Modeling tuberculosis patients using loglinear models

• Published in: Journal of Physics: Conference Series
• Main Author: Nasseruddin N.; Mahmud Z.; Md Isa K.A.
• Format: Conference paper
• Language: English
• Published: Institute of Physics Publishing 2019
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85076105682&doi=10.1088%2f1742-6596%2f1366%2f1%2f012109&partnerID=40&md5=9b7115f56b16fb14b6f983ffc52e3ea5

Tuberculosis (TB) patients usually get infected by tuberculosis bacteria called Mycobacterium tuberculosis which can spread from one person to another through air. TB cases normally infect the lungs (Pulmonary TB) and in some other cases it infects other organs such as brain, heart, spine, lymph nodes, adrenal gland and intestines (Extrapulmonary TB). However, patients who are infected with tuberculosis bacteria may have complications due to other factors that can delay the process of healing. Using loglinear modelling, this study will examine whether types of TB is strongly associated with patient's age, gender, whether or not they smoke, have HIV and into drug use. Records of 477 TB patients obtained from a previous related study are used in the modelling process where five loglinear models are compared to find the most parsimonious model that will determine the strength of the association between the variables. Likelihood ratio G 2, p-values, expected values and standardized residuals are used in the analysis. Based on the comparisons of the loglinear models, homogeneous model is found to be the most parsimonious compared to independence and joint models. The analysis based on the homogenous loglinear model show that that the estimated odds of patients who smoke is 2 times more likely to have a Pulmonary TB compared to the non-smokers. Also patients who are drug users are 8 times more likely to have a Pulmonary TB compared to patients who are non-drug users. Hence, a recommendation for the healthcare providers is to give more attention to patients in these categories. © Published under licence by IOP Publishing Ltd.