Background: The use of intensive medical care near end of life is often questioned because of potential burden to patients, their families, and society. Efforts to moderate intensive end-of-life care may be facilitated by early identification of those at greatest risk for receiving such care. Objective: To examine factors associated with intensive end-of-life medical care utilization in nursing home residents with severe cognitive impairment. Design: Retrospective review of existing Medicare data: 1998-2001 Minimum Data Set (MDS), Medicare Denominator, MedPAR, and hospice files. Methods: Subjects were Minnesota and Texas nursing home residents from rural and urban counties (USDA metro-nonmetro continuum codes: 0-2 urban, 6-9 rural), who had severe cognitive impairment and who died during 2000-2001. Hospice and managed care enrollees were excluded. High medical care users were defined as subjects with.7+ intensive care unit (ICU) days in the last 90 days of life. Measures of end-of-life medical care utilization intensity included tube feeding on the last MDS report, number of hospital and ICU days, and total hospital charges during the study period. Results: The study population included 1494 nursing home residents who were hospitalized within 90 days prior to death; 82 (5%) met the high medical care user criteria. In multivariable analysis: urban location (p < 0.001), lack of do-not-resuscitate directive (p = 0.002), non-white race (p = 0.021), and having 3+ comorbidities (p = 0.021) were independently associated with high medical care utilization. Conclusions: Urban nursing home location and lack of do-not-resuscitate directives were the strongest predictors of high medical care utilization near the end of life.