Related Experiment Video
Updated: May 25, 2026
07:31
Implementation of a Real-Time Psychosis Risk Detection and Alerting System Based on Electronic Health Records using CogStack
Published on: May 15, 2020
Leveraging machine learning algorithms and explainable AI for predicting mental health disorder treatment at the
Daniela Candanedo1, Edmund Agyemang2, Farhana Chaudhry2
1Department of Epidemiology, Celia Scott Weatherhead School of Public Health and Tropical Medicine at Tulane University, New Orleans, LA, USA.
Acta Psychologica
|May 23, 2026
Summary
Machine learning models effectively predict workplace mental health treatment needs. SHapley Additive exPlanations (SHAP) identified work interference and family history as key predictors, aiding early intervention strategies.
Area of Science:
- Occupational Health
- Artificial Intelligence
- Public Health
Background:
- Workplace mental health disorders present a significant global challenge, impacting productivity.
- Early and accurate prediction is crucial for timely interventions.
Purpose of the Study:
- To evaluate machine learning (ML) algorithms for predicting mental health treatment needs in the workplace.
- To enhance model interpretability using explainable AI (XAI) techniques.
Main Methods:
- Six ML algorithms (logistic regression, random forest, gradient boosting (GB), categorical boosting (CB), support vector machine, neural network) were assessed using the Open Sourcing Mental Health (OSMH) Dataset.
- SHapley Additive exPlanations (SHAP) were employed to identify key predictive features.
- The top 10 features from CB and GB models were combined for final model development.
Main Results:
- SHAP-enhanced CB and GB models demonstrated superior performance in predicting mental health outcomes.
- The CB model achieved 0.8452 accuracy, 0.8651 F1 score, and 0.9101 AUC.
- The GB model achieved 0.8254 accuracy, 0.8417 F1 score, and 0.9070 AUC.
- Work interference and family history were identified as the most influential positive predictors, while coworker support and help-seeking behavior were negative predictors.
Conclusions:
- Interpretable ML models, particularly CB and GB enhanced with SHAP, show significant potential for supporting targeted workplace mental health interventions.
- These models can facilitate early detection and intervention for employees.
- Findings should be considered within the context of pre-pandemic data and self-reported information.
