Occupational Mental Health: An Investigation of Risk Indicators Using Interpretable Machine Learning Techniques.

Abstract

Objective: The aim of the study was to apply interpretable machine learning to identify key factors influencing work-related mental health cases to support early intervention.

Methods: Using 1117 records from Brazil's Notifiable Diseases Information System for the period from 2007 to 2022, five machine learning models were developed to classify mental health cases as mild or severe. SHAP analysis was employed to rank and interpret the most influential predictors.

Results: The decision tree model achieved 82.9% accuracy (92 of 111 cases classified, including 83 of 85 severe cases), while the support vector machine reached 82.0% accuracy (91 of 111 correct, including 84 of 85 severe). Key determinants included work removal, protective measures, and regional factors. High-risk occupations comprised energy/water operators, legal professionals, and engineers.

Conclusions: Interpretable machine learning models effectively predict mental health outcomes, revealing actionable sociodemographic and occupational risk factors for targeted interventions.