Surveillance Methods for Monitoring HIV Incidence and Drug Resistance

Disease surveillance is the continuous collection, analysis, and interpretation of health-related data. Information gained from routine HIV disease surveillance is vital to national program managers deciding to implement new prevention or treatment programs. In this dissertation, we describe methods for estimation of HIV incidence and the prevalence of HIV drug resistance.;HIV incidence estimation is critical for identifying at-risk populations for targeted interventions and measuring the effectiveness of these interventions. We provide an in-depth literature review of the available options for estimating HIV incidence, including cross-sectional assays. Next, we describe a novel cross-sectional assay for HIV incidence estimation that discriminates between recent and long-term infections on the basis of within-host viral diversity. Diversity is measured using a version of Shannon's entropy that we adapt to improve discriminatory ability. These adaptations include breaking the within-host sequence alignment into smaller sections to allow for more nuanced detection of within-host variability, and we suggest an algorithm for adjusting for multiple HIV infections using clustering methods.;HIV drug resistance surveillance guides national programmatic managers identifying effective treatment regimens for HIV-infected individuals in their countries. We describe a large-scale consulting project with the World Health Organization to redesign their guidance for pre-treatment and acquired HIV drug resistance surveillance in low- and middle-income countries. Our consulting work prompted a variety of interesting statistical questions that we address in a series of papers. We describe a novel method for calculating sample sizes for two-stage clustered surveys in which the finite population correction can be applied. Our method results in a sometimes dramatic decrease in sample size while still achieving the desired precision. We introduce a novel acquired HIV drug resistance outcome for measuring viral load suppression that incorporates information on patient loss-to-follow-up. This outcome has increased epidemiological utility over previously used outcomes. Finally, we evaluate methods for confidence interval estimation for proportions measured in surveys and provide recommendations for their use.