An integrated approach to processing WHO-2016 verbal autopsy data: the InterVA-5 model

Area of Science:

• Public Health
• Biostatistics
• Epidemiology

Background:

• Verbal autopsy (VA) is crucial for determining causes of uncertified deaths, addressing significant global mortality uncertainty.
• The World Health Organization (WHO) provides international standards for VA interviews and cause categories.
• Computer models are essential for efficiently processing VA data and standardizing cause assignment.

Purpose of the Study:

• To introduce the InterVA-5 model, designed to align with the WHO-2016 verbal autopsy standard.
• To create a harmonizing model capable of processing data from WHO-2016, WHO-2012, and Tariff-2 formats.
• To generate standardized cause-specific mortality profiles for diverse populations.

Main Methods:

• Developed InterVA-5 building upon the InterVA-4 model.
• Utilized a training dataset from the Population Health Metrics Research Collaboration (PHMRC) and expert input for the knowledge base.
• Evaluated the model using test datasets from PHMRC and the Afghanistan National Mortality Survey.

Main Results:

• InterVA-5 demonstrated high concordance with hospital-assigned causes (CCC 0.92 for children, 0.86 for adults).
• The model showed excellent performance in harmonizing different input formats (CCC up to 0.97 for WHO-2016/WHO-2012).
• Evaluated datasets included approximately 75% of input items from WHO-2016, WHO-2012, and Tariff-2 formats.

Conclusions:

• InterVA-5 performs well and effectively harmonizes various input formats for cause-of-death assignment.
• The model is a valuable resource for global cause-specific mortality measurement and evaluation.
• Future re-versioning of InterVA-5 is anticipated as more WHO-2016 data becomes available.