Demographic surveys and censuses often record pregnancy-related deaths, defined as those occurring during the maternal risk period (pregnancy, delivery, and six weeks postpartum), but do not include cause of death. This study presents a method for estimating obstetric mortality from pregnancy-related deaths data. Calculations are based on multiple-decrement life tables, and data needed are simply age-specific fertility and mortality rates that are commonly available in Demographic and Health Survey (DHS) or census data, and an estimate of the relative risk of death from nonobstetric causes during the maternal risk period. The method is tested on 59 DHS surveys from Africa. Results show that, on average, less than half of the pregnancy-related deaths are attributable to obstetric causes. This proportion varies with the level of mortality and fertility, and in particular with the prevalence of HIV in the population.