Study design, settings and participants
Institution-based cross-sectional study was conducted from March to April, 2017 at Felege- Hiwot Referral Hospital which is found in Bahir Dar city administration, Northwest Ethiopia. Bahir Dar is a capital city of Amhara region and located 565 kms from Addis Ababa, the capital city of the country. Felege Hiwot referral hospital is the public hospital in the city. According to six month performance report of the hospital, it serves the population in the region and those from Beneshangul Gumuez population as a referral center. It provides promotion, preventive, curative and rehabilitative services to them. In outpatient chronic follow up department approximately 220 adult diabetes mellitus patients are seen weekly(21).
All diabetes patients aged greater than or equal to 18 years old and who visited the outpatient chronic follow up department for one and above year were participating in the study. On the other hand, those who were critically ill and unable to participate in the interview were excluded.
Sample size determination and Sampling procedure
The required sample size of the study was determined using a single population proportion formula. The prevalence of poor dietary practice among diabetic patients (P) was taken as 51.4% [16] and with the assumptions of: a 95% of confidence level and 5% margin of error. Based on this assumption and adding 10% non-response rate the required sample size was 404.
Systematic random sampling technique was employed to recruit a total of 404 study subjects which is based on the number of diabetes mellitus outpatients who had an appointment during data collection period in the hospital (referring from registration book). These patients were divided for representative sample size and the individual approached through calculating sampling interval (K= N/n); Where N is the total number of diabetic patient who had appointment during data collection period (880) and n is the calculated final sample size (404).
Finally, the determined samples for each date was achieved through exit interview every 2ndinterval within four weeks of working days. The starting point was selected randomly by lottery method to take as a starting patient for the interview.
Based on the decision to collect data over the course of one month, the patient coming to the clinic for a follow-up service was interviewed until the total sample size reached and one study subject was recruited only once.
For in-depth interview, purposive sampling was employed to select few participants among all diabetes patients who were on treatment follow-up at the Felege Hiwot Referral Hospital. Because of the experience of living with DM was considered very fundamental for effective deliberation on dietary practice experience, only patients who had DM at least five years and age greater than 18 years of both male and female diabetes patients were recruited to participate in the study. We tried to include patients believed to be interactive, open minded and those who were willing to participate in the study until information saturation.
Data collection procedures and quality control
Data were collected by interviewing eligible participants using a pretested and structured questionnaire. Patients were given an orientation on the protocol and specific details concerning participation in the study. Data were collected by using a pretested structured questionnaire developed by reviewing different related articles that assessed diabetes risk factors, demographic characteristics, wealth status, duration of diabetes, behavioral and social factors (10, 12, 16, 18-20, 22-30). Measurements of wealth index were developed from 2011 Ethiopian demographic and health survey (30).
Dietary adherence of DM patients was assessed by using questionnaires developed by Amelmal Worku’s modified form of the eight-item Morisky medication adherence scale (MMAS-8) (16, 18, 31, 32). It had 11 components and was computed by taking the mean value to classify the respondents follow good and poor dietary adherence. Accordingly, respondents were allowed to choose correct answers by indicating whether a given statement is Yes or No. Finally, the dietary practice of respondents was scored and computed for dietary practice variables.
Social support was measured using the Oslo-3 Social Support Scale (OSS-3) with three questions. The response categories were assessed independently for each of the three questions, and a sum score was created by summarizing the raw scores. The sum score scale ranging from 3–14, which was then operationalized into “poor support” 3–8, “moderate support” 9-11and “good support”12–14[37].
In case of qualitative data, an interview-guide questions were prepared in Amharic (local language of respondents) to elicit information from participants about their experiences, views that influenced their dietary practice. Participants were asked a series of open-ended questions that reviewed all topics systematically and encouraged participants to share their perspectives. The participants who were willing to interview were appointed for an interview at a venue of their choice. The interview was conducted face to face by the principal investigator. Audio recorder and transcribed verbatim were applied during the interview. The interviewer was probed participants to obtain additional information on interesting topics that emerged and to clarify ambiguous comments.
To ensure quality of data, pre-test was conducted using 5% of the sample on diabetes patient who were not included in the main study area before two weeks to assess instrument simplicity, flow and consistency. Amendments were made accordingly after the pre-test. Data were collected by four trained, diploma nurses and two supervisors (BSc nurse). Additionally, the completeness of the questionnaire was checked by supervisors and principal investigators every day and incomplete questionnaire was discarded. For qualitative data quality was maintained by daily data transcription. That means documenting the data both in written form and in audio recording and translates the concepts directly in word to word.
Data analysis
The collected quantitative data were coded and entered into the computer using Epi data version 3.1. Then it was exported into the windows of Statistical Package for Social Science (SPSS) version 20 for data analysis. Household wealth index was determined from asset data using principal component analysis (PCA). First, variables were coded between 0 and 1, and then the variables entered and analyzed using PCA and those variables which have commonality values greater than 0.5 were used to produce factor scores. Next, the produced factor scores were computed to produce a common factor score. Finally, the common factor score was categorized into three categories (tertile) as (poor, middle and rich). The frequencies, percentage, mean and standard deviation were described using tables and graphs. Logistic regression was applied to assess the association between dependent and explanatory variables. Both Bivariate and multivariable analysis were applied to determine factors affecting dietary practice. Those independent variables at P–Value of 0.2 in the bivariate analysis were fitted to the multivariable analysis. The degree of association was interpreted by using adjusted odd ratios with 95% confidence intervals and P-value less than 0.05 was considered statistically significant.
In qualitative data analysis, the investigator was transcribed the audio-records on the same day as the completion of the interviews enabling him to capture observations of the non-verbal points by linking the audio-recorded interviews, field notes, and the researcher’s memory of the event. Data gathered from the transcription of the interviews was organized by individual questions with the corresponding verbatim answers from the participants. The instigator was passed the following steps for analysis. 1)each recorded interview was downloaded into the researcher’s laptop; 2) recorded interviews were downloaded to discuss creating a backup copy, 3) a verbatim translation of the transcripts from Amharic into English was done by the researcher. The translation was checked by listening to the recorded interviews again whilst reading the computer files, 4) after completing the transcriptions, computer folders, download disks, and transcribe, and translate notes were labeled using identification number, date and place of the interview to make the connection back at any time when the researcher was in need, and to ensure confidentiality. The responses were coded line by line to find commonly used words or descriptions of experiences, which was labeled to characterize the answers and provide illustrative quotations reflecting these descriptions.