Predictors of loss to follow-up among adult tuberculosis patients in Southern Ethiopia: a retrospective follow-up study | BMC Public Health

Study setting and period

Data for this study were extracted from March 1 to March 30, 2020, from records of TB patients who enrolled in TB care in four public health facilities (three health centers and one primary hospital) between June 20, 2016, and June 07, 2019, in Gibe Woreda, Hadiya zone, southern Ethiopia. Gibe Woreda has a total population of 141,061, among whom 50.5% were female [20]. The public health facilities provide services in various outpatient and inpatient departments and disease prevention and control activities. Diagnosing and treating TB in Ethiopia is based on Ethiopia’s national TB treatment guidelines [21]. This study was conducted in the Primary Health care Units including the health center and primary hospital. When this study was conducted, Gibe Woreda had three health centers and one primary hospital. The facilities were Homecho Primary Hospital, Omochora Health Centre, Megacho Health Centre, and Amboro Health Centre. Homecho Primary Hospital and Omochora Health Centre independently perform TB diagnosis and initiate directly observed treatment, short-course (DOTS). Whereas, the Megacho and Amboro Health Centres receive TB confirmed referral cases from the nearest highest facilities for DOTS.

Study design and eligible population

A retrospective follow-up study was conducted based on the record reviews of patients enrolled on first-line TB treatment under DOTS between June 20, 2016, and June 07, 2019. In this study, all adult TB patients (≥15 years old) who had a known TB outcome and registered on TB logbooks in public health facilities in Gibe Woreda were considered eligible. Patients whose records did not include treatment outcomes or whose patient cards indicated they have transferred out to another health facility were excluded. The study cohort was then categorized into two groups based on the main exposure variable to LTFU, which was the distance that patients had to travel from their permanent residential address to reach a health facility providing TB care. We selected distance traveled as the main exposure variable based on the evidence that it strongly predicted LTFU from TB care [13].

Ethical considerations

All methods in the study were performed following the relevant guidelines and regulations, e.g., the Declaration of Helsinki. The ethical approval is obtained from the Haramaya University College of Health and Medical Sciences Institutional Health Research Ethics Review Committee (IHRERC) with a reference number IHRERC/025/2020. As the study was based on secondary data, on behalf of the patients who received TB care in the selected facilities, informed consent was obtained from the health facility heads to access patient records. Data were collected anonymously to ensure confidentiality.

Sample size determination and sampling techniques

The sample size was estimated based on a Log-rank test comparing two survival curves using Schoenfed’s method in Stata version 16. Adult TB patients who traveled a distance of over 10 km [13] to reach the nearest health facility to receive TB care constituted the exposed group; those who traveled a distance fewer than 10 km constituted the control groups. The assumptions considered were 5% significance level, 90% power, 1:2 allocation ratio (exposed to the unexposed group), and a reference hazard ratio of 1.4 [13]. Accordingly, the minimum calculated sample size was 397 (149 in the exposed group versus 248 in the unexposed group). Adding 5% for incomplete records separately for each group, the minimum sample size considered was 418 (157 in the exposed versus 261 in the unexposed group). After assigning the minimum calculated sample size proportionally to each of the selected public health facilities based on the size of TB patients enrolled during the study period, a simple random sampling method was used to select individual records of TB patients to be reviewed (Fig. 1).

Flow chart depicting how tuberculosis patients’ records from each facility are selected. Based on the number of TB patients enrolled during the study period, June 2016 to June 2019, the sample size was allocated proportionally to the studied health facilities (one primary hospital and three health centers). The distance patients had to travel to a health facility to receive care was the primary exposure variable to the outcome variable, loss to follow-up. TB = Tuberculosis

Data collection procedure and measurement of study variables

The Data collection tools are developed based on the variables available in the registration logbook and patient card. The data extraction formats were developed based on the variables recorded in these sources. Two diploma nurses in each of the TB clinics collected the data, and the overall data collection was supervised by one of the investigators. The data extraction tool included variables on socio-demographic characteristics (age, sex, residence, marital status, occupational status, educational status, and religion), distance traveled, functional status, type of TB, previous TB treatment, DOTS follow-up center, treatment outcome, ART medication status, family support, and nutrition support.

Distance traveled (the average distance that a patient from the same sub-district traveled) to reach the nearest health facility to receive TB care irrespective of the means of transport was coded as 1″ < 10 km” and 2″ ≥ 10 km”. The measured values of this variable are recorded in the registration logbook. As health professionals working in the TB caregiving facility frequently went out for supportive supervision to their catchment sub-districts (Kebeles), they have valid information on how far a certain village is located from a given health facility.

Diagnosis and treatment of TB were made following to the Ethiopian National Guidelines on TB, Drug-resistant TB, and Leprosy [22]. The guideline suggests using sputum microscopy as a primary diagnostic tool in the absence of Xpert. Accordingly, a TB patient is said to have a smear-positive pulmonary TB (PTB⁺) if s/he has positive acid-fast bacilli (AFB) results for at least one or two initial sputum specimens by direct microscopy. Diagnosis of smear-negative pulmonary TB (PTB⁻) was established when a patient having symptoms suggestive of TB had two AFB negative test results by direct microscopy, and no response to a course of broad-spectrum antibiotics, and radiological abnormalities consistent with pulmonary TB, and decision by a clinician to treat with a full course of anti-TB or patient whose diagnosis is based on culture positive for Mycobacterium tuberculosis. When a patient had culture-proven or histopathologic evidence from a biopsy or strong clinical evidence that TB has affected body organs other than the lungs, s/he is diagnosed with extra-pulmonary tuberculosis (EPTB) and a physician decides to treat the patient with a full course of anti-TB therapy. Patients receive daily rifampicin, pyrazinamide, isoniazid, and ethambutol for 2 months (intensive phase) followed by daily rifampicin and isoniazid for 4 months or more (continuation phase). Five mutually exclusive treatment outcomes are documented including cured (confirmed smear-negative in the last month of treatment and on at least one previous occasion), treatment completed (a patient completed treatment and had no evidence of failure but without records to evidence cure), treatment failure (a patient whose sputum smear or culture is positive at month 5 or later during treatment), died (patient who dies during TB treatment), and LTFU (a patient who has been on TB treatment for at least 4 weeks and whose treatment was interrupted for eight or more consecutive weeks). Treatment success was finally defined as a sum of cured and completed treatments [22].

Nutrition support was guided by the cut-off points for body mass index (BMI) indicated in the National Guidelines [22]. Nutritional support is recommended for adults identified at admission with severe acute malnutrition (SAM) (BMI < 16 Kg/m2) and moderate acute malnutrition (MAM) (16 Kg/m² ≥ BMI < 17 Kg/m²) and had TB/HIV co-infections. Plumpy nut (an energy-dense fortified therapeutic food designed for the treatment of SAM) or plumpy sup (an energy-dense fortified therapeutic food designed for the treatment of MAM) based nutrition intervention was provided for 6 months (Plumpy nut for the first 3 months and Plumpy sup for another 3 months) for TB patients with SAM and Plumpy sup for 3 months for TB patients who had MAM at baseline [22].

Data management and statistical analysis

Data were checked for completeness and consistency and entered into Epi Data version 3.1 and exported to Stata version 16.0 for data management and statistical analysis. Descriptive statistics were computed to summarize findings and reported using numerical summary measures for continuous variables and percentages for categorical variables. Data for dates are collected and documented in the Ethiopian calendar. These have been changed to the Gregorian calendar by adding 2806 days to the dates documented in the Ethiopian calendar. The origin of time was the date of admission and the endpoint was the date event occurred or the date censoring occurred. After declaring our data in Stata as survival data, we obtained person-time months of observation, incidence rates, incidence rate ratio, compared these values by the grouping variable, distance traveled to health facilities. The event of interest was LTFU and other censored outcomes include cure, treatment completion, treatment failure, and death. The Kaplan-Meier survival curves together with the log-rank test were presented to display whether there was a significant difference in survival probability among adult TB patients by the distance they traveled to a health facility to receive TB care. The cumulative survival probabilities are presented in Life Table. To identify variables that best predicted LTFU, the Cox regression model was used. Before running the bivariable model, we run the null model to see how each variable improved the model compared to the null model. A variable with P < 0.25 and added improvement to the model prediction was entered into a multivariable Cox proportional hazards regression model. The Cox proportional hazard regression assumptions were checked using the Schoenfeld residual test which tests the correlation between residuals and the survival time; P-value greater than 0.05 indicates that the proportional hazard regression assumption was met. Using the post-estimation ‘phtest’ command in Stata, we obtained P-values well above the significance test both for the global test (chi2 = 11.98, P-value = 0.447) and also for each of the covariates (the minimum and maximum P-values obtained were 0.078 versus 0.948). All statistical tests that compared survival curves and predicted LTFU were declared significant at P-value< 0.05.