The case for verbal autopsy in health systems strengthening.

The quest for robust, aff ordable, consistent methods for fi lling global health information gaps is familiar. But frequently the focus is on how the international community can extract population-based data, rather than how countries and communities can use robust methods to understand and act on their own health and disease burdens. This week, researchers are convening in Vancouver, Canada, for the 4th Global Symposium on Health Systems Research organised by Health Systems Global. At the same time, the INDEPTH Network (the International Network for the Demographic Evaluation of Populations and Their Health) is meeting in Kampala, Uganda. INDEPTH and Health Systems Global are distinctive groups committed to research for action. This commitment is surprisingly rare. Health researchers tend to avoid the worlds of policy and politics, prioritising technical sophistication over conceptual depth, and emphasising scientifi c productivity but often with limited attention to uptake among non-academics and local decision makers. Health systems research has emerged in response, as a fi eld working with and for those who organise and deliver services, rather than in a process that is separate to it. In this context, we refl ect on verbal autopsy (VA) as a means to connect data and action, as well as contributing to national and global knowledge. VA is a pragmatic approach used in many low-income and middle-income countries for determining causespecifi c mortality levels where registration of deaths is otherwise incomplete. VA uses a structured interview with fi nal caregivers on the medical signs and symptoms of the deceased prior to death, which are interpreted to assign probable medical causes. WHO has led an international process of standardising interview protocols and VA cause-of-death categories based on the International Classifi cation of Diseases to facilitate cross-national analyses, and automated models such as InterVA and Smart VA off er pragmatic approaches to processing VA data that are cheap, consistent, and amenable to widespread application by operators with school-level skills. VA has also been adapted for use on mobile devices (fi gure), off ering considerable operational effi ciencies. At the national level, following widespread use in research environments, VA has become a recognised source of mortality data where other methods are not used. The two most promising avenues to scale up VA are in enhancing civil registration and vital statistics (CRVS) systems and sample registration systems, as highlighted in The Lancet Series Who Counts and Counting Births and Deaths. Several African and Asian countries have included VA in sample registration systems, generating useful information on cause of death patterns, as well as on mortality levels and trends. Whether and how VA contributes to national CRVS systems is ultimately a matter for countries themselves. For now, WHO VA standards, InterVA, and SmartVA off er eff ective approaches to tracking causes of death that are amenable to cost-eff ective scale-up, off ering new opportunities for evidence-based services. Accumulating methodological and substantive evidence bases will be key to understanding the potentials and obligations of VA for mortality registration within and beyond the health system. By virtue of the deaths it investigates, VA is also an opportunity to better understand social exclusion from access to health systems. Social autopsy (SA) is a related technique that seeks to understand the detail of how and why deaths occur in particular contexts. SAs collect and analyse information on the circumstances of mortality to identify social and health systems failures related to deaths. Alongside medical causes of death, categorising the circumstances of mortality at scale is important for understanding mortality.