Objectives: To use place-of-service (POS) codes in the Medicare hospice claims files to document where elderly hospice users with cancer die. Design: Retrospective cohort study. Setting: Surveillance, Epidemiology, and End Results (SEER) cancer registry areas. Participants: Elderly Medicare beneficiaries who died of lung, breast, colorectal, or pancreatic cancer in 2007 and 2008 (N = 46,037). Measurement: Use of hospice, place of service at death (home, nursing home, hospital, inpatient hospice, other), length of stay in hospice. Results: Two-thirds of the beneficiaries used hospice. Younger, male, black, Asian, and unmarried beneficiaries and those enrolled in fee-for-service Medicare or from areas with lower income were less likely to use hospice. Hospice enrollment also varied significantly according to SEER registry. Thirty percent of the hospice users were not receiving home-based care at the time of death, and 17% were enrolled for less than 3 days. Factors associated with hospice death in the home mirrored those associated with hospice use. Individuals dying in hospitals (odds ratio (OR) = 5.13, 95% confidence interval (CI) = 4.63–5.69), inpatient hospice (OR = 1.86, 95% CI = 1.70–2.02), and nursing homes (OR = 1.19, 95% CI = 1.10–1.28) had greater odds of a short hospice stay (≤7 days) than those dying at home, after controlling for all other measured factors, whereas those dying in nursing homes had greater odds of long stays (>180 days) (OR = 1.46, 95% CI = 1.28–1.67). Conclusion: New hospice POS codes are useful for understanding place of death for hospice users. Hospice deaths cannot be assumed to happen at home.
- place of death