Hospices have been expected to reduce health expenditures since their addition to the US Medicare benefit package in the early-1980s, but the literature on their ability to do so is mixed. The contradictory findings noted in previous studies may be due to selection bias and the period of cost comparison used. Accounting for these, this study focuses on the length of hospice use that maximizes reductions in medical expenditures near death. We used a retrospective, case/control study of Medicare decedents (1993-2003, National Long Term Care Survey screening sample) to compare 1819 hospice decedents, with 3638 controls matched via their predicted likelihood of dying while using a hospice. Variables used to create matches were demographic, primary medical condition, cost of Medicare financed care prior to the last year of life, nursing home residence and Medicaid eligibility. Hospice use reduced Medicare program expenditures during the last year of life by an average of $2309 per hospice user; expenditures after initiation of hospice were $7318 for hospice users compared to $9627 for controls (P<0.001). On average, hospice use reduced Medicare expenditures during all but 2 of hospice users' last 72 days of life; about $10 on the 72nd day prior to death, with savings increasing to more than $750 on the day of death. Maximum cumulative expenditure reductions differed by primary condition. The maximum reduction in Medicare expenditures per user was about $7000, which occurred when a decedent had a primary condition of cancer and used a hospice for their last 58-103 days of life. For other primary conditions, the maximum savings of around $3500 occurred when a hospice was used for the last 50-108 days of life. Given the length of hospice use observed in the Medicare program, increasing the length of hospice use for 7 in 10 Medicare hospice users would increase savings.