.png){kind=icon}
.png){kind=icon}
Return
Development and Internal Validation of an Explainable Machine Learning Model for Compassion Fatigue Risk Stratification Among Clinical Nurses in China
.png){kind=icon}
OA
AI
M
C
DOI：10.2147/rmhp.s617883.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Meng Chen,1 Chengcheng Che2 1Department of Gastroenterology, Shengjing Hospital of China Medical University, Shenyang, Liaoning, 110004, People’s Republic of China; 2Department of Emergency Medicine, Shengjing Hospital of China Medical University, Shenyang, Liaoning, 110004, People’s Republic of China Correspondence: Chengcheng Che, Department of Emergency Medicine, Shengjing Hospital of China Medical University, No. 36 Sanhao Street, Heping District, Shenyang, Liaoning, 110004, People’s Republic of China, Email 77702147@cmu.edu.cn Background: Compassion fatigue (CF) is a significant occupational challenge among nurses and is associated with adverse workforce and patient-care outcomes. In China’s demanding healthcare system, identifying nurses at elevated risk of CF may help inform targeted support strategies. This study aimed to develop and internally validate an explainable machine learning (ML)-based model for CF risk stratification among clinical nurses. Methods: A cross-sectional survey was conducted among 969 clinical nurses in Liaoning Province, China. CF status was classified using established Professional Quality of Life Scale questionnaire cutoff criteria. A hybrid approach combining the Boruta algorithm and Least Absolute Shrinkage and Selection Operator regression was employed. Eight ML algorithms were developed and compared. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), accuracy, recall, and F1-score. Shapley Additive exPlanations (SHAP) analysis was used to interpret the optimal model and quantify the contribution of important risk factors. Results: Based on questionnaire-defined criteria, 56.2% of participants were classified as having elevated CF symptoms. Among the evaluated algorithms, the Naïve Bayes (NB) model demonstrated the best overall performance, achieving an AUC of 0.924 (95% confidence interval [CI]: 0.894– 0.954) in the testing set. It also showed favorable calibration and potential net benefit. SHAP analysis indicated that social support, work engagement, and mindfulness were important protective factors, whereas exposure to workplace violence, frequent night shifts, prolonged daily working hours, and department assignment were important risk factors associated with elevated CF symptom classification. Conclusion: The NB-based model demonstrated strong discrimination and interpretability for stratifying nurses at elevated risk of CF within this study population. The findings highlight potentially modifiable factors associated with elevated CF symptoms and may support targeted occupational health strategies. External and prospective validation studies are still needed before broader implementation in clinical or administrative settings. Keywords: compassion fatigue, machine learning, naïve Bayes, risk stratification, SHAP, clinical nurses
Keywords:
compassion fatigue
machine learning
naïve Bayes
risk stratification
SHAP
clinical nurses
Journal
IF:
2
Papers:
269
Citations:
4.1K
Organization
Cited Papers