HARVEST Dashboard to see a single representative figure for seven key HIV/AIDS metrics by population and location globally.
Metadata Viewer Dashboard to see metadata on the HARVEST Dashboard content along with comparative global distribution and figures for each of the three main project data sources.
You can also access data directly by reaching out to us at POP.International@census.gov.

What are the HARVEST Dashboards?

The HARVEST Dashboard project is funded by the President’s Plan for Emergency AIDS Relief (PEPFAR). Managed and overseen by the U.S. Department of State, PEPFAR has saved over 20 million lives and prevented millions of infections since 2003. PEPFAR’s response to the HIV/AIDS pandemic has become increasingly data-driven during the past decade. PEPFAR is active in many countries experiencing the highest burden from HIV/AIDS but assembling a global picture of the pandemic requires combining PEPFAR data with data from other sources. The first sources chosen are reports published by an authoritative national organization, such as the Ministry of Health or National AIDS Coordination Council (what we call “Country Direct” data). The second is the AIDSinfo database published by UNAIDS.

The backend of these Dashboards is the HARVEST database where Country Direct data are coded and entered alongside PEPFAR and UNAIDS data, which are loaded programmatically into the system. The Dashboards create a global picture for each indicator based on a relatively simple precedence order algorithm.

Why Was this Project Started?

It is hoped that the Dashboards can serve as advocacy tools for improving the coverage of observed HIV/AIDS data available to the public, and that HARVEST can serve as a data repository for researchers and policy analysts. This project was founded to answer two primary questions.

Q1 What do the data assembled directly from national governments say?

While many countries make their own data available, and regional data repositories exist, no other central repository for global HIV/AIDS data collected directly from national bodies exists. Uniquely, we seek to collect these data directly from country sources and bring it into one central location for easy and effective use in a standardized format.

Q2 Where does the data from national governments, PEPFAR, and UNAIDS agree or disagree?

To address this question, we put that country data in direct comparison with data from the two most prominent global authorities on HIV/AIDS data (PEPFAR and UNAIDS). In doing so, we seek to highlight areas where the existing data from authoritative sources agree or disagree and where each source may focus.

Methodology

Data Types

There are three types of data available in the dashboards: observed, estimated, and imputed.

Observed data are based on case reporting generated from health information reporting systems. Observed data are counts for which no attempt has been made to adjust for over- or under-reporting. While reporting from these systems may undercount the true value of some indicators, observed data are simple to understand.

Estimated data are generated from models, which combine multiple sources of observed data using mathematical relationships. Researchers use models to generate a more accurate or complete value for an indicator or when an indicator is functionally impossible to measure. An example of the latter is the number of people living with HIV, which would require testing 100% of the population. More information on the observability of each indicator is given in the Metadata Viewer Dashboard.

Imputed data are derived when a missing value unambiguously completes an otherwise complete demographic sex or age disaggregation for an indicator for a particular geography. Imputed data may also be derived when data for a complete geographic set of subnational units is present, but a national total is missing. The explicit imputation and aggregation procedure is outlined below.

Each data point in the Dashboards has a data type. The Dashboards emphasize observed values rather than estimates generated from models. When both observed and estimated data are present for a given location, metric, and population then the observed data point is displayed. Imputed values are only displayed if neither an observed nor an estimated value are available.

Data Sources

Country Direct data are reported from national authoritative sources, such as the Ministry of Health or National AIDS Coordination Council. These data are manually entered and coded into HARVEST by Census Bureau analysts. Country data may be observed or estimated based on data availability for a given indicator. Census Bureau analysts code data as observed or estimated based on either metadata published with a report or the context of the report section in which an indicator is found. For example, if case and laboratory systems are mentioned in the content of PLHIV on treatment who attain viral suppression, a third 95 coverage indicator would be marked as observed.

PEPFAR data are available to the public through Panorama Spotlight. These data are loaded directly into the HARVEST database from the Clinical Cascade – Results by Age and Sex table available from https://data.pepfar.gov/dashboards under MER Downloadable Data. PEPFAR data are reported at the site level based on paper and electronic patient records and used for program monitoring and to track progress meeting the 95-95-95 goals. Since PEPFAR data are aggregated from medical records, they are considered observed data in the Dashboard system.

UNAIDS data comes from the AIDSinfo database maintained by UNAIDS. UNAIDS produces modeled estimates annually using the best available epidemiological and programmatic data. Modelled estimates are required for some figures; it is impossible, for example, to count the exact number of people living with HIV when some people do not yet know their status. In addition, counting the number of people who are newly infected with HIV or who have died from AIDS-related causes would be logistically impossible. Country teams use UNAIDS-supported software, such as EPP/Spectrum, to develop estimates annually.

Precedence Procedure

The HARVEST Dashboard attempts to present the most complete global picture available on HIV/AIDS. Doing so requires presenting data from different sources in the same visualizations. To determine which data point is represented for each indicator-geography-population combination, a precedence procedure was applied. Precedence is determined by filtering for source type, data type, and record year—in that order—specifically that data published by authoritative national sources is preferred over PEPFAR data, which is preferred over UNAIDS data.

Within Country Direct data, observed data is preferred over estimated data, which is preferred over imputed data. Within each statistic type, newer data is preferred over older data, within the five-year window of data considered in HARVEST for each Dashboard update. If no Country Direct data are available within HARVEST, observed data are preferred over estimated data, which is preferred over imputed data. Within each data type, PEPFAR data are preferred over UNAIDS data. Finally, for each, newer data are preferred over older data.

Country Direct data are expected to be the most complete since they should include data aggregated from all healthcare delivery sites, which contrasts with PEPFAR data that are based on sites receiving PEPFAR support. UNAIDS data are nationally representative estimates, however the HARVEST Dashboard prioritizes observed data over estimated data. The prioritization is based on the simplicity of observed data to understand and the intended audience of the HARVEST Dashboard. Though as elsewhere stated, the number of people living with HIV and any indicator with PLHIV in the denominator must be adjusted even in countries with advanced surveillance and reporting systems.

The first and second filters overlap somewhat, given the nature of PEPFAR data and UNAIDS data, which are likely to be observed and modeled respectively. Essentially, this means that UNAIDS estimates are used only when necessary to present the most complete possible global picture for an indicator or because an indicator is estimated by definition.

Imputation and Aggregation Procedure

Missing count values may be imputed when the missing value unambiguously completes an otherwise complete demographic sex or age disaggregation for an indicator for a particular geography. For sex—male, female, and total values comprise a complete set*. For age—aged less than 15 years, greater than or equal to 15 years, and total values comprise a complete set. Values can be imputed for missing totals as well as the components of totals. For example, a total is imputed (through addition) if male, female, and unknown are reported as disaggregates. An unknown value is not required for a set to be considered complete but is included in calculations when present. A disaggregate can also be imputed through subtraction in the case a total is reported and the other disaggregates are available.

Where national totals are otherwise missing, national totals are imputed from subnational totals in a subset of countries where those subnational areas can be considered to cover nearly all of the national territory. Indicators reported as percentages are also not imputed.

Note that there is a relationship between the three data types in that PEPFAR and Country Direct data may be used as inputs to UNAIDS modeled estimates and that countries may in turn publish UNAIDS estimates. Even so, differences should be expected when comparing subnational components to national aggregates by indicator when moving between data sources due to the different universes of data collection between the sources—Country Direct (theoretically encompassing all service delivery sites in the country) and PEPFAR (encompassing only PEPFAR-funded sites). The size of the difference is proportional to the share of HIV/AIDS-service delivery in a country that is PEPFAR supported—as the PEPFAR share of service delivery increases, difference between Country Direct and PEPFAR should decrease. UNAIDS data are modeled, which will lead to differences with Country Direct and PEPFAR although for certain variables, UNAIDS estimates are considered the standard.

*Some countries disaggregate HIV/AIDS-related data using a non-binary sex classification that include gender identity. Such reporting is rare, so visualization is limited to Female, Male, and Unknown—the dis-aggregations available by sex in PEPFAR data systems.

Indicator Definitions

The definitions used in both Dashboards come from the PEPFAR Monitoring, Evaluation, and Reporting Indicator Reference Guide, and the U.N. Global AIDS Monitoring Guidance. Data sourced from PEPFAR or UNAIDS are automatically matched to their definitions. Data sourced from country reports are matched to an indicator definition by a U.S. Census Bureau analyst. The HARVEST data model allows for fine age disaggregation (five-year age bands) and sex categories of female, male, and unknown. To maximize comparability with publicly available PEPFAR and UNAIDS data, the Dashboards are limited to coarse age disaggregation (over or under age 15) and reporting sex using male, female, and unknown categories.

What is the Scope of Data Available in HARVEST?

At the time of the most recent release on March 1, 2023, the HARVEST database contained 209,030 data points which represent 8,731 unique combinations of location, age, sex, and metric.

Globally, this includes data for 174 national areas, 470 subnational areas, 13 metrics (7 of which are included in the Dashboards).