This study aimed to identify key risk factors associated with high-somatization risk among frontline nurses responding to infectious diseases, utilizing a robust machine learning approach. Data were collected from 222 nurses in South Korea, including sociodemographic factors, work-related conditions, job stress, and somatic symptoms. Of the participants, 13.1% had high levels of somatization. Machine learning models, including Logistic regression, XGBoost, and the Random Forest algorithm, were developed to predict high-somatization risk. The models were evaluated based on their mean performance across ten CV folds. While XGBoost demonstrated the highest mean AUC (0.734), the Logistic Regression model was prioritized for interpretation due to its superior performance in identifying the high-risk group. Cross-model feature importance analysis revealed that Infection Anxiety and Delayed Staffing Assignment were the 2 most consistent and influential predictors across both models, followed by factors like satisfaction with salary/bonus and daily PPE hours. These findings suggest that integrating machine learning into occupational health assessments can facilitate the early identification of frontline nurses at high risk of somatization, enabling targeted workplace interventions. This study highlights the potential for proactive support strategies to enhance nurse well-being and maintain health care system resilience during future infectious disease responses.
Shin et al. (Tue,) studied this question.