Population Health Improvement (PHI) Core
Funded Grant
Overview
Affiliation
View All
Overview
description
Abstract We propose a Population Health Improvement (PHI) Core that will support the use of contemporary artificial intelligence/machine learning methods and advanced causal inference techniques (e.g., quasi-experimental designs, target trial emulation) in large population-level datasets to leverage these analytic approaches in diabetes translation research. SoCal CDTR Members will have access to state-of-the-science tools and approaches to increase the validity and impact of their work to address our PHI Core theme of “Applying Cutting-Edge AI/ML Methods and Causal Inference Techniques to Advance Diabetes Translation and the Health of All Americans with or at Risk for Diabetes.” Recent stagnation in reducing type 2 diabetes prevalence and the rise in diabetes-related complications highlight the need for innovative strategies translating research into effective interventions. Growing amounts of data and computational capacity are driving novel data-driven methods that can enable important insights and accelerate diabetes translational research. The use of such methods is highlighted in NIDDK’s 2024 Artificial Intelligence (AI) in Precision Medicine for Diabetes and other Chronic Diseases Workshop and emphasized in the NIH’s 2025 Data Science Strategic Plan. However, many investigators lack sufficient knowledge or skills to use these techniques. Recognizing this gap, the proposed PHI Core will support the use of these methods and techniques, building on a strong track record of consultation and collaboration across multiple technical areas (e.g., simulation, causal inference, biomedical informatics, data science). The proposed PHI Core Leads bring expertise in causal inference techniques, biomedical AI/ML research, and study design for evaluating data-driven methods. The PHI Core also includes a multidisciplinary group of Subject Matter Experts who will provide expertise in application of AI/ML in different settings, data acquisition, and healthcare costs and utilization. The Specific Aims of the PHI Core are to: 1) Establish methodological infrastructure supporting innovative applications of AI and ML-based causal inference in diabetes translational research. This infrastructure will provide an integrated computing environment optimized for advanced AI coupled with comprehensive healthcare datasets including electronic health records (EHRs) and claims data; 2) Provide technical expertise, consultation, and collaborative support in implementing data-driven models to elucidate predictors of diabetes-related outcomes and ML-based heterogeneous treatment effects of interventions; 3) Advance the application of advanced causal inference methods, such as quasi-experimental methods and target trial emulation method, using both RCT and observational data in diabetes research, and 4) Foster methodological innovation through training, education, and partnerships with diabetes researchers. The PHI Core will develop and implement innovative methodological approaches for translational diabetes research, with emphasis on improving implementation of evidence-based interventions in different settings.
