Determinants of Time to Treatment Dropout among Tuberculosis Patients in Buno-Bedele and Illu Ababora Zones, Oromia Regional State, Ethiopia


 Tuberculosis is a chronic infectious disease caused by Mycobacterium tuberculosis. It typically affects the lungs (pulmonary tube) but, can affect other parts of the body as well (extra pulmonary tube). The study was aimed to investigate the determinants of time to drop out of treatment for TB patients. Secondary data was used from 375 TB patients of the selected health stations and hospitals at Buno-Bedele and Illu Aba Bora Zones. The response variable for this study was the survival time (Time to dropout the treatment among TB patients) measured in days and the covariates were gender of the patient, marital status, HIV co-infection, Phase of TB treatment, TB type, TB category, Previous TB history, HIV Co infection, Anemia and Physical inactivity. Descriptive statistics, Kaplan-Meier Estimation method, Semi-parametric survival models and parametric survival models were used for the analysis of time to TB treatment dropout dataset. From 375 patients who started TB treatments about 24.8% dropout and 75.2% censored at the end of the study and the median survival time of TB patients were 199 days. The Log-rank results showed that marital status, HIV co infection, Diabetic mellitus, Cancer and Anemia cases had significant difference between the survival experience at 5% level of significance, whose different levels have an impact in the survival time of TB patients; whereas Sex, Phase of TB treatment, TB type, TB category, previous TB status, co-morbidity, and physical inactive had not significant difference between the survival experience at 5% level of significance. Finally, the result of Cox-proportion hazard model showed that, age, HIV co-infection and Anemia had a significant effect on tuberculosis patients during the study period.


Introduction
Tuberculosis (TB) is a chronic infectious disease caused by Mycobacterium tuberculosis (MTB).
TB typically affects the lungs (pulmonary tube) but, can affect other parts of the body as well (extra pulmonary tube). The Global tuberculosis report showed that TB now ranks above HIV as a leading cause of death worldwide [2]. An estimated incidence for the year 2015 of TB in Ethiopia was 191/100,000 population. In addition, as the national surveys on the burden of TB epidemic showed 31% -41% of TB patients are HIV positive [3]. The previous study shows that the prevalence of delayed presentation for HIV care among TB/HIV co-infected patients was 59.9 %. The study also shows tobacco non-users of TB/HIV co-infected participants were also 50 % less likely to present late for HIV care compared to tobacco users. The relative odds of delayed presentation among Tb/HIV co-infected patients with ambulatory and bedridden functional status was higher than with working status [4]. The aim of this study is to identify risk factors that affect Survival Time to Drop out treatment among TB patients in case of Buno-Bedele and Illu Aba Bora zones, Oromia, Ethiopia.

Methods:
Secondary data was used from 375 TB patients of the selected health stations and hospitals at Buno-Bedele and Illu Aba Bora Zones. The response variable for this study was the survival time (Time to dropout the treatment among TB patients) measured in days and the covariates were gender of the patient, marital status, HIV co-infection, Phase of TB treatment, TB type, TB category, Previous TB history, HIV Co infection, Anemia and Physical inactivity.

Survival function
In order to estimate the survival function, the estimator proposed by Kaplan and Meier takes into account for censoring by adjusting the number of subjects at risk, Where ( ) denote the distinct ordered times of Drop out and, and denote the number of events and the number of individuals still at risk at time , respectively.

Comparison of Survival Curves
Kaplan-Meier method for estimating survival curves and the log-rank test for comparing two estimated survival curves are the most frequently used statistical tools in medical reports on survival data.

Log-rank test
The log rank test is a non-parametric test for comparing two or more independent survival curves. The log rank test statistic for comparing two groups is given by: Where: is the number of rank ordered event times, 1 is the observed number of events in group one at event time , 1 = 1 − is the expected number of events corresponding to 1 , 1 is the number of individuals at risk in group1 just prior to event time , 2 is the number of individuals at risk in group 2 just prior to event time , is the variance of the number of events d 1i at time t i , n i and d i are the number of individuals at risk and number of vascular complication in both groups ( i.e., group 1 and group 2) just prior to event time t i , respectively.

The Cox ('Semi-Parametric') Proportional Hazards Model
It is a survival analysis regression model, which describes the relation between the event incidence as expressed by the hazard function and a set of covariates. Mathematically, the Cox model is written as; where the hazard function h(t) is dependent on (or determined by) a set of p covariates (x1, x2, …, xp), whose impact is measured by the size of the respective coefficients (b1, b2,.., bp). The term h0 (t) is called the baseline hazard, and is the value of the hazard if all the xi are equal to zero (the quantity exp (0) equals 1). The't' in h (t) reminds us that the hazard may (and probably will) vary over time.

Multivariable Cox Proportional Hazard Regression Analysis
Multivariable Cox PH analysis (by using stepwise selection process) including all the potential risk factors that had a P-value of less than or equal 0.25 in single covariate Cox PH analysis.
There are three covariates were significant at 5% level of significance. Hence; we have a final multivariate model which includes the three covariates namely: Age, HIV co-infection and Anemia are the risk factor for the dropout of TB patient or these variables significantly affects the survival of TB patients. See table S1.

Test of the assumption of proportional hazard
From the table S2, the overall global test for covariates was not satisfying the assumption of proportional hazard model, hence p-value is larger than 5%. The global is no significant at 5% level of significance it means that the proportional hazard assumption is satisfied.

Interpretation of results of Final Model of Cox regression model
The interpretation from the results of the final model which consists of the main effects is based on the hazard ratios. Consequently, the interpretation of covariates that are included in the final proportion hazard model of TB dropout patients is as follows.

Parametric Regression Analysis
For the data on TB patients, the parametric models were fitted. The common applicable criterion to select the model is the AIC statistic proposed by Akaikie (1983). From table S3, the Weibul regression model has the least AIC value which shows that the Weibul regression model well fitted to data TB patients.

Discussion
This research was conducted to identify predictors of drop out of treatment among TB patients. TB patients who had anemia were 1.79 more likely to anti-TB treatment drop out than those TB patients who had not anemia. This finding supports the study conducted by SW Lee et al. [8].
They confirmed that anemia is a common hematological abnormality in patients with TB.
Because TB-associated anemia is usually mild and resolves with anti-TB treatment, close observation is sufficient without other cause of the anemia.

Conclusions
The main aim of this study was investigating the determinants of time to drop out of treatment

Declarations
Funding: No funding was obtained for this study.

Consent for publication: Not applicable
Competing of interests: The authors declare that they have no competing interests Availability of data and materials: We can provide the dataset that has been used to do this study up on reasonable request.
Authors' contribution: All authors contributed equally to the study. WE conceived the idea, DG; AM contributed in the design analyses and interpretation, WE the corresponding author drafted the manuscript. All authors read and approved the final manuscript.