Development of a verbal autopsy tool for investigating cause of death: the Kanchanaburi project

Abstract

The study described the development of a verbal autopsy (VA) tool, in which a relative or caretaker of a descendent was interviewed about symptoms or events preceding the death, to identify the underlying cause of death (COD). The tool applied in this study transforms detailed VA questionnaires via an algorithm into a probable COD. To test the validity of the tool, preliminary matching was carried out between COD from two independent sources, from expert review of hospital records and from the VA tool algorithm. Validity testing was carried out in 2003 by examining 921 cases of hospital deaths in Kanchanaburi province occurring in 2002. After excluding VA interview non-response and insufficient medical record cases, 652 COD cases were compared. The test indicated a tolerable degree of validity and suggested revisions that helped to improve the questionnaire. The revised version has been coded as a computer program to substitute for the voluminous paper questionnaire in hand-held computers for convenient use.