Social determinants of health to predict health-related quality of life in diabetes patients: explainable machine learning approaches

Comprehensive Summary

This research presented by Wang et al. examined the application of social health determinants on quality of life in diabetes patients and the use of machine learning models to calculate higher risk diabetes correlated with lower social drivers of health. A cross sectional survey was administered to diabetic adults (n = 4946) across the United States that utilized data from the 2017 Behavioral Risk Factor Surveillance System (BRFSS). The Social determinants of health covered (n = 23) included socioeconomic, demographic, physical factors, etc. ML algorithms namely, RF, SVM, DT, KNN, XGBoost, AdaBoost, LightGBM, and CatBoost were utilized to process the dataset. Results indicated that machine learning model AdaBoost had the best performance with an AUROC of approximately 0.70, measuring how accurately the model could distinguish between a high risk patient vs. low risk. Wang et al. proposed that validation be performed on external datasets in future unlike the internal validation gained from a single-time point.

Outcomes and Implications

This use of artificial intelligence in medicine could increase awareness of the adverse effects social determinants of health can have on health statutes in patients. Physicians may be better able to treat individuals based on risk of bad health outcomes that are derived from key features such as demographics, socioeconomic factors, behavioral factors, etc. Not only that, but resources provided in the healthcare system could be distributed more effectively and efficiently based on risk factors and need determined by machine learning models.