National sources of vital status information: extent of coverage and possible selectivity in reporting.

The completeness of death reporting and characteristics of deaths not found by the Social Security Administration, Internal Revenue Service, Veterans Administration, and National Death Index were investigated in a follow-up study (1965-1983) of mortality among 18,313 randomly selected Vietnam-era veterans. Overall, 97% of all known deaths identified by these means had been found by the National Death Index (for the years 1979-1983), 83% by the Social Security Administration files, 80% by the Veterans Administration file, and only 23% by the Internal Revenue Service file. Coverage by the Veterans Administration file differed considerably by time period of death; 28% of deaths occurring in 1965-1971 appeared in the Veterans Administration system compared with 87% in 1972-1983. Social Security Administration coverage varied somewhat over time-87% in the period from 1965-1981 and 64% in 1982-1983. Deaths not found by the National Death Index or the Veterans Administration file occurred more frequently among those with certain characteristics, such as nonwhite racial background, nonhonorable discharge, and low rank at discharge. There were no such differences for deaths found by the Social Security Administration. Importantly, these characteristics were strong predictors of subsequent mortality. Thus, in studies of mortality, there is a potential for bias when certain reporting sources are used to the exclusion of others.