Improved Methods for Estimating HIV Prevalence and Incidence

UNAIDS Spectrum/EPP 2014 tool to estimate epidemic trends.

Urban Uganda using the UNAIDS Spectrum/EPP 2014 software to estimate epidemic trends.

    
The international response to HIV relies on reliable estimates of the size of the epidemic and recent trends. These estimates are used for international resource mobilisation, national programme planning, evaluating the effectiveness of existing programmes, and policy planning and resource allocation. Current approaches to generating national HIV epidemic estimates in SSA use time series of HIV prevalence among pregnant women attending antenatal clinics (ANC) and prevalence among adults 15–49 y in nationally representative household surveys to estimate national HIV epidemic prevalence. Natural HIV progression and numbers on antiretroviral treatment (ART) are used to infer HIV incidence from this prevalence curve. In this work package we aim to understand if incorporating additional information into this process affects the estimates of HIV prevalence and incidence.

Background

The methods used to derive these estimates are continually being improved as new data become available. Many countries in in sub-Saharan Africa now have two or more nationally representative household surveys. Incorporating age-specific HIV prevalence data from successive surveys offer the potential to improve estimates of age-specific HIV incidence patterns. In South Africa, modelling has illustrated that incorporating age-specific HIV prevalence and all-cause mortality information into model estimates generated different historical HIV incidence trends, and may lead to better projections of future epidemic trends. Moreover, assuming that HIV prevalence trends among pregnant women are representative of all adults may result in biased estimates as the age-distribution of HIV shifts to older ages in the ART era.

This HIV Modelling Consortium work package focuses on improving the accuracy, precision, and quantification of uncertainty for national HIV epidemic estimates in sub-Saharan Africa.

Aims

To test whether incorporating additional information affects estimates of HIV prevalence and incidence in sub-Saharan Africa. Four issues will be addressed:
(1) calibrating to age-specific HIV prevalence data
(2) incorporating uncertainty about the natural HIV progression and survival
(3) including all-cause mortality information
(4) accounting for biases in prevalence trends among pregnant women.

Approach

Three or four focus countries in southern and eastern Africa will be selected to develop and test methods addressing the four topics described above. Appropriate country surveillance data will be gathered from countries and other resources (e.g. UNAIDS). Mathematical and statistical modelling to develop, test, and apply new methods will be conducted by HIV Modelling Consortium partners at Imperial College London and the Harvard School of Public Health.