Saving mother's lives: programs that work.

Maternal mortality is a complex problem requiring complex responses. Nevertheless, every intervention must operate through one of 3 pathways: preventing pregnancy, preventing complications, or preventing death when obstetric complications occur. We describe interventions following each pathway and assess their evidence base. In general, the more specific the intervention (such as procedures) the stronger the evidence. Broad interventions ("programs" for example) have a weaker evidence base although evidence is accumulating. The potential for robust evidence for effective programs is limited by epidemiologic design-the logistic difficulties of randomization, blinding, and single impact variables.