AIM-AHEAD Research Spotlight Series: Showcasing Innovative Research Across the Consortium

Responsible Health AI Lab (RHAIL)

Martine De Cock, PhD
University of Washington Tacoma
North/Midwest Hub

Expanding access to care and biomedical data remains a key challenge in healthcare, particularly for populations facing barriers to support and treatment adherence. Through the AIM-AHEAD PAIR Program (Cohort 1), Dr. De Cock has contributed to the Responsible Health AI Lab (RHAIL), which focuses on developing artificial intelligence (AI) approaches to support patients and caregivers while improving access to health data.

Study Design and Methods

RHAIL’s work spans two primary areas: conversational agents and synthetic data generation. One focus involves developing AI-powered chatbots to support health-related needs. This includes the “Care for Caregivers Online” (COCO) project, a mobile application that integrates a GPT-based chatbot (“cocobot”) to guide family caregivers using principles from Problem-Solving Therapy through structured, goal-oriented conversations.

A related effort applies similar approaches to tuberculosis (TB) care in collaboration with the University of Washington, focusing on chatbot-supported digital interventions to address challenges with treatment adherence.

In parallel, RHAIL explores synthetic data generation to address limitations in access to biomedical data. Supported by National Science Foundation (NSF) projects, including a NAIRR demonstration with Sage Bionetworks and a CISE-funded collaboration with the University of Central Florida, this work uses real-world data to develop models that can generate artificial datasets for broader research use.

Key Findings

Across these efforts, the work highlights the potential of AI-driven tools to support both caregivers and patients. Conversational agents can guide users through health-related challenges, including caregiving support and symptom-related questions, while structured prompting approaches shape how these systems assess needs and generate responses.

Synthetic data efforts further underscore ongoing challenges in data accessibility. Approaches that generate artificial data from real-world sources may help expand the availability of research datasets while addressing existing barriers to access.

Implications for Research and Clinical Practice

Together, these projects illustrate how AI can be applied to improve both care delivery and data access. From supporting caregivers and treatment adherence to enabling broader use of biomedical data, this work reflects continued efforts to develop scalable, responsible AI applications in healthcare.

Recognition and Next Steps

Dr. De Cock’s work through RHAIL has contributed to five publications and has been recognized through the CAMDA Health Privacy Challenge for advances in synthetic data generation. Ongoing and recent efforts are supported by multiple federal grants, including awards from the National Science Foundation (NSF) and the National Institutes of Health (NIH), which support the continued development of AI-driven tools for healthcare applications.

Machine Learning for Survivors of Congenital Heart Disease

John Valdovinos, PhD
California State University Northridge 
West Hub

Congenital heart disease (CHD) requires ongoing monitoring to support long-term health outcomes, yet many patients experience gaps in care over time. Prior research has identified several factors associated with loss-to-follow-up, including lack of insurance coverage, limited access to specialized adult CHD care, gaps in patient knowledge, and challenges with transition readiness. Through participation in the AIM-AHEAD PAIR Program (Cohort 1), Dr. John Valdovinos has explored how artificial intelligence and machine learning (AI/ML) can be used to better understand care patterns and identify patients at risk of disengaging from routine cardiovascular care.

Study Design and Methods

Dr. Valdovinos’ research draws on data from the All of Us Controlled Tier Dataset (v8), including 6,459 patients with varying levels of CHD severity, ranging from atrial septal defects to more complex conditions such as hypoplastic left heart syndrome. Guided by life-course concepts, the study examines patterns of healthcare utilization using machine learning methods, including clustering and risk-scoring models, to characterize care trajectories.

The study also incorporates digital health data to better understand patient behavior outside of clinical settings. A subset of 537 CHD patients with Fitbit data is used to assess physical activity measures such as step count, active minutes, and heart rate, with comparisons to a non-CHD cohort of 36,071 individuals. Device adherence is considered to account for differences in wearable usage and data completeness.

Key Findings

Analyses highlight variation in healthcare access and utilization among individuals with CHD, including differences in reported barriers to care. Survey data suggest that experiences with healthcare access may differ across patient groups, underscoring the importance of both clinical and non-clinical factors.

Wearable data further point to differences in device use and activity patterns between CHD and non-CHD populations, with adherence influencing the availability of usable data.

Implications for Research and Clinical Practice

By integrating clinical, survey, and wearable data, this work demonstrates the potential of AI/ML to identify patterns associated with care utilization and disengagement among CHD patients. These insights help inform approaches for identifying patients at risk of loss-to-follow-up and supporting more personalized, continuous care.

This research also contributes to broader efforts to incorporate digital health data into clinical research and to better understand how behavioral and access-related factors shape long-term outcomes in CHD populations.

Recognition and Next Steps

Dr. Valdovinos’ work has resulted in one published manuscript and an additional submission, with findings presented at venues including the IEEE EMBS Conference, the AIM-AHEAD Annual Meeting, and international symposia. Ongoing efforts include grant activity, with one submission and additional proposals in progress.

This work has also supported training and mentorship across multiple levels, including undergraduate, post-baccalaureate, and graduate students in engineering and psychology.