Study Setting
This study was conducted in a random set of patients from all MDR-TB treatment initiation sites in Ethiopia. The country’s population was estimated at 102 million in 2017, with 84% being rural [17]. In 2018, there were 281 public hospitals, 3,622 health centers, and 16,660 health posts in the country. All the hospitals and health centers provided TB diagnosis and treatment services and 65% of health posts provided DOTs service for drug susceptible TB patients [5].
Till 2014, there were 14 MDR-TB TICs (one in Tigrai, three in Amhara, five in Oromia, two in Southern Nations and Nationality Peoples’ Regional States (SNNPR), two in Addis Ababa and one in Dire Dawa City Administrations) found in Ethiopia [6].
The study was conducted using records of a random subset of the patients who started MDR-TB treatment between 2009 and 2014 (Figure 1).
Study Design and Population
A health facility-based retrospective cohort study design was used. The study population was all pulmonary MDR-TB patients who started treatment between January 1, 2009 and December 31, 2014 in all MDR-TB TICs in Ethiopia. Confirmed pulmonary MDR-TB patients based on culture and DST or Genexpert or Line Probe Assay and with positive base line culture were included in the study. Patients who transferred in were excluded from the sample.
Sample Size and Sampling Techniques
The sample size was calculated considering a 95 % Confidence level (Zα/2) at 1.96; 31.6% unfavorable treatment outcome (p) [18]; 2.5% degree of precision (d) ; total study population (N)=1559 and finite population correction [19]. Based on this, the calculated sample size was 751.
Sampling Technique
All the 14 MDR-TB TICs found in Ethiopia from 2009 to till the end of 2014 were included in the study. These TICs were dispersed over the country with 1 found in Tigrai, 3 in Amhara, 5 in Oromia, 2 in Southern Nations and Nationality Peoples’ Regional States (SNNPR), 2 in Addis Ababa and 1 in Dire Dawa City Administrations [6]. The sample size was proportionally distributed over each TIC based on their patient load. These 14 TICs also hosted patients from the other five regions in the country that did not have TICs. A sampling frame for each TIC was prepared to select the 751 patients using the simple random sampling method from the MDR-TB register. The sampling procedure is pictured in Figure 2.
Measurements
The main outcome variable of this study was treatment outcome (categorized as favorable, unknown or unfavorable). The independent variables included were socio-demographic characteristics (age, sex and place of residence) and clinical conditions [type or form of TB (smear positive or smear negative TB), HIV/AIDS status, category of TB patients (new, return after lost to follow up, treatment failure, relapse and other), weight, presence of a TB treatment supporter, treatment regimen (new versus retreatment), having co-morbidities other than HIV/AIDS, BMI, bacilli load, degree of drug resistant and X-ray findings].
Operational Definitions /definition of terms
MDR-TB: a strain resistant to at least Rifampicin or both Rifampicin and Isoniazid;
Intensive phase treatment outcome: The outcome at which injectable agent was discontinued and the patient put on an oral continuation regimen;
Favorable treatment outcome: The outcome at which the patient was culture converted and alive at the end of the intensive phase.
Unknown treatment outcome: The outcome at which sputum culture not done or sputum culture sample sent but no feedback or no result or information to assign culture converted or not;
Unfavorable treatment outcome: included lost to follow up, died, not evaluated, treatment terminated and culture not converted at the end of the intensive phase;
Sputum conversion: defined as two consecutive negative smears or cultures from samples collected at least 30 days apart.
Data collection
A pretested and structured record review checklist was used to collect the data from the MDR-TB register and treatment card. Where needed, the ART register was reviewed to complete missing information from the MDR-TB register for those HIV positive patients enrolled to chronic HIV/ART care. Data were collected by two teams each consisting of one supervisor and two data collectors. The data collectors were trained nurses and the supervisors were trained BSc public health graduates. To maintain data quality, the collected data were submitted daily to the supervisors for verification with feedback provided the following morning.
Statistical Analysis
Data were entered into and cleaned using Epidata version 3.02 and analyzed using Statistical Package for Social Sciences (SPSS) version 20.0. Exploratory data analysis was carried out to check the level of missing values and presence of influential outliers. Multi-co linearity, normality was also checked for continuous variables. The normality of the data was checked using a histogram. For continuous and normally distributed data mean and standard deviation were reported and otherwise median and inter quartile range were reported. For categorical variables, frequencies and percentage were reported. The median duration of the intensive phase treatment and the median time of sputum culture and smear conversion were computed.
Finally, multivariable multinomial logistic regression analysis was done to identify independent factors associated with MDR-TB intensive phase treatment outcomes. For the purpose of selecting potential candidate variables, a multivariable model was constructed for variables having a P value <0.25 in the bivariate analysis [20]. Statistical significance was considered with two sided P-values of 0.05 and 95% Confidence Intervals (CI).