An Institutional based cross-sectional study was conducted with adult TB patients in the continuation treatment phase from March 9 to May 30, 2019. It was supplemented with qualitative data to explore the lived experiences and perspectives of the healthcare providers on patient adherence during the continuation phase.
Study setting and participants
The study was conducted in 9 districts of Central Gondar Zone and Gondar town administration in Northwestern Ethiopia. Namely, Gondar Zuria, Tach Armachio, Wegera, East Dembia, West Belesa, Takusa, East Belesa, Alefa, and Gondar town administration were included in the study. The study area consists of 74 functional public health facilities (Health centers and Hospitals) serving approximately 2.9 million people in the area. Primary healthcare units that involves Health posts, Health Centers and district hospitals are mandated to provide TB treatment and care. In each health facility, a TB clinic has been established with at least one TB focal healthcare provider to regularly manage cases and follow their treatment. TB focal healthcare provider is a healthcare provider that received additional short-term training on TB and permanently assigned to TB Clinic.
The study population included all adult (≥18 years) TB patients enrolled in the continuation treatment phase as eligible participants. Patients with Multi-drug-resistant Tuberculosis (MDR-TB) and Extensively Drug-resistant Tuberculosis (XDR-TB) were not included in the study due to the distinct treatment period and approaches. All healthcare providers who were assigned as TB focal person in TB clinics were eligible for the key-informant interview.
Sample size and sampling techniques
We calculated the sample size using a single population proportion formula with assumptions, including, proportion of patients adherent to TB medications (P = 66%) during the continuation phase ; margin of error (d = 5%), and 1.5 design effect. Considering the finite population and 10% non-response rate, the final sample size was 331 TB patients selected from 22 health facilities in the 9 districts. Facilities were randomly selected stratifying by urban (town administrations) and rural settings. All eligible TB patients were included during the study period. Nine TB focal healthcare providers were purposively selected for key-informant interview from nine different health facilities.
Data collection tools and procedures
Description of interviewer-administered questionnaire and key-informant interviews
In this study, socio-demographic characteristics, disease characteristics, TB treatment-related knowledge, and attitude, provider-patient relationship using below and above the average score as a cut-off values. Family wealth quantiles were also included as predictor variables. We used an interviewer-administered questionnaire for socio-demographic, behavioral, and treatment adherence related data using Amharic (local language). But the treatment related information was taken from TB unit registers. The internal consistency of the tool was checked through piloting, hence, the Cronbach’s α value found to be 0.769, which was approaching well with the cut of value . The content and face validity were also evaluated by six senior domain experts selected from TB treatment centers, TB experts in health offices, and behavioral researchers in research and teaching institutes. We trained and employed six data collectors and two supervisors. We assessed patients’ TB treatment-related knowledge using 8 questions that include, TB curability, how to confirm cure of the disease, TB treatment period, refilling time and adherence, feeding practice, side-effects, and treatment supporter’s role. Those patients who scored above the median value were determined as good knowledge [17, 19, 29, 30]. Similarly, a 4 attitude questions that include, patient trust on TB medication, medication related misconceptions, value for medication and belief of cure were used to assess attitude towards TB treatment [19, 20]. Those patients who score above the median value of attitude questions were determined as good attitude towards TB treatment. In order to measure provider-patient relationship, we used a 6 questions that include ways of communication,  were adapted from literatures. The family wealth quantiles (lowest, second, middle, fourth, highest) was constructed from multiple items including household assets, services, and facilities . Then for simplicity of analysis, we merged into three outcomes (poor, middle and rich) by categorizing below middle quantiles into poor, middle quantiles as it is and above the middle quantiles into rich.
For the qualitative data, a semi-structured key informant interview guide was used to explore the existing provider-patient relationship and support during the continuation treatment phase. Participants were asked mainly about their opinion on which TB treatment phase is the risk for non-adherence? Why non-adherence to TB treatment was high on the specified phase? How was the communication and relationship between patients and healthcare providers? And how effective was the community-based treatment support system? The key-informant interview was recorder for audio after getting consent from each participants.
Description of assessment of the outcome
To assess adherence to TB treatment during continuation phase we deployed a short (11 questions) version of the Adherence to Refill and Medication Scale (ARMS) . The original ARMS tool consists of 12 questions with two subscales, 8 questions about medication-taking, and 4 questions about refilling . One of the items, “How often do you forget to take your medicine when you are supposed to take it more than once a day?” was not relevant to TB medication, since TB pills often are taken once a day. Each of the items were structured as a Likert scale with responses of “none,” “some,” “most,” or “all” of the time, which were given values from 1 to 4.
We used a Principal Component Analysis (PCA) technique to compute the family wealth index quantile separately for urban and rural depending on assets and services specific to the urban and rural population .
Adherence to TB treatment was measured using the short (11 questions) version ARMS values that range from 1= “none of the time” to 4=“all of the time”. One item was reverse coded then the overall adherence score ranges from 11 to 44. The lower scores indicate better adherence and the higher score represents a higher level of non-adherence, items were asking about how frequently failed to adhere to specific elements . The scales were further transformed into dichotomous outcomes using the recommended classification (scored 11 as adhered and > 11 as non-adhered) [32, 33].
We used the Variance Inflation Factor (VIF) to check the multicollinearity effect among predictor variables . We conducted a single-level analysis using binary logistic regression to identify factors that are associated with adherence to TB treatment during the continuation phase. The Crude Odds Ratio (COR) and Adjusted Odds Ratio (AOR) with 95% CI and p value < 0.05 were computed using STATA version 14 software to determine statistical significance of the association between predictor and outcome variables.
For the qualitative data analysis, we transcribed the audio data into the Amharic language by experts in the field and translated into English by fluent speakers. After familiarization with the transcript, we assigned codes inductively and deductively and categorized into themes. We examine patterns, relationships, contradictory responses, and gaps in understanding in each theme. Quotes were selected and presented for norms of the participants’ shared perceptions.