Study design, setting, and population
A multicenter institution-based cross-sectional study was conducted from September 2010 to December 2017 among bacteriologically confirmed pulmonary DR-TB patients in the four DR-TB Treatment Initiating Centers (TICs) of Amhara regional state of Ethiopia. Amhara region is the second populated region in Ethiopia which contributes to the highest number of DR-TB in the country. A total of nine treatment initiating center were available in the Amhara regional state during the study period. From those nine TICs, the University of Gondar Hospital, Borumeda Hospital, Debre-Markos Hospital, and Woldia Hospital were selected for the study. These TICs were the oldest in the region, which contributes 90% of the DR-TB burden in the region. All the selected four TIC was evenly distributed in all corners of the region. These four TICs were giving DR-TB diagnosis and treatment services not only to the Amhara region which they exist but also to the nearest region around them including Tigray, Afar, and Benishangul-Gumuz regional populations. The source population for the study was all the DR-TB patients in the Amhara region and the study population was all bacteriologically confirmed pulmonary DR-TB patients in the selected four TICs. In this study, the DR-TB patient was defined as a patient with a bacteriologically confirmed result resistant to one of these; rifampicin resistant (RR), MDR, or Pre-XDR) as confirmed by using either of the Xpert/MTB Rif test or Line Probe Assay (LPA) test or solid/liquid culture media.
Variables of the study and measurement
The dependent variable in this research was the level of sputum smear grading. After education and demonstration of patients on how to give quality sputum, three sputum samples (spot-morning-spot) were collected from all patients before they start ant-TB treatment. All the sputum were stained for acid-fast bacilli (AFB) and examined using a 100 times magnified objective lens of the Ziehl-Neelsen (ZL) method. The degree of sputum AFB positivity or grading was assigned to one of the five categories (negative, scanty, 1+, 2+, and 3+). The sputum bacillary load of was graded as negative when there were no tuberculosis bacilli per 100 field of observation, as scanty when there was one to nine bacilli per 100 field of observation, as plus one (1+) when there was ten to ninety-nine bacilli per 100 fields of observations, plus two (2+) when the observed bacilli were one to ten per field, and three plus (3+) when the number of bacilli was greater than ten per field of observation. The independent variables for this study were patients’ sociodemographic, behavioral, and clinical characteristics. Patients sex (categorized as male and female), age in years (ordered as less than 25, 25 to 44, and above 44), residency (categorized as urban and rural), educational status (ordered as no formal education, primary, secondary, and certificate and above), occupation (categorized as private business for those who had and run their own business, employed for those who had employed at the governmental or none governmental institutions, and unemployed for those individual with no job or no continues source of money), marital status (categorized as married, never married, divorced/widowed/separated), housing condition (dichotomized as homeless and having home, those patients who live in the street or lacks permanent night sleeping area were labeled as homeless), treatment supporter (dichotomized as yes or no), history of cigarette smoking (characterized as smoker and none smoker, we say smoker when the patient has a documented history of cigarettes smoking in their medical chart irrespective of the dose and duration), alcohol drinking (labeled as yes or no, patients who had a documented history of alcohol intake in their chart were grouped as yes), and khat chewing (grouped as yes or no, khat was a stimulant substance and patients was labeled as chewer when they have a documented history of khat chewing in their medical history) were the used sociodemographic and behavioral variables, while patients functional status (classified as working, ambulatory, and bedridden), anatomical site of tuberculosis (dichotomized as pulmonary and disseminated, when the TB affects only the lung parenchyma we consider as pulmonary but if it involves both the lung parenchyma and any other site we classified as disseminated), comorbidity (labeled as yes or no), Human Immunodeficiency Virus (HIV) coinfection (dichotomized as positive or negative), Body Mass Index (BMI) (ordered as < 16 kg/m2, 16 to 18.49 kg/m2, and > 18.5 k gram per meter square (kg/m2), which was calculated dividing patients weight in kilo gram to patients high in meter square), and the number of tuberculosis treatment history was the used as clinical variables.
Data processing and statistical analysis
Data was entered into Epi-data 4.2.0.0 to minimize entry error and exported to SPSS 20 statistical software for further data management and analysis. Frequency and percent were used to describe discrete variables whereas we used to mean with Standard Deviation (SD) for normally distributed and median with Interquartile Range (IQR) for skewed continuous data. The bivariable and multivariable ordinary logistic regression model was fitted. The assumption in ordinary logistic regression (proportional odds) was checked at − 2 Log-Likelihood (−2LL) by using the Chi-Square test. The null hypothesis in the proportional odds assumption stated that the effects of any explanatory variables are the same across response categories. We declare that the assumption is satisfied when the test statistics for proportional odds has a p-value of > 0.05. In the case of this study, the p-value for the proportional odds assumption test was 0.3222 and we accept the null hypothesis. The pseudo R2 values were determined to compare whether the fitted model with the explanatory variables significantly predicts the outcomes than the intercept at -2LL. The final model goodness-of-fit was assessed by using Pearson’s and deviance Chi-square statistics. These chi-squared statistics are intended to test whether the observed data are consistent with the fitted model. The null hypothesis was the model is fit while the alternative hypothesis was the model is not fit. We accept the null hypothesis when the Pearson’s and deviance chi-square statistics has a p-value of above 0.05. In this study, the p-value of Pearson’s and deviance chi-square statistics was 0.582 and 0.990, respectively. Those variables with a p-value of less than 0.2 in the bivariable model were fitted in the multivariable model and considered statistically significant when the p-value was less than 0.05.