Machine-learning models predicting osteoarthritis associated with the lead blood level.

Area of Science:

• Environmental health
• Epidemiology
• Biostatistics

Background:

• Lead is a significant environmental pollutant in industrialized nations.
• Lead exposure is identified as a risk factor for osteoarthritis (OA) in elderly women.

Purpose of the Study:

• To compare the performance of various machine learning (ML) algorithms in predicting OA prevalence associated with lead exposure.
• To evaluate the effectiveness of ML models in identifying OA risk from population-based survey data.

Main Methods:

• Utilized data from 2224 women aged 50+ from the Korea National Health and Nutrition Examination Surveys (2005-2017).
• Developed and tested five ML algorithms: logistic regression (LR), k-nearest neighbor, decision tree, random forest, and support vector machine.
• Split data into 70% training and 30% testing sets for model generation and validation.

Main Results:

• All tested ML algorithms demonstrated acceptable predictive accuracy for OA prevalence.
• The logistic regression (LR) model achieved the highest accuracy.
• LR model exhibited the greatest area under the receiver operating characteristic curve, indicating superior predictive performance.

Conclusions:

• Machine learning models are effective tools for predicting the risk of osteoarthritis associated with lead exposure.
• Population-based survey data can be successfully utilized with ML techniques to assess environmental health risks like lead exposure and OA.
• Logistic regression emerged as the most accurate model for this specific prediction task.