Fig. 1

Flow chart of maternal deaths included in the study. Our final analysis included 109 deaths. VA was performed for 106(92.9%) deaths and medical files of 91(83.5%) women could be traced. Piloting our approach was done based on seven maternal deaths which were later excluded from analysis. Of the 132 deaths, 10 were community deaths and no clinical records were available. The recording of one death was so minimal that no information to trace the family was available. Three facility deaths were identified in the field in which two of them were identified during visits to the community and one was reported by the district health office but not reported by the routine regional MDSR system. Out of the 8 deaths that could not be traced for VA, 4 were because the demographic information was not sufficient to trace the family in the villages. The other 4 deaths were reported in the regional MDSR data but, there was no record in facility data. It was later revealed that these were suspected maternal deaths that were reported anonymously to the region but the regional office did not follow them up to confirm whether they were maternal deaths (Fig. 1)