Interventions in health policy and care management have the potential to reduce COVID-19 infections and deaths, particularly if they can be targeted to the most vulnerable populations such as patients with Alzheimer's Disease and Related Dementias (ADRD). In this supplement proposal, we compile information and develop tools that can accelerate and target such interventions. The first aim is to identify the medical and socioeconomic characteristics of people that make them most vulnerable to COVID-19. We create cohorts of patients with ADRD and for other vulnerable populations based on their Fall 2019 characteristics, and follow them through 2020 to identify those at greatest risk of both “direct” COVID-19 outcomes (e.g., critical illness, mortality) and “indirect” increases in non-COVID outcomes. The second aim is an ambitious proof of concept: using natural experiments to shed light on novel drugs to treat or prevent COVID-19 with a particular focus on drugs most heavily used by ADRD patients (e.g., anticholinesterase inhibitors). We will develop and apply a machine learning approach to test the potential effect of drug classes on COVID-19, measured by diagnosis, hospitalization, ICU admission, and death. This supplement is made possible by a unique opportunity: Access to near-real-time Medicare claims data (one-month lag), which CMS appears willing to make available through an expedited data use agreement. The application will supplement an ongoing program project on Improving Health Outcome for an Aging Population, whose overarching aim is to better understand health trends and disparities, determinants of health, and approaches to improving health for an aging population in an evolving landscape.