Our goal was to use cross-sectional national mortality data to provide
a multivariable statistical analysis of the factors that contribute t
o the decision of whether an autopsy will be performed. The identifica
tion of determinants of the autopsy is an important prerequisite for f
inding cost-effective alternatives for arresting or reversing the decl
ine of autopsy rates in the circumstances in which the autopsy can con
tinue to make a crucial contribution to clinical medicine and public h
ealth. The source of the data was 1986 National Center for Health Stat
istics (Washington, DC) mortality data tapes for Kentucky, Maryland, M
innesota, and Washington for the 1986 calendar year. Separate multiple
logistic regressions were conducted on these data on a state-by-state
basis, with a total of 139,063 individual mortality records as the un
it of analysis. The dependent variable in all models was autopsy (yes/
no). Odds ratios for selected explanatory variables were estimated for
all four states, and the relative contribution of each explanatory va
riable was studied in a detailed analysis of one state. In general, th
e following independent variables had a statistically significant posi
tive relationship with whether an autopsy will be performed: male sex;
nonwhite ethnicity; death due to ill-defined or unknown cause; death
due to accident, suicide, or homicide; presence of a nationally recogn
ized medical center in the county of death; and death occurring in a s
tandard metropolitan statistical area. In general, the following indep
endent variables had a statistically significant negative relationship
with whether an autopsy will be performed: older age at death; higher
income level of the decedent; death in a nursing home; death at home;
and residency in the county of death. The two most important variable
s influencing the autopsy decision were age at death (especially old a
ge) and death due to accident, homicide, or suicide.