Probabilistic methods for verbal autopsy interpretation: InterVA robustness in relation to variations in a priori...

Area of Science:

• Public Health
• Epidemiology
• Biostatistics

Background:

• InterVA is a probabilistic model for interpreting verbal autopsy (VA) data.
• It calculates causes of death based on reported symptoms and a priori probabilities.
• The study investigated the sensitivity of InterVA to variations in these probabilities.

Purpose of the Study:

• To assess the impact of altering a priori probabilities on InterVA's interpretation of VA data.
• To determine if changes in probabilities affect the characterization of population mortality composition.

Main Methods:

• A priori probabilities in the InterVA model were systematically modified.
• Six model variants were created by altering 25% and 50% of probabilities by 1-3 logarithmic steps.
• 1,000 VA records from South Africa were analyzed using the original and modified InterVA models.

Main Results:

• Cause-specific mortality fractions (CSMFs) remained functionally similar across all model variants.
• Even substantial variations in a priori probabilities did not lead to significantly different public health conclusions.
• The rank order of causes of death was consistent across all tested InterVA versions.

Conclusions:

• InterVA is a robust model for verbal autopsy data interpretation.
• Significant variations in a priori probabilities do not substantially alter InterVA-derived results.
• Physician-derived a priori probabilities are likely sufficient for global use of InterVA, particularly in areas lacking death certification.