Details
Date: May 23, 2024
Location: Storrs, CT
Abstract
Tuberculosis (TB) remains a significant global health challenge, and Brazil exemplifies the complexities of controlling this infectious disease. Reliable estimates and forecasts of TB incidence rates are crucial to guide public health policies. This study focuses on the high-burden municipalities of Eastern Rio Grande do Sul, Brazil. We propose a novel spatiotemporal model based on the Hausdorff-Gaussian process to analyze TB incidence data. This model incorporates spatial dependence dictated by the Hausdorff distance, allowing it to “borrow strength” from neighboring municipalities and generate more reliable estimates, particularly for smaller areas. Our analysis has two primary goals. First, we aim to generate accurate TB incidence estimates by incorporating municipality-specific characteristics through covariates and a spatiotemporal random effect. The model delivers trustworthy expected incidence rates, consequently allowing for calculating standardized incidence ratios (SIRs). Second, our model offers predictive capabilities, forecasting TB incidence rates one year ahead to support proactive public health planning. We demonstrate our model’s effectiveness and competitive performance against other specialized areal data models. The insights gained from this study can guide policymakers in developing effective TB control strategies.