Modeling prognostic heterogeneity in multinational tuberculosis outcomes using quantile regression on multimodal...

This study investigates the heterogeneity of tuberculosis (TB) treatment outcomes across 75 countries by applying quantile regression modeling to a comprehensive multimodal dataset from Kaggle, encompassing socioeconomic, demographic, and comorbidity factors collected annually. Our objective is to move beyond average-centric analyses and characterize the full conditional distribution of TB incidence and treatment success, thereby revealing critical determinants at both high- and low-risk extremes. By integrating features such as HIV prevalence, GDP, healthcare expenditure, and comorbidity rates, the quantile regression framework facilitates nuanced risk stratification, uncovering complex interactions between predictors and outcome variability. Model evaluation using mean absolute error, R-squared, and quantile loss confirms robust predictive performance and enhanced interpretability relative to traditional linear regression approaches. Key findings highlight the outsized influence of HIV prevalence and healthcare spending on unfavorable TB quantiles, while favorable outcomes correlate more strongly with higher GDP and expanded health infrastructure. These insights provide actionable guidance for policymakers and global health stakeholders seeking to allocate resources efficiently and design targeted interventions for diverse epidemiological profiles. The primary contribution of this work is a statistically rigorous, distribution-aware approach for evaluating and explaining multinational TB prognoses, advancing both methodological and practical understanding of global TB disparities.