Latent profiles of workplace violence exposure and predictive modeling of sleep disorders among emergency department nurses

Workplace violence and sleep disorders are common among emergency department (ED) nurses, yet heterogeneity in violence exposure and its relationship with sleep disorders remains poorly understood. This nationwide cross-sectional study included 1540 Chinese ED nurses. Latent profile analysis identified violence exposure patterns, while LASSO, 7 machine learning algorithms, and SHAP were used for sleep disorder prediction and model interpretation. Three violence profiles were identified: low-level (87.0%), moderate multi-form (10.3%), and high multi-form exposure (2.7%). Sleep disorders affected 59.3% of participants. Random forest showed the highest predictive performance (ROC AUC = 0.624). SHAP analysis showed greater predictive contributions from emotional exhaustion, cynicism, and occupational characteristics than from violence exposure profiles. Workplace violence exposure among ED nurses was heterogeneous. Burnout-related and occupational characteristics provided stronger predictive information for sleep disorder status than violence profiles. External validation and longitudinal studies are needed to assess generalizability and temporal relationships.