Dimension-level conditional associations among coach–athlete relationship quality, psychological strain, and burnout in Chinese competitive athletes

Introduction Total scores do not indicate whether associations among coach–athlete relationship quality, psychological strain, and athlete burnout are diffuse or concentrated in specific dimensions. This study aimed to identify dimension-level conditional associations, evaluate bridge stability, and examine how robust these associations are to observed sample heterogeneity. Methods In this cross-sectional study, 564 Chinese competitive athletes completed the Coach–Athlete Relationship Questionnaire (CART-Q), Athlete Psychological Strain Questionnaire (APSQ), and Athlete Burnout Questionnaire (ABQ). Total-score correlations and a seven-node Spearman extended Bayesian information criterion graphical least absolute shrinkage and selection operator (EBICglasso) network were analysed following assessment of measurement quality. The complete nine-node network was examined as a supplementary analysis. Estimation uncertainty and stability were assessed using 1,000 non-parametric bootstrap samples and case-dropping analysis. Covariate-adjusted networks, exploratory subgroup permutation tests, and sensitivity analyses were also conducted. Results Better relationship quality was associated with lower psychological strain ( r = −0.255) and burnout ( r = −0.353), whereas psychological strain was positively associated with burnout ( r = 0.636; all Benjamini–Hochberg (BH)-adjusted p < 0.001). Performance concerns–emotional/physical exhaustion was the only cross-community edge whose bootstrap interval excluded zero [weight = 0.275, 95% interval (0.190, 0.354)]. Correlation-stability coefficients were 0.70 for bridge strength and 0.60 for bridge expected influence. The core association was retained in both the adjusted and supplementary nine-node networks. No subgroup comparison reached statistical significance after Holm adjustment. Conclusion Performance concerns–emotional/physical exhaustion was the clearest cross-domain conditional association, but the cross-sectional network did not establish a mechanism or intervention target. Longitudinal and multilevel replication is needed.