Prevalence of and factors associated with work-related musculoskeletal disorders among orthopedic nurses: a cross-sectional study with exploratory interpretable machine-learning modeling

Objective This study aimed to estimate the prevalence of musculoskeletal symptoms and work-related musculoskeletal disorders (WMSDs) among orthopedic nurses, identify demographic, occupational, and lifestyle characteristics associated with WMSDs, and explore whether routinely collected characteristics could discriminate nurses with WMSDs from those without. Method A single-center cross-sectional study was conducted among nurses at a tertiary orthopedic specialty hospital in China between March and November 2023. Conventional analyses included multivariable logistic regression, modified Poisson regression with robust standard errors, and negative binomial regression. Logistic regression as a benchmark, elastic-net logistic regression, random forest, XGBoost, and support vector machine were compared using repeated 5-fold cross-validation with fold-specific preprocessing. Prediction performance was assessed using AUC with bootstrap 95% CIs, PR-AUC, Brier score, calibration intercept and slope, and descriptive threshold metrics; TreeSHAP was used for model interpretation. Result Among 948 nurses with valid questionnaires, 654 reported musculoskeletal symptoms in the previous 12 months, irrespective of work attribution (68.99%). After excluding 44 participants with uncertain work-related attribution, 604/904 participants were classified as having work-related WMSDs (66.81%). Each 1-point increase in the strict ergonomic risk score was associated with WMSDs in logistic regression (adjusted OR = 2.631, 95% CI 2.039–3.432; p < 0.001) and modified Poisson regression (adjusted PR = 1.283, 95% CI 1.203–1.369; p < 0.001). The highest observed AUC was 0.752 (95% CI 0.721–0.791) for the domain random forest model; the domain XGBoost model had the highest PR-AUC (0.841). Conclusion Musculoskeletal symptoms and WMSDs were common among orthopedic nurses in this single-center sample. Occupational exposure and postural load were associated with WMSDs. The exploratory prediction models showed limited-to-moderate discrimination and warrant prospective multicenter external validation before practical application.