Table 1.

Evidence for meeting GATHER checklist for global health estimates.

GATHER checklist item (abbreviated)	Evidence of compliance	Status
Define indicators, population and time period of estimates	The estimates and their metadata are available from http://aidsinfo.unaids.org under the sections ‘People living with HIV’, ‘New HIV infections’, 'AIDS-related deaths', 'Treatment' and 'Elimination of mother to child transmission'.	Met
Funding sources	UNAIDS; US Government; Global Fund for AIDS, Malaria and Tuberculosis; Bill and Melinda Gates Foundation	Met
How data inputs were accessed	Data are compiled by country HIV estimates teams from program records and surveillance systems	Met
Inclusion and exclusion criteria	All survey and surveillance data are used in each country unless there are known data quality issues, such as, very short time series (surveillance data), or response rates are so low that findings are likely to be biased (national surveys). All available program data are used (number of ART and PMTCT patients) after validation by country teams and global partners	More detail required than can be provided here
Data sources and references, diagnostic methods, sample size	Prevalence and incidence trends are based on surveillance and survey data. Surveillance data include HIV prevalence measured at ante-natal clinics as well as surveillance conducted among key populations (sex workers, clients, men who have sex with men, people who inject drugs) and any additional groups relevant to a country’s epidemic for which data are available. Sample sizes are typically 300–500 per site. National surveys of HIV prevalence are available for some countries and usually include sample sizes of 5000–45,000. The sample sizes for each survey or study are included in the software. Testing is conducted according to standard HIV testing protocols with confirmatory tests for HIV positives	Met
Identify and describe any categories of input data that have potentially important biases	Numbers of people receiving antiretroviral medicines or therapy could be undercounted if not all sites have reported by the time of estimates development or might be double-counted if de-duplication efforts are not feasible or not conducted. Prevalence trends among pregnant women are not necessarily representative of the total population. Mortality data used to calibrate incidence curves might be under-reported due to stigma related to AIDS deaths. Survey measures of prevalence may be biased if refusal rates are high or if testing algorithms were not accurately implemented	More detail required than can be provided here
Additional data sources	Parameters for generalized epidemics are based primarily on demographic surveillance sites in eastern and southern Africa. Survival on ART is based on data from IeDEA consortium sites which might not be representative of all public providers	More detail required than can be provided here
All data inputs are available	Complete files with all input data can be downloaded from unaids.org. For those country files not publicly available, country team contact details can be requested from UNAIDS	Met
Conceptual overview of the methods	See Spectrum Quickstart guide and Spectrum manual at www.avenirhealth.org. Models are also fully described in published articles: [29–31,39–42,44,71]	Met
Description of the steps	Quickstart guide describing steps to create the estimate files is available from www.avenirhealth.org	Met
How were candidate models evaluated	See meeting reports at www.epidem.org.http://www.epidem.org/sites/default/files/reports/TechnicalRefinementsforSpectrum%202013.pdf	Met
Results of model evaluations and sensitivity analyses	See meeting reports at www.epidem.org.http://www.epidem.org/sites/default/files/reports/TechnicalRefinementsforSpectrum%202013.pdf	Met
Methods for calculating uncertainty	The following articles describe the methods for calculating uncertainty: [30,39]	Met
How to access code	Code can be requested from Avenir Health and East–West Center.	Being documented
Access to results	See http://aidsinfo.unaids.org	Met
Access to uncertainty results	See http://aidsinfo.unaids.org	Met
Interpretation of results in addition to other evidence and changes	See latest UNAIDS publications at www.unaids.org	Met
Limitations of the estimates	As of mid-2016 a selection of the limitations include: Incidence by age and sex is based on assumptions from cohort studies and might not reflect the situation in individual countries. For concentrated epidemics estimates of sizes of key populations may not match the population in which surveillance is conducted, making it difficult to estimate the total population of people living with HIV. In generalized epidemics information on prevalence among children is only available for a few countries against which to validate and compare the modelled estimates. Information on distribution of ART by age and sex is limited in most countries. Fertility patterns among women living with HIV (especially in concentrated epidemics) are not available, reducing the ability to accurately estimate children exposed to and potentially infected with HIV	More detail required than can be provided here