Older Medicare beneficiaries with disabilities are a high-cost Medicare population and use substantial amounts of health care. A distinctive feature of this population is their use of long-term care (LTC), which can be provided in a facility (e.g., nursing facility, assisted living facility) or the community. This dissertation examined the effect of facility LTC versus community LTC on Medicare expenditures and service use among fee-for-service Medicare beneficiaries from 2000-2009. I examined facility LTC's effect on: Medicare expenditures for inpatient and physician services (Aim 1); Medicare expenditures for sub-acute services (skilled nursing facility, home health, and hospice) (Aim 2); and emergency department (ED) visits, observation days, and likelihood of 30-day or 60-day hospital readmission (Aim 3). To illuminate the role of competing mortality, I implemented most models as survival-adjusted models. To account for potential unobservable differences among those who used community versus facility LTC, I used an instrumental variables approach in all models. The primary dataset for the dissertation was the Medicare Current Beneficiary Survey, supplemented with public data on LTC supply. Incident facility LTC users had substantially lower adjusted cumulative survival. In Aim 1, this higher mortality for facility LTC users decreased cumulative Medicare expenditures on both inpatient and physician services, because individuals were progressively less likely to survive and use services. However, non-significant effects on intensity of service when individuals were alive yielded a non-significant total effect on each expenditure type. In Aim 2, higher mortality again decreased expenditures. However, facility LTC increased intensity of sub-acute service use when alive. The combined effects translated to significantly higher sub-acute expenditures for facility LTC users in early months, though the countervailing effects resulted in a non-significant total effect in later months. In Aim 3, facility LTC increased counts of ED visits, largely due to differences in service intensity when alive. Facility LTC had a non-significant effect on likelihood of 30-day or 60-day readmission. Together, this dissertation suggests LTC facilities affect residents' healthcare use. Payment and benefit design should seek to reward the integration of LTC and healthcare delivery into more appropriate and efficient use of Medicare-covered services for vulnerable older adults.