Using Human Resources Data to Predict Turnover of Community Mental Health Employees: Prediction and Interpretation of...

Area of Science:

• Health Services Research
• Applied Machine Learning
• Human Resources Management

Background:

• Employee turnover is a significant challenge in community mental health centers.
• Predicting and mitigating turnover is crucial for maintaining service continuity and quality.

Purpose of the Study:

• To apply machine learning (ML) methods to predict mental health employee turnover within 12 months.
• To identify key predictors of turnover in this specific workforce using human resources data.

Main Methods:

• Utilized human resources data from 621 employees at a community mental health center.
• Applied six ML models: logistic regression, elastic net, random forest (RF), gradient boosting machine (GBM), neural network, and support vector machine.
• Employed graphical and statistical tools for interpreting predictive relationships and interactions.

Main Results:

• Random Forest (RF) and Gradient Boosting Machine (GBM) demonstrated superior predictive performance (AUC > 0.8).
• Identified significant predictors including past work years, work hours, wage, age, and employee type.
• Discovered unique predictors for mental health staff, such as training hours and proportion of clients with schizophrenia.

Conclusions:

• Machine learning effectively predicts mental health employee turnover using HR data.
• Identified predictors, including nonlinear and interactive effects, offer insights for developing targeted retention strategies.
• Further research is warranted to refine predictive models for mental health workforce retention.