Application Of Cox Proportional Hazards Model In Case Of Tuberculosis Patients In Kisumu County, Kenya
Tuberculosis (TB) remains a major global health problem. In 2013, World Health Organization estimated 8.6 million people to have developed TB and 1.3 million died from the disease including 320 000 deaths among HIV-positive people. Kenya is ranked 15th up from 13th position among the 22 high TB burden countries worldwide which contributes 80% of the global TB. The main objective of this study was to describe the pattern of time of treatment interruption in TB patients in Kisumu County Kenya. The specific objectives were to compare Survival functions for two groups of Tuberculosis patients and identify significant predictors of loss to follow-up (LTFU) among the TB patients. This was a retrospective cohort study based on TB patients that were registered in the unit TB registers. Kisumu County was randomly sampled for the study from the top ten high TB burden counties in Kenya; Secondary data was collected from documents of all TB cases registered from January to December 2013 in the health facilities. This study focused on time of loss to follow-up for TB patients enrolled on treatment in the health facilities. Kaplan-Meir estimator and Cox proportional hazard model have been used for the analysis and model building. Logrank tests and Wilcoxon tests were used for comparison of survival data. Cox model was able to provide the estimates covariates and their effect on loss to follow-up of Tuberculosis patients after adjustment for other explanatory variables. From the 1,275 patients in this study a total of 107(8.4%) interrupted treatment and 1,168 (91.6%) were censored. The median time of treatment interruption was 2.167 months (65 days).This analysis shows that most of the Tuberculosis patients are lost to follow up during the intensive phase of treatment which is within the first two months. The Log rank and Wilcoxon test were not significant in LTFU survival experience between the various categories for covariates HIV status, weight, gender and type of TB at alpha significance level of 5% (p<0.05). We find significant differences in survival experience of the patients in different categories of age; patients with more than 35 years of age had better survival rates compared to those who were younger. Results of the proportional hazards Cox regression analysis of TB patients revealed that the covariates age, weight and Tuberculosis patient type were significant factors associated with treatment interruption for TB patients. There is need to strengthen follow up of patients on TB treatment especially during the intensive phase and until completion of treatment.