To analyze diabetes mellitus-related mortality among elderly and the rate of undereporting of diabetes mellitus as a cause of death when statistical data on diabetes exclusively on the underlying cause of death are considered. A total of 2.974 death certificates of elderly people living in a housing project in the city of Rio de Janeiro were revised. The study period was 1994. Of them, 291 deaths were due to diabetes mellitus, 150 as the underlying and 141 as the secondary cause of death. The proportion of deaths where diabetes was stated as the underlying cause in relation to the total of diabetes deaths was calculated globally and categorized by sex and age groups. Of the 291 deaths studied, 138 (47.4%) were men and 153 (52.6%) were women. Mortality rates showed a continuous age increase and were higher among men, though sex difference was smaller when only the underlying cause was considered. It was found a higher proportion of deaths (22%) occurring at home. Overall rate of diabetes deaths as the underlying cause was 51.5%, with higher rates seen in women. The analysis of mortality statistics based exclusively on the underlying cause of death can yield misleading profiles due to unrandomized underreporting. There is a need of further studies with diabetic elderly cohorts for a more accurate mortality analysis in this population group.