Follow-up surveys for census estimates of maternal mortality: experiences from Bolivia and Mozambique

Maternal mortality measurement through special census questions will be a common practice in the 2010 census round. To check this information or make it cause-specific, some countries have experimented with follow-up surveys containing verbal autopsies or triangulation with administrative data. However, follow-up studies can be costly and not without complications. In order to assess the benefits, two such experiences are discussed in detail (Bolivia, 2002; and Mozambique 2007–2008) and two others mentioned more briefly (Islamic Republic of Iran, 1996; and Democratic People’s Republic of Korea, 2008). In the former, several problems were apparent. In Mozambique, the follow-up survey used a cluster sample of 4.5% of deaths, from all causes. This design was adequate for the more common causes, but not for maternal mortality. Another problem was the large proportion of invalid cases (35.1%, plus 16% not verifiable) and the likelihood that there was a similar proportion of omitted deaths. The Bolivian census generated many invalid cases and missing ages, due in part to the flawed design of the questionnaire. This overburdened the follow-up, so that only 15% of the census deaths of women of reproductive age unrelated to pregnancy could be investigated. Once the false positives were eliminated, the results seem consistent with Growth Balance analyses, but the many classification errors compromise confidence in the results. Despite this mixed record of outcomes, it is believed that carrying out a limited number of similar studies in the current census round may be valuable, if appropriate lessons are learned from these experiences.