ABSTRACT
Sexual and gender minorities (SGM) face unique health issues, but studies on SGM health are scarce. In
particular, limited data are available among SGM individuals on age-related conditions such as Alzheimer’s
disease (AD) and related dementias. AD is a fatal degenerative disease with a diverse range of risk factors,
ranging from clinical characteristics to social determinants of health (SDoH). AD patients often progress from
cognitively unimpaired to (possible) mild cognitive impairment (MCI), followed by increasing severity of
dementia with AD clinical syndrome. Nevertheless, evidence suggests there exists heterogeneity in the
progression to AD through multiple intermediate stages. Characterizing the different AD progression pathways
and the associated risk factors is crucial for risk stratification and prevention. On the other hand, the
proliferation of large clinical research networks (CRNs) with real-world data (RWD), including electronic health
records (EHRs), claims, and billing data among others, offers opportunities for generating real-world evidence
(RWE) that will have direct translational impacts on AD prevention and care in the SGM populations.
Nevertheless, there are a number of key research and methodological gaps in using RWD for studying AD in
SGM, including the lack of (1) validated computable phenotypes (CP) and natural language processing (NLP)
tools that can accurately define the SGM populations and extract key patient characteristics and outcomes
(e.g., MoCA scores to determine severity), (2) consideration of the heterogeneity in AD and its progression
pathways, and (3) consideration of AD disparities in SGM populations, especially structured on both individual-
and contextual-level SDoH. Responding to NOT-AG-21-050, we propose to analyze large collections of RWD
in the OneFlorida+ and INSIGHT networks, two CRNs contributing to the national Patient-Centered Clinical
Research Network (PCORnet), to: (1) create real-world longitudinal SGM and AD cohorts that can be followed
by virtue of routine clinical care, (2) model the heterogeneity in AD progression with novel federated machine
learning methods, and (3) examine SGM disparities in AD outcomes (i.e., onset and progression pathways)
and in the causal paths via which AD clinical risk factors and SDoH impact these AD outcomes. Our project is
novel and will have direct translational impact as it provides concrete RWE to fill the knowledge gaps by
examining whether AD disparities exist between SGM (and SGM subgroups) and non-SGM, and identifies
potentially actionable AD risk factors and SDoH significant to SGM and their disparities. The success of this
project will fill important gaps in our knowledge of AD risk and progression pathways in the SGM populations,
and establish a framework for creating RWD-based virtual cohort, which can inform national pragmatic trials
across PCORnet for future SGM aging clinical studies.