A comparison of three verbal autopsy methods to ascertain levels and causes of maternal deaths in Matlab, Bangladesh

Background Verbal autopsies have been widely used iu determine the levels and causes of maternal death but few studies have assessed the reliability of various methods, Methods We compared the levels and causes of maternal mortality in three data sources from Matlab, Bangladesh: (1) maternal deaths identified through a unique demographic surveillance system (DSS); (2) maternal deaths identified as a result of a previous detailed investigation into the levers and causes of maternal mortality: and (3) maternal deaths identified in the current special study. All studies used lay reporting, but differed in terms of the nature of the study, the sex of the interviewer, the format of the questionnaire and the procedure to derive the diagnosis. Results There were substantial disagreements between the routine reporting and the special studies. The DSS identified 67.2% of all deaths occurring during pregnancy or within 42 days postpartum (82.3% of direct obstetric deaths, 70.0% of deaths due to induced abortions ansi 42.4% of indirect obstetric deaths). Extending the definition of maternal deaths to 90 days postpartum increased the numbers of maternal deaths between 1987 and 1993 from 174 to 196. The two special studies also disagreed in the ascertainment of the causes of maternal deaths and yielded different cause of death distributions; the proportion of direct obstetric deaths (excluding abortion) was 50.4% in the current system compared to 44.5% previously (P = 0.001). Conclusions This study confirms the known difficulties in the ascertainment of the levels and causes of maternal mortality. The large disparities in the levels and causes of maternal mortality using three different methods of lay reporting in a population with an almost complete vital registration system add to the growing concern about the inaccuracies in the measurement of maternal mortality.