OBJECTIVES: This research investigates the impact of specific diagnoses morbidity and comorbity, physician utilization, and physician costs on race differences in rates of mortality among the Medicare elderly. METHODS: CMS billing data from physician record files across five years and Medicare Enrollment Data Base (EDB) data for 6.3 years are examined. We use population data from TN for African Americans (N=61,178), and compare this to a 10% random sample of Caucasians in TN (N=54,643). The total N for the analyses is 115,821, and the overall mortality rate is 36.8%. Cox's proportional hazard models are estimated with survival duration measured in days. Models focus on morbidity and comorbidity between: 1) cardiovascular disease; 2) diabetes; 3) mental illness; and 4) stroke. We investigate a variety of linear and non-linear specifications for the influence of physician service utilization and costs. RESULTS: The effect of physician service utilization on mortality is U-shaped. Regular physician visits significantly decrease mortality, while infrequent visits and high frequency utilization are associated with increased risk for mortality. Patterns of physician utilization decreases race differences in mortality rates from 18.1% to 8.2%. Second, we find that costs have a linear positive impact on mortality rates net of service utilization and diagnosis. Third, we find that race differences in diagnoses of diabetes and mental comorbidity with diabetes further reduce the disparity in mortality rates-from 8.2% to 3.3%. CONCLUSIONS: Motivating and enabling African Americans to make physician visits is key to reducing race differences in mortality. Race differences in patterns of physician utilization are responsible for more than half of the race disparity in mortality outcomes. Costs of care are associated with poorer survival rates. Higher priced care does not yield better outcomes. Rather, costly care is associated with higher rates of death and is likely an indicator of problem severity.