Spatial Clustering And Bayesian Modeling For Some Characteristics Of Tuberculosis Patients With Environmental Risk Factors

Abstract

Background: Tuberculosis (TB) remains a major public health concern in India, with varying patterns across demographic groups and regions. Identifying risk factors and spatial clustering is crucial for improving prevention and control strategies.

Objective: This study aimed to evaluate the association of sex, diabetes status, and HIV status with TB outcomes using both frequentist and Bayesian logistic regression approaches, and to identify spatial clustering of TB cases in Dhemaji district, Assam.

Methods: A total of 1,340 TB cases (pulmonary and extra-pulmonary) reported between 2018 and 2020 were analyzed. Frequentist logistic regression and Bayesian logistic regression (with vague priors and MCMC sampling) were applied to assess risk factors. Spatial analysis was conducted to detect significant clusters of TB cases at the prefecture level.

Results: Males and individuals with diabetes had significantly higher odds of TB outcomes, while HIV status was not significantly associated. Frequentist analysis indicated that males had 23.6% higher odds (OR = 1.237, 95% CI: 0.956–1.599) and diabetics 10.8% higher odds (OR = 1.108, 95% CI: 0.820–1.497), though confidence intervals included the null for both predictors. In contrast, Bayesian analysis identified these associations as statistically credible (Sex: OR = 1.197, 95% CrI: 0.077–0.451; Diabetes: OR = 1.077, 95% CrI: 0.037–0.382). HIV status showed no meaningful effect in either model. Descriptive findings revealed pulmonary TB was more common among males (53.21%) compared to females (16.49%), with a significant sex-based difference. Spatial analysis identified clustering in Silapathar and Dhemaji, indicating localized high-risk areas.

Conclusion: Sex and diabetes status emerged as significant predictors of TB outcomes, whereas HIV status showed no association. The Bayesian approach demonstrated greater sensitivity in detecting risk factors than the frequentist model. Spatial clustering highlights priority areas for intervention. These findings underscore the importance of targeted screening for males and diabetics and focused resource allocation in identified hotspots to strengthen TB prevention and control efforts in Dhemaji district.