Theory-based nutrition education intervention through male involvement improves the dietary diversity practice and nutritional status of pregnant women in rural Illu Aba Bor Zone, Southwest Ethiopia: A quasi-experimental study

Maternal undernutrition is a major public health problem that disproportionately affects women in low-income countries. Despite attempts to address maternal nutritional needs, Ethiopia still has a high rate of undernutrition. Hence, this study aimed to evaluate the effect of theory-based nutrition education through male engagement on dietary practice and the nutritional status of pregnant women. A pretest−posttest quasi-experimental study was conducted among 403 pregnant women selected from 22 kebeles of Illu Aba Bor zone, Southwest Ethiopia from July to December 2019. A pre-tested, structured interviewer-administered questionnaire was used for data collection. A qualitative 24-h dietary recall was used to assess dietary diversity, and the Mid-Upper Arm Circumference was used to assess nutritional status. The intervention effect was evaluated using difference-in-difference, generalized estimating equation, and linear mixed-effects models. The mean dietary diversity score differed significantly between the couple group, women-alone and the control group (p < 0.001). According to the multivariable generalized estimating equations model, couples were 3.9 times; adjusted odds ratio (AOR) = 3.91, 95% CI: (2.57, 6.88) and women alone were 2.8 times; AOR = 2.86, 95% CI: (2.17, 3.88) more likely to consume a diverse diet than the control group. The nutritional status of the women in the couple group improved significantly by the end of the intervention (p < 0.001). This study showed that involving males in nutrition education intervention was effective in improving the dietary diversity practice and nutritional status of pregnant women. The findings imply the need for targeting couples in designing nutrition education interventions. Pregnant women in rural communities of Illu Aba Bor Zone, Southwest Ethiopia, participated in this community‐based pretest−posttest nonequivalent control quasi‐experimental study. The Zone is one of the 21 zones of Oromia National Regional state located at a distance of 600 km from Addis Ababa in the southwest direction. The Zone is divided into 14 districts, with 23 urban and 263 rural kebeles (the lowest administrative unit) and a total population of 934,783, of which 467,553 are males and 465,792 are females, and lies between 340 52′12″E to 410 34′E longitudes and 70 27′40″N to 90 02′10″N latitude. Agriculture is the predominant means of livelihood in the Zone. The intervention started in July 2019 among pregnant women in the first and early second trimesters of pregnancy (Figure 2). Flowchart of the progress of the study participants through the phases of the intervention Pregnant women in their first and early second trimesters (up to 16 weeks of gestation) were included in the study. We enrolled all consenting pregnant women in their first 16 weeks of pregnancy because the intervention was to be implemented before the delivery period. Furthermore, pregnant women who had lived in the study area for at least 6 months were included to maintain homogeneity in access to nutrition‐related information and health services. Pregnant women with the diagnosis of serious health problems requiring a special diet and nutritional needs were excluded. G*Power 3.0.10 (Faul et al., 2007) software was used to calculate the sample size considering a 95% confidence level, group of three, a power of 88% for F test, 5% margin of error and effect size of 0.2. Further considering 10% allowance for nonresponse and design effect of 2, the final sample size was 373. However, the sample size calculated for the baseline survey as part of this project described elsewhere (Tsegaye et al., 2020) was larger. Hence, larger sample size was considered. In the first stage, four districts were selected. In the second stage, these districts were stratified into three according to the geographic direction and proximity to each other. The districts were assigned to either the control group or one of the two intervention groups. Then, kebeles were selected at random from the districts that were selected. Pregnant women were enumerated using house‐to‐house visits in all of the selected kebeles, and the pregnant women who fulfilled the inclusion criteria were included in the study (Figure 2). Due to the nature of the intervention, the participants and intervention implementers were not blinded to the allocation. In this study, TPB was used to identify the psychosocial processes (behaviour mediators) that lead to intervention outcomes, both for evaluation and to explain changes in dietary behaviour. The study's goal was not to test the theoretical model; rather, it was to provide guidance on the types of variables and processes that may be important in shaping maternal health behaviours and should be addressed in the intervention. The intervention aimed to improve personal and normative nutrition actions (Armitage & Conner, 2001) towards diversified diet sustenance as a result of positive nutrition attitudes and behaviour control. Participants were assigned to one of the three study arms: Couple group: husband and wife received health education together; women alone: pregnant women received health education alone and the control group: received no education. It was implemented once a month for three sessions over 3 months. Each education session lasted for 45−60 min. The language of communication during the intervention delivery was the local language (Afan Oromo). Training and educational resources were developed based on the WHO recommendation of community Infant and Young Child Feeding Counselling Package (UNICEF, 2013) and Federal Ministry of Health training manual for trainers (Federal Democratic Republic of Ethiopia, 2013). Furthermore, the formative assessment at the beginning of the project guided the development of the tool. The intervention package included (1) a training guide for the nutrition educators, (2) a manual for the nutrition educators, (3) leaflets with the core messages for the pregnant women's families and (4) counselling cards. The intervention protocol was pilot tested in a similar setting outside the study area for 2 weeks and modification was made based on the result of the pilot testing. Two public health professionals with a qualification of Bachelor of Science and two qualified in Master of Public Health were recruited as educators and supervisors of the intervention process, respectively. Educators were selected based on their previous experience in giving education services. A 5‐day intensive training was given to the educators and supervisors using the training manual. Nutrition education was given to the couples (pregnant women and their husbands together) and to the pregnant women alone. The two intervention groups were placed in groups and received nutrition education every month for 3 consecutive months. Direct (teacher‐directed) and interactive (discussion, sharing) teaching approaches guided the strategy for conducting the nutrition education sessions. The core contents were: (1) describing nutrients in foods and their role in promoting good health, (2) the consequences of inadequate nutrient intake, as well as vulnerability to the severity of the consequences of inadequate nutrient intakes and (3) key recommendations for including a variety of foods to increase intake of essential nutrients. In addition, adherence to the iron/folic acid supplement and iodized salt was also emphasized in the nutrition education. Then, the participants were encouraged to make implementation goals. One health extension worker (HEW) was assigned to each intervention village to provide counselling and support to mothers. The HEWs were trained before going to the field. They made home visits once per month for 3 months of the intervention period to counsel the pregnant women to adopt recommended practices following the nutrition education. The counsellors were provided with counselling cards. Achievements and unresolved difficulties during the visits were discussed during supervisory or monitoring visits. All pregnant women seeking ANC received nutrition education from the care providers, following a national standard protocol. Pregnant women from both the control and the intervention groups had access to this service. The fidelity of the intervention was evaluated using best practice recommendations developed by the National Institutes of Health (NIH) Behavioral Change Consortium (Bellg et al., 2004). The recommendations cover (i) study design: establishing procedures to monitor and reduce the potential for contamination between active treatments or treatment and control, as well as to measure dose and intensity; (ii) provider training: standardization of training to ensure that all providers are trained in the same manner; (iii) treatment delivery: behavioural checklists to ensure that providers adhered to the treatment protocol and (iv) treatment receipt: through supervisory visits to the study area and performance evaluation meetings with the nutrition educators and HEWs. A pre‐tested structured interviewer‐administered questionnaire adapted from different literature (Ajzen, 2019; Coates & Swindalea, 2007; CSA & ICF, 2016; Kennedy et al., 2010) was used to collect the data. The questionnaire was prepared in English and translated to the local language (Afan Oromo) and then translated back to English by language experts to keep its consistency. Training was given to the data collectors and supervisors on data collection and anthropometric procedures. A pretest was done on 5% of the sample to assess the applicability of the instrument and was verified accordingly. There were three phases to the data collection process. In the first phase, a list of pregnant women who met the criteria (sampling frame) was compiled by conducting a home‐to‐home enumeration (census) in each of the selected kebeles. During the registration of the pregnant women, the district, kebele, ‘Got’ (further subdivisions of kebeles), ‘Gare’ (the smallest community structure combined to form Got), and name of household head were considered for the subsequent follow‐up visits. The first date of the last menstrual period and a pregnancy test were used to screen the pregnant women. Second, baseline data on TPB components, food taboos, sociodemographic and economic characteristics, 24‐h recall on dietary diversity, household food security status, nutritional status (MUAC), obstetric history, health service‐related factors and household assets to compute wealth index were collected. In the third phase, end‐line data on TPB components, 24 h recalls, household food security status and other variables were collected after the intervention process. Women's attitudes, subjective norm, perceived behavioural control and intention towards diverse diet consumption were all assessed as outcome‐related psychological data. A 5‐point response Likert scale was used to assess the attitude, subjective norm, perceived behavioural control and behavioural intention. A colour‐coded scale was used using different colours with red indicating strongly disagree, yellow indicating disagree, light blue indicating neutral, light green indicating agree and green indicating strongly agree. Women's attitudes towards eating a variety of foods were examined using 15‐item questions (e.g., attitudes towards increasing food consumption during pregnancy, attitudes towards eating one more meal daily, and attitudes towards taking iron/folic acid every day). The perceived behavioural control was examined using three items. Intention to consume a diversified diet was also assessed using five questions. Additionally, subjective norms towards dietary diversity practice were assessed using three questions. Individual women's dietary diversity (WDD) was assessed using a 24‐h qualitative dietary recall. WDD is a score based on nine food groups that are recommended as a qualitative measure of micronutrient adequacy of a diet (Kennedy & Ballard, 2011). Participants were asked to recall everything they ate and drank in the previous 24 h, both inside and outside the home. In addition, they were asked to recall any snacks they had eaten in between main meals. Consumption of a food item during the reference period was assigned a ‘1’, while nonconsumption was assigned a ‘0’. The foods were then divided into nine food groups: (1) starchy staples; (2) dark green leafy vegetables; (3) vitamin A‐rich fruits and vegetables; (4) other fruits and vegetables; (5) organ meat; (6) meat and fish; (7) eggs; (8) beans, nuts and seeds and (9) milk and milk‐related products. The definition of the sum of food groups consumed during the reference period was used to calculate a dietary diversity score. The dietary diversity score was converted into tertiles, with the highest tertile defining ‘high’ dietary diversity score and the two lower tertiles defining ‘low’ dietary diversity score (Belachew et al., 2013). As there is minimal change in mid‐upper arm‐circumference (MUAC) during pregnancy, it is considered to be a better indicator of nutritional status of pregnant women than body mass index (BMI) (Tang et al., 2016) because pregnancy‐related weight gain affects the reliability of using BMI to assess pregnant women's nutritional status. MUAC measurement was taken on the left arm of participants to the nearest 0.1 cm using flexible and nonstretchable measuring tapes following the standard procedures (Tang et al., 2016). The Household Food Insecurity Access Scale (HFIAS) that has been validated and used in other developing countries was used to assess household food security status (Coates & Swindalea, 2007). The tool has nine questions enquiring about the household's previous month's experience of three domains of food insecurity: feeling the uncertainty of food supply, insufficient quality of food, and insufficient food intake and/or its physical consequences. An HFIAS score ranging from 0 to 27 was calculated for each participant and then households were categorized into different levels of food security (food‐secure and mildly, moderately, or severely food‐insecure) according to the recommendation (Coates & Swindalea, 2007). The Ethiopian Demographic and Health Survey's wealth constructs were used to determine household wealth status which included type and number of livestock (cow/oxen, chickens, goat, sheep and horses); ownership of improved sanitation (latrine) facility; type of cooking fuel; housing condition (main construction of house wall, floor and roof); the number of sleeping rooms; ownership of chair; and ownership of a working mobile phone (CSA & ICF, 2016). The wealth index was constructed using principal component analysis. The assumptions of the principal component analysis of overall sampling adequacy (Kaiser−Meyer−Olkin measurement of sampling adequacy > 0.5), Bartlett’s Test of Sphericity (p ≤ 0.05), having commonality > 0.5 and not having the complex structure correlation of ≥0.40 were checked. The latent factor reflecting the wealth index generated using principal components analysis was then classified into wealth quintiles. The affective attitudes were assessed as women’s self‐evaluation at baseline and at study endpoint that is, before and after the intervention using the same questions. To assess behaviour controllability, the women were asked if they experienced challenges with sustaining a diverse diet. Subjective norm was assessed as individual versus family involvement in sustaining a diverse diet by asking the woman to evaluate the influence of self‐initiative over family support on the sustenance of the diverse diet. The field supervisors provided onsite support for data collectors daily. All completed questioners were collected by their respective supervisors and checked overnight. Morning sessions were arranged to provide feedback on issues of concern identified from fieldwork on the next day. Doubtful interviews and anthropometric measurements were repeated by the supervisors. The logbook was used for monitoring the data collection procedures. The intensive training and supportive supervision were important aspects of quality assurance efforts. Data were entered into Epidata version 3.5.1 and exported to SPSS version 23 for analysis. The study population was described using summary statistics of mean and percentages based on the study outcomes, sociodemographic characteristics and other factors. The Pearson correlation analysis was used to investigate the relationship between the dietary diversity score, nutritional status of pregnant women and the TPB constructs. Analysis of variance (ANOVA) was used to compare means between the control and intervention groups. A post hoc test (Tukey HSD test) was employed to establish the level of significance of values between and within groups when ANOVA was statistically significant. Statistical significance was declared at a p value of less than 0.05. Difference‐in‐differences (DID) was used to estimate the average treatment effect on the treated by comparing the difference across time in the differences between outcome means in the control and treatment groups. This technique was used to control for unobservable time and group characteristics that confound the effect of the treatment on the outcome. The effect of the intervention on dietary diversity was assessed using the generalized estimating equation (GEE) with a binary logit function. First, we performed correlation on all structures, and the Quasi‐Information Criteria was the same for all correlation structures; therefore, we used an interchangeable correlation structure. The model was run to account for the clustered data and observational correlation within subjects. While fitting the model to account for the effects of various confounding variables, the unstructured covariance matrix was taken into consideration. Sociodemographic and socioeconomic characteristics, household food security status, time, intervention, time and intervention interaction were analysed. The effect of the intervention was assessed using time and the intervention interaction. The odds ratio was calculated along with a 95% confidence interval (CI). Statistical significance was determined at a p value of less than 0.05. A linear mixed‐effects model was used to estimate the effect of the intervention on changes in the nutritional status of pregnant women over time. Before fitting the model, the normality assumption of the outcome variable (MUAC) was assessed by using the Shapiro−Wilk’s test and the test showed that the assumption was satisfied (p > 0.05). We used the Akaike information criterion (AIC) to assist in selecting the appropriate statistical model. We chose the model that demonstrated the minimum AIC. Variables with p < 0.2 in the bivariate linear mixed regression model were selected as candidate variables for the multivariable linear mixed model analysis. The intervention's effectiveness was assessed by examining the interaction between time and the intervention.