Estimating causes of death where there is no medical certification: evolution and state of the art of verbal autopsy

Figures & data

Figure 1. Timeline of milestones in the evolution of verbal autopsy standards.

Figure 2. 2022 WHO VA Instrument revision process.

Table 1. Summary of features for the 2022 WHO VA Instrument

General Features
Instruments, software and training materials	Available for download from WHO (footnote)
Deployment versions	Paper and Tablet (ODK)
Languages	English, Arabic, French, Kiswahili, Portugese, Spanish
General identification and context indicators	44
Age specific modules	Neonatal: 0–27 completed days (Under 4 weeks)
	Child: 28 completed days to 11 years (4 weeks to 11 y
	Adult: >11 years (12 years and above)
Number of indicators^a	Neonatal 80; Child 133; Adult 179
Median time for interview^b	Neonatal 19 mins; Child 27 mins; Adult 32 mins
Health service use during the fatal illness	Included
Health care treatment & experience before death	Included
Open narrative checklist	Included
Open narrative text	Included, located at the start of the interview
Open narrative audio	Included, located at the start of the interview
Status of Civil Registration of Death	Included
Status of Medical Certificate of Death	Included
Compliance with UN Statistics^c	Yes
Batched analytics	Yes
Mapped to WHO ICD-10 & ICD-11 cause list	Yes
Mapped to IHME GBD cause list	Yes
Country applications of 2016 WHO VA in CRVS as of 2021	Bangladesh, Colombia, Ethiopia, Ghana, Kenya, Morocco, Mozambique, Rwanda, Senegal, Tanzania, Thailand, Zambia, and Zimbabwe

https://www.who.int/standards/classifications/other-classifications/verbal-autopsy-standards-ascertaining-and-attributing-causes-of-death-tool

^a.Indicators managed by skip patterns. Categories overlap and are not mutually exclusive.

^b.Based on the WHO 2016 instrument. Source: Mishra, V. (2017). Verbal Autopsy: Comparative analysis of three verbal autopsy algorithms with the WHO 2016 verbal autopsy questionnaire. MSc. Thesis, SwissTPH, University of Basel.

^c.UN Fundamental Principles of Official Statistics compliance requires the use of software that provides open exchange of data and data processing techniques.

Table 2. Target causes of death for the 2022 WHO VA Instrument

Stillbirths	2 causes
Neonatal	7 causes
Maternal	12 causes
Communicable	17 causes
Non-communicable	22 causes
External (Injury)	11 causes
Total	71 with 64 discrete causes (overlapping categories)

Table 3. Comparison of specific features of diagnostic algorithms

Features	InterVA5	InSilicoVA	SmartVA/Tariff
Computing platform compatibility	Windows MacOs Linux	Windows MacOs Linux	Windows only
Number of indicators used by algorithm	304	304	211
Exact implementation/replication in openVA^a	Yes	Yes	No
Implementation without training dataset	Yes	Yes	No
Can produce instantaneous results for a single death	Yes	No	Yes
Only significant symptoms used at individual level	No	No	Yes
Accounts for absence of symptoms	No	Yes	No
Accounts for missing symptoms	No	Yes	No
Provides distribution of probabilities for each cause for a single death	Yes	Yes	No
Provides measure of uncertainty for individual cause assignments	No	Yes	No
Direct estimation of cause-specific mortality fractions	No	Yes	No
Provides a distribution of probabilities for each CSMF	No	Yes	No
Provides uncertainty measure for cause-specific mortality fractions	No	Yes	No

^aSource: Samuel J. Clark, openVA development team. www.openva.net.

Figure 3. Map of applications of WHO Verbal Autopsy in research surveys, demographic surveillance (circles) and in national CRVS systems (country shading).