Based on the Atomic Bomb Casualty Commission/Radiation Effects Research Foundation series of over 5,000 autopsies, we examined death certificate accuracy for 12 disease categories and assessed the effect of potential modifying factors on agreement and accuracy. The overall percentage agreement between death certificate and autopsy diagnoses was only 52.5%. Although neoplasms had the highest detection rate, almost 25% of cancers diagnosed at autopsy were nevertheless missed on death certificates. Confirmation and detection rates were above 70% for neoplasms and external causes of death only. Confirmation rates were between 50 and 70% for infectious diseases and heart and other vascular diseases. Detection rates reached a similar level for infectious, cere- brovascular, and digestive diseases. Specificity rates were above 90% for all except the cerebrovascular disease category.
Overall agreement decreased with increasing age at death and was worse for deaths occurring outside of hospital. There was some suggestion that agreement improved over time, but no indication that radiation dose, sex, city of residence, or inclusion in a biennial clinical examination program influenced agreement. Since the inaccuracy of death certificate diagnoses can have major implications for health research and planning, it is important to be aware that their accuracy is low and that it can vary widely depending on cause, age, and place of death.