplementary Material 1. Supplementary Material 2. Supplementary Material 3. Acknowledgements Not applicable. Abbreviations Confidence Intervals Flipped Classroom Medical Subject Headings Odds Ratios Risk Ratios Standardized Mean Difference Authors’ contributions M.C. and Y.W. conceived and designed the study. Y.W. performed the literature search, study selection, and data extraction. M.C. and Y.W. analyzed the data. M.C. and Y.W. wrote the manuscript. All authors reviewed and approved the final version of the manuscript. Funding This work was supported by the Promotion and Application Project Fund of Sichuan Provincial Health Care Committee for Cadres (Grant No. Chuan Jian Yan 2025-202). Data availability All data analyzed in this study were extracted from previously published studies and are included in this article and its supplementary materials. Declarations Ethics approval and consent to participate This study was conducted and reported in accordance with the Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) guidelines. As this study is a systematic review and meta-analysis based on previously published data, no ethical approval or informed consent was required. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1. Supplementary Material 2. Supplementary Material 3. Data Availability Statement All data analyzed in this study were extracted from previously published studies and are included in this article and its supplementary materials.