2022-04-02-世界卫生组织-Revision_of_the_2016_WHO_verbal_autopsy_instrument_report_59页_1mb

WHO 2016 Verbal Autopsy (VA) Instrument Revision Summary

1. Revision Objectives

• Reduce questionnaire length and duration by eliminating redundant or poorly performing questions.
• Improve clarity, logistics, and response validity.
• Maintain diagnostic performance for cause-of-death (COD) assignment using existing algorithms (Tariff 2.0, InterVA, InSilicoVA) and physician review.

2. Methodology

• Mixed-Method Approach:
  • Quantitative Analysis: Used global VA data (28,427 deaths from 13 low/middle-income countries). Analyzed response patterns, questionnaire validity, and question importance using:
    • TMLE (Targeted Maximum Likelihood Estimation).
    • Entropy coefficients to measure association with CODs.
  • Qualitative Analysis: Cognitive interviews in Zambia and Morocco to assess question interpretation and respondent recall.
• Expert Consultations: Workshops with VA experts, clinicians, and field teams to refine changes.

3. Key Changes

• Number of Questions: Reduced by 18% (from 40 to 33 questions).
• New Questions Added: 10 new questions for COD assignment (see Appendix D).
• Improvements:
  • Restructured sequences for injuries, maternity, and maternal sections.
  • Added hints and clarifications to 26% of questions.
  • Simplified terminology for translations.

4. Enumerated Issues & Resolutions

• Tobacco Use: Eliminated redundant questions; focus on daily use and duration.
• Convulsions: Dropped redundant questions; grouped question sequences.
• Vaccinations: Removed indirect questions; focus solely on prior immunizations.
• Unconsciousness: Streamlined sequences and dropped redundant questions for neonates/children.
• Maternal Deaths: Simplified logic for late maternal death capture.

5. Data Support

• Global Sample: 28,427 deaths (50% female, 77% adult, 13% child, 10% neonate).
• Validation: Altered items expected to retain algorithm performance while improving efficiency.

6. Future Directions

• Release a beta version for field testing (April 2022).
• Adapt software for algorithm compatibility with revised tool.
• Continuous monitoring to assess impact of removed/modified questions.

Conclusion

The 2022 VA instrument balances parsimony and reliability. Redundancies eliminated and clarity improved aim to enhance COD assignment without biasing outcomes.