Study settings
The study was embedded into the Haramaya Health Demographic Surveillance and Health Research Centre (HDS-HRC), which was established in 2018. The HDS-HRC located in Haramaya district. Haramaya District, located 500 km away from the capital city, Addis Ababa to the east. Haramaya district consists of 33 kebeles (the lowest administrative unit in Ethiopia). HDS-HRC covers 12 rural kebeles which is representative and randomly selected by considering geographic and environment issues. In HDS-HRC 2306 pregnant women were followed. The district has mixed farming, with the major cash crop being khat (Catha edulis Forsk) [13]. The study was conducted from January 5 to February 12, 2021.
Study design and population
A community based cross-sectional study was conducted. All pregnant women living in the district constituted the source population; whereas all pregnant women who lived in the selected kebeles for at least 6 months during the study period were the study population. The sample size was determined using single and double population proportion formulas with their corresponding assumption, and the largest sample was considered. However, this study is part of a larger longitudinal study which obtained birth outcome information in pregnant women; thus, the sample size which was used in this study was calculated from the larger study that included 475 pregnant women. Birth outcome (Low birthweight) was the main outcome was used in calculating the sample size. Sample size was determined from Charan and Biswas study [14].
Where, n = sample size, Z α/2 = 1.96 at type 1 error of 5%, Z β = 0.84 at 80% power, P1 = LBW in pregnant women with no gestational HTN (22.9%), P2 = LBW in pregnant women with gestational HTN (10.37%), p1-p2 = difference in prevalence of low birth weight between pregnant women with no gestational HTN at birth and those with gestational HTN and p = pooled prevalence = (p1 + p2)/2 [15].. A study among pregnant women in the area reported LBW prevalence of 21.0%% (LBW DD). Hence, we proposed that LBW in pregnant women with gestational HTN would be 22.9%, while those with gestational HTN would remain 10.37%,. Hence, p1 = 22.9%, p2 = 10.37%, their proportions being p1 = 0.229 and p2 = 0.1037, and p = (0.229 + 0.1037)/2 = 0.16635.
Using the above descriptive, the sample size n = 2(1.96 + 0.84) 2 × 0.16635(1–0.21)/(0.229–0.1037) 2, n = 2.152/0.01, equal 216 was calculated, which implied we needed to recruit 216 participants in each arm of the study (half in the with gestational HTN group and a half in the no gestation HTN group) making 432 participants showing a significant association between gestational HTN and LBW. Nevertheless, by adding 10% non-response rate to 475 participants were included. After constructing a sampling frame from the HDS-HRC database, simple random sampling was applied to the eight randomly selected kebeles and then the eligible women were selected using computer generated lottery method.
Data collection and measurement
Data was collected through interview administered questionnaires by trained research assistants. The questionnaire contained data on socio-economic, obstetric, maternal perception, food consumption, dietary knowledge, attitude, and practices of pregnant women. The questionnaire was initially prepared in the English language and was translated to the local language (Afan Oromo) by an individual with good command of both languages. It was also pre-tested on 10% of the sample in Kersa District before data collection. In addition, mid-upper arm circumference (MUAC) was measured to assess nutritional status.
Four measures were used to measure the dietary practices of pregnant women including dietary diversity, food variety, animal source foods consumption, and frequency of meals. In this context, the formerly validated food frequency questionnaire (FFQ) containing 27 of the most commonly lists of food items consumed by the district community was used to assess dietary diversity of the study participants [16,17,18,19,20]. Additionally, this validated FFQ was used to assess dietary diversity of the participants [21, 22]. Initially, the list of food items was established based on consultation of key informants living in the study area, who knew the culture, local language, and foods typically consumed. Then the food frequency questionnaire was pretested on 10% of the sampled pregnant women in the district who were not included in the main study and necessary modifications were made based on the observations. In addition, pretested food frequency questionnaires were carried out on 10% of the sampled pregnant women of the district not included in the main study. Necessary modifications were made before actual implementation to generate data. Finally, to measure the consumption of each food per day, per week or per month for the FFQ in the past 3 months to consider the difference of dietary consumption within a day of a week to take the concept into account. However, we considered the greater difference of dietary practice in the local community over the day of the week, the intake of each food item per day [6, 23] was not taken as a cut-off point to label consumers. In doing so, pregnant women were defined as “consumer” of a food item if they had consumed those items at least once over a period of a week [21, 24].
Validity and reliability were assessed for each construct by using the common factor analysis with oblique (Promax) rotation and were censored by the use of already validated FFQ from a similar study. The reliability or internal consistency of each scale was assessed using Cronbach‘s alpha values as the reliability estimates and ranged from 0.67 to 0.86. A Cronbach‘s alpha of 0.7 was generally considered acceptable (Nunnally, 1978). In this study, the Cronbach‘s alpha value was 0.76 as well as the test-retest method was used to determine the reliability of the instruments during the pretest. A test-retest correlation coefficient of 0.76 (CI: 0.61–0.82) was computed from the two sets of data and found to be adequate. A value of + 0.80 or greater was generally would indicate good internal consistency.
The food items in the FFQ were grouped into ten food groups. These are: cereal, white roots and tubers, pulse and legumes, nuts and seeds, dark green leafy vegetables, other vitamin A-rich fruits and vegetables, meat, fish and poultry, dairy and dairy product, egg, other vegetables, and other fruits [25]. The sum of each food group that the pregnant women consumed over a period of 1 week were calculated to analysis the dietary diversity score (DDS). Furthermore, dietary diversity score was converted into tertiles, and the highest tertile used to label “high” dietary diversity score whereas both lower tertiles combined were defined as” low” dietary diversity score. Food variety score (FVS) is the frequency of individual food items consumed in the reference period of the study. Therefore, it was estimated by the intake of 27 food items by each individual over 7 days [21],with maximum of FVS fourth. Finally, the mean FVS of pregnant women was calculated and those of them with FVS greater than the means were labeled as having “high” food variety score whereas those with FVS lower than the means were defined as having “low” FVS. Furthermore, consumption of foods from animal source (ASF) was estimated by counting the frequency of each food from animal sources that pregnant women ate over a reference period. Animal source foods score was also converted into terciles and the highest tercile used to label as “highs, while the two lower terciles combined were defined as “low” ASF.
Data quality assurance
Two days of rigorous and extensive training with the final version of the questionnaires was given to each data collector and supervisor prior to pre-test. Collected data was checked by supervisors before being sent to the data entrée on daily basis. We pre-tested the questionnaires on 10% of the sampled pregnant women of the kersa district, that were not included in the main study, and modification was done based on the pre-test observations. The supervisors kept the alleyway of the field procedures and checked the completed questionnaires daily to approve the accuracy of data collected, and the research team managed the overall work of data collection.
Data processing and analysis
Data were double entered using EPiData version 3.1 software. Data were cleaned, coded, and checked for missing and outliers, for further analysis exported to STATA version 14 (College Station, Texas 77,845 USA) statistical software. The outcome variable was dichotomized as dietary practice = 1 (appropriate) and dietary practice = 0 (inappropriate). Thus, the Poisson regression analysis model with robust variance estimate was fitted to identify predictors of the dietary practice of women. For multivariable analyses, only variables that displayed a p < 0.25 in the bivariate analyses were entered in the adjusted model. The backward regression was fitted with selected socio-economic and fertility-related variables. The results are presented as adjusted prevalence ratios (APRs) with 95% CI. The statistical level of significance was set at alpha = 5%. The explanatory variables were examined for multi-collinearity before taking them into multivariable models using correlation matrix for the regression coefficients, using the standard errors, and variance inflation factor value. Possible interactions between covariates were tested. Akaike’s information criterion (AIC) and Bayesian information criterion (BIC) were used to test for model fitness.
To estimate the economic level of the families, a wealth index was employed. The wealth dispersion was generated by applying principal component analysis. The index was calculated based on the ownership of latrine, selected household asset, quantity of livestock, and source of water used for drinking, that was to 41 household variables (Supplementary file 1). Nutritional knowledge of the women was gauged through 16 nutritional knowledge questions on the feature of nutrition needed in their course of pregnancy. Lastly, the highest tertile was defined as having “Good” nutritional knowledge and the two lower tertiles were labeled as “Poor” nutritional knowledge. The maternal attitude was evaluated with 12 Likert scale questions using PCA. The factor scores were totaled and classified into tertiles (three parts), and the highest tertile was defined as having a “Favorable” maternal attitude and the two lower tertiles were characterized as “Unfavorable” maternal attitude. The maternal perceived vulnerability of malnutrition was evaluated with 10 Likert scale questions using PCA. The factor scores were totaled and classified into tertiles (three parts), and the highest tertile was defined as having a perceived vulnerability “Yes” and the two lower tertiles were characterized as “No” maternal perceived vulnerability. Similarly, perceived severity of malnutrition, perceived benefit to healthy nutrition perceived barrier to healthy nutrition and perceived self-efficacy to control malnutrition during pregnancy were calculated by using their composite questions. Women’s autonomy was evaluated by seven validated questions which were adopted from the Ethiopian demographic health survey [22]. For each response to a question, the response to each question was coded as “one” when the decision was made by the pregnant women alone or jointly with their husband, otherwise “zero”.
Ethical consideration
All methods of this study were carried out in accordance with the Declaration of Helsinki-Ethical principle for medical research involving human subjects. Ethical approval letter was obtained from Haramaya University Institutional Research Ethics and Review Committee (IRERC) with a reference number of (IHRERC/266/2020) before the commencement of data collection. Written informed consent to participate was obtained from participants and legally authorized representatives “of minors below 16 years of age and illiterates” and their privacy and confidentiality were maintained. All personal identifiers were excluded, and data was kept confidential and used for the proposed study only.
Operational definition
Meal frequency
Is defined as how many times a day peoples eat or several daily eating occasions.
Appropriate dietary practices
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 Low FVS or low DDS or low ASF consumption [6, 23].
Wealth index quintile
Was computed from the wealth score of the households by PCA and the composite was ranked by quantile. Quintile was used to label the household’s wealth status to poorest, poor, middle, rich, and richest category.