Background: Unacceptably high levels of maternal deaths still occur in many sub-Saharan countries and the health systems may not favour effective use of lessons from maternal death reviews to improve maternal survival. We report results from the analysis of data from maternal death reviews at Bugando Medical Centre north-western Tanzania in the period 2008–2012 and highlight the process, challenges and how the analysis provided a better understanding of maternal deaths.

Methodology: Retrospective analysis using maternal death review data and extraction of missing information from patients’ files. Analysis was done in STATA statistical package into frequencies and means ± SD and median with 95 % CI for categorical and numerical data respectively.

Results: There were 80 deaths; mean age of the deceased 27.1 ± 6.2 years and a median hospital stay of 11.0 days [95 % CI 11.0–15.3]. Most deaths were from direct obstetric causes (90); 60 % from eclampsia, severe pre-eclampsia, sepsis, abortion and anaesthetic complications. Information on ANC attendance was recorded in 36.2 % of the forms and gestation age of the pregnancy resulting into the death in 23.8 %. Sixty one deaths (76.3 %) occurred after delivery. The mode of delivery, place of delivery and delivery assistant were recorded in 44 (72.1), 38 (62.3) and 23 (37.7 %) respectively.

Conclusion: Routine maternal death reviews in this setting do not involve comprehensive documentation of all relevant information, including actions taken to address some identified systemic weaknesses. Periodic analysis of available data may allow better understanding of vital information to improve the quality of maternity care.