Evaluation of U.S. Mortality Patterns at Old Ages Using the Medicare Enrollment Data Base

We evaluate the consistency of age-specific mortality patterns for older Americans using the Health Care Financing Administration´s (HCFA’s) Medicare Enrollment Data Base (EDB), a data set that includes over 30 million records. To do this, we compare the mortality curves across regions within race and sex groups, and we compare national mortality curves of white men and women relative to African American and Puerto Rican men and women. The Medicare EDB is a promising source of age-specific mortality data at oldest ages. Both the period and cohort age at death data for whites in the Medicare EDB appear to be very consistent, at least up to age 95, perhaps higher for women. Above age 100 the patterns become extremely inconsistent. However, questions remain about the age-specific mortality patterns of African Americans, especially the rates for men. The African-American mortality crossover is found in every period and cohort comparison of the national population. This crossover pattern is remarkably consistent, occurring at ages 85 or 86 in every cohort and period comparison. This data set is not sufficient to determine whether the African -American mortality crossover is real or due to age misreporting. And the considerable inconsistency across regions also creates concern about these data. When aggregated at the national level, the mortality curves for African American women may be okay up to age 90. In general, we conclude that further examination of the African American data is necessary before using them for substantive analysis.