Objective: To trace all maternal deaths at a tertiary East African university hospital with a systematic registration of all births.
Design: Descriptive study.
Sample: One hundred and nineteen cases of maternal death which occurred in the period from 2000 to 2007 (including).
Methods: Identification through the birth registry and separate manual tracing of all case records. Account of practical problems concerning identification of cases and analysis of time trends, mothers' domicile, occurrence by phase of pregnancy, birth and puerperium, and diagnoses.
Results: There was considerable under-reporting of deaths in the medical birth registry. Twenty of 119 mothers died before 23 weeks' gestational age, most of them of unsafe abortion. Other prevalent direct causes of death were hemorrhage, eclampsia and other hypertensive complications. HIV/AIDS was primary cause in 20 cases.
Conclusion: Even with relatively complete ascertainment of births, single hospital-based medical birth registries have limitations in studies of maternal deaths. They may identify risks among women who arrive for delivery at the hospital, but are not well suited for estimation of total maternal mortality within the hospital walls. This would require additional data. Extending the birth registry monitoring system to all health institutions with obstetrical services in a region will give more reliable estimates to be followed over time and serve as a basis for regular auditing, to the benefit of mothers and their children.