Modelling Tuberculosis Mortality Incidence among Farmers in Benue State Using Count Data Regression Models

Abstract

Tuberculosis (TB) remains a major public health concern in many parts of the world, including Benue State, Nigeria, where agricultural communities are particularly vulnerable. This study aims to model the monthly mortality incidence of tuberculosis (TB) among farmers in Benue State, Nigeria, focusing on serologically confirmed, active, severe, recovered, and mortality cases using count data regression models. Three count data regression models: Poisson Regression (PR), Negative Binomial Regression (NBR), and Generalized Poisson Regression (GPR) were employed to predict TB-related mortality based on these variables. Secondary data from the Benue State Epidemiological Unit, Makurdi, spanning from January 2010 to December 2023, served as the basis for analysis. The study found the presence of over-dispersion in the Poisson Regression model which necessitated the use of NBR and GPR. Model performance was evaluated using -2 Log-Likelihood (-2 logL), Akaike Information Criterion (AIC), and Bayesian Information Criterion (BIC). Among the three competing models, NBR provided the best fit, with a -2 logL value of -901.92, an AIC of 1203.85, and a BIC of 1223.70, effectively addressing the over-dispersion in the data. The analysis identified confirmed, active, and severe TB cases as significant predictors of TB-related mortality in Benue State. Additionally, a strong negative and significant relationship was observed between recovered cases and mortality, indicating that an increase in recoveries correlates with a decline in TB-related deaths. The study recommends that policymakers and researchers should prioritize the Negative Binomial Regression model for TB analysis, enhance TB case management and treatment adherence, improve TB data collection, and design targeted interventions to reduce severe cases and increase recovery rates among vulnerable populations like farmers.