Assessing HIV testing data from healthcare facilities in Uganda

Joseph Ouma is a bioinformatician and SSACAB PhD Fellow based at the University of Witwatersrand, Johannesburg, in South Africa. SSACAB (the Sub-Saharan African Consortium for Advanced Biostatistical Training (SSACAB) is one of the eleven Developing Excellence, Leadership, and Training in Science in Africa (DELTAS Africa) programmes. DELTAS Africa funds Africa-based scientists to amplify the development of world-class research and scientific leadership on the continent while strengthening African institutions. DELTAS Africa is implemented through the AESA Platform. AESA (The Alliance for Accelerating Excellence in Science in Africa) is a funding, agenda-setting, programme management initiative of  the a partnership of the African Academy of Sciences (AAS), the African Union Development Agency (AUDA-NEPAD), founding and funding global partners, and through a resolution of the summit of African Union Heads of Governments. DELTAS Africa is supported by Wellcome and the United Kingdom Foreign, Commonwealth and Development Office (FCDO formerly DFID).

Summary

The study seeks to obtain more accurate HIV prevalence estimates for districts in Uganda using available data sources with no additional data collection costs. The purpose is to increase the precision of HIV prevalence estimates in an environment that is decentralized and where districts are empowered to make service delivery decisions.  Combining health facility and population survey data leads to a 28% improvement in the precision of the estimate, hence more reliable information is available for decision making at the district level.

Background

In Uganda, HIV prevalence nationally is 6.2%; an estimated 1.2 million people are living with HIV. Population surveys show that HIV distribution varies from 3.1% to 8.0% among regions. This large variation in Uganda and other countries in Africa, where HIV is endemic and resources are limited, suggests that interventions cannot be limited to high risk areas, especially because in smaller governance areas, data is scarce.

Population-based surveys and health facility testing data are commonly used to obtain indicator estimates to monitor the epidemic. However, when estimates are required at lower geographical levels such as districts, the utility of such efforts has limitations. Population-based surveys are designed to provide accurate estimates at first (national) or second (regional) governance levels. The survey sample sizes at the third (e.g. district) or lower governance level are small and therefore produce less precise estimates. Alternative health facility testing data is biased because it reflects only individuals who seek care, and therefore cannot be generalized to the general population.

To overcome these limitations, health facility and survey data can be combined, complementing each other to obtain more accurate prevalence estimates for decentralized geographic populations.  These combined approaches, such as Hybrid Prevalence Estimation (HPE), can be applied if the compatibility/comparability of the data sources is known.

Spatial methods, such as kernel density implemented in PrevR package (PrevR is a statistical function/model implemented in the R statistical software and it is used for HIV prevalence estimation), have been used to derive estimates at lower administrative levels. These models, however, assume continuous or similar prevalence across areas, which does not reflect real settings. Estimates from PrevR and other spatial-based approaches also use only data from the target geographical area, which often have small survey sample sizes, leading to less reliable estimates for decision making.

This study computed and compared HIV prevalence among individuals surveyed in the 2011 Uganda AIDS indicator survey. The population was stratified into those who reported to have visited a health facility and were tested for HIV there, and those receiving HIV testing in the community. The study also computed the propensity of testing for HIV in a health facility, information which was used to combine the population survey and health facility testing dataset to obtain district level HIV prevalence estimates for all districts in Uganda.

Description of study

The study seeks to obtain more accurate HIV prevalence estimates for districts in Uganda using available data sources with no additional data collection costs. The purpose is to increase the precision of HIV prevalence estimates in an environment that is decentralized and where districts are empowered to make service delivery decisions.

The study analyzed population survey data from the Uganda AIDS Indicator Survey (2011) to determine the proportion of people tested for HIV, and, of those, the proportion tested in a health facility. The computed HIV prevalence ratio for those tested in a health facility to those tested in a community setting used Katz methodology for each of the demographic subpopulation groups. We assessed the factors associated with testing for HIV in a health facility and those associated with testing for HIV in a community setting.

Outcomes of the study

A two-fold increase in HIV prevalence was found among facility testers compared to non-facility testers. The difference was even greater if respondents were male, aged 15-19 years or 40-49 years, never married and/or had no sexual partners in the two months preceding the population survey. Prevalence of testing in a health facility may have been lower for these population subgroups due to their poor health-seeking behavior.

Prevalence of facility testing was higher among females, married/cohabiting individuals and those who have only one sexual partner, although HIV prevalence within this population subgroup was lower.

Although HIV positivity has been found to be higher among females compared to males in several studies, distinguishing by testing venue found that females test at health facilities are less likely to test positive compared to males, while females who test in a community setting have a higher positivity rate compared to males tested in the same setting.

Lessons

While HIV prevalence is higher among those who are tested at health facilities compared to those who do not test at health facilities, it is higher yet in specific population subgroups accessing health facilities for health care. These subgroups also have lower HIV testing rates. It is therefore clear that although prevalence among individuals who access health facilities for health care due to ill health, antenatal or ill health among family members is higher, there are HIV prevalence variations among those accessing health facilities. This information can be incorporated into HIV screening algorithms in clinics.

It is also important to continually review health facility testing data as well as population survey datasets to ensure that information upon which service delivery decisions are made is current. Using health facility testing data alone may not be consistent with the actual distribution of HIV in the general population.

Factors associated with HIV positivity for those tested in a health facility are different from factors associated with HIV positivity among those tested in a community setting. The likelihood of HIV positivity is lower for facility testers if they are female, residents of rural areas, have a secondary level of education and/or are aged 15-19 years. These factors are not associated with HIV positivity for those tested in a community setting.

Impact

• HIV prevalence information from health facility data cannot be used independently to explain the status of the HIV epidemic in the general population. Population survey data are more informative, although these have to be used in combination with health facility testing data if precise estimates are needed at decentralized levels.
• Understanding of the variations in prevalence by population subgroup is important for the implementation of appropriate interventions even among individuals accessing HIV testing in the same setting. While health facility data shows higher prevalence compared to community level testing data, specific population subgroups require emphasis on HIV screening to minimize missed HIV diagnoses at health facilities.
• Combining health facility and population survey data leads to a 28% improvement in the precision of the estimate, hence more reliable information is available for decision making at the district level.