To construct the risk prediction model for pediatric nurses' effort-reward imbalance based on interpretable machine learning,and to compare the predictive performance of different models.To explain the results of the optimal model using SHAP interpretation.

Using the convenience sampling method,a total of 414 pediatric nurses from 6 hospitals in Anhui province,Shanxi province,Jiangxi province and Hunan province in June 2025 were selected as the research subjects.They were randomly divided into training set and validation set at a ratio of 7∶3.The Chinese Nurse Stressor Scale and the Effort-Reward Imbalance Questionnaire scale were used for investigation.LASSO regression was employed to screen the characteristic variables and identify the important predictors.The important predictors were incorporated into the machine learning model to construct three risk prediction models for pediatric nurses' effort-reward imbalance:Logistic regression model,Extreme Gradient Boosting model,and Random Forest model.The areas under the receiver operating characteristic curves(AUC),accuracy,sensitivity,and F1 score of the models were compared to evaluate the predictive performance of the models and select the optimal model.The SHAP explanation was used to interpret the optimal model.

LASSO regression identified three important factors:the number of night shifts per month,workload and time allocation,and educational background.The AUC values of the three prediction models(Logistic regression model,Extreme Gradient Boosting model,and Random Forest model) were 0.725,0.890,and 0.903 respectively,with accuracies of 0.673,0.794,and 0.801,sensitivities of 0.421,0.731,and 0.813,and F1 scores of 0.547,0.773,and 0.798.The SHAP explanation results showed that the importance ranking of the influencing factors were the number of night shifts per month,workload and time allocation,and educational background.

The risk prediction model for pediatric nurses' effort-reward imbalance based on interpretable machine learning constructed by Random Forest has better performance than Logistic regression and Extreme Gradient Boosting model.Personalized predictions should be made based on the number of night shifts each nurse takes per month,workload and time allocation,and educational backgroun.It provides a reference for the early identification of the imbalance between effort and reward for the nurses and the formulation of personalized intervention measures.