The Best of All Worlds: Assigning Causes of Death to Verbal Autopsy Data Using a Harmonized Medley of Algorithms

Samuel J. Clark, The Ohio State University
Jason R. Thomas, The Ohio State University
Richard Li, Yale University
Tyler McCormick, University of Washington, Seattle

The analysis of verbal autopsy data is complicated by the need to select an algorithm or reconcile different cause assignments made by different algorithms. We propose a harmonized method that uses three different methods -- Tariff, InterVA, and InSilico -- within a single algorithm, allowing each method to contribute to the estimation. We will compare the results using both simulated and observed data, and use standard methods (such as the chance-corrected CSMF accuracy) to assess the relative performance of the different methods.