Software
Tools for processing verbal autopsy data
Several open tools interpret verbal autopsy interviews automatically. This page summarises the most widely used ones and links to their official sources — nothing is hosted or distributed from interva.net.
InterVA4 and InterVA5 (R)
Open-source R implementations of the InterVA models, maintained within the openVA project and published on CRAN. InterVA4 implements the InterVA-4 model aligned with the 2012 WHO instrument; InterVA5 implements InterVA-5 aligned with the 2016 WHO instrument.
- InterVA4 — CRAN
- InterVA5 — CRAN
- Official InterVA project pages (peterbyass.com)
InSilicoVA
A Bayesian hierarchical model for probabilistic cause-of-death assignment, published by McCormick et al. in the Journal of the American Statistical Association (2016). Unlike models with a fixed symptom–cause matrix, InSilicoVA can learn the relationships between symptoms and causes from the data themselves. Available as an R package: InSilicoVA — CRAN.
SmartVA and the Tariff method
SmartVA-Analyze, developed by the Institute for Health Metrics and Evaluation (IHME), is a desktop application that implements the Tariff 2.0 method: causes are ranked by additive "tariff" scores derived from how strongly each symptom points to each cause. The original Tariff method was published by James et al. (2011) and refined as Tariff 2.0 by Serina et al. (2015). Official sources: IHME verbal autopsy tools. An R replication of the Tariff method is available as Tariff — CRAN.
openVA
openVA is an R package that provides a single, standardised interface to the major open-source VA algorithms — InterVA4, InterVA5, InSilicoVA, Tariff, and the NBC method — with common data formats and comparison/visualisation utilities. It is described in Li et al., The R Journal (2023). The project site is openva.net.
Choosing a tool
All of these tools accept data structured around the WHO VA instruments, and all produce individual-level cause assignments that can be aggregated into population cause-specific mortality fractions. Comparisons in the literature generally find broadly similar population-level performance, with differences driven by the input data and cause lists rather than the algorithms alone. For head-to-head evidence, see the comparative studies in Research.
Frequently asked questions
Which verbal autopsy tool should I choose?
All major tools accept data structured around the WHO VA instruments and produce comparable population-level results. If you work in R, the InterVA packages or openVA are the natural starting point; SmartVA-Analyze suits users who prefer a desktop application. Comparative studies generally find differences driven more by input data and cause lists than by the algorithms themselves.
Do I need to know R to use these tools?
For InterVA4/5, InSilicoVA, Tariff, and openVA — yes, they are R packages distributed via CRAN. SmartVA-Analyze from IHME is a standalone desktop application that requires no programming.
What input data do the tools expect?
Interviews collected with the WHO verbal autopsy instruments (2012 or 2016 versions for most tools), typically exported from ODK or similar data-collection systems in the standard WHO question format. See the methodology page for the instrument versions.
Can I compare results across algorithms?
Yes. The openVA package provides a single interface to InterVA4, InterVA5, InSilicoVA, Tariff, and NBC with common data formats and comparison utilities — it is the standard way to run and compare several algorithms on the same dataset.
Does interva.net host or distribute any of this software?
No. This page is an overview only; all links point to the official sources (CRAN, IHME, openva.net, and the InterVA project pages). Always download from those sources and follow their licences and citation requirements.
Disclaimer. interva.net does not develop, host, or distribute any of this software. Names are referenced descriptively; always download tools from the official sources linked above and follow their respective licences and citation requirements.