Spatial and Prediction Models for Addressing Challenges in Pediatric Tuberculosis Control and Care
Date of Award
Doctor of Philosophy (PhD)
Tuberculosis (TB) is among the leading causes of global mortality among children <5 years. Each year, over 1 million TB cases occur among children <15 years worldwide, and nearly one quarter of those children die; approximately 80% of those deaths occur among children <5 years. Alleviating the burden of pediatric TB and mortality requires 1) enhanced efforts to prevent transmission to children and 2) treating more children for TB. Targeting resources to children with a known TB exposure has been a cornerstone of the public health response to prevent transmission and detect cases early. Infectious adults must be diagnosed and treated earlier to prevent transmission to their child contacts. Modeling studies suggest that targeting community-level active case-finding to areas with high local transmission intensity may demonstrate population-level reductions in TB incidence. However, obtaining conclusive evidence of concentrated transmission requires access to spatial and genomic data, which is often only available under research conditions in high TB-incidence settings. In chapter 1, I use Bayesian spatial modeling methods to probe routinely collected, age-disaggregated TB notification data to demonstrate that overrepresentation of young child cases co-locate with areas of high local transmission intensity, identified by molecular evidence of transmission from a prospective cohort study in the same setting. This finding suggests that the use of models that leverage widely available notification data should be explored as tools to target case-finding and treatment efforts in high-transmission locations to maximize the direct and indirect benefits of active screening approaches. In chapter 2, I leverage data from a large prevalence survey to investigate a poorly understood form of TB that may frustrate symptom-based active case-finding efforts. Given that modeling estimates suggest that 96% of global childhood mortality due to TB occurs among children not receiving antituberculosis treatment, identifying and treating more cases of pediatric TB provide an opportunity to reduce child mortality. Diagnostic tools for pediatric pulmonary TB are limited by paucibacillary disease in children as well by resource constraints in many high TB-incidence settings. This contributes to poorer treatment outcomes through missed diagnoses and treatment delays. In chapter 3, I describe the analysis of a cohort of children being evaluated for TB from Cape Town, South Africa to demonstrate that a majority of antituberculosis treatment-decisions could be made using clinical evidence alone, without the need for additional diagnostic testing. In chapter 4, I describe the assembly of a large cohort of pediatric TB diagnostic evaluation data sourced from multiple geographically diverse, high TB-incidence settings to develop a prediction model for TB and investigate its validity and generalizability. As part of this work, I describe efforts in partnership with the World Health Organization to operationalize the prediction model as a treatment-decision algorithm to guide the evaluation of children with presumptive pulmonary TB.
Gunasekera, Kenneth Suranga, "Spatial and Prediction Models for Addressing Challenges in Pediatric Tuberculosis Control and Care" (2022). Yale Graduate School of Arts and Sciences Dissertations. 599.