Artificial neural network modelling using complete blood count for the early detection of black lung disease among Indonesian coal miner: A retrospective longitudinal study.

BACKGROUND

Black lung disease remains a major occupational health problem among coal miners and is frequently diagnosed at an advanced stage, limiting opportunities for prevention. This study aimed to develop a concise and accessible early detection model using routine complete blood count (CBC) parameters.

METHODS

A retrospective longitudinal analysis was conducted using annual health examination data from 807 Indonesian coal miners collected between 2013 and 2021. An artificial neural network (ANN) model was developed and internally validated, with further validation using Cox proportional hazard regression.

RESULTS

The cumulative incidence of black lung disease was 13.9%. Eosinophil and monocyte levels consistently emerged as the most influential predictors in both models, with adjusted hazard ratios of 1.286 (95% CI: 1.254-1.319) and 1.136 (95% CI: 1.034-1.247), respectively. The ANN demonstrated high internal predictive performance.

CONCLUSION

These findings indicate that routinely collected CBC parameters, particularly eosinophils and monocytes, may serve as practical biomarkers for early identification of black lung disease in high-risk occupational populations. Integrating CBC-based screening into occupational health surveillance could strengthen early detection and prevention strategies.