Does sampling at service sites (markets) produce valid estimates of maternal mortality? Examining selection bias in a new approach to sampling using the sisterhood method
Jacqueline S Bell, University of Aberdeen
Ouedraogo Moctar, IMMPACT Project (Centre Muraz)
Ann E Fitzmaurice, University of Aberdeen
Nicolas Meda, Centre Muraz
Wendy J. Graham, University of Aberdeen
Sampling at Service Sites (Markets) is a new approach that uses non-probability sampling and the sisterhood method to estimate maternal mortality. Here we report on experience in Burkina Faso, where the representativeness of the sample was explored by comparing demographic and socio-economic characteristics of the respondents with those of women measured in a concurrent census in the same area, and these were found to be similar. The census data provided an opportunity to determine which respondent characteristics are most closely associated with maternal mortality, and therefore most relevant in assessing how representative the sample was for estimating mortality indicators. Although household maternal deaths and sisterhood maternal deaths were found to be associated with different sets of characteristics, the two estimates of MMR obtained in the census (for respondents’ households and for sisters) were similar. These findings confirm the usefulness of this approach for resource poor countries, with limited information systems.