Essays in HIV: Mapping the distribution of marginalized groups and analysis of ART treatment effectiveness across patient groups
Restricted (Penn State Only)
- Author:
- Sanei, Sanam
- Graduate Program:
- Statistics
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- February 21, 2024
- Committee Members:
- Bing Li, Professor in Charge/Director of Graduate Studies
Le Bao, Chair & Dissertation Advisor
Runze Li, Major Field Member
Lingzhou Xue, Major Field Member
Qiushi Chen, Outside Unit & Field Member - Keywords:
- HIV/AIDS Epidemiology
Key Population
Size Estimation
Generalized Linear Mixed-Effect Models (GLMM)
Bayesian Modeling
Mapping HIV Prevalence
Case-Control Sampling
Hidden Markov Models (HMM) - Abstract:
- Achieving the new goal of three zeros set by UNAIDS (zero new HIV infections, zero AIDS-related deaths, and zero discrimination by 2030) is still a major challenge in global health. My dissertation work focuses on the development of statistical models that lead to a better understanding of HIV/AIDS epidemic and potentially improve current HIV/AIDS-related health policy. In the first project, we develop generalized linear mixed-effect models to estimate the female sex worker population at the grid-cell level across countries of Eastern and Southern Africa. Size estimation of the key populations like female sex workers (FSW), men who have sex with men (MSM), and drug users who inject the drug is important for controlling the spread of HIV and treating HIV patients more effectively since these key populations have a higher HIV prevalence than the general population. It is considerably hard to get the size and location of key populations because of social stigma and discrimination toward these populations. We proposed a procedure for variable selection among hundreds of spatial covariates from Demographic and Health Surveys (DHS) data. We show that with our proposed models and identified demographic variables, we can obtain a reasonable estimation of female sex workers’ distribution. In the second project, we propose a subsampling procedure to decrease the computation time of zero-inflated models. Zero-inflated models are increasing in popularity and are vital to a variety of applications and disciplines. Performing variable selection, estimating parameters, and diagnosing model fit for zero-inflated models is often prohibitively slow, especially for large datasets and Bayesian models. We show that we can consistently estimate the intercept and slope parameters of both the zero and conditional models. Performance is evaluated using a spatial presence-only dataset related to the number of female sex workers in four countries in sub-Saharan Africa. The third project discusses the recovery of immune function in HIV patients by a hidden Markov model. Antiretroviral therapy (ART) is the main treatment for HIV disease to reverse the progression of the disease. The number of CD4 cells (a type of white blood cell) is a key indicator of the progression of HIV. CD4 cell counts are subject to substantial variability due to measurement errors of imprecise measurement techniques and natural short-term variations within the immune system. This variability leads to a noisy representation of the true underlying disease state. We devise a hidden Markov model (HMM) framework to model the recovery of immune function in patients on ART and compare the treatment effectiveness based on patients’ characteristics and time being on treatment.