Behavior change interventions delivered through interpersonal communication, agricultural activities, community mobilization, and mass media increase complementary feeding practices and reduce child stunting in Ethiopia

Background: Appropriate infant and young child feeding practices are critical for optimal child growth and development, but in Ethiopia, complementary feeding (CF) practices are very poor. Alive & Thrive (A&T) provided intensive behavior change interventions through 4 platforms: interpersonal communication (IPC), nutrition-sensitive agricultural activities (AG), community mobilization (CM), and mass media (MM). Objectives: The aim of this study was to evaluate the impact of A&T intensive compared with nonintensive interventions (standard nutrition counseling and agricultural extension service and less intensive CM and MM) on CF practices and knowledge and child anthropometric outcomes. Methods: We used a cluster-randomized evaluation design with cross-sectional surveys among households with children aged 6-23.9 mo [n = 2646 at baseline (2015) and n = 2720 at endline (2017)]. We derived difference-in-difference impact estimates (DDEs) and conducted dose-response and path analyses to document plausibility of impacts. Results: At endline, exposure to IPC was 17.8-32.3%, exposure to AG was 22.7-36.0%, exposure to CM was 18.6-54.3%, and exposure to MM was 35.4% in the intensive group. Minimum dietary diversity and minimum acceptable diet increased significantly in the intensive group but remained low at endline (24.9% and 18.2%, respectively). Significant differential declines in stunting prevalence were observed (DDE: -5.6 percentage points; P < 0.05) in children aged 6-23.9 mo, decreasing from 36.3% to 22.8% in the intensive group. Dose-response analyses showed higher odds of minimum dietary diversity (OR: 3.3; 95% CI: 2.2, 4.8) and minimum meal frequency (OR: 1.9; 95% CI: 1.4, 2.6) and higher height-for-age z score (HAZ) (β: 0.24; 95% CI: 0.04, 0.4) among women exposed to 3 or 4 platforms. Path analyses showed a strong relation between AG and egg consumption, which led to increased child dietary diversity and HAZ. Conclusions: Delivery of social and behavior change interventions using multiple platforms was feasible and effective, resulting in improvements in CF practices and child stunting within a 2-y period. There is a need for continued efforts, however, to expand intervention coverage and to improve CF practices in Ethiopia. This trial was registered at clinicaltrials.gov as NCT02775552. A&T is an initiative to save lives, prevent illness, and contribute to healthy growth and development through improving IYCF practices. In phase I (2009–2014), A&T operated in Bangladesh, Ethiopia, and Vietnam, reaching millions of children <2 y old through large-scale social and behavior change communication interventions and achieving substantial gains in IYCF practices (13–16). The focus of phase II (2015–2017) in Ethiopia was to operationalize the Government of Ethiopia's National Nutrition Plan in one region, Amhara, to improve IYCF practices using a multisectoral approach. In 3 western zones of Amhara, A&T with Save the Children as its implementing partner worked with government health extension workers (HEWs), health development team leaders (HDTLs; a cadre of community health volunteers), and agricultural extension workers to deliver IYCF messages through interpersonal communication (IPC) and promote nutrition-sensitive agricultural activities (AG) to benefit children <2 y old. In intensive intervention areas, HEWs provided IYCF-focused counseling during health post visits and home visits and conducted food demonstrations, HDTLs provided IYCF-focused messaging during home visits, and agricultural extension workers promoted AG activities such as designating a chicken whose eggs are prioritized for a child <2 y old in the household and prioritizing vegetables from home gardens for those children (no inputs were provided as part of the program). The Ethiopian Orthodox Church priests and leaders delivered community mobilization (CM) activities such as sermons about adequate child feeding during religious fasting periods, which are common and extensive in the region, and enhanced community conversations about IYCF were led by community-based organizations. In nonintensive areas, HEWs and HDTLs provided standard nutrition counseling and food demonstrations as feasible, without additional implementation support from A&T; agricultural extension workers provided standard agricultural services; and little or no IYCF-focused CM activities were held. However, there was some spillover of activities, as A&T tools and materials were being adopted by government, nongovernmental organizations, and other stakeholders in the country. The mass media (MM) component, implemented in both intensive and nonintensive areas, consisted of a regional broadcast of radio drama called “Sebat Mela” (translated as “Seven Wisdoms”), which included 12 episodes with stories that aligned with A&T's IYCF messages, associated jingles, and testimonials of model mothers. In intensive areas with limited access to radio, supplemental activities were conducted, including broadcasting the radio drama through mobile vans with speakers and utilizing traveling performers to enact parts of the drama. Thus, A&T used 4 different platforms—that is, IPC, AG, CM, and MM—to deliver behavior change interventions to targeted beneficiaries. The intensive group received all the interventions; the nonintensive group received standard IPC and AG and less intensive CM and MM. These platforms and the specific interventions were developed based on the experiences and lessons from A&T phase I in Ethiopia. We used a cluster-randomized, nonblinded impact evaluation design with repeated cross-sectional surveys to assess the impact of the A&T intensive intervention package compared to a nonintensive program. A cross-sectional household survey was conducted at baseline (2015) and exactly 2 years later (2017) in the same communities in households with children <2 y old. This article presents findings on the primary outcomes for the evaluation (i.e., the WHO-recommended core CF practices) and the secondary outcomes of maternal knowledge about CF and stunting prevalence among children aged 6–23.9 mo. Sample size was calculated to detect differences in the primary outcomes of CF practices for children aged 6–23.9 mo, between the 2 intervention groups, considering an α of 0.05, a power of 0.80, and an intraclass correlation of 0.02, and estimated baseline prevalence of the primary outcomes. Assuming a baseline prevalence of 5.1% for minimum dietary diversity, we estimated that a total sample of 2700 children aged 6–23.9 mo (1350/group) was sufficient to detect a minimum of 6-pp difference in the proportion of children achieving minimum dietary diversity. With a baseline prevalence of 55.8% for minimum meal frequency, this total sample size was also sufficient to detect a minimum of 10-pp difference in the proportion of children achieving minimum meal frequency. A cluster was defined as a rural woreda (district). Among the total of 41 woredas in the 3 western zones of Amhara, Save the Children selected 20 woredas as possible A&T intensive areas on the basis of being first- or second-level agriculturally productive areas, not participating in the Productive Safety Net Program (a national cash and food transfer program targeted to chronically food insecure households), geographic proximity, size, and other operational aspects to ensure homogeneity across the sample. We stratified randomization by zone, and the woredas were randomly assigned to either the intensive (10 woredas) or the nonintensive (10 woredas) intervention by use of computer-generated pseudo-random numbers. All communities within an allocated woreda received the same intensive or nonintensive interventions. Households in the intensive and nonintensive areas were not explicitly informed about the results of the randomization. There was no blinding of the intervention at the level of service delivery. The primary outcomes were CF practices in children aged 6–23.9 mo, and the secondary outcomes were maternal CF knowledge and prevalence of stunting among children aged 6–23.9 mo. CF practices were measured based on the indicators recommended by WHO (23). Five CF indicators were examined: 1) minimum dietary diversity (defined as the consumption of foods from ≥4 of 7 food groups in the previous 24 h); 2) minimum meal frequency (defined as the frequency of consuming foods as appropriate for age and BF status); 3) minimum acceptable diet (defined as BF, achievement of the minimum dietary diversity, and age-appropriate minimum meal frequency); 4) consumption of iron-rich or iron-fortified foods; and 5) timely introduction of solid, semisolid, or soft foods (24). The CF indicators were constructed based on maternal previous 24-h recall of foods consumed. For the total number of food groups consumed, all liquids and foods consumed by the child during the previous day were classified into 7 food groups based on a standardized method (24). Maternal CF knowledge was assessed based on mothers’ responses to a set of 12 questions about CF. Items were validated in a previous study (25) and adapted for our study context. Each knowledge item was given a score of 1 (correct) or 0 (incorrect), and the sum was used as the CF knowledge scores (scale: 0–12). Anthropometric data were collected by using a standardized method (26). Locally manufactured collapsible length boards, which were precise to 1 mm, were used to measure the recumbent length of children. Weights of the children were measured using electronic weighing scales that were precise to 100 g. Weight and length were converted into height-for-age z scores (HAZs), weight-for-age z scores (WAZs), and weight-for-height z scores (WHZs), according to the WHO child growth standards (27). Stunting, underweight, and wasting were defined as <−2 SD for HAZ, WAZ, and WHZ, respectively. Exposure to the 4 different intervention platforms described previously (i.e., IPC, AG, CM, and MM) was defined by measures of exposure to individual interventions as follows: IPC exposure measured by mothers’ report of receiving IYCF messages during HEW home visit, health post visit, or HDTL home visit in the last 3 months; AG exposure measured by receipt of messages about raising a “baby's chicken” or a “baby's vegetable garden”; CM exposure measured by participation in a food demonstration, enhanced community conversation, or a priest sermon about child feeding and fasting; and MM exposure measured by hearing of the Sebat Mela radio program. Exposure to multiple platforms was defined as any combination of platforms exposed, with a range of 0–4 platforms. Differences in sample characteristics at baseline and endline between the 2 intervention groups were tested using linear regression models (for continuous variables) or logit regression models (for categorical variables), accounting for geographic clustering (28). For analyses of impact, we derived difference-in-difference impact estimates (DDEs) using fixed-effects regression models that assessed differences between the intensive and nonintensive groups over time (29). We present pure intention-to-treat DDEs; adjusted DDEs that control for geographical clustering and child age and sex; and models fully adjusted for geographic clustering, child age and sex, baseline characteristics that were different between groups, and characteristics that changed differentially over time. Dose–response analyses were conducted using regression models with exposure variables constructed from individual interventions and 4 categories of platforms (IPC, AG, CM, and MM). We also conducted path analyses to examine the linkage between intervention exposure to child HAZ through minimum dietary diversity. To test the accuracy of self-reported outcome measures, we measured social desirability to assess and account for potential bias in our main impact estimates on CF practices. Social desirability, the tendency of respondents to act in a manner that is viewed favorably by others, was measured with the use of a scale based on a subset of 5 items adapted from Reynolds’ short forms of the Marlowe–Crowne social desirability scale (30). Data analysis was performed using Stata 15 (Stata Corporation). Approval for the study was obtained from the institutional review boards of the Ministry of Science and Technology in Ethiopia and the International Food Policy Research Institute. All mothers of study children were provided with detailed information about the study at recruitment. Verbal informed consent was obtained from mothers prior to their participation in the survey.