Academic Commons

Theses Doctoral

Population-level HIV risk and combination implementation of HIV services

Philip, Neena M.

Background:
HIV transmission is greatly reduced when antiretroviral treatment (ART) suppresses an infected person’s HIV viral load. It is unclear, however, whether the contextual risk of incident HIV is optimally reduced by widespread individual-level suppression of HIV viral load alone or in combination with other HIV prevention services. HIV service coverage and community norms can influence risk in small area geographies; and contextual factors, like gender inequality and stigma, may foster environments conducive to HIV transmission. Yet, the relationship between places with high HIV levels and the clustering of area risk factors is unknown.

The goal of this dissertation is to learn if and how a geographically focused combination implementation strategy could reduce population-level HIV risk. Analyses explored whether small area risk profiles explain area differences in HIV. The guiding hypothesis is that in high HIV prevalence settings, low HIV service uptake in a geographically defined area increases the prevalence of high HIV viremia, leading to greater HIV transmission and incident HIV.

Methods:
A systematic review was conducted to examine the association between population-level measures of HIV viral load and incident HIV infection in generalized and concentrated epidemics. Publications were English, peer-reviewed articles published from January 1, 1995 through February 15, 2019 that explicitly defined HIV viral load and assessed outcomes of HIV recency, incidence, seroconversion, or new diagnosis. Studies sampled general or key populations through population-based surveillance registries, household-based enumeration, cluster sampling, or respondent driven sampling. Descriptive statistics summarized review findings.

The Swaziland HIV Incidence Measurement Survey (SHIMS) data were used for the remaining analyses. Using a two-stage cluster-based design, a nationally representative, household-based sample of adults, ages 18-49 years was enrolled from December 2010 to June 2011 in Eswatini. Consenting adults completed an interview and received home-based rapid HIV testing and counseling. All seropositive samples were tested for HIV viral load using the COBAS AmpliPrep/Taqman HIV-1 Test, v 2.0. Adults testing HIV-seronegative were enrolled in a prospective cohort for the direct observation of HIV seroconversion, completing an interview and home-based rapid HIV testing six months later.

Multi-level latent class modeling was performed to identify statistically significant combinations of HIV risk factors and to classify the combinations into small area risk profiles. In the cross-sectional sample, linear regression with robust standard errors assessed the correlation between area profiles and places with high levels of uncontrolled HIV infection, or HIV core areas, measured by the area prevalence of detectable virus (≥20 copies/milliliter) among HIV-positive adults and among all adults, regardless of HIV status. In the prospective cohort, generalized linear regression of longitudinal data assessed the association between area profiles and places prone to new HIV infections (i.e., HIV susceptible areas), measured by area-level HIV seroconversions.

Results:
The systematic review found an evidence base primarily of lower quality studies and inconsistent HIV viral exposure measurement. Overall findings supported a relationship between increasing levels of suppressed HIV in HIV-infected populations and fewer new infections over time. Better quality studies consistently showed higher population viremia (i.e. HIV viral quantity among all persons, regardless of HIV status) associated with HIV incidence in high prevalence populations; population viral load (i.e., HIV viral quantity among only HIV-positive persons) did not show an association with incident HIV in high prevalence, general populations and was inconsistent in key populations.

To determine whether area risk profiles can pinpoint HIV core areas, latent class modeling was used to categorize 18,172 adults into one of six HIV risk types. The risk typology, classified through unique combinations of HIV service uptake and sexual risk behaviors, conveyed an adult’s propensity for HIV transmission and/or acquisition risk. The model next identified the area-level composite prevalences of HIV risk types; estimated the three most frequent, unique composite combinations; and categorized them into area risk profiles characterizing HIV risk: low-moderate acquisition risk, moderate acquisition/transmission risk, and high acquisition/transmission risk. The high acquisition/transmission areas comprised the largest proportions of highest risk transmission and acquisition types. The prevalence of detectable viremia progressively increased from low-moderate acquisition, moderate acquisition/transmission, and high acquisition/transmission profiles [17.7%, 25.4%, and 35.1%, respectively]. When compared with low-moderate acquisition areas, the prevalence of detectable viremia was 7.4% [p<.001] higher in moderate acquisition/transmission areas and 17.1% [p<.001] higher in high acquisition/transmission areas. The prevalence of detectable viral load significantly decreased from low-moderate acquisition to moderate acquisition/transmission areas [76.6% versus 68.7%, p<.001], and was significantly higher in high acquisition/transmission areas by 7.3% [p<.001], when compared with low-moderate acquisition areas.

To determine whether area risk profiles can predict HIV susceptible areas, a total of 18,172 adults were surveyed of which 4396 [24%] had detectable viremia. 11,880 [96%; n=12,357] HIV-seronegative adults enrolled in the prospective cohort and 11,155 [94%] of them completed an endline visit. Four area profiles were identified, defined by unique patterns in prevalence of HIV viremia and of sexual risk behaviors. The proportion of HIV susceptible areas progressively increased from Profiles A, B, C, and D [14.3%, 21.8%, 24.6%, and 30.8%, respectively]. HIV susceptible areas were more than twice as likely to occur in Profile D than Profile A environments [RR 2.13, 95% confidence interval (CI) (1.13, 4.00); p=0.02]. Profile D areas had prevalences of unknown partner HIV status and detectable viremia at 28% and 24%, respectively. In contrast, Profile A areas had prevalences of only 8% with unknown HIV status and 31% with detectable viremia.

Conclusion:
This dissertation shows that geographic risk profiles can explain differences in population-level HIV outcomes. Risk factors spatially cluster in predictable, meaningful combinations that can inform an area typology of HIV risk. The co-location of adults predisposed to greater HIV risk may heighten levels of uncontrolled HIV infection, thereby creating potential area sources of ongoing transmission; however, the concurrent levels of other risk factors may have more influence in reducing population-level incidence than previously considered. A composite indicator of contextual HIV risk may reveal places core to HIV transmission and susceptible to HIV acquisition. Such area profiles may help identify the combination of locally specific risk factors that readily promulgate HIV and better inform the design of place-based HIV intervention packages to enhance current strategies towards global HIV control.

Geographic Areas

Files

This item is currently under embargo. It will be available starting 2021-10-02.

More About This Work

Academic Units
Epidemiology
Thesis Advisors
Mathema, Barun
Degree
Ph.D., Columbia University
Published Here
October 5, 2020