Caveats of this approach include the following:

    1)  

    Quality of data varies by race and ethnicity in both death certificate data and Census data. The quality of data for the white and black populations are good in contrast to that for other groups. American Indian is estimated to be underreported on the death certificate by more than 20% (probably greater than this) and Asian and Pacific Islander is underreported by about 10%. With respect to population data, perhaps, the most serious problem is the 33% increase in the American Indian population between 1960 and 1990 that reflects increased preference to self-report as an American Indian.

    2)  

    Age-specific death rates, especially those for the elderly, are underestimates.

    3)  

    Hispanic origin and race are separate items. We picked the best rates going across these two variables, but show the deaths for both Hispanic origin and race.

    4)  

    Hispanic origin was not reported on all State death certificates in this time period. The rates used in the calculation were based on 46 States and the District of Columbia. The observed deaths count was inflated to a national total by using the reciprocal of the fraction of the Hispanic population living in these 46 States and D.C.

    5)  

    The number of events, particularly at young ages for certain minority groups, are small. This would affect the stability of the rates. To counter that tendency, we used data for the 1990–95 time period. In the summary table, we have expressed the number of deaths in terms of a single year of data.

    6)  

    The reduction for all cancer sites combined differs depending on whether the values are summed across specific cancer sites or is done separately. The reduction is greater when the values are summed.



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement