Related Experiment Videos
Development and Internal Validation of an Explainable Machine Learning Model for Compassion Fatigue Risk
1Department of Gastroenterology, Shengjing Hospital of China Medical University, Shenyang, Liaoning, 110004, People's Republic of China.
Risk Management and Healthcare Policy
|July 27, 2026
Summary
A machine learning model identified nurses at high risk for compassion fatigue (CF). Key protective factors include social support and mindfulness, while workplace violence and long hours increase risk. This aids targeted support strategies for nurses.
Area of Science:
- Nursing
- Occupational Health
- Machine Learning
Background:
- Compassion fatigue (CF) significantly impacts nurses, affecting workforce stability and patient care.
- Identifying at-risk nurses in China's demanding healthcare system is crucial for developing targeted support.
- This study focused on creating an explainable machine learning (ML) model for CF risk stratification in clinical nurses.
Purpose of the Study:
- To develop and internally validate an explainable ML model for stratifying compassion fatigue risk among clinical nurses.
- To identify key risk and protective factors associated with compassion fatigue in nurses.
- To inform the development of targeted occupational health strategies for nurses experiencing CF.
Main Methods:
- A cross-sectional survey of 969 clinical nurses in Liaoning Province, China, was conducted.
- A hybrid approach using Boruta algorithm and LASSO regression identified significant variables.
- Eight ML algorithms were developed and compared, with the Naïve Bayes model selected for its performance and interpretability via SHAP analysis.
Main Results:
- Over half (56.2%) of nurses reported elevated compassion fatigue symptoms.
- The Naïve Bayes model achieved high performance (AUC=0.924) in predicting CF risk.
- Social support, work engagement, and mindfulness were protective factors; workplace violence, night shifts, long hours, and department type were risk factors.
Conclusions:
- The developed Naïve Bayes model effectively stratifies nurses at risk for compassion fatigue with strong interpretability.
- Findings highlight modifiable factors that can inform occupational health interventions.
- Further external and prospective validation is recommended before widespread implementation.