BACKGROUND: Medical care at the end of life is often expensive and ineffective. OBJECTIVE: To explore associations between primary care and hospital utilization at the end of life. DESIGN: Retrospective analysis of Medicare data. We measured hospital utilization during the final 6 months of life and the number of primary care physician visits in the 12 preceding months. Multivariate cluster analysis adjusted for the effects of demographics, comorbidities, and geography in end-of-life healthcare utilization. SUBJECTS: National random sample of 78,356 Medicare beneficiaries aged 66+ who died in 2001. Non-whites were over-sampled. All subjects with complete Medicare data for 18 months prior to death were retained, except for those in the End Stage Renal Disease program. MEASUREMENTS: Hospital days, costs, in-hospital death, and presence of two types of preventable hospital admissions (Ambulatory Care Sensitive Conditions) during the final 6 months of life. RESULTS: Sample characteristics: 38% had 0 primary care visits; 22%, 1-2; 19%, 3-5; 10%, 6-8; and 11%, 9+ visits. More primary care visits in the preceding year were associated with fewer hospital days at end of life (15.3 days for those with no primary care visits vs. 13.4 for those with > or = 9 visits, P