MISCLASSIFICATION OF RACE IN CARDIOVASCULAR DISEASE MORTALITY DATA

Abstract

Graber and colleagues found an ominous deterioration in the quality of race data for American Indian deaths in Maine.1 Specifically, errors because of coding and data entry increased from 0% during 1978–1982 to 33% in 1993–1997; errors in reporting increased from 3% to 22% during the same periods. Part of the problem of race misclassification is being resolved. Coding errors have been reduced as a result of new procedures introduced in April 2003 and have been used by a growing number of states, including Maine, in conjunction with the National Center for Health Statistics (NCHS). The new procedures, in which NCHS codes and edits race/ethnicity data and then quickly returns the data to the states for additional processing, are described on the NCHS Web site (http://www.cdc.gov/nchs/data/dvs/multiple_race_documentation_5-10-04.pdf). Graber et al. did not mention misclassification of race in the denominators of death rates. Errors in the denominators tend to offset, somewhat, errors in the numerators, as shown in an NCHS study of misclassification of race/ethnicity at the national level.2 The authors present 2 types of death rates, those based on “any mention” of cardiovascular disease and those based on “immediate” cause of death. Do the authors mean underlying cause rather than immediate cause? Underlying cause of death is the standard basis for tabulating cause-of-death data as recommended by the World Health Organization.3 While the new NCHS procedures reduce errors introduced during data processing, they do not resolve errors resulting from misreporting of race/ethnicity on death certificates, which Graber et al. show account for almost half the racial/ethnic misclassifications for American Indian deaths in Maine. Errors in reporting can be reduced only by training funeral directors, who obtain information on race/ethnicity from informants or by observation, to be more accurate. Until funeral directors are better trained, NCHS and the states will have solved only half the problem of misclassification of race/ethnicity in mortality data. Graber and colleagues make an important contribution by demonstrating a relationship between state budget resources and data quality. Maine’s race/ethnicity data deteriorated in part because of relaxed quality control standards owing to budget constraints. During the past decade, federal budgets for vital statistics, which support state operations, have failed to keep up with inflation, while—as the authors point out—state budgets, too, have been reduced, with consequences for data quality. The public health community needs to ensure that the state and federal governments provide adequate resources to maintain the data sets used to monitor the health of our citizens.