Scholarlink

Background &amp; objectives <p>Social determinants of health (SDoH) play a crucial role in shaping health outcomes and quality of life, particularly among individuals living with chronic conditions such as diabetes. Our aim in this study was to develop and validate ensemble ML models that can predict HRQoL in patients with diabetes using social determinants of health (SDoH).</p> Methods <p>This cross-sectional study leveraged data from the 2017 Behavioral Risk Factor Surveillance System (BRFSS), targeting adult participants aged 18 years and older with diabetes. Twenty-three predictors, including demographic, socioeconomic, physical, behavioral, mental, and SDoH factors were assessed. The ML model’s performance was examined by calculating precision, recall, F1-score, false positive rate, and the area under the receiver operating characteristic curve (AUROC). Data analysis was performed from August 2024 to September 2024.</p> Results <p>Among 4946 participants, 2381 (48.14%) were aged 65 years or older, and 2815 participants (56.91%) were female. The study population consisted of 142 Asians (2.87%), 538 Blacks (10.88%), 237 Hispanics (4.79%), and 4029 White individuals (81.46%). The AdaBoost model achieved the highest AUROC at 0.70 (95% CI: 0.66–0.73), significantly surpassing the other models. The performance metrics of the AdaBoost model were as follows: accuracy, precision, recall, F1 score, and false positive rate were 0.65 (95% CI: 0.62–0.68), 0.68 (95% CI: 0.64–0.72), 0.73 (95% CI: 0.69–0.76), 0.70 (95% CI: 0.67–0.73), and 0.27 (95% CI: 0.23–0.30), respectively.</p> Conclusion <p>In this cross-sectional study, we developed and internally validated an ML-based model to identify patients with diabetes at risk for lower HRQoL. The AdaBoost model achieved moderate discriminative performance but showed promise as a screening and prioritization tool to support more personalized and timely interventions. These findings suggest that ML-based risk stratification may complement existing clinical approaches by enabling earlier identification of at-risk individuals; however, external validation and calibration optimization are needed before routine clinical implementation.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

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

  • 期刊论文

  • 发表日期：

    2026年2月2日

查看全文
  • Yao-Chin Wang,
  • Md. Mohaimenul Islam,
  • Arinzechukwu Nkemdirim Okere,
  • Tahmina Nasrin Poly,
  • Ming-Chin Lin

摘要

Background & objectives

Social determinants of health (SDoH) play a crucial role in shaping health outcomes and quality of life, particularly among individuals living with chronic conditions such as diabetes. Our aim in this study was to develop and validate ensemble ML models that can predict HRQoL in patients with diabetes using social determinants of health (SDoH).

Methods

This cross-sectional study leveraged data from the 2017 Behavioral Risk Factor Surveillance System (BRFSS), targeting adult participants aged 18 years and older with diabetes. Twenty-three predictors, including demographic, socioeconomic, physical, behavioral, mental, and SDoH factors were assessed. The ML model’s performance was examined by calculating precision, recall, F1-score, false positive rate, and the area under the receiver operating characteristic curve (AUROC). Data analysis was performed from August 2024 to September 2024.

Results

Among 4946 participants, 2381 (48.14%) were aged 65 years or older, and 2815 participants (56.91%) were female. The study population consisted of 142 Asians (2.87%), 538 Blacks (10.88%), 237 Hispanics (4.79%), and 4029 White individuals (81.46%). The AdaBoost model achieved the highest AUROC at 0.70 (95% CI: 0.66–0.73), significantly surpassing the other models. The performance metrics of the AdaBoost model were as follows: accuracy, precision, recall, F1 score, and false positive rate were 0.65 (95% CI: 0.62–0.68), 0.68 (95% CI: 0.64–0.72), 0.73 (95% CI: 0.69–0.76), 0.70 (95% CI: 0.67–0.73), and 0.27 (95% CI: 0.23–0.30), respectively.

Conclusion

In this cross-sectional study, we developed and internally validated an ML-based model to identify patients with diabetes at risk for lower HRQoL. The AdaBoost model achieved moderate discriminative performance but showed promise as a screening and prioritization tool to support more personalized and timely interventions. These findings suggest that ML-based risk stratification may complement existing clinical approaches by enabling earlier identification of at-risk individuals; however, external validation and calibration optimization are needed before routine clinical implementation.