HIV Estimates Accuracy Tool

Tools for public health

29 Nov 2018

HIV remains one of the most important public health concerns in the European Union and European Economic Area. Accurate data are therefore crucial to appropriately direct and evaluate public health response.

The ECDC HIV Platform tool now provides a newer and integrated version of the HIV Modelling and HIV Estimates Accuracy Tools. The HIV Platform Tool can be accessed here: https://www.ecdc.europa.eu/en/publications-data/hiv-platform-tool

The HIV Estimates Accuracy Tool is an end-user application that uses statistical methods to calculate adjusted estimates from HIV surveillance data taking into account the issues of missing data and reporting delay. The tool accepts case-based surveillance data for HIV containing a minimum required set of variables.
Version: 1.0.0

For experienced R users, a CRAN-like repository is set up for installing the tool as an R package in R GUI or RStudio.

Go to the repository of R packages

The tool can be installed using standard R commands executed in R console:

Type the following command:

> install.packages("hivEstimatesAccuracy", repo = "http://www.nextpagesoft.net/hiv-estimates-accuracy/repo/")
and press Enter. This downloads and installs latest version of the tool and all its dependencies.
Once R is done with installing, the tool will run with the following command:

> hivEstimatesAccuracy::RunApp()
The user can update the tool with the following command:

> update.packages(repo = "http://www.nextpagesoft.net/hiv-estimates-accuracy/repo")

3. An offline Windows x64 deployment package with R environment embedded

The last option for local deployment of the tool is to use a deployment package, which includes all required software and R packages.

Go to the Windows x64 deployment package

Download the deployment package (201 MB download size).
Unpack the file to a folder.
After unpacking, a new folder will appear called ’hivEstimatesAccuracy’. Double-click on the ’hivEstimatesAccuracy.bat’ file. This will open the tool in the default web browser. Close the browser window when finished with the file.

This offline package can only run on 64-bit versions of Microsoft Windows 7, 8, and 10.

Data integrity verification

MD5 checksum: 8913cda2188a2692b0b3204710abf7de

SHA1 checksum: ef4ae437359448b9814c315391960f4b7efd9316

Prerequisites

For option 1 an active Internet connection is required. Relatively recent versions of web browsers such as Chrome, Firefox, Internet Explorer, Edge and Safari with support for JavaScript enabled are recommended.

For option 2, both R engine and Pandoc must be installed, with LaTeX preferably installed as well. Option 3 (an offline Windows x64 deployment package) includes the required software and does not require additional prerequisites.

Required software:

a) R engine performs all calculations

b) Pandoc converts Markdown documents (report sources) to HTML, LaTeX and/or Word. It is delivered with RStudio, so if one is already using RStudio, there is no need to install it separately.

Optional software:

c) LaTex has various alternatives, including Miktex, TexLive and TinyTex that generates PDF reports. If it is not installed, outputting the main report to PDF will fail.

Background

Missing data are a well recognised problem within surveillance systems. When values for certain variables are missing and cases with missing values are excluded from analysis, it may lead to biased and potentially less precise estimates. Reporting delay, the time from case diagnosis to notification, can lead to problems when analysing the most recent years given that information on certain cases or variables may not have been collected yet due to national reporting process characteristics. These phenomena are common in disease surveillance and also apply to HIV.

The tool performs multiple imputations for the missing values of a set of variables (age, gender and CD4 count) using joint multivariate normal models and extensions or full conditional specification (also known as multiple imputation by chained equations). Additionally, the tool allows for correction of delays in reporting through reverse time hazard estimation. The adjustments may be used separately or in combination.

The outputs include results in the form of a report containing tables and graphs, and datasets in various file formats, in which the corrections have been incorporated and are ready for further analysis.

Data Preparation

The tool accepts case-based surveillance data for HIV containing a minimum required set of variables. The support files types are rds., txt, csv, xls and xlsx (uncompressed and ZIP archives).

Several attributes/variables are required by the tool to run the adjustments. The upload file must contain all these attribute/variable names. Different variable names are accepted, but the variables must be coded in a specific way.

Manual

A complete instruction manual explains the basics of the ECDC HIV Estimates Accuracy tool and includes technical and methodological details on how to use he tool and calculate estimates. The manual can also be consulted to interpret tool outputs and aid in the selection of certain parameters.

Surveillance and monitoring