Abstract

BackgroundVerbal autopsies (VAs) have been widely used to determine cause of death for research in developing countries, but their use in vital registration systems has been limited due to the interview time of VA instruments (VAIs). This study aimed to create a shortened VAI to address this issue by using the Population Health Metrics Research Consortium (PHMRC) gold standard VA validation dataset and the Tariff method to systematically reduce the size of the PHMRC VAI. MethodsThe Tariff method calculates a score for each symptom–cause pair to measure strength of association between the two. The effect on performance after low-ranking symptoms were deleted from the VAI was measured using cause-specific mortality fraction (CSMF) accuracy and chance-corrected concordance (CCC) as test statistics at population and individual levels, respectively. FindingsWith negligible effect on performance, we were able to reduce the number of items from 183 to 105 in the adult VAI, from 127 to 72 in the child VAI, and from 149 to 76 in the neonatal VAI. CSMF accuracy in adults decreased from 0·770 (95% CI 0·766–0·775) in the full VAI to 0·770 (0·765–0·775) in the reduced VAI, in children from 0·785 (0·776–0·789) to 0·783 (0·776–0·787), and in neonates from 0·827 (0·818–0·835) to 0·818 (0·812–0·827). Mean CCC in adults decreased from 50·5% (95% CI 50·3–50·6) to 49·7% (49·5–49·8), in children from 52·3% (51·5–52·5) to 51·9% (51·6–52·2), and in neonates from 48·2% (47·9–48·5) to 47·5% (47·2–47·7). InterpretationWe have been able to reduce the number of items by over 40% in the PHMRC VAI while retaining its performance characteristics, thus removing a major impediment to its routine use in vital registration systems. This shortened VAI is currently being piloted to determine the effects of reduced item number on survey time. FundingThe Health Information Systems Knowledge Hub of the University of Queensland, a strategic partnership initiative of the Australian Agency for International Development.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call