openVA: An Introduction

Zehang 'Richard' Li
16 March 2016

WHO VA Working Group meeting

A software that provides

  • Open-source codes
    • Available on CRAN and Github
    • Complete update history
  • Flexible data input and manipulation
    • Works with WHO/InterVA, PHMRC, and customized data format
  • Multiple algorithm support
    • Currently supports three major algorithms
    • Consistent syntax to run all models in one place
    • Easy visualization and post-process

Existing software

Software Platform Access
King-Lu R package Deprecated from CRAN
  • SmartVA
  • Tariff package
  • Open-Sourced Tariff
  • Windows
  • R package
  • R scripts
  • Compiled, proprietary software
  • Open-source, on CRAN
  • Upon request from author
  • InterVA-4
  • InterVA4 package
  • Windows
  • R package
  • Compiled, proprietary software
  • Open-source, on CRAN
InSilicoVA R package Open-source, on CRAN
NBC R scripts Upon request from author

Data input

  • InterVA standard format

    • 245 items
    • ''Y'' = Yes; '' '' = No
  • PHMRC gold-standard data

    • 252 items
    • Including categorical and continuous variables
    • Require pre-processing
  • Binary spreadsheet

    • Customized set of symptoms and indicators

Data output

  • All software output:

    • Cause-of-death assignment (1st, 2nd, 3rd, …)
    • Likelihood of each COD assignment
  • Sometimes output:

    • Complete probability distribution of COD for each death
  • Never output:

    • Cause-Specific Mortality Fraction

What does openVA do

  • Standardize input data

    • Internal check of WHO format data
    • Easy download of PHMRC gold-standard data
    • Easy conversion of customized data
  • Fit models

    • Simple model fitting
    • Many customization options
  • Process output

    • Quick summary
    • Organize output for CSMF, COD assignment, etc.
    • Flexible visualization

Supported methods

  • InterVA v4.02
  • InterVA v4.03
  • InSilicoVA
  • Tariff
  • NBC (to be included)

Supported data input format

  • WHO 2012
  • PHMRC
  • Any dichotomized data
  • Need training data?
Model WHO 2012 PHMRC Customized data
InSilicoVA No Optional * Yes
InterVA No Optional * Yes
Tariff Yes Optional * Yes


*Training data is optional for PHMRC format input since the PHMRC gold-standard dataset can be used to train the model.

openVA workflow

Step 1. Read data from CSV