Measuring HIV incidence, mortality, and the cascade of HIV care in Kenya

Human immunodeficiency virus (HIV) presents challenges to effective measurement and control, including the frequently asymptomatic or nonspecific nature of initial HIV infection, slow disease progression, life-long infection, as well as the stigma associated with an incurable sexually transmitted infection. HIV can be monitored through a testing and treatment cascade that starts with diagnosis, followed by antiretroviral treatment (ART) initiation, and finally, viral suppression. Ultimately one hopes to achieve a state of epidemic control, which has been defined multiply in terms of reductions in incidence and mortality, or in the ratio of new infections to deaths, or new infections to prevalent infections. This thesis provided an overview of the epidemiology of HIV in Kenya, with a focus on issues with measurement and modeling of incidence, mortality and indicators of program achievement. After an overview of HIV epidemiology in Chapter 1, we presented trends in HIV incidence in western Kenya, a high HIV prevalence region, during a period of rapid expansion of HIV treatment and voluntary medical male circumcision, the two current most effective biomedical forms of HIV prevention (Chapter 2). The study demonstrated the value of estimating HIV incidence in high-burden populations that are severely affected by HIV and will thus have a disproportionate influence on the control of the epidemic at the national level. In Chapter 3, we summarized the Spectrum HIV epidemic model’s application to Kenya, with a focus on sub-national estimation using ancillary tools, and how these estimates evolved through three cycles of models from 2015–17, showing that improvements in model input led to more plausible fits. We also compared the HIV prevalence trends in routine antenatal care to a small-area population-based cohort within the region of interest. We investigated trends in antiretroviral treatment (ART) use in Kenya and South Africa between 2007 and 2012 (Chapter 4), and found significant increases in ART use in Kenya and South Africa, respectively. While in South Africa it seemed disparities in ART access had decreased over the period, in Kenya there was evidence of increasing disparities by residency and household wealth. Using data collected in the 2012 Kenya AIDS Indicator Survey, we analyzed the impact of adjustments to self-reported HIV status using biomarkers for antiretroviral exposure and viral load suppression (Chapter 5), resulting in adjustments of approximately 10 percentage points with antiretrovirals or 15 percentage points with undetectable viral load (17 points with both) versus self-report alone. We found adjusted HIV positivity for adult deaths in Nairobi of 20.9%, higher than predicted from models (Chapters 6 & 7). Documentation of HIV status was poor among the HIV-infected decedents, as was ART use among those previously diagnosed. In the high HIV-prevalence region of Kisumu, 23.1% of deaths of all ages were due to HIV, with an HIV mortality rate of 251 per 100,000 population (Chapter 8). The methods presented in this thesis can be used more broadly to improve HIV measurement so that countries can gather relevant data to scale up effective HIV programs.