AI2Equity: AI Integrating SDOH Data to Advance Health Equity in Cardiovascular Risk Prediction
Abstract
Disparities in cardiovascular disease (CVD) risk exist among racial/ethnic minorities and lower socioeconomic
groups, where social determinants of health (SDOH) and structural social factors substantially contribute. The
underlying clinical, social, and intersectional structural factors are interconnected and complex. However, most
risk prediction models do not account for these factors and their complex interactions with clinical factors.
Machine learning (ML)-based approaches have been applied to this important problem, but most are limited in
terms of equity, SDOH integration, methodology advancement, interpretability enhancement, and generalizability.
To address these limitations, we will build an AI-powered, equitable, interpretable, actionable, and
generalizable CVD risk model to inform social and clinical interventions and reduce disparities among
racial/ethnic minorities and socially disadvantaged populations. With experts in AI/ML, clinical primary care,
SDOH and health equity, and biostatistics, our multidisciplinary team is well-suited to carry out this important
and timely study. We will apply advanced deep learning ML approaches to predict CVD risk and identify social
and clinical risk factors in diverse patient populations from one national community health center research
network (OCHIN) and two academic hospitals. First, we will develop a social-ecological AI model to improve
health equity (AI2Equity) in CVD risk prediction by integrating clinical factors and multi-domain, multi-level social
and structural factors into an advanced deep learning architecture (Aim 1). Second, we will enhance the
AI2Equity model’s fairness and interpretability through algorithmic optimization, explainable AI, and qualitative
study (Aim 2). Third, we will assess and improve AI2Equity’s generalizability across multiple healthcare
systems/settings and simulate its effectiveness compared to traditional CVD risk prediction tools (Aim 3). This
project provides a solid foundation for future clinical trials involving CVD prevention. The study aligns with NIH’s
commitment to “Leveraging the potential of AI/ML to accelerate the pace of biomedical innovation while
prioritizing and addressing health disparities and inequities.” Clinical decision support powered by AI2Equity,
leveraging social and clinical factors and their interactions, can better guide resources, promote health equity,
and steer multi-component interventions (e.g., clinical intervention, social support, and policy changes) in CVD
prevention. It will allow for more equitable CVD prevention interventions in racial/ethnic minorities and
socially disadvantaged groups and has the potential to reduce disparities.
