Hospital statistics of causes of death for developing countries may be biased when the utilization of hospital services is low or selective. Using Addis Ababa, Ethiopia as an example, we argue that hospital data can be useful for demonstrating general cause-specific mortality patterns. In addition, a comparison of hospital statistics with data from a surveillance of burials allows for the identification of weaknesses in health services provision. We find a low level of hospital services utilization during terminal illness. Despite similarities in the cause of death structure in the different data sources, hospital statistics under-estimate the prevalence of infectious diseases. In addition, we identify an important gender bias in the utilization of health services in the direction of males being significantly more likely to die in medical facilities than females.