Background
Major changes in health policy, health service delivery, specific protocols, guidelines and recommendations for the management of common causes of maternal death have been developed in South Africa since the advent of the current democratic government. However, maternal mortality ratio remains high. The scientific community has conducted numerous studies on maternal mortality in South Africa; save for an analysis of the causes of maternal deaths, stratified by province. This study examines the geographic distribution of maternal causes of death in South Africa.

Methods
A pooled cross-sectional dataset for the years 2002–2006 retrieved from the vital registration database of Statistics South Africa was used to analyse maternal causes of death. About 8773 maternal deaths between 10–55 years were analysed using frequency tables, cross-tabulations and logistic regression. Maternal mortality ratios (MMR), odds ratios (OR) and 95% confidence intervals (CI) were used to analyse provincial disparities.

Results
MMR was highest in the Free State (286/100 000) and lowest in the Western Cape (87/100 000). Tuberculosis (10.4%) was the leading single indirect cause of maternal deaths while hypertensive disorders (9.1%) were the leading direct cause of death. KwaZulu-Natal women had a significantly higher risk of dying from sepsis (aOR=3.1,95% CI=1.2-7.9). North West women had the lowest risk of dying from hypertensive disorders (aOR=0.4,95% CI=0.2-0.7). The risk of dying from complications of labour was lowest for Gauteng women (aOR=0.4,95% CI=0.1-0.9). The 30–34 years age group had a significantly high risk (aOR=2.5,95% CI=1.6-4.0) of dying from abortion while the 25–29 years age group had a significantly higher risk of dying from maternal infectious diseases (aOR=2.3,95% CI=1.3-3.9). The 40–44 years age group had a significantly higher risk of dying from haemorrhage (aOR=2.3,95% CI=1.3-3.9 and the 45+ age group from other maternal diseases (aOR=3.3,95% CI=1.2-8.5) and miscellaneous direct causes (aOR=4.1,95% CI=1.7-9.9) respectively.

Conclusions
The study shows great variations in the distribution and causes of maternal deaths by age and provincial level. Poorer provinces had lower MMR than the better off provinces. The provincial variations in the leading causes of death indicate the importance of targeted interventions at sub-national level.