A Joint Modeling Approach for Analysis of Longitudinal Body Weight and Sputum Status of Tuberculosis patients in Jimma University Specialized Hospital

Abstract

Tuberculosis is a major public health problem even though it is treatable and curable. Weight and sputum conversion during anti tuberculosis (TB) treatment period is an important component and they have been described as a useful marker to assess the progress of TB patients’. Objective:-The objective of this study is to fit a joint model in which both the longitudinal weight and sputum status are studied to investigate their joint evolution and identify the risk factors for the body weight and sputum status of tuberculosis patients in Jimma University specialized Hospital during six months diagnosis period. Method: The data for this thesis were obtained from a retrospective study from TB patients registered between 2011 and 2013. The following statistical models were considered: linear mixed model for the separate body weight analysis, generalized linear mixed model for sputum status and a joint model with correlated random effects was fitted to simultaneously study the evolution over time of a longitudinal body Weight and Sputum status. The estimation of the model parameters was done by maximum and restricted likelihood and maximum likelihood based on adaptive Gaussian hermite Quadrature as implemented in the SAS procedure NLMXED. Result: The overall proportion of tuberculosis patients during follow up time having positive and negative sputum status is 39.3% and 60.7% respectively. Based on the data exploration the mean change of body weight has a linear relation with time. From the separate linear mixed model all covariates (types of TB, age, dose) are significant and their interaction by time were the risk factors for the body weight of TB patients. In case of separate generalized linear mixed model age, types of TB, dose and time have a significant effect on the sputum status of TB patients. Similar covariates were significant in the joint model of body weight and sputum status and estimates were found to be very close to separate analysis. But, the joint model yields higher precision and allows for quantifying the association between outcomes and association between the outcomes in this joint model was negative (ρ =-0.698, p=0.0001). Conclusion: The results of the separate and joint models almost the same. When the joint model is compared with the separate model, it is both the most parsimonious model and also fits the data better than the separate model. The joint model showed that the body weight and positive sputum status are inversely related each other.