Construction and evaluation of a predictive model for high mental workload risk among clinical nurses: a cross-sectional study

Abstract

Aim: To develop and validate a predictive model for identifying clinical nurses at high risk of mental workload (MWL) using a nomogram-based approach, grounded in the Job Demands–Resources theoretical framework. Methods: Guided by the Job Demands–Resources model, a total of 826 clinical nurses were recruited from nine tertiary hospitals. Data were collected through standardized questionnaires assessing demographic characteristics, mental workload, and psychosocial factors including emergency response capacity, perceived social support, and coping style. A total of 16 variables were screened using least absolute shrinkage and selection operator (LASSO) regression. Seven significant predictors were then entered into a binary logistic regression model and used to construct a nomogram. Model performance was assessed using the area under the curve (AUC), calibration curves, Hosmer–Lemeshow test, and 10-fold cross-validation. Results: Seven variables were identified as independent predictors of high mental workload: gender, salary satisfaction, frequency of night shifts, turnover intention, emergency response capacity, perceived social support, and negative coping style. The nomogram demonstrated good discriminative ability in both the training (AUC = 0.796, 95% CI: 0.741–0.852) and validation cohorts (AUC = 0.793, 95% CI: 0.757–0.830). Calibration curves showed strong agreement between predicted and observed outcomes. The C-index derived from bootstrap resampling was 0.761, while 10-fold cross-validation yielded a mean C-index of 0.771, indicating robust internal validity and consistent performance. Conclusion: A validated nomogram was developed to predict the risk of high mental workload among clinical nurses. The model exhibited favorable discrimination, sound calibration, and consistent internal reliability, offering an effective means for identification and focused intervention.