DEPARTMENT OF HEALTH & HUMAN SERVICES
Public Health Service
Centers for Disease Control and Prevention
National Center for Health Statistics
6525 Belcrest Road Hyattsville, Maryland 20782
September 16, 1998
M. Alfred Haynes, M.D., M.P.H.
Committee on Cancer Research Among Minorities and the Medically Underserved
Institute of Medicine
2101 Constitution Avenue, N.W. Washington, D.C. 20418
Dear Dr. Haynes:
I am writing in response to your letter of July 24, 1998 in which you asked for an analysis of cancer mortality based on race. Please find enclosed a statement of the methodology we used including several caveats, a summary table, and a detailed table demonstrating the reduction in cancer deaths that could be achieved if differences by race or ethnicity could be eliminated.
A few notes on what we've done: You asked us if we could illustrate the number and percent of deaths that would be avoided if all race or ethnic groups shared the same mortality risks for specific cancer sites. You also asked that we consider that the ''best" mortality experience be the one that all groups could, at least hypothetically, experience.
We departed somewhat from the methodology you recommended in the letter with respect to sex, rates, and race:
First, in consultation with NCI, we performed the calculations separately by sex because of the major differences in cancer experience by sex.
Second, we used age-adjusted death rates to select a "best" group, and age-specific death rates to calculate the "excess" deaths.
Third, we combined both the Asian and Pacific Islander and American Indians into a total Asian and Pacific Island group and a total American Indian group. These groups were formed to stabilize estimates, to use groups for which we have estimates of the degree to which race has been misreported, and to use the most recent data available, which is only available for the broad groups.
Even with these steps, we do have some serious reservations about the procedures used and resulting conclusions because race and ethnicity-specific death rates are known to be substantially too low for certain racial/ethnic groups. The consequence is that excess deaths presented here may be substantially greater than would be the case if the rates were not a function of these errors. Indeed, we are most concerned about using the American Indian rates as a baseline or "best rate" group. Given the errors in reporting, it is difficult to believe that other groups could achieve these rates. The two kinds of errors in reported death rates include serious misclassification of race and ethnicity and misreporting of age, particularly among the elderly. These problems occur on the death certificate and also on censuses and surveys.