Risk Prediction for the Development of Hyperuricemia: Model Development Using an Occupational Health Examination...

Area of Science:

• Metabolic diseases
• Occupational health
• Predictive modeling

Background:

• Hyperuricemia (HUA) is a significant metabolic disease in China, posing a considerable health burden.
• Steelworkers represent a population at risk for metabolic diseases, necessitating targeted health management strategies.

Purpose of the Study:

• To develop and compare predictive models for hyperuricemia incidence in steelworkers.
• To evaluate the clinical applicability of different machine learning models for HUA risk prediction.

Main Methods:

• A retrospective cohort study involving 2992 steelworkers with baseline (2017) and follow-up (2019) surveys.
• Development and comparison of Logistic regression, Convolutional Neural Network (CNN), and XG Boost models to predict HUA incidence.
• Evaluation of model performance using discrimination (ROC AUC), calibration (Brier score), and clinical applicability metrics.

Main Results:

• The XG Boost model demonstrated superior performance with an ROC AUC of 0.806 and a Brier score of 0.095 in the training set.
• XG Boost achieved higher sensitivity (81.5%) and specificity (86.8%) compared to Logistic regression and CNN models.
• The XG Boost model exhibited greater clinical applicability for predicting HUA onset risk in the steelworker population.

Conclusions:

• The XG Boost model is a highly effective tool for predicting hyperuricemia onset risk in steelworkers.
• Machine learning approaches, particularly XG Boost, offer significant advantages over traditional methods like Logistic regression for HUA risk assessment.
• These findings support the implementation of advanced predictive analytics for occupational health surveillance and disease prevention.