Autopsies have long been viewed as a biased source of information with regard to the mortality statistics that prevail in the hospital and community. This bias could be of either a demographic or clinical nature. Our objective was to define the autopsy characteristics from the National Cancer Institute of Mexico and determine how representative they were of hospital mortality. Age, sex, place of residence, socioeconomic status, and tumor location were obtained from the Hospital Mortality Registry (1985 and 1987). During these two years, 451 deaths were registered; in 290 of these cases (64.3%), an autopsy was performed. Discrepancies of 9.8 percent were found between autopsy diagnosis and mortality registry data. Our results indicate that autopsy examinations effectively reflect total hospital mortality, and represent a useful tool for epidemiological cancer studies in Mexico. Furthermore, we believe that mortality statistics should be based on autopsy results.