Peer-reviewed article

Development and Validation of a Nomogram-Based Risk Stratification Model for Moderate-to-High Frequency Workplace Violence Among Chinese Pediatric Nurses: A Cross-Sectional Study.

Xuemei Ding, Huichen Zhang, Mengting Liu, Xiaoyu Pan, Qingmei Ding et al. · 1 Jan 2026

Abstract

BACKGROUND: Workplace violence (WPV) against pediatric nurses remains a major occupational concern, but practical tools for concurrent risk stratification are limited. OBJECTIVE: To develop and validate a nomogram-based model for concurrent risk stratification of moderate-to-high frequency WPV among Chinese pediatric nurses, and to assess class imbalance using the synthetic minority over-sampling technique (SMOTE) as a complementary analysis. METHODS: This multicenter cross-sectional study included 592 pediatric nurses from 10 tertiary hospitals in Shandong, China. Electronic questionnaires assessed 24 variables across individual, interpersonal, organizational, and policy levels. Participants were randomly divided into training and test sets at a 7:3 ratio. Primary and complementary logistic regression models were developed using the original and SMOTE-resampled training sets, respectively. Models were evaluated using discrimination, calibration, decision curve analysis, risk stratification, and sensitivity analyses. RESULTS: In the primary model, hospital level, nurse-child-caregiver communication conflicts, tolerance of violence, occupational stress, and burnout were independently associated with moderate-to-high frequency WPV. The model showed stable discrimination, with area under the curve (AUC) values of 0.794 in the training set and 0.808 in the test set, and acceptable calibration in both datasets. The optimal cutoff was 0.195, which effectively stratified nurses into low- and high-risk groups in both datasets (p < 0.001). In the SMOTE-based analysis, fair or poor health status, delayed hospital attitudes to WPV, and blaming employees additionally became significant. This model achieved comparable discrimination, with AUC values of 0.814 and 0.805 in the resampled training and original test sets, respectively, and higher sensitivity, but poor calibration in the test set, particularly at higher predicted probabilities. CONCLUSIONS: The primary nomogram may serve as a practical decision-support tool for concurrent WPV risk stratification, whereas the SMOTE-based model should be regarded as a complementary analysis. These findings may help nursing managers identify nurses who currently require prioritized prevention and support.