Study area and period
The study was conducted in West Gojjam Zone from May to August 2018. The zone is one of the eleven zones in Amhara Region. A region is the first-level administrative division in Ethiopia whereas zone is the second-level administrative division of the country. Zones are further divided into a number of woredas. Woredas (the third-level administrative divisions in the country) further divided into kebeles (the smallest administrative unit in the country).
West Gojjam Zone has fifteen woredas with a total population of 2,641,240, of which 50.7% were females. The number of estimated pregnant women was 61,072. Teff, maize, millet, bean, pea, grass pea, pepper, barley, wheat, cabbage, collard green, tomato, potato, papaya, mango, and avocado are foodstuffs commonly produced in the zone. Cattle, chicken, sheep, and goat are the common live-stocks in the study area (23).
A community-based cross-sectional study was done among pregnant women. Quantitative data complemented with qualitative method. A quantitative data collection followed by a qualitative method to explore and explain the result of the quantitative data.
Source population and study population
All pregnant women in the zone were the source population, whereas pregnant women in selected kebeles were the study population.
Inclusion criteria: Since dietary practice is affected by the local culture, pregnant women who lived in the kebele at least for six months were included in the study. Gestational age before 16 weeks was another inclusion criterion because this is a baseline study of a cluster-randomized controlled community trial which was registered in Clinical Trials.gov; its ID number is NCT03627156. Title of the trial is “the effect of guided counseling in improving selected nutrition outcomes of pregnant women in West Gojjam Zone, Ethiopia: A cluster randomized controlled community trial.”
Exclusion criteria: pregnant women who had confirmed/diagnosed hypertension and/or diabetes mellitus were excluded from the study because nutrient requirements and dietary practices of hypertensive and diabetes cases are different from their counterparts.
Sample size determination
The sample size was calculated using single and double population proportion formulas with their respective assumptions. The sample size calculated by a single population proportion formula using the following assumptions gave the largest sample size used in this study: 95% confidence level, the proportion of appropriate dietary practice from the previous study was 34.5% (13), 5% marginal error, 5% non-response rate, and design effect 2. The calculated sample size was 712.
Cluster sampling technique was used to select pregnant women. From the fifteen woredas in the zone, eight of them had nutrition intervention program implemented by a nongovernmental organization. Therefore, these eight woredas were excluded from the study because their dietary practice may not be representative of the dietary practice of all pregnant women. From the seven woredas, three were selected by simple random sampling (SRS) method. Again, sample clusters (kebeles) were chosen using the SRS method. Ten clusters from Bahir Dar Zuria Woreda, six clusters each from South Achefer and Burie Zuria Woreda were selected based on proportional to the size allocation.
House to house survey was conducted to identify eligible women in selected clusters and all of them were included in the study. The women’s pregnancy status was assessed by enquiring her last menstrual period and confirmed with a pregnancy test.
Data collection and measurements
Data were collected by interviewer-administered questionnaires. The questionnaire includes socio-demographic variables, obstetrics history, dietary related variables, food security status, maternal perception, and socio-cultural issues. Six female nurses and three male public health professionals were recruited as data collector and supervisor, respectively. Additionally, three female laboratory technicians were recruited to do pregnancy tests. Data collectors administered the questionnaire through a face-to-face interview at the participant’s home. To maintain optimal privacy of the mothers, other family members didn’t have free access to the place where the interview was conducted.
Food frequency questionnaire (FFQ), which was taken from previously validated questionnaires (24-26), containing 54 food items was used to gather dietary data. The questionnaire was also validated after assessment of locally available foodstuffs. For the food frequency questions, the women were asked about the frequency of consumption of each food per day, per week or per month in the prior 3 months by taking the variation of dietary intakes within days of the week into consideration (25, 26).
Food items of the FFQs were grouped into nine food groups: 1. cereals, roots, and tubers; 2. Vitamin-A-rich fruits and vegetables; 3. other fruits; 4. other vegetables; 5. legumes and nuts; 6. meat, poultry, and fish; 7. fats and oils; 8.dairy; and 9. Eggs (24).The consumers of a food item were defined as consumption of a food item at least once a week (26). The number of food groups the women ate within a week were counted to analyze dietary diversity score (DDS).
Food variety score (FVS) was computed by counting the individual food items the women consumed within a week. Then, the mean FVS was analyzed. Utilization of animal source food (ASF) was assessed by counting the frequency of each animal source foods the women took within the days of a week. Finally, the frequency of ASF consumption was divided into terciles (three parts). Dietary diversity score, FVS, ASF consumption and frequency of meal were used to assess dietary practices.
Food security status was assessed using 27 questions adapted from household food insecurity access scale which was previously validated for use in developing countries (27). Food secure households experienced fewer than the first 2 food insecurity indicators. Whereas, a household which experienced from 2-10, 11-17 and > 17 food insecurity indicators were considered as mildly, moderately and severely food insecure households, respectively.
Wealth index of the household was determined using Principal Component Analysis (PCA) by considering latrine, water source, household assets, live-stock, and agricultural land ownership. The responses of all non-dummy variables were classified into three parts, and the highest one was coded as 1 and the two lower values were given code 0. In PCA those variables having a commonality value of greater than 0.5 were used to produce factor scores. Lastly, the score for each household on the first principal component was retained to create the wealth score. Quintiles of the wealth score were created to categorize households as poorest, poor, medium, rich and richest.
To determine edible crops and vegetables production, each crop and vegetable species cultivated by the household were counted. Then, classified into three parts and the highest value was labelled as high production and the two lower values were considered as low production. In the study area, khat, eucalyptus, “Gesho” (a local plant used to make tella (local beer) and areki), pepper and onion are the common cash crops. Count of these cash crops was used to decide cash crop production.
The total ownership of live-stock was measured by Tropical Livestock Units (TLUs). Currently, there is no TLU index created specifically for Ethiopia, therefore, the indexes for Tropical Africa was used in this study. The TLUs were calculated using the following weighted index factors: cattle= 0.7, horses= 0.5, mules=0.5, donkey= 0.5, sheep= 0.1, goats= 0.1, chickens= 0.01(28).
Women’s autonomy was assessed using eight questions. For each question, code one was given when a decision was made by the woman alone or jointly with her husband, otherwise zero. The mean was used to classify a woman’s decision making power (16)
Maternal knowledge on diet during pregnancy was assessed using 12 questions. Code one was given when the response was correct, or else zero. The attitude was assessed by twenty Likert scale questions using PCA. The factor scores were summed and ranked into terciles (three parts). Then the highest tercile was labeled as a favorable attitude, if not unfavorable attitude.
Subjective norms, intention, perceived susceptibility, severity, benefit, and barriers were assessed using their respective composite questions. Mean was computed and women who score above the mean for each variable were categorized as having subjective norms, intention, perceived susceptibility, severity, benefit, and barriers, otherwise no.
Quality Assurance Mechanisms
To maintain quality of the research, the questionnaire was adapted from standard data collection instruments and it was pretested. Data collectors, supervisors and laboratory technicians were trained for three days. Data collection procedure was supervised by the supervisors and principal investigator.
The data collection team holds a daily meeting and feedback was given daily. Double entry and verification were done by the principal investigator. Important assumptions were checked using the standard procedures. The dietary practice of pregnant women was the dependent variable for the quantitative data whereas independent variables were socio-demographic, enabling, predisposing, reinforcing and obstetric factors.
Data processing and analysis
Quantitative data were edited and coded manually. Then, entered into Epi Info version 7.2.2 software and exported to SPSS version 23 software for cleaning and analysis. The frequency of meal, FVS, ASF consumption, DDS and dietary practice were determined.
First, a generalized linear mixed model was fitted to account the contribution of cluster-level variables but the result showed that the model was not appropriate to analyze this data. Thus, bivariate and multivariable logistic regression analyses were used to assess predictors. Hosmer-Lemeshow Goodness-of-fit-test was done to check model fitness. Correlation between independent variables was checked using the Pearson Correlation Coefficient. P-value < 0.2 was used as a cut-off point to select variables for the final model. Backward elimination was used and P-value < 0.05 was considered statistically significant.
Definitions of terms
Food is any nutritious substance that people eat or drink
Appropriate dietary practice: when women had at least four meals daily, good FVS, high DDS and high ASF consumption whereas it was inappropriate when women had less than four meals daily or poor FVS or low DDS or low ASF consumption (25).
High DDS: tertiles were calculated from food groups and the highest tertile was considered as a high DDS, whereas the rest two lower tertiles were taken as a low DDS (26).
Food variety score: women who had above the mean food variety were considered as having good FVS, otherwise, having poor FVS (25).
Utilization of ASF: the highest tertile for ASF consumption was considered as the high frequency of ASF consumption whereas the two lower tertiles were taken as the low frequency of ASF consumption (26).
Knowledge on the maternal diet: women who score >=75% and >50% from knowledge questions were considered as having high and medium knowledge score, respectively. Or else labeled as having low knowledge score (12).
The dietary practice is observable action of dietary habit which is the lived experience of an individual. Therefore, a phenomenological qualitative study was conducted using key-informant interviews and focus group discussions (FGDs). Three FGDs involving 6-12 participants in each group were conducted in Amharic (local) language. Twelve FGD participants were mother-in-lows. Eight FGD participants were husbands, and six were health professionals.
Seventeen key-informant interviews were carried out among pregnant women, health professionals, nutrition officers (nurses and pharmacist in their profession), community leaders and one-to-five network member leaders (one-to-five network is a lower level governmental structure consisting of five people. One leader monitor, mobilize and train five members about health, nutrition and developmental related issues). All participants were Orthodox Christians and Amhara in Ethnicity. The respondents’ age ranged from 25-55 years. Educational status of the participants ranged from having no formal education to a primary degree.
A typical case and extreme sampling techniques were used to select the study participants for the key-informant interviews. FGD participants were chosen using a homogeneous sampling technique.
Qualitative data were collected using semi-structured interview and FGD guides, prepared by reviewing existing literature and the local context. Moreover, interview and FGD guides were revised based on identified gaps during interviewing or facilitating. During data collection, tape-recorder was used in addition to field notes. All interviews were conducted by the principal investigator at a convenient time in public meeting halls in the village. The interviews took from 30 to 45 minutes. The FGDs were also facilitated by the principal investigator and took from 45minutes to 1hour. Note taker who has participated in qualitative data collection before this research was recruited to take notes during the interviews and FGDs.
Transcription was done on the same day of data collection. During the interpretation of findings, individuals’ detailed views of the meaning of events were considered. To assure trustworthiness key-informant interviews were triangulated with FGDs and quantitative data were triangulated with qualitative method.
First, the researchers repeatedly listened to tape-recorded interviews and FGDs. Then, verbatim transcription was done in Amharic language. Transcribed interviews and FGDs were translated from the Amharic to English. Finally, data were entered into the QDA MINER LITE version 2.0.2 software for analysis. Codes were given for similar ideas using the software and cross-checked between researchers, used to establish analytical categories. Then, the meanings were organized into themes for thematic analysis. Finally, the result was presented in narratives in triangulation with the quantitative results and illustrations.
Institutional Review Board of Bahir Dar University approved the study. Permission was obtained from West Gojjam Zone and each woreda administrators. Written consent (fingerprint for women who cannot read and write) was secured from the study participants. Confidentiality was maintained by excluding personal identifiers from the data collection form and keeping the data in a locked board.