The role of nutrition-sensitive agriculture combined with behavioral interventions in childhood growth in Ethiopia: An adequacy evaluation study

Objective: The study aimed to investigate the role of nutrition-sensitive and specific interventions along with nutrition education on child stunting during the first 1000 days in Ethiopia. Methods: An adequacy evaluation study was used to see changes between the baseline and end-line data after following for 1 year. A sample of 170 mother-child pairs who had a 1-year followed up was used to detect differences. We performed structural equation modeling to elucidate changes in feeding behaviors, socioeconomic status, water, sanitation and hygiene on child linear growth. Furthermore, the independent effect of covariates on child linear growth was handled using a general linear model. Results: A total of 170 and 270 mother-child dyads were interviewed at baseline and end-line surveys, respectively. After about 1 year of intervention, the annual rate of stunting prevalence declined from 29.3% (95% confidence interval [CI] = 18.6, 42.7) to 16.4% (95% CI = 10.7, 24.2). There was a significant change in the mean of length-for-age Z-score which changed from −1.18 to −0.45 (P <.034). Adjusting for the different constructs of the health belief model, child sex, age, feeding behaviors, and dietary diversity, one egg consumption per day was responsible for the most significant variability explained (36%) for stunting reduction. Conclusions: Sustainable access to egg consumption for children below 2 years experienced a substantial reduction in childhood stunting. A combination of nutrition-sensitive agricultural and direct nutrition interventions along with behavioral-based education is a sustainable strategy in reducing and preventing child growth from faltering in the early life stages. We established multidisciplinary experts from agriculture, health, communication, and social protection sectors to bring a behavioral change to prevent malnutrition and to have healthy individuals. Given the causes of malnutrition by its nature are multifaceted and intertwined 13 ; we planned to design the intervention that integrates both nutrition‐specific and nutrition‐sensitive agriculture activities. Hence, the intervention process and the Health Belief Model (HBM) 14 provided the conceptual framework for the design and development of detailed compartments of the intervention packages. The intervention framework integrates theory, practical evidence, and verifiable findings from reports and information collected from the target population to develop culturally appropriate and empirically sound interventions. 15 The six steps of intervention mapping were properly done to evaluate whether the goals of the intervention were achieved or not. Then we evaluated the intervention performance and objective alignment by contextualizing local area features regarding knowledge, behavior, predisposing, and enabling factors (existing opportunities and threats) that attribute to wide practices of sub‐optimal feeding patterns, deep‐rooted malnutrition, and morbidities. A conceptual framework showing the inter‐linkage between the expected outcomes and mediating factors for the improvement of child growth was modified and developed. 13 , 16 , 17 The best practical evidence from the literature 18 , 19 , 20 and a summary of theoretical methods 21 were used in the selection of existing methods and techniques. The HBM was used as the focus of behavioral change during the selection of behavior‐based methods and strategies which has been widely employed in health and nutrition behavior change studies. 22 We implemented the intervention in two selected Kebeles (namely, Abaso Kotu and Kolla Motie) with high malnutrition burden and rural food in‐secured areas of South Wollo Zone, Ethiopia. Lactating mothers with their children below the age of 12 months were recruited to receive the intervention packages and followed up for 12 months. All lactating mothers with infants younger than 12 months and pregnant women and lived for at least 6 months were included in the study. All of the participants who fulfilled the inclusion criteria were registered for the intervention. We carried out adequacy evaluative intervention, which is recommended to compare the outcome in the target group with either previously defined goal or with changes observed in the target group following the intervention program, 23 , 24 to measure the contributions of our intervention in improving and sustaining maternal and child nutrition. The baseline survey was carried out in May 2019 and the end line survey was taken in April/May 2020 after 12 months of the period but with the same season. The intervention was designed, by the end of the intervention program, to achieve the reduction of stunting by 20%. The sample size was determined using STATA version 15 by considering the percentage of change in length‐for‐age Z‐score (LAZ) from a study in Malawi. 25 The baseline mean LAZ score was −1.81 with a SD of 1.1, and the end line SD and correlation were 1.15 and 0.65, respectively. We had also assumed that 95% of confidence level, 80% of power and 10% of loss to follow up. Then the final sample size was found to be 170 mother‐child dyads. During the baseline survey, a minimum of the above sample size was enrolled. However, any woman who delivered after the baseline survey was also registered and provided the intervention packages. As a result, during the end‐of‐intervention survey, there were 270 mother‐child dyads. The interventions were implemented in the first 1000 days plus of critical lifespans. To induce lasting behavior change, ensure continuity, and enhance the linkage between communities and health facilities, secondary target groups consisted of farmers, family members, women development army (WDA), agricultural development army or agent (ADA), health extension workers (HEW), and Kebeles administrators had been incorporated in the intervention. The nutrition education package was composed of four components, namely (a) education and counseling of mothers or caregivers along with the provision of vegetable seeds and egg‐laying pullets, (b) training of HEWs, ADA, farmers and WDA about diversified food production via organic agriculture, modification, complementary food preparation, and feeding behaviors. (c) Bimonthly home visits and supervision of community‐based nutrition mentors (HEWs and ADA), (d) sensitization workshops, meetings, food festivals, and experience‐sharing through the community‐field visits. Draft versions of the package were discussed with ‐district health and nutrition experts, village health committees and fine‐tuned by the research team. The intervention primarily applied nutrition and health education intervention that aimed to change feeding behavior, hygiene conditions, improving livelihood status, diversified diet, and nutrition. Frontline agents (HEWs and ADA) compiled all lists of program beneficiaries and called to participate in nutrition education which was voluntary and consent was sought from the caregivers before the commencement of the sessions. The intervention included five sessions (see Table S1) with both group training and demonstration. The education was designed based on the findings obtained from the baseline survey and reshaped in accordance with recommendations from the Food and Agricultural Organization (FAO), 26 World Health Organization (WHO), 27 and United States of Children's Fund (UNICEF) 28 training resources. We developed the training materials both in the local language (Amharic) and the English language. Key messages on organic production, breastfeeding, complementary feeding, hygiene, and nutritious diet were published and distributed in the form of leaflets, brochures, posters, and manuals and the food pyramid was used to describe local available food groups customized from the national nutrition strategies of Ethiopia 29 and during each session, every participant was given posters, leaflets, and manuals to be used at their residential address. We made individual home‐based follow‐up sudden visits to assess the adoption of all participants to the training given following each session, to reinforce appropriate practices, and to correct mal‐practices. The follow‐up visits were conducted by the researcher and respective health workers and developmental agents from the intervention villages. In addition to nutrition education, we also provided at least six egg‐laying pullets, five types of vegetable seeds (cabbage, tomato, carrots, lettuce and spinach), and varieties of complementary foods (legumes, nuts, and some cereals), which are not produced in the areas but have the potential for intensified production in these environs. The success of the nutrition education package was evaluated via baseline and end‐line survey findings. Extensive process evaluation was also being performed to document the reach, dose, and fidelity of the intervention. The baseline survey questionnaire, with additional questions targeting the nutrition education intervention, was administered during interviews with the caregivers. We recruited six data collectors and trained them on how to gather the information. The questionnaire comprised of socio‐demographic characteristics, the socioeconomic conditions, food security status, morbidity occasions, child breast and complementary feeding practices, maternal behavior regarding child feeding, water, sanitation and hygiene (WASH), and child anthropometric indices. The socio‐economic status of the households was assessed by computing the wealth index based on the housing condition, the main source of drinking water, types of latrine, the main type of fuel, land ownership, and household assets (electricity, television, radio, watch, phone, bed, car, wagon, bicycle, bank account, and goat/sheep/cow/ox/horse). We performed principal component analysis to classify the economic status into quintiles. 30 The change in food security status was also measured by the validated tool of the FAO guideline to categorize as food secured, mildly, moderately, and severely food secure. 31 , 32 To measure the difference in the child and maternal hygiene before and after the intervention, we assessed WASH components using nine‐item questions with a score of one for those who fulfilled recommendations and zero for those who did not practice it properly. The WASH index was computed using the principal component analysis to divide the hygienic condition of children and mothers into terciles. 33 The contributions of the intervention on maternal or caregivers' behavioral change on feeding practice were assessed using the HBM, 34 which contained six components: perceived susceptibility, perceived severity, perceived benefit, perceived barriers, cue to action, and self‐efficacy. To identify any observed change in behavior in the six constructs, we ran confirmatory factor analysis and judged based on the differences detected from the baseline findings. The child dietary diversity score (CDDS) and minimum dietary diversity for woman (MDD‐W) were determined from the 24‐hours recalls. 35 , 36 The CDDS score ranges from zero to eight, and children who eat foods from five or more food groups daily were met the minimum recommended dietary diversity. 35 The variety of foods consumed by the children was further analyzed in detail using 14 food items. We also computed age‐specific infant and young child feeding (IYCF) indicators such as early initiation of breastfeeding, exclusive breastfeeding (EBF), initiation of complementary food, minimum meal frequency (MMF), minimum dietary diversity (MDD), minimum acceptable diet (MAD), and the proportion of children who consumed iron‐rich food, continued breastfeeding and malpractices related to child feedings. MMF is defined as the proportion of breastfed and non‐breast‐fed children 6 to 23 months of age who receive solid, semi‐solid, or soft foods (including milk feeds for non‐breast‐fed children) the minimum number of times or more the previous day. The MMF was defined as two times for breastfed children aged 6 to 8 months, three times for breastfed children aged 9 to 23 months, and four or more times for non‐breastfed children aged 6 to 23 months. MAD is a composite indicator calculated from two fractions: breastfed children 6 to 23 months of age who had at least the minimum dietary diversity and minimum meal frequency during the previous day; and non‐breast‐fed children 6 to 23 months of age who received at least two milk feedings and had at least the minimum dietary diversity not including milk feeds and the minimum meal frequency during the previous day. 35 MDD‐W was determined using a guide recommended by Food and Nutrition Technical Assistance Version‐3 (FANTA‐III) from grains, roots, and tubers; pulses; nuts and seeds; dairy; meat, poultry, and fish; eggs; other Vitamin A‐rich fruits and vegetables; dark leafy greens and vegetables; other vegetables; and other fruits. 36 The woman consumed five and above food groups per 24‐hours of duration met the minimum recommendation. To assess child nutritional status, weight, mid‐upper arm circumference (MUAC) and length were measured during the baseline, monthly on the subsequent follow‐up periods, and end‐line survey. Weight was taken using a well‐calibrated and adjusted salter scale in the kilogram to the nearest 10 g and one digit after decimals. Recumbent length in centimeters was measured with a portable wood‐made sliding board and recorded to the nearest 1 mm. MUAC was taken at the mid‐point of the right arm of each child as recommended by WHO. Each measurement was obtained twice and results were averaged. If results were >3% discrepant, then a third measurement was obtained. Results were converted to Z‐scores using WHO Anthro. 37 Data were managed via digitalized electronic devices; stored as a form of comma‐delimited file and exported to STATA/SE version 15 (StataCorp LP, College Station, Texas) for analysis. Frequency distributions were done to identify outliers. The descriptive statistics were done separately for the baseline and end‐line data sets (meaning 170 mother‐child pairs at baseline and 270 mother‐child pairs at end line) (Table 1). However, for other statistical tests (paired t test, χ 2 test, structural equation modeling (SEM) and general linear model (GLM) paired data set was used to detect the change. Changes in socio‐demographic characteristics, food groups consumed, feeding behavioral change between the baseline and end‐line were tested using a Paired t test for continuous variables, and χ 2 test for nominal variables. Individual dietary diversity scores of children and women were calculated according to WHO specifications of 8 and 10 food groups, respectively. 35 , 36 Percentage distributions of target intervention participants in rural District of Dessie, North Central Ethiopia 2020 Reliability analysis was performed for composite variables such as wealth index, WASH index, maternal behavior on child feeding and anthropometric measurements. Cronbach’s alpha was used to assure the internal consistency of items which was greater than 0.7 for each composite variable. We also checked coefficient of variations, coefficient of reliability and technical error of measurement (TEM) for evaluating the validity of anthropometric measures. Anthropometric indices such as weight‐for‐length Z‐score (WLZ), length‐for‐age Z‐score (LAZ), weight‐for‐age Z‐score (WAZ), Body mass index‐for‐age Z‐score (BAZ) and mid‐upper arm circumference‐for‐age Z‐score (MUACZ) were generated using WHO Anthro version 3.2.2 growth standard. 31 Biologically implausible values based on WHO‐recommended cutoffs at 6 SD were eliminated. The prevalence of stunting, wasting, and underweight were compared between the baseline and end‐line. Changes in the mean score of the LAZ, WLZ, and WAZ were tested using a paired t test. The current average annual rate of reduction of stunting (AARR) was determined as AARR = 1 − (P t+n /P t ). Where P t+n is the latest prevalence of stunting after 1 year, P t is the starting year prevalence of stunting, and n is the number of years between them. SEM was used to predict the status of child stature; path analysis (confirmatory factor analysis [CFA]) was utilized to assess the direct and indirect relationships of the observed and unobserved variables of health belief constructs of HBMs with child growth. The analysis was managed using AMOS 23. The health beliefs, feeding behaviors, and hygiene conditions were investigated for the presence of a mediator effect on child malnutrition. The degree of correspondence between the conceptual model and actual data was evaluated using a good‐of‐fit test. The cut‐off criteria to consider the model a good fit to the data included CFI >0.90, TLI >0.90, RMSEA and a standardized root mean square residual (SRMR) <0.06. 38 There were modest increases in the factor loadings of items that fulfill the assumption of CFA, each of them was greater than 0.7 (GFI = 0.913, CFI = 0.97, TLI = 0.96, RMSEA = 0.048 and SRMR = 0.036). Higher factor scores of observed and unobserved variables were seen with end‐line survey result which has a relatively higher number of participants. General linear models with random intercepts and robust standard errors were used to assess for continuous repeated LAZ‐score of children between the baseline and end‐line data. The effects of covariates for the difference in child LAZ (T1‐T2) were evaluated through this model with the repeated measures of analysis of variance (ANOVA) at 95% of confidence level and coefficient of determination. The statistical significance level was declared at a P‐value of less than .05.