The Goerlitz Autopsy Study is a population-based autopsy study, conducted in 1987 in the municipality of Goerlitz, population 78,484, in former East Germany. It is unique in that 1,023 (96.5 percent) of the 1,060 subjects who died in the municipality over a period of one year were investigated by full autopsy. An underlying cause was assigned to every death, using the procedures recommended by the International Classification of Diseases, Injuries and Causes of Death, ninth revision, German edition. Assignment of cause of death was done independently by the attending clinician and the pathology team who performed the autopsy. The data were analyzed through cross-classification of deaths by death certificate diagnosis and autopsy-based diagnosis. Sensitivity and positive predictive values were calculated for the death certificate diagnoses, assuming that the autopsy findings represent the correct reference set. Overall, 47 percent of diagnoses on death certificates differed from those based on autopsy and, for 30 percent of the subjects, the difference crossed a major disease category. The proportion of disagreement was higher for deaths occurring in nursing homes and among the very old, but was unrelated to gender. In the death certificates, diseases of the circulatory system and endocrine and metabolic disorders were over-represented, whereas infectious diseases, neoplasms, and respiratory, digestive, and genitourinary diseases were under-represented. Neoplasms in the death certificates are characterized by relatively high sensitivity and predictive values, even though these were still unsatisfactory, given the chronicity of the clinical course and the opportunities available for correct diagnosis.(ABSTRACT TRUNCATED AT 250 WORDS)