Natural disasters have increased in frequency and severity in recent decades, a trend projected to intensify under climate change. The risks these disasters pose to individual health and well-being are heightened among older adults with age-related disabilities or chronic illness, such as those living with Alzheimer’s disease or dementia. Public policies regarding disaster preparedness and response serve an important role in avoiding or reducing the health impacts of disasters and improving individual health resiliency. The optimal design of these policies depends crucially on quantifying the short- and long-run health risks imposed by disasters, precisely identifying the most vulnerable populations, and understanding the underlying mechanisms that drive and mediate outcomes. However, much remains unknown about these important dimensions. This project will use individual-level administrative Medicare data on all elderly (65+) and long-term disabled beneficiaries from 1992–2017 to estimate the short- and long-run effects of a broad range of natural disasters – such as hurricanes, tornadoes, and floods – on health, health care use, and health-related expenditures among the elderly. The study will compile a comprehensive and spatially detailed database of US natural disasters from a variety of data sources. The Medicare data provide the 9-digit ZIP code of residence, allowing researchers to pinpoint the location of victims more precisely than prior research. The project will track individuals living in disaster-affected areas before the disaster regardless of where they move, and a control group of unaffected elderly will be used to establish how outcomes would have evolved in the absence of the event. The project will also use a new machine learning approach to identify vulnerable subpopulations based on individual health characteristics (e.g., pre-existing chronic conditions), demographic traits (e.g., age, race, sex), and local economic and public health circumstances (e.g., per capita income, prevalence of obesity, community participation in the National Flood Insurance Program's Community Rating System). The project will also assess several continuity-of-care mechanisms that could moderate or exacerbate post-disaster mortality and disease impacts, including damage to health care facilities, loss of access to one’s physician, and health plan network restrictions. This analysis will be performed for the whole population of beneficiaries and for subpopulations thought to be particularly vulnerable to continuity-of-care disruptions, like adults with multiple pre-existing chronic conditions. Finally, the project will estimate the extent to which certain characteristics (e.g., health care availability, local government expenditure, and healthy behaviors) in areas where victims relocate to following a disaster can explain their health outcomes.