Research question
To be able to implement effective interventions a greater understanding of the complex multi-level nature of the HIV epidemic is necessary. The project uses a multi-level framework to better understand the causal pathways of HIV infection (both in terms of acquisition and transmission of infection) in a rural by identifying and quantifying important environmental, community, household and individual-level determinants of HIV incidence and prevalence. Particular emphasis is placed at the level of the local community which is seen as being vital both to understanding the spread of the epidemic and to effective prevention efforts.
The project has the following objectives:
1. To understand causal pathways of infection by identifying and quantifying community-, household- and individual-level determinants of HIV acquisition in a rural South African population
2. To measure the spatial variation in HIV prevalence and incidence and quantify the impact of these variations on risk of acquisition of infection
3. To understand mechanisms of epidemic spread by identifying and quantifying key community-level determinants of HIV incidence and prevalence, including measuring the impact of access to and utilization of ART services on community level HIV prevalence and incidence, and identifying policy-relevant predictors of HIV prevalence and incidence
Fit of the research question within the Africa Centre research strategy
The project is an integral part of the HIV epidemiology research agenda (HIV dynamics and population impact) at the Africa Centre; it is co-funded by the NIH/NICHD (1R01-HD058482).
Data sources and methods
We use data from all of the sources available at the Africa Centre (HIV surveillance, demographic surveillance, socioeconomic household survey, geographical information system, antiretroviral treatment programme). Our analytical approaches utilize all of the advantages of data collected by the Africa Centre, in particular the location of HIV seroconversion events both in time and space and associations of seroconversions with demographic, socioeconomic, behavioural, and treatment factors. We use survival analysis and a multitude of regression techniques in conjunction with selection models and multi-level methods.
Findings
1. To investigate causal pathways of HIV infection, it is important to understand (and, if necessary control for) time trends in HIV incidence. We used data (563 new HIV infections observed in 16,256 person-years) from the population-based longitudinal HIV surveillance at the Africa Centre for Health and Population studies to test whether HIV incidence in the study community changed from 2003 through 2007. We find that HIV incidence has neither increased nor increased but remained constant over the time period at the high level of 3.4 per 100 person-years (95% confidence interval 3.1-3.7).
We used data from a questionnaire on sexual partnership patterns conducted in 2003/4 from 2,699 males to construct community-specific measures of partnership concurrency and average numbers of lifetime partners. We then used a parametric Weibull regression model to investigate the impact of community partnership patterns on an individual’s hazard of acquisition of HIV infection between 2005 and 2009 (after controlling for multiple individual risk factors). Whilst numbers of lifetime partners in a community strongly influences risk of acquisition of infection(p=0.03), our results showed no evidence that living in a community with high levels of partnership concurrency had any impact on an individual’s risk of acquiring HIV (p>0.5).
2. We applied for the first time two complimentary spatial analytical techniques (kernel smoothing and spatial cluster detection using the Kulldorff spatial scan statistic) to investigate the geographical patterns and clustering of HIV infections at a truly local level. The results reveal considerable geographical variation in local HIV prevalence (range = 6 – 36%) within this relatively homogenous population and provide clear empirical evidence for the localised clustering of HIV infections. Three high-risk, overlapping spatial clusters (RR=1.34-1.62) were identified by the Kulldorff statistic along the National Road (p≤0.01), whilst three low-risk clusters (RR=0.2-0.38) were identified elsewhere in the study area (p≤0.017).
3. To inform the placement and design of HIV prevention interventions and ART access, we analyzed the distributions of all people who HIV-seroconverted from 2003 through 2008. We found that more than 60% of all seroconversions occur in people below the age of 25 and more than 80% occur before the age of 30. 50% of seroconversions occur in people who are neither in school nor employed. The HIV epidemic is clustered in households with an HIV prevalence in households with at least one HIV-positive member of more than 40%, vs. an overall HIV prevalence in the community of approximately 20%. Half of all seroconverters accessed voluntary counselling and testing (VCT) before seroconversion. We further find that 32% of recently HIV-infected, 30% of non-recently HIV-infected, and 39% of HIV-negative individuals knew of ART. Of the recently infected individuals who had heard of ART, 43% did not know where to access ART, but 32% stated that they “personally know somebody who takes ART”. HIV prevalence in homesteads of recent seroconverters was 40%, about twice the HIV prevalence in the general population. In 57% of homesteads of recently HIV-infected individuals there was at least one other HIV-positive person. However, individuals knowledge of ART was not significantly associated with the number of HIV-positive members of their homesteads (p=0.759).
One of the key objectives within objective 3 of the grant is to attempt to quantify the impact of ART on epidemic dynamics. To achieve this we will use a GIS methodology to delineate communities with high and low levels of ART coverage respectively (given the underlying characteristics of the community and HIV prevalence). We will then use a multi-level Weibull regression to compare HIV incidence among repeat testers in high versus low ART coverage communities controlling for multiple individual characteristics. Our analysis of ART coverage has already revealed significant heterogeneity in ART coverage related in particular to distance to nearest clinic. The analysis requires the 2009 HIV sero-surveillance data to be statistically viable.
Policy implications
Our findings contribute substantially to the understanding of distal and proximate determinants of the HIV acquisition and transmission in a community in rural South Africa where HIV incidence has remained at very high levels in the recent past. The findings inform the design, placement, and targeting of prevention interventions as well as the potential to integrate HIV treatment with prevention programmes.
