Verbal Autopsy: Methods in Transition

Abstract

Understanding of global health and changing morbidity and mortality is limited by inadequate measurement of population health. With fewer than one-third of deaths worldwide being assigned a cause, this long-standing dearth of information, almost exclusively in the world's poorest countries, hinders understanding of population health and limits opportunities for planning, monitoring, and evaluating interventions. In the absence of routine death registration, verbal autopsy (VA) methods are used to derive probable causes of death. Much effort has been put into refining the approach for specific purposes; however, there has been a lack of harmony regarding such efforts. Subsequently, a variety of methods and principles have been developed, often focusing on a single aspect of VA, and the resulting literature provides an inconsistent picture. By reviewing methodological and conceptual issues in VA, it is evident that VA cannot be reduced to a single one-size-fits-all tool. VA must be contextualized; given the lack of "gold standards," methodological developments should not be considered in terms of absolute validity but rather in terms of consistency, comparability, and adequacy for the intended purpose. There is an urgent need for clarified thinking about the overall objectives of population-level cause-of-death measurement and harmonized efforts in empirical methodological research.

Returning home to die or leaving home to seek health care? Location of death of urban and rural residents in Burkina Faso and Senegal

“…In case of disagreement, the VA is reviewed by a third physician who assigns a consensus diagnosis. Otherwise, the cause of death is categorized as ‘indeterminate’ [30]. The most widely used automated method on INDEPTH sites is the one based on Bayes’ theorem implemented in the InterVA-4 software [31].…”

The Integrated Model for e-Health Data Collection and Sharing under Distributed Environments in Tanzania

“…The VA information is vital for public health, decision making, health sectors reviews, planning and resource allocation as well as program monitoring and evaluation [24]. Also, the cause of death statistics is useful to understand which disease kill and how many people die [24][25][26].…”

Measuring maternal, foetal and neonatal mortality: Challenges and solutions

“…(41,42) In VA, family members or caregivers (lay reporters) of the deceased are asked about the signs and symptoms occurring before the death. Symptom data from VA interviews are then interpreted by physicians or by automated methods.…”

“…Symptom data from VA interviews are then interpreted by physicians or by automated methods. (41,43) VA has some validity for causes of neonatal death in low-resources settings, however its performance is generally worse for fetal deaths. (44)(45)(46)(47) VA performs better at identifying overall maternal deaths when compared to identifying direct causes of maternal death.…”