Mining the risk: early cardiovascular detection in workers.

Area of Science:

• Occupational health
• Machine learning
• Cardiovascular disease risk assessment

Background:

• Cardiovascular disease (CVD) is a leading global cause of death.
• Existing cardiovascular risk (CVR) tools are limited for unique populations like miners.
• This study utilizes machine learning (ML) and longitudinal data to predict CVR progression in occupational settings.

Purpose of the Study:

• To develop ML models for predicting cardiovascular risk progression using accessible clinical markers.
• To assess the utility of Body Mass Index (BMI) and Blood Glucose (BG) as CVR proxies in mining workers.
• To address limitations of current CVR assessment tools in extreme working conditions.

Main Methods:

• Retrospective longitudinal analysis of 89,045 Chilean mining workers' health data.
• Modeling transitions between defined BMI and BG categories using successive visit pairs.
• Applying ML techniques (XGB, RF) with stratified cross-validation and hyperparameter tuning.
• Evaluating model performance using AUC, accuracy, sensitivity, and specificity.

Main Results:

• ML models achieved high accuracy in predicting BMI transitions, with AUC up to 0.95 for severe progression (morbid obesity).
• Blood glucose (BG) transition prediction showed actionable results, with AUC up to 0.83 for progression to diabetes.
• Models demonstrated good generalization and consistency, with minimal evidence of overfitting.

Conclusions:

• ML models effectively predict clinically relevant BMI and BG risk transitions from occupational health data.
• Longitudinal data and scenario-based evaluation enhance ML model performance for CVR assessment.
• This approach offers potential for improved CVR assessment and preventive decision-making in high-risk working populations.