Reporting delays of deaths with AIDS in the United States.

The 1993 change in the AIDS case definition used in the United States has disrupted AIDS incidence trends. Death incidence is not disrupted by the change, but, like AIDS incidence, must be corrected for reporting delays. This article uses data from the national AIDS surveillance system to examine reporting delays of deaths with AIDS and finds (a) by mid-1994, death counts after two quarters of subsequent reporting seem to be 82% complete; (b) beginning in mid-1991, there is a significant trend toward faster reporting in more recent quarters, but this trend would be masked if very short delays were used in estimation; (c) persons dying in the quarter of diagnosis tend to have longer delays; (d) delays are more variable than assumed by standard methods, making uncertainty estimation difficult; and (e) delay corrections provided with the public data set may be too pessimistic.