ethnicity predicts variations in health within the Black population (David and Collins, 1997; Fruchter et al., 1985).

The U.S. census also collects data on White ethnic subgroups, but the extent to which ethnicity predicts variations in health for the White population has not been systematically explored in recent health research.

In creating mortality statistics, the numerator for death rates for any given group comes from absolute counts based on death certificates. An undercount in the numerator suppresses death rates for the subject group. A growing body of evidence indicates that funeral home directors and other officials who record racial status on death certificates misclassify a relatively high proportion of Native Americans or Alaska Natives, Hispanics, and Asians or Pacific Islanders as White. This has serious implications for the quality and accuracy of mortality data trends for these populations (Hahn, 1992). Miscategorization undercounts death rates for these groups and slightly inflates rates for Whites.

Sorlie et al. (1992) compared race as self-reported during personal interviews for the Current Population Survey with race as recorded on the self-reporter’s death certificate. They found very high agreement for Blacks and Whites; however, 26 percent of Native Americans or Alaska Natives, 18 percent of Asians or Pacific Islanders, and 10 percent of Hispanics were classified as another racial category on the death certificate; most were classified as White. A study of mortality data for American Indian infants found that 28 percent were misclassified as another race on the death certificate (Kennedy and Deapen, 1991). Another study of data in a cancer surveillance registry found that 40 percent of cancer patients registered with IHS as Native Americans or Alaska Natives were identified as another race in the registry (Frost et al., 1992). For Native Americans or Alaska Natives, misclassification on the death certificate also appears to vary by cause of death; Native Americans or Alaska Natives who die alcohol-related deaths are more likely to be correctly coded than those who die as a result of some other major chronic illness (Frost et al., 1994).

The denominator for mortality statistics comes from census data. Obviously, inaccuracy in those data also limits the quality of these health-related statistics (Notes and Comments, 1994). A denominator based on an undercount inflates the rate in exact proportion to the undercount. Although the overall undercount for the U.S. population is relatively small, it is much higher for Blacks than it is for Whites. Evaluations based on demographic analyses suggest that there is a net census undercount of 11 to 13 percent for all of the 10-year age groups for Black males between the ages of 25 and 64 (National Center for Health Statistics, 1994a). Thus,