Correcting for Verbal Autopsy Misclassification Bias in Cause-Specific Mortality Estimates

Area of Science:

• Public Health
• Epidemiology
• Biostatistics

Background:

• Verbal autopsies (VAs) are crucial for determining cause of death (COD) in low- and middle-income countries.
• Computer-coded verbal autopsy (CCVA) algorithms often misclassify COD, leading to biased cause-specific mortality fraction (CSMF) estimates vital for health policy.
• Accurate mortality data is essential for effective public health interventions and resource allocation.

Approach:

• This manuscript reviews current practices and challenges in raw COD predictions from CCVA algorithms.
• It provides a comprehensive guide on applying the VA calibration approach to correct for misclassification bias.
• The study utilizes the calibratedVA software for accurate CSMF estimation.

Key Points:

• CCVA algorithms exhibit misclassification, impacting the reliability of raw COD data.
• A calibration method leveraging CCVA misclassification rates can adjust VA-based CSMF estimates.
• The calibratedVA software offers a practical tool for implementing this correction method.

Conclusions:

• The VA calibration approach effectively corrects misclassification bias in COD estimates.
• Accurate CSMFs are achievable through calibration, improving health policy decision-making.
• The study demonstrates the application of calibratedVA using real-world data from Mozambique for child and neonatal deaths.