To the Editor.— Although the accuracy of death certificates has been questioned, 1-3 they continue to be the primary source of mortality statistics. Death certificates may seriously underestimate the prevalence of certain diseases. 4 This may be especially true for alcoholism. 1-5 In an ongoing study of alcoholism, mortality was confirmed in 246 of 1,289 (19%) known alcoholics five to eight years after treatment of alcoholism or its complications. Autopsies were performed on 102 patients (43.6%). Two raters compared the death certificates and the available autopsies of 50 of these patients to determine the validity of death certificate information, with special regard to alcohol-related diseases or conditions. Using the autopsy as the standard for determining the cause of death, both raters agreed that only 60% of the death certificates gave an accurate general epidemiological description of the patient's death. In the 40% of certificates, where discrepancies occurred between the two