Modeling the Mortality Incidence of Diabetes Mellitus among Farmers in Benue State Using Count Data Regression Models
David Adugh Kuhe *
Department of Statistics, Joseph Sarwuan Tarka University, Makurdi, Benue State, Nigeria.
Bem Gura Ahua
Department of Mathematics and Computer Science, Rev. Fr. Moses Orshio Adasu University, Makurdi, Benue State, Nigeria.
*Author to whom correspondence should be addressed.
Abstract
Aims: This study aimed to model monthly diabetes mellitus cases among farmers in Benue State, Nigeria, focusing on clinically confirmed, active, severe, and mortality cases using count data regression models.
Study Design: The study adopted a retrospective quantitative design using count data regression modelling. Three models—Poisson regression (PR), negative binomial regression (NBR), and generalised Poisson regression (GPR)—were used to model and predict diabetes-related mortality from confirmed, active, and severe diabetes cases.
Place and Duration of Study: Secondary data were obtained from the Benue State Epidemiological Unit, Makurdi, Benue State, Nigeria, covering the period from 1 January 2010 to 30 December 2023.
Methodology: Monthly surveillance data on confirmed, active, severe, and mortality cases were analysed using PR, NBR, and GPR models. Overdispersion was assessed to determine the suitability of PR. Model performance was evaluated using the −2 log-likelihood (−2 LogL), Akaike information criterion (AIC), and Bayesian information criterion (BIC).
Results: The PR model exhibited overdispersion, supporting the use of NBR and GPR. The NBR model recorded the lowest AIC (150.12) and BIC (158.47), whereas the GPR model recorded the lowest −2 LogL (146.78). Based on the information criteria, NBR was selected as the most appropriate model. Confirmed, active, and severe cases were statistically significant positive predictors of diabetes-related mortality.
Conclusion: The NBR model provided the preferred fit on the basis of AIC and BIC and appropriately accommodated the dispersion in the count data. The findings support strengthening rural healthcare services, community-based diabetes education, surveillance and monitoring systems, evidence-informed health planning, and the capacity of local healthcare workers to prevent and manage diabetes-related complications and mortality among farmers in Benue State.
Keywords: Diabetes mellitus, diabetes-related mortality, count data, poisson regression, negative binomial regression, generalised poisson regression, overdispersion, farmers, Benue State, epidemiological surveillance