The openVA Toolkit for Verbal Autopsies

Area of Science:

• Public Health
• Biostatistics
• Medical Informatics

Background:

• Verbal autopsy (VA) is crucial for determining causes of death (COD) in areas lacking vital statistics systems.
• Deaths outside medical facilities often lack official documentation, necessitating VA surveys.
• Existing VA methods have incompatible implementations, hindering systematic analysis and comparison.

Purpose of the Study:

• To introduce the openVA package, a standardized R framework for verbal autopsy data analysis.
• To provide a unified platform for applying and comparing various cause of death (COD) inference algorithms.
• To facilitate systematic analysis, model fitting, comparison, and visualization of VA data.

Main Methods:

• Development of the openVA package, an open-source R implementation of widely used VA methods.
• Support for diverse data input/output formats and customizable cause-symptom associations.
• Demonstration of a comprehensive pipeline for VA data analysis within the R environment.

Main Results:

• The openVA package offers the first standardized framework for VA data analysis.
• It ensures compatibility with all openly available VA methods and data structures.
• The package streamlines the process of model fitting, summary, comparison, and visualization.

Conclusions:

• The openVA package significantly improves the systematic analysis and comparison of verbal autopsy methods.
• It addresses the incompatibility issues of previous VA model implementations.
• This framework enhances the reliability and comparability of cause of death (COD) estimations from VA data.