BACKGROUND: Despite global health efforts to reduce maternal mortality, rates continue to be unacceptably high in large parts of the world. Feasible, acceptable, and accurate postmortem sampling methods could provide the necessary evidence to improve the understanding of the real causes of maternal mortality, guiding the design of interventions to reduce this burden. METHODS AND FINDINGS: The validity of a minimally invasive autopsy (MIA) method in determining the cause of death was assessed in an observational study in 57 maternal deaths by comparing the results of the MIA with those of the gold standard (complete diagnostic autopsy [CDA], which includes any available clinical information). Concordance between the MIA and the gold standard diagnostic categories was assessed by the kappa statistic, and the sensitivity, specificity, positive and negative predictive values and their 95% confidence intervals (95% CI) to identify the categories of diagnoses were estimated. The main limitation of the study is that both the MIA and the CDA include some degree of subjective interpretation in the attribution of cause of death. A cause of death was identified in the CDA in 98% (56/57) of cases, with indirect obstetric conditions accounting for 32 (56%) deaths and direct obstetric complications for 24 (42%) deaths. Nonobstetric infectious diseases (22/32, 69%) and obstetric hemorrhage (13/24, 54%) were the most common causes of death among indirect and direct obstetric conditions, respectively. Thirty-six (63%) women were HIV positive, and HIV-related conditions accounted for 16 (28%) of all deaths. Cerebral malaria caused 4 (7%) deaths. The MIA identified a cause of death in 86% of women. The overall concordance of the MIA with the CDA was moderate (kappa = 0.48, 95% CI: 0.31-0.66). Both methods agreed in 68% of the diagnostic categories and the agreement was higher for indirect (91%) than for direct obstetric causes (38%). All HIV infections and cerebral malaria cases were identified in the MIA. The main limitation of the technique is its relatively low performance for identifying obstetric causes of death in the absence of clinical information. CONCLUSIONS: The MIA procedure could be a valuable tool to determine the causes of maternal death, especially for indirect obstetric conditions, most of which are infectious diseases. The information provided by the MIA could help to prioritize interventions to reduce maternal mortality and to monitor progress towards achieving global health targets.