Back to search

Boston University

Statistical methods for understanding tuberculosis epidemiology among key populations

Abstract

dc:description.abstract

Tuberculosis (TB) is a preventable and curable disease but continues to be a significant global health challenge. Disease burden is heterogeneous and often marginalized populations are disproportionately impacted. Controlling the TB pandemic depends critically on identifying and reaching individuals at greatest risk for TB disease for testing and treatment. Understanding TB epidemiology among high-risk groups typically facing social and economic disparities is key to designing tailored interventions to decrease TB incidence. In this dissertation, we develop statistical methods that are motivated by the need to understand TB epidemiology among key populations defined as groups of individuals with elevated disease burden and/or transmission risk. Reaching and engaging key populations in TB research is critical for informing the design of interventions that will have the greatest impact on reducing TB burden. Respondent-driven sampling (RDS) is a link-tracing sampling design that was developed to sample from populations that can be challenging to reach through traditional sampling approaches. Statistical methods for inferring associations with RDS data is an underdeveloped area that requires further attention if we aim to leverage RDS to learn about disease dynamics in key populations. We develop a semi-parametric randomization test for RDS data to test for the association between two variables that accounts for the dependence introduced by RDS study design. We test the performance of our method using simulation studies. We apply our approach to test for associations between variables that are important for understanding TB epidemiology among people who smoke illicit drugs (PWSD) using data collected from an RDS study in Worcester, South Africa. In high TB burden settings, it is most common for individuals to receive TB testing and diagnosis once they are symptomatic and present to a clinic. It is well documented that barriers to accessing care cause delays in diagnosis and treatment as well as more opportunities for transmission in the community. To understand care seeking behavior we must first understand the various phenotypes of disease severity that are found in clinic settings at time of diagnosis. Current approaches to classifying disease severity are inadequate for understanding the spectrum of TB disease. We develop a data-driven approach to classify TB disease severity using information from various data sources collected at time of diagnosis such as symptoms, chest radiography, and sputum burden. We present a model that is motivated by the statistical framework of the event-based model (EBM) to identify different disease severity classes in the TB disease spectrum. We test our method using simulation studies. We apply our method to TB data from a cohort of individuals initiating TB treatment in a rural community in the Western Cape, South Africa and correlate predicted disease severity class with time to TB sterilization. Finally, we develop a software package for the implementation of our model for TB disease severity alongside a tutorial for researchers to apply our method in practice.

Degree

thesis:*
Name thesis:degree_name
Doctor of Philosophy
Level thesis:degree_level
doctoral
Discipline thesis:degree_discipline
Biostatistics
Grantor
Boston University
Year dc:date.issued
2024

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Malatesta, Samantha Ashley
Advisor dc:contributor.advisor
  • White, Laura F..

Subjects

dc:subject × 1

Rights

Language dc:language.iso
en_US

Identifiers

dc:identifier.*
Handle dc:identifier.uri
https://hdl.handle.net/2144/53098

Chain of custody

source
Harvested from
Boston University
Base URL
open.bu.edu/oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
related terms
citation

Malatesta, Samantha Ashley. Statistical methods for understanding tuberculosis epidemiology among key populations. doctoral thesis, Boston University, 2024. https://hdl.handle.net/2144/53098