Comparative Efficacy of Interventions for Burnout in Physician Residents: A Systematic Review and Network Meta-Analysis of Randomized Controlled Trials.

BACKGROUND: Burnout during residency is common and is associated with adverse outcomes for clinicians and patients; however, the comparative efficacy of intervention approaches remains unclear. We aimed to compare the efficacy of different categories of interventions aimed at reducing burnout among physician residents. METHODS: We conducted a systematic review of randomized controlled trials (RCTs) in physician residents (International Prospective Register of Systematic Reviews [PROSPERO]: CRD42025610352). Searches were conducted from February 19 to 26, 2025 in PubMed, Cochrane CENTRAL, PsycINFO, Web of Science, and Ichushi-Web.Interventions were classified into six categories: coaching, reflective intervention, intensive care unit short shift schedule, mind-body intervention, skills-based stress management, and information provision. Outcomes were emotional exhaustion (EE), depersonalization (DP), reduced efficacy (RE), and global burnout (GB). We performed a Bayesian random-effects network meta-analysis and reported standardized mean differences (SMDs) with 95% credible intervals (CrI); negative SMD values indicate improvement. Interventions were ranked using the surface under the cumulative ranking curve (SUCRA). Uncertainty was assessed using posterior effect estimates with 95% CrI and ranking probabilities. Certainty was assessed using Grading of Recommendations Assessment, Development and Evaluation (GRADE) summary of findings tables. RESULTS: A total of 20 RCTs (n = 2,616) were included; risk of bias was mostly rated as some concerns (5% low, 60% some concerns, and 35% high risk). Effects were generally small, and estimates were imprecise. By SUCRA, reflective interventions ranked highest for EE and DP (EE: SMD -0.515 [95% CrI, -1.068 to 0.041]; DP: SMD -0.432 [-0.988 to 0.133]), whereas coaching ranked highest for RE and GB (RE: SMD -0.142 [-0.538 to 0.262]; GB: SMD -0.451 [-1.180 to 0.195]). Certainty was predominantly low to very low. CONCLUSIONS: The comparative superiority of any single intervention category remains unclear. Reflective interventions, which support guided reflection on clinical work, and coaching showed comparatively favorable signals for some burnout domains, but estimates were imprecise and of low certainty. Findings may inform cautious prioritization and highlight the need for pragmatic, adequately powered trials.