A machine learning approach to health equity: identifying underdiagnosed chronic diseases in US adults using national health and nutrition examination survey (2015-2018, 2021-2023)
1
Department of Health and Kinesiology, University of Illinois Urbana-Champaign, Champaign, United States
Popul. Med. 2026;8(Supplement Supplement 1):
ABSTRACT
INTRODUCTION:
Chronic diseases remain the leading causes of morbidity worldwide.1 Machine learning (ML) has demonstrated strong performance as an effective public health approach in identifying individuals diagnosed with chronic diseases in previous studies;2,3 however, few studies have focused on undiagnosed individuals who have limited access to healthcare despite having the condition. This study aims to apply ML models to identify underdiagnosed chronic diseases in US adults and to conduct stratified analyses to assess health equity across population subgroups.
METHODS:
This cross-sectional, nationally representative study used data from the National Health and Nutrition Examination Survey (NHANES) 2015-2018 and 2021-2023. We developed classification models to predict underdiagnosed chronic diseases, including diabetes, hypertension, and high cholesterol. The models incorporated 13 predictors, including socio-demographics, healthcare access, insurance, and financial status. Nine algorithms were trained using 10-fold cross-validation, and model performance was further evaluated through stratified analyses by participants’ insurance status (yes versus no).
RESULTS:
The diabetes (N = 13980), hypertension (N = 10741), and high cholesterol (N = 10525) datasets included 344, 1046, and 779 underdiagnosed individuals, respectively. The ML models achieved moderate discrimination across diseases (best Area Under the Curves (AUCs): 0.72 for diabetes, 0.73 for hypertension, and 0.68 for high cholesterol). Stratified analyses showed that models achieved higher AUCs among uninsured individuals with hypertension, while exhibiting similar AUCs across insurance groups for diabetes and high cholesterol.
CONCLUSIONS:
Our study finds that the ML models demonstrated moderate ability to identify underdiagnosed chronic diseases using a limited set of features. However, challenges remain in reliably identifying diabetes and high cholesterol among uninsured individuals. Future work needs to focus on improving model performance and algorithmic reliability to contribute to a more equitable and fair healthcare system.
Chronic diseases remain the leading causes of morbidity worldwide.1 Machine learning (ML) has demonstrated strong performance as an effective public health approach in identifying individuals diagnosed with chronic diseases in previous studies;2,3 however, few studies have focused on undiagnosed individuals who have limited access to healthcare despite having the condition. This study aims to apply ML models to identify underdiagnosed chronic diseases in US adults and to conduct stratified analyses to assess health equity across population subgroups.
METHODS:
This cross-sectional, nationally representative study used data from the National Health and Nutrition Examination Survey (NHANES) 2015-2018 and 2021-2023. We developed classification models to predict underdiagnosed chronic diseases, including diabetes, hypertension, and high cholesterol. The models incorporated 13 predictors, including socio-demographics, healthcare access, insurance, and financial status. Nine algorithms were trained using 10-fold cross-validation, and model performance was further evaluated through stratified analyses by participants’ insurance status (yes versus no).
RESULTS:
The diabetes (N = 13980), hypertension (N = 10741), and high cholesterol (N = 10525) datasets included 344, 1046, and 779 underdiagnosed individuals, respectively. The ML models achieved moderate discrimination across diseases (best Area Under the Curves (AUCs): 0.72 for diabetes, 0.73 for hypertension, and 0.68 for high cholesterol). Stratified analyses showed that models achieved higher AUCs among uninsured individuals with hypertension, while exhibiting similar AUCs across insurance groups for diabetes and high cholesterol.
CONCLUSIONS:
Our study finds that the ML models demonstrated moderate ability to identify underdiagnosed chronic diseases using a limited set of features. However, challenges remain in reliably identifying diabetes and high cholesterol among uninsured individuals. Future work needs to focus on improving model performance and algorithmic reliability to contribute to a more equitable and fair healthcare system.
| eISSN: | 2654-1459 |
We process personal data collected when visiting the website. The function of obtaining information about users and their behavior is carried out by voluntarily entered information in forms and saving cookies in end devices. Data, including cookies, are used to provide services, improve the user experience and to analyze the traffic in accordance with the Privacy policy. Data are also collected and processed by Google Analytics tool (more).
You can change cookies settings in your browser. Restricted use of cookies in the browser configuration may affect some functionalities of the website.
You can change cookies settings in your browser. Restricted use of cookies in the browser configuration may affect some functionalities of the website.
