Core E: Genetic and Policy Data Core

GENETIC AND POLICY DATA CORE REVISED SUMMARY TIME-AD will be supported by a specialized Genetics and Policy Data (GPD) Core focusing on causal identification based on quasi-experimental sources of variation, including genotypes, and state and Kaiser Permanente (KP) systems policies. The GPD Core addresses the unique data access and statistical modeling issues for quasi-experimental methods, whereas the Analytics Core focuses on methods that rely on control of confounders. We will use genetic and policy data primarily as instrumental variables (IVs). The GPD Core will identify genetic and policy data needed for IV analyses in each Project, develop code to create genetic and policy IVs and estimate effects, implement over-identification and other tests to evaluate the validity of each instrument, and provide documentation on the construction, motivation, and quality of each instrument. The quality metrics will include strength of the association between each instrument and the relevant endogenous variables, evidence regarding potential violations of the exclusion restriction assumption, and methods for over-identification tests. The GPD Core will work closely with each of the projects and with the Analytics Core. This specialized expertise in IV methods will facilitate comparing evidence on exclusion restriction violations and help construct the IVs in a way that avoids weak IV bias or uses weak-IV robust methods, are valid across heterogeneous populations, and address contexts which may modify the effects of the IVs on each exposure. The Core will handle IVs using genetic variants associated with alcohol use, depression, and sensory impairment; policy IVs based on state excise taxes and minimum legal drinking age; and health systems-based IVs leveraging changes in the KP antidepressant formulary and the introduction of Screening, Brief Intervention and Referral to Treatment (SBIRT) protocols in adult primary care. We will also develop polygenic scores for Alzheimer’s Disease (AD) for multiple ancestry groups, which will be used in bias-detecting or “reverse” Mendelian Randomization. Bias-detecting Mendelian Randomization will be used to evaluate and quantify reverse causation bias from incipient AD to each of the four exposure groups in TIME-AD (alcohol use, depression, sensory impairments, social isolation). The GPD Core will provide each project with code to construct multiple IVs to use for over-identification tests and evaluation of non-linear dose response effects. The GPD Core team includes experts on genetic and policy data, quasi-experimental methods, and health disparities in the NIH framework; staff will be responsible for ensuring cross-project compatibility, documentation, and resource sharing. Our team is familiar with interdisciplinary traditions for IV methods, drawing from economics, causal inference, and the more recent Mendelian Randomization literature. The specialized GPD Core will foster cross- fertilization between genetic and policy IV approaches, and adoption of the most rigorous approaches for implementation and validation of IV analyses to evaluate strategies to prevent AD/ADRD.