Study Setting and Participants
This study was registered as Clinicaltrials.gov #NCT02638571, and the protocol, along with baseline results, have been published . Briefly, this cluster-randomized intervention trial for community-based nutrition education was done in 12 kebeles/villages, selected from two districts of the Sidama Zone, Southern Ethiopia. Kebeles were randomly assigned to the intervention and control groups after stratification by districts using the lottery method, as the prevalence of child malnutrition and number of children was different in each district. A total of 772 mothers with children aged 6-15 months were recruited initially at the baseline. The total number of participants at baseline was 771 as one child was excluded due to not fulfilling the inclusion criteria. At the midpoint of data collection, the total number was 692 (354 in the intervention group and 338 in the control group), and at end point it was 621 (307 in the intervention group and 314 in the control group). Mothers who have apparently healthy breastfeeding infants aged 6-15 months who were permanent residents in the area included in the study. Children who were receiving supplemental or those that were severely or moderately malnourished and had started therapeutic food were excluded from the study. Children who started therapeutic feeding excluded because their weight gain or improved situation would not show the effect of the intervention. The study was not blinded, because the districts were far apart (it was not possible to walk between kebeles and back in one day), did not share markets, health centers, and health posts, and study personnel did not overlap between areas.
The Intervention and Education Materials
Key messages were developed based on the Theory of Planned Behaviour (TPB) and Health Belief Model (HBM) principles [18,19]. Health Extension Workers, two of whom were located in each kebele, were provided with nine months of additional nutrition education, along with the usual health education. The HEWs provided the mothers in the intervention group with five main lessons. An intervention with recipe demonstrations on preparation of porridge for complementary feeding using germinated pulse and cereal was given once a month and repeated again after midpoint (4.5 month) data collection. In addition, participating mothers in the intervention group were counseled by HEWs during house-to-house visits. The additional messages delivered to the intervention group were not included in the usual health education delivered to the control group. The control group received the usual health education provided in the area, which is mainly based on the essential nutrition action messages.
All HEWs are trained for 1 year before deploying for their services in their local community. They trained on Family Health as one of the training packages where a general nutrition education covered. For this study, a Training of Trainer (TOT) manual was used to provide additional training on pulses to HEWs in the treatment kebeles but not in control kebeles. This manual was developed and used by the Canadian International Food Security Research Fund (CIFSRF) for the "Scaling-up Pulse Innovations for Food and Nutrition Security" project . Key messages included in the TOT manual were the importance of consuming food from all food groups and dietary diversification; the benefits of pulses; household pulse processing and preparation techniques, and the need to prepare and cook a variety of pulse-based dishes, including pulse-cereal mix complementary food. HEWs were trained for three days with demonstrations. At the same time, HEWs in the intervention group had refresher training in communication and counseling skills. In addition, HEWs were trained to use a quick guide when counseling mothers during house-to-house visits . In the control sites, HEWs continued to provide routine health education. These HEWs had not been specially trained in using pulses in complementary food.
Before the intervention was introduced, the training material and counseling poster were pre-tested on purposively selected mothers to assess whether the content and format were realistic, understandable, culturally appropriate, visually appealing, and motivating. These mothers from the Hawassa Zuria district, who did not participate in the actual study, were provided with a half-day education and their understanding of the messages was assessed through discussion. Each picture on the poster was also assessed for its cultural acceptance.
The KAP of mothers regarding pulse consumption and feeding practices were collected at the baseline, midpoint, and end point of the intervention period. A standardized questionnaire was used to assess the mothers' intentions to use cereal-pulse incorporated complementary food. Theory of Planned Behavior  and the Health Belief Model (HBM) was used to frame questions to assess the KAP of mothers based on the guidelines of Macias and Glasauer .
Dietary Diversity and Growth Assessment
Using a structured questionnaire, the mothers were asked about the type and number of meals consumed by their young children in the previous 24 hours . In addition, the Dietary Diversity Score (DDS) for each child was calculated based on the World Health Organizations (WHO) guidelines for measuring individual dietary diversity scores, using the following food groups to calculate the DDS: 1) grains, roots, and tubers; 2) legumes and nuts; 3) dairy products (milk, yogurt, cheese); 4) flesh foods (meat, fish, poultry, and liver/organ meats); 5) eggs; 6) vitamin-A rich fruits and vegetables; and 7) other fruits and vegetables . The response of mothers was recorded as “Yes” if they said the child ate the particular food and “No” if they said the child did not eat the food. The answer “Yes” was recorded as 1 and “No” recorded as 0 and a sum of the total number of food groups consumed was calculated. The mothers also asked to estimate the amount of the food the child eats using locally used equipment for each child and the proportion of children consuming four or more food groups per day was determined. In addition to the number of meals, the frequency of the children's pulse consumption was assessed using a frequency questionnaire to evaluate monthly consumption of pulses.
The anthropometry of the children was taken at baseline, midpoint, and end point using standardized techniques . In brief, weight was measured using an electronic scale (Seca 770), and the children were draped in a light cloth of known weight during the measurement. The recumbent length was measured to the nearest 0.1cm using the Shorr measuring board. The Middle Upper Arm Circumference (MUAC) of the left arm of young children was measured using arm circumference insertion tape. All anthropometric measurements were entered and analyzed using WHOAnthro version 3.2.2.
Assessment of Socio-Demographic Characteristics
Data on the socio-demographic characteristics of the participants, including those of the participants' household, such as age, gender, ethnicity, income, and KAP of mothers, were assessed using a standard questionnaire adopted from previous studies [15,25] with modifications. To assess the food insecurity of the households in the study area, a standardized questionnaire adapted from Food and Nutrition Technical Assistance (FANTA), the "Household Food Insecurity Access Scale," was used . As suggested by Ballard et al., 2011, only the last three questions of the nine included to analyze food insecurity. These questions have been validated in low-income countries to measure household hunger. These three questions comprise the Household Hunger Scale (HHS). Food insecurity was assessed with a recall period of the last four weeks (30 days) prior to the data collection.
Household wealth status was measured by an asset-based (non-monetary) wealth index adopted from CSA . During data collection, each participating household reported assets owned and other housing and sanitation-related characteristics. These included ownership of a radio, TV, mobile phone, TV, and bicycle, access to electricity, and quantity of livestock, land size, and level of income. Housing characteristics used in the wealth index calculation include the dwelling's structure, number of rooms and bedrooms, and ownership (whether it is privately owned or rented). Each household received a score of 1 or 0 depending on whether it had the particular asset (1= yes and 0=No). Each binary variable was then weighted by the inverse of the proportion of households that owned the particular item or had the particular characteristics .
Haricot Beans for Women's Empowerment in Household Decision Making
Researchers associated with the larger project funded by CIFSRF attended an intervention nutrition education and demonstration session just prior to the midpoint data collection where mothers explained that although they understood the benefits of feeding their children pulses, they could not fully provide pulses as complementary food to their young children due to a shortage. At this time (late May and early June 2016), much of the population were affected by flooding that occurred due to an extended drought in the area. These climatic changes had prevented planting and/or reaping of pulses during the first harvest. The researchers decided, after the midpoint data collection to provide each of the mothers in the intervention group with a single gift of a two kg bag of quality haricot bean seed and a two kg bag of fertilizer to plant during the June-July planting season. Both intervention and control groups received pulse seeds and fertilizer for future planting from the study, with controls receiving seeds after the end point.
The women agreed to plant the seeds after a training session. Agriculture experts from Hawassa University's College of Agriculture (partner institution) trained 386 mothers for one day on techniques of planting, applying fertilizer, and weeding. The mothers in the control group were later provided with one kg of haricot bean seed at the end of end point data collection. The provision of a small amount of a new variety haricot bean seed was meant to enable smallholder female farmers to improve their wellbeing and that of their families.
Data were entered into SPSS version 20 software. Chi square and repeated measures Analysis of Variance (ANOVA) were used to investigate relationships between the pre- and post-intervention data on KAP of mothers, and growth and DDS of their children. ANOVA was used to compare means between the control and intervention groups, and when ANOVA was statistically significant, a post hoc test (Tukey HSD test) was used to determine the level of significance of values between and within groups. A value of p < 0.05 was considered as statistically significant.