Advancing Data Science Approaches to Address Health Disparities Through Artificial Intelligence (AI), and Machine Learning (ML), and Community-Engaged Research

Post Date: July 20, 2026

Expiration Date: July 20, 2028

Purpose: this topic will support the development, implementation, and evaluation of community-engaged AI/ML interventions that convert routinely collected clinical and community-linked data into timely actions to improve screening completion, treatment adherence, disease control, and continuity of care in populations experiencing health disparities.

Background:

AI/ML is sufficiently advanced to support community-engaged interventions for populations experiencing health disparities. These models are most useful when used to activate specific, measurable actions rather than autonomous clinical decisions. Recent advances in electronic health record (EHR)-based prediction, natural language processing, multimodal data integration, and mobile and remote monitoring makes it feasible to identify patients at high short-term risk of missed screening, uncontrolled chronic disease, medication interruption, avoidable acute care use, and loss to follow-up

The strongest near-term opportunity is not diagnosis alone, but interventions targeting the right person to the right outreach, service, and follow-up at the right time.

For health disparities science, AI/ML should be embedded in clinical and/or community settings where implementation can improve uptake of evidence-based care. High-value use cases include, but are not limited to:

• Using clinic, pharmacy, and remote monitoring data to prompt community health worker outreach for uncontrolled hypertension
• Identifying patients overdue for colorectal, cervical, or breast cancer screening and routing them to mailed or mobile screening options
• Predicting diabetes treatment interruption and prompting refill support, nutrition counseling, and home glucose monitoring
• Flagging pregnancy and postpartum patients at rising risk for hypertension, depression, or missed visits to activate nurse navigation and telehealth follow-up. 

The main question is no longer whether AI/ML can generate accurate predictions in retrospective datasets. The critical question is whether engaging community in AI/ML systems can improve real-world outcomes when prospectively integrated into care delivery, and workflows that community organizations and health systems can sustain. Research should therefore move beyond model development alone and test complete intervention pathways on whether outcomes improve, including: 

• What data are required?
• Which predictions are actionable?
• What service is triggered?
• Who delivers it?
• How patients respond? 

Also of interest are use cases with short feedback loops, clear operational workflows, and measurable clinical endpoints.