EXPLORING TIME-DEPENDENT FACTORS IN THE SURVIVAL OF TUBERCULOSIS PATIENTS: A COX PROPORTIONAL HAZARDS ANALYSIS

Abstract

Background: This research paper is attempted to identify time dependent covariates effects of Tuberculosis patient’s database using Cox proportional Hazard Model. The study of time-to-event data with censoring and variables has seen considerable application of the Cox proportional-hazards regression model. Over time, the covariates' values could fluctuate. At the 5% level of significance, the P value suggests that there is no difference between such Model 1 results and that the survival times are same in the male and female groups.

Methodology: The sources of data were collected from National Institute of Research in Tuberculosis (NIRT), Chennai (ICMR). 1237 tuberculosis patients were enrolled for the study, which consisted of a randomized clinical trial with three distinct treatment regimens and a control group. Each regimen was given a six-month course of treatment. Sputum conversion time (from positive to negative) during the course of treatment is the relevant event. The total number of events was 1062, of which 85.8% were cases that experienced the event of interest, which is defined as cases that became diseased throughout the treatment period (6 m), and the total number of censored observations was 175 (14.2%).

Result: In this research paper, initially 1236 samples of patients were considered with six month duration of study period. The patients are treated as the event of interest is smear sputum conversion time into positive to negative during the observation period. Hereafter the number of event occur 1062 patients and remaining patients has been censored.  Model 2 results of gender not significant and sputum culture is statistically significant. Hazard ratio value is two groups; the values are 1.075 and 0.762 respectively. It’s labelled as high positive and low positive sputum culture status. Model 3, the significant value at the 5% level of significant is P = 0.162, it represents the value from the product term's coefficient for the interaction between gender and sputum culture status.

Conclusion: The P value indicates that there is no significant interaction effect in gender and sputum culture, then remove this noisy model. In addition, point estimate coefficient of gender is proved the hazard ratio HR=1.091 it shows that there is no interaction effect of gender. Finally, the hazard ratio in the variable of sputum culture status HR= 0.847 and conclude the hazard for the high positive group is 0.86 times the hazard for the low positive status group.  As a result, we discover that interaction has a minimal value, which making it a better model than crude and adjusted Estimate model.