Project Abstract 1.2M people are living with HIV (PWH) in the US. With advances in antiretroviral therapy, the survival of PWH has increased; half of PWH are over the age of 50. Given the number of older PWH is steadily increasing, this has become a new population of interest in aging research. Older PWH face a 60% increased risk of dementia with different risk profiles for neurological disorders compared to the general population. Specifically, disparities in dementia diagnoses and care are deeply rooted in social determinants of health (SDoH), yet, the dementia risk in PWH has not been well-characterized considering disparities including SDoH—a growing concern in HIV- aging research. Currently, there is no cure for dementia, thus, it is vital to develop strategies for early recognition of dementia and provide interventions on modifiable factors early to delay its onset. Developing an early warning system (EWS) using risk prediction models enables the detection of PWH at a high risk of dementia, supporting timely biobehavioral interventions. However, no such EWS exists for PWH. The proliferation of real-world data (RWD) such as electronic health records (EHRs) and claims data, leveraging artificial intelligence (AI), particularly machine learning (ML), offers unique opportunities to generate real-world evidence (RWE) for HIV- aging research. This proposal creates a cohort of older PWH with all-cause dementia including Alzheimer’s disease (AD) and AD-related dementias (ADRD) using a unique RWD source—OneFlorida+ network (20M patients from Florida, Georgia, and Alabama) integrated with both individual- (e.g., education, social cohesion) and contextual- (e.g., neighborhood characteristics) SDoH. Built upon this unique resource, our study has two objectives: (1) examine disparities in the risk of dementia among PWH≥50 by leveraging large-scale RWD linked with SDoH data and (2) prototype a prediction model for an EWS that identifies PWH at a high risk of dementia. The Specific Aims are: (1) developing computable phenotypes and natural language processing methods and tools to systematically extract key characteristics and relevant outcomes using RWD; (2) developing ML-based prediction models and examining disparities in the risk of dementia among older PWH; and (3) applying a user- centered design approach to prototype an EWS for detecting older PWH with high risk of dementia. Findings will serve as the foundation for R01 submissions focused on the expansion of EWS as an AI-driven clinical decision- support tool optimizing early detection of PWH at a high risk of dementia and the development of biobehavioral interventions supporting the control of ‘realistically modifiable’ SDoH factors for delay of dementia onset that can be used in PWH’s everyday life. Mentors are committed to the candidate’s training, each providing unique expertise to the research and training plan. This K01 application, consistent with the NIA’s mission, will support the candidat...