Workload and Diagnostic Safety in Radiology Residents: A Recovery-Strain Mediation Analysis

Diagnostic safety in radiology may be influenced by occupational conditions beyond the point of image interpretation. This study examined whether workload was associated with diagnostic safety-related events directly or through sequential effects on recovery impairment and psychological strain. MATERIALS AND METHODS: In a nationwide cross-sectional survey of 3729 radiology residents from 407 training institutions in mainland China, we evaluated structural pathways linking workload, recovery impairment, psychological strain, and recent self-reported diagnostic safety-related events. Workload, recovery impairment, and psychological strain were operationalized as standardized composite scores. Generalized structural equation modeling with a weighted least squares mean- and variance-adjusted estimator was used to estimate direct and indirect associations.

Workload was positively associated with recovery impairment (β = 0.45, P < .001), and recovery impairment was strongly associated with psychological strain (β = 1.63, P < .001). Psychological strain was associated with 1-month diagnostic safety-related events (β = 0.39, P = .03), whereas direct paths from workload and recovery impairment to events were not statistically significant. The serial indirect pathway from workload through recovery impairment and psychological strain to safety-related events was statistically significant (P = .03). In the 3-month sensitivity analysis, estimates followed the same direction as the primary model but were smaller in magnitude and did not reach statistical significance. Findings were consistent across individual workload indicators and training stages.

Among radiology residents in China, workload was associated with recent self-reported diagnostic safety-related events primarily through impaired recovery and psychological strain rather than through a direct workload effect. These findings support a recovery-strain framework for understanding diagnostic vulnerability during training and suggest that protecting restorative capacity may be relevant to diagnostic safety interventions.