Dissertation/Thesis Abstract

Improving Decision-Making about HIV Treatment and Prevention in the United States: Model-Based Approaches
by Gonsalves, Gregg Steven, Ph.D., Yale University, 2017, 94; 10633246
Abstract (Summary)

Background The AIDS epidemic in the United States is still the most serious one in the developed world. Though progress has been made in combating the epidemic, there is more to be done in terms of optimizing HIV prevention and treatment. This dissertation explores model-based approaches that may be used to answer, in part, three specific questions for decision makers about HIV services: where along the HIV care continuum is it best to direct efforts to improve clinical outcomes; how to choose among different geographic locations to improve mobile HIV testing, and; what effects does the timing and magnitude of responses to HIV outbreaks have on their costs and trajectories?

Methods To address the first question, I used data from the Centers for Disease Control and Prevention and the North American AIDS Cohort Collaboration on Research and Design from 2009-2012 to estimate the distribution of time spent in and dropout probability from stages in the continuum of HIV care. I used these estimates to develop a queueing model for the expected number of patients found in each stage of the cascade.

To assess how to improve the detection of new cases of HIV infection, I conducted simulations to assess four alternative approaches to mobile HIV testing in three hypothetical geographic zones. The approaches are distinguished from one another by how they manage the tradeoff between exploration and exploitation in zone selection and how they process the information obtained from previous days of testing. They include: 1) Thompson sampling (TS), an adaptive Bayesian search algorithm; 2) an explore-thenexploit (ETE) strategy; 3) a strategy using only prior information; and; 4) a performance benchmarking strategy with access to perfect information.

Finally, to explore the costs and epidemiologic trajectories of nascent HIV epidemics, I developed a simple stochastic model of an outbreak in a small population and used this approach to analyze the costs associated with implementation of a comprehensive contact-tracing, syringe exchange, and antiretroviral treatment (ART) intervention in Scott County, Indiana, the site of a recent HIV outbreak among people who inject drugs. I examined the effects of an intervention initiated in March 2015, when the major state response began and compared them to those of a hypothetical response initiated at the start of the outbreak in November 2014.

Results The queueing model estimates that individuals spend an average of about 3.1 months following HIV diagnosis before being linked to care, or dropping out of care with a probability of 8%. Those who link to care wait an additional 3.7 months on average before getting their second set of laboratory results (indicating retention in care) or dropping out of care with probability of almost 6%. Those retained in care spent an average of almost one year before achieving viral suppression on antiretroviral therapy or dropping out with an average probability of 13%. For patients who achieved viral suppression, the average time suppressed on ART was 4.5 years.

Comparisons of alternative mobile HIV testing strategies indicated that TS outperformed ETE 63% of the time, with 15% more new cases identified on average than ETE. This was within 90% of the benchmark established by the strategy with perfect information. Using last year's prevalence as prior information performed poorly compared to the other strategies. In sensitivity analyses, TS outperformed ETE in almost all circumstances.

In assessing the response to the Scott County HIV outbreak, a hypothetical intervention in November 2014 using contact tracing and syringe exchange efforts at the levels used in the actual response in March 2015 resulted in a 14% decrease in total mean costs. Starting these programs earlier with an enhanced response make a greater impact on costs, with earlier introduction of an expanded syringe exchange program having the most dramatic economic effects. As syringe exchange coverage is increased, further reductions in costs are gained. Earlier intervention, particularly with expanded syringe exchange, slows the growth of epidemic, but does not stop it unless coverage of syringe exchange is in excess of 90%.

Conclusions These model-based approaches suggest that: 1) HIV interventions will be most effective if they focus on more rapidly identifying newly infected individuals, and increasing the fraction of them retained in care who achieve viral suppression; 2) Thompson sampling should be further investigated for use in HIV testing programs and; 3) an earlier and more robust response to the outbreak in Scott County, Indiana would have substantially reduced total costs of the epidemic.

Indexing (document details)
Advisor: Cleary, Paul D.
Commitee:
School: Yale University
School Location: United States -- Connecticut
Source: DAI-B 78/11(E), Dissertation Abstracts International
Source Type: DISSERTATION
Subjects: Public health
Keywords: Decision-Making, HIV / AIDS, Model-Based, Operations Research, Prevention, Treatment
Publication Number: 10633246
ISBN: 9780355105384
Copyright © 2019 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy
ProQuest