Mining the Risk: Early Cardiovascular Detection in Workers

Ricardo Jorquera¹Guillermo Droppelmann^2*Max Dollmann¹Gonzalo Blanco¹Ignacio Ahumada¹Alfonso Lira³Felipe Feijoo^3*

• ¹Workmed, Santiago, Chile
• ²Blackmind, Santiago, Chile
• ³Pontificia Universidad Catolica de Valparaiso, Valparaíso, Chile

Cardiovascular disease (CVD) is the leading cause of death worldwide. Although tools exist to assess individual cardiovascular risk (CVR), they often fall short in unique populations such as miners, who work under extreme conditions. To address these limitations, this study applies machine learning (ML) and longitudinal data to predict risk progression using accessible clinical markers. Body mass index (BMI) and blood glucose (BG) were chosen as proxies because they are affordable, routinely measured in occupational health checks, and responsive to metabolic stresses common in mining. Methods: A retrospective longitudinal analysis of 89,045 Chilean mining workers (420,966 pre-employment exams; 2021–2024) was conducted. For each worker, successive visit pairs were formed to model transitions between clinically defined BMI and BG categories. Four binary outcomes were specified per biomarker: any upward transition; adjacent upward transition; obesity→morbid obesity / prediabetes→diabetes; and any transition ending in morbid obesity/diabetes. ML techniques were built to assess transitions across scenarios. We applied a stratified 70/30 train–test split, repeated 7-fold cross-validation, random hyperparameter search (AUC objective), and downsampling within folds to address imbalance. Performance in the imbalanced test set was summarized by AUC, accuracy, sensitivity, and specificity with 95% CIs. Model correlation was assessed using Pearson correlations of predicted probabilities. Results: Predicting BMI transitions was highly accurate. Best performance occurred for severe progression (Scenario 4: any transition ending in morbid obesity), where XGB achieved AUC 0.95 and accuracy 0.91, with high sensitivity and specificity. For broader BMI transitions (Scenarios 1–3), models remained reliable (AUC 0.84–0.87). BG transitions were harder but actionable. Strongest results were for progression to diabetes (Scenario 4), with RF reaching AUC 0.83 (95% CI: 0.82–0.90) and accuracy 0.76; other BG scenarios yielded AUC 0.71–0.77. Cross-validation closely matched test performance. Pairwise probability correlations were typically >0.90 for BMI and >0.80 for BG in severe scenarios, indicating good generalization and no overfitting. Conclusion: ML models effectively predict BMI and BG risk transitions using occupational health data. Longitudinal visit pairs and scenario-based evaluation enhance accuracy and generalization. These findings highlight the potential of this approach to improve CVR assessment and support preventive decision-making in high-risk workers.