Psychological and psychosocial correlates of teaching-related burnout dimensions among residency programme faculty: a multicentre cross-sectional study

Abstract Background Burnout is multidimensional, but whether its psychological and psychosocial correlates differ across dimensions requires direct testing. We examined perceived stress, psychological symptoms, sleep quality, resilience, and general social support in relation to emotional exhaustion, cynicism, and professional efficacy among residency programme faculty. Methods This multicentre cross-sectional survey included 478 residency programme faculty from 15 hospitals, of whom 99.0% were from Guangdong Province. Blockwise regression models were fitted to 471 complete cases, with hospital fixed effects and inference adjusted for clustering by hospital. Standardised coefficients were compared across outcomes using 10,000 hospital bootstrap replicates. The moderation analyses were exploratory, and Benjamini–Hochberg correction was applied within testing families. Wild cluster and site influence analyses assessed robustness to the limited number and unequal sizes of clusters. Results In bivariate analyses, anxiety and depressive symptoms and poorer sleep quality were correlated with greater emotional exhaustion and cynicism and lower professional efficacy. In adjusted models, perceived stress was associated with greater emotional exhaustion (β = 0.457) and cynicism (β = 0.369) and with lower professional efficacy (β = −0.199). Resilience was associated mainly with higher professional efficacy (β = 0.369). General social support had small associations with lower cynicism (β = −0.101) and higher professional efficacy (β = 0.096), although the cynicism estimate was less consistent across methods. After the coefficient directions were aligned, perceived stress was more strongly associated with emotional exhaustion and cynicism than with professional efficacy, with no clear difference between the first two dimensions. Resilience was more strongly associated with professional efficacy than with the other dimensions. The exploratory interaction between perceived stress and general social support for cynicism was small (β = −0.093, BH-adjusted q = 0.008; ΔR² = 0.008), and its inferential support varied when hospitals were omitted individually. Conclusions Perceived stress was associated with all three burnout dimensions but less strongly with professional efficacy, whereas resilience was associated primarily with professional efficacy. General social support showed small direct associations, and the exploratory interaction should not be interpreted as an established buffering mechanism. Longitudinal studies using teaching-specific measures are needed. Generalisability beyond the participating hospitals, particularly outside Guangdong Province, is limited.