Inconsistencies in coding of race and ethnicity between birth and death in US infants. A new look at infant mortality, 1983 through 1985.

OBJECTIVE: To ascertain the consistency of the racial and ethnic classification of US infants between birth and death and its impact on infant mortality rates.
SUBJECTS: All US infants born from 1983 through 1985 who died within a year.
DESIGN: We used the national linked birth/infant-death computer tape, augmented with information on infants' race and ethnicity at death, to compare the coding of race and Hispanic ethnicity at birth and at death. We also assessed infant mortality rates by race and ethnicity as defined (1) by the standard algorithm and (2) by the rule that, beginning in published tabulations for 1989, assigns newborns the race of their mothers. Finally, we estimated infant mortality rates based on consistent coding of race and ethnicity at birth and death.
RESULTS: Inconsistency in the coding of race is low for whites (1.2%), greater for blacks (4.3%), and greatest for races other than white or black (43.2%). Most infants reclassified at death (87.3%) are classified as white at death. Inconsistency in coding is lower for non-Hispanic whites (3.5%) and non-Hispanic blacks (3.3%) than for Hispanic populations (30.3%). Compared with the standard algorithm for calculation of infant mortality, consistent definition at birth and death produces rates 2.1% lower for whites, and higher for all other groups--3.2% for blacks, 46.9% for American Indians, 33.3% for Chinese, 48.8% for Japanese, 78.7% for Filipinos, and 8.9% for Hispanics.
CONCLUSIONS: The coding of race and ethnicity of infants at birth and death is remarkably inconsistent, with substantial impact on the estimation of infant mortality rates. A need exists to reconsider the nature and definition of race and ethnicity in public health.