Predicting workplace absenteeism using machine learning: a pilot study in occupational health.

Area of Science:

• Occupational Health
• Data Science
• Machine Learning in Healthcare

Background:

• Workplace absenteeism poses significant challenges to organizational productivity and employee well-being.
• Current absenteeism management strategies are largely reactive, necessitating predictive models for proactive interventions.

Purpose of the Study:

• To develop and validate machine learning models for predicting workplace absenteeism patterns.
• To identify risk factors associated with prolonged absence for evidence-based occupational health interventions.

Main Methods:

• Utilized Random Forest and Gradient Boosting algorithms on a Brazilian company's absenteeism dataset (2007-2010).
• Included demographic, clinical (BMI, ICD-10 absence reasons), and occupational factors in the analysis.
• Excluded statistical outliers (>30 hours) to focus on typical absence patterns.

Main Results:

• Random Forest classification model achieved 84% accuracy (AUC=0.89) in distinguishing prolonged absences.
• Random Forest regression model predicted typical absence duration (R²=0.13, RMSE=3.93h, MAE=2.37h).
• Key predictors included absence reason, BMI, and workload intensity, with significant interactions.

Conclusions:

• Machine learning models are feasible for predicting prolonged absenteeism and identifying at-risk employees.
• These models can support personalized health interventions and resource allocation in occupational health.
• External validation and ethical considerations regarding privacy and fairness are crucial for implementation.