Statistical count models for predicting the number of treatment dropouts in Pediatric Tuberculosis patients

Tuberculosis (TB) is a bacterial infectious disease that remains a challenging and unexplored public health concern in pediatric patients. This study aims to fit a suitable count-based regression model and identify the key factors associated with treatment dropout in paediatric TB patients. The retrospective data of 2086 pediatric TB patients was obtained from the Nikshay database at the Department of TB and Respiratory Diseases of Sir Sunderlal Hospital, Banaras Hindu University, Varanasi, Uttar Pradesh, India. The analysis utilized various count data models, including Poisson, Negative Binomial, Zero-Inflated, Hurdle and Zero- truncated models. Model performance was assessed using Akaike Information Criterion, Bayesian Information Criterion and log likelihood values which confirmed the models’ effectiveness.The findings suggest that some key factors, such as “Missed followup” and “Contact tracing not done”, as well as patients’ failing to access the nutritional and financial support provided by the National Tuberculosis Elimination Programme, are significant factors contributing to patients’ non-adherence to TB treatment. These insights can inform public health strategies aimed at increasing support services, enhancing TB control programs, and contributing to broader TB elimination efforts.