Nutrition education in Southeast Sulawesi Province, Indonesia: A cluster randomized controlled study

Abstract This study evaluated the impact of a nutrition education intervention on child feeding practices and children's nutritional status. Using a randomized controlled trial, we conducted an intervention for 6 months among caregivers with children aged 6–17 months in two subdistricts of Kendari, SE Sulawesi Province, Indonesia. In all, 22 integrated health posts were randomly assigned to an educational intervention or control group with 266 participants in both groups. Participants in the intervention group attended four nutrition classes and received a monthly home visit by cadres (community volunteers), whereas participants in the control group only received standard monthly health care at the health post. The primary study outcome was children's dietary diversity scores (DDSs). Mixed model analysis was conducted to examine the intervention effects on DDS and children's growth adjusting for clustering within subvillages. The study showed the educational intervention had a significant effect on children's DDS. Children in the intervention group had a larger DDS compared with children in the control group (Beta [mean difference] = 0.34, 95% CI: 0.02 to 0.66, P = 0.038). The intervention effect on height‐for‐age z‐score (HAZ) could not be shown (Beta = 0.24, 95% CI: −0.06 to 0.56, P = 0.112). However, stunting prevalence remained stable in the intervention group but increased in the control group. These results indicated nutrition education delivered through nutrition classes combined with regular home visits by cadres as influencers provided a great potential to be adopted to complement other nutrition programmes in community health centres.


| INTRODUCTION
In Indonesia, child stunting remains a significant problem and raises serious concerns because of its high prevalence. Indonesia Basic Health Research (IBHR) reported that stunting prevalence in the last 10 years was from 30% to 37%, indicating stagnant progress. Twenty provinces in Indonesia showed a stunting prevalence higher than the national average. SE Sulawesi Province, where the present research was conducted, has been listed among the top 7 provinces in terms of the stunting rate, with a stunting prevalence of 42.6% in 2013 (Indonesia Health Ministry, 2007. The complementary feeding period from 6 to 24 months of age is characterized by the highest nutrient requirements to support optimal child growth and development (Dewey, 2016). Insufficient quantities, less varied food types and low feeding frequency adversely affect children's nutritional status. In Indonesia, stunting and underweight prevalence among children aged 0-24 months has been shown to rise with increasing age (Julia, 2009), indicating a reflection of either poor nutrition during the complementary feeding period or other factors including infection, low maternal height and education, poor sanitation and hygiene and low socio-economic status (Bardosono, Sastroamidjojo, & Lukito, 2007;Beal, Tumilowicz, Sutrisna, Izwardy, & Neufeld, 2018;Sandjaja et al., 2013;Torlesse, Cronin, Sebayang, & Nandy, 2016). Indonesia Food Consumption Survey in 2014 revealed deficient vegetables and fruits consumption among children under 5; the proportion of consumed vegetables was 48%, whereas fruits was under 10%. This survey reported that consumed animal food among children was also poor (<20%), even for seafood (Indonesia Health Ministry, 2014). Recent IBHR data reported that only 46.6% children aged 6-23 months met the World Health Organization's (WHO's) recommendation for minimum dietary diversity (MDD). The situation was worst in the age group 6-11 months as only 29.8% were fed according to recommendations (Indonesia Health Ministry, 2019).
Little is known about the factors related to stunting in SE Sulawesi, but poor feeding practices characterized by low quantity and quality of food during the vulnerable development stage are important factors associated with undernutrition here and in other regions of Indonesia. In 2018, IBHR reported that children aged 6-23 months achieving MDD totalled only 24.7% in this region (Indonesia Health Ministry, 2019). A systematic literature review on the potential determinants of child feeding practices among Indonesian children showed that at the individual level, lack of maternal nutrition knowledge, beliefs and perceptions was the barrier to achieving optimal feeding practices (Blaney, Februhartanty, & Sukotjo, 2015). A case control study in Konawe District in SE Sulawesi showed that mother's nutrition knowledge and child feeding practices were significantly associated with underweight status in children under five (Tasnim, Mwanri, & Dasvarma, 2018).
The Indonesian government provides basic health and nutrition services that reach communities from various economic statuses at subvillage levels through integrated health service posts (Posyandu).
Posyandu is designed to provide five main services consisting of registration activity, weighing, recording the result of weighing, nutrition counselling and providing certain health services, such as immunizations, Vitamin A distribution and food supplementation. All services are conducted by health volunteers (cadre) except for the last services that provided by midwifes or nutrition officers. However, from our observation in the research location, nutrition counselling services were rarely provided. The limited attention to counselling efforts is also a problem for other Posyandu in Indonesia. This is related to a lack of or low knowledge levels and skills in nutritional counselling from cadres (Khomsan et al., 2014;Nazri et al., 2015) and limited service time to divert all activities on Posyandu Day (Februhartanty, 2005).
According to the Indonesian Ministry of Agriculture Report in 2018, Kendari in SE Sulawesi is categorized as a region with a high food security index reflecting the adequate availability and affordability of food (Food Security Agency, 2018). The research site is a coastal area with abundant seafood production and is located in the centre of SE Sulawesi where villages have no barrier accessing markets (BPS, 2019). Its environment supports dietary diversity for children.
However, our formative research in the location revealed that mother's knowledge and skills in preparing complementary feeding were limited and only followed family and local community habits.
Mothers commonly introduced a single dish such as empty porridge (rice porridge without additional foods), bananas or instant cereals to their children at complementary feeding time. Providing a single dish lasts until a child reaches 8 months or even continues until 1 year.
Mother's nutritional knowledge is a crucial factor influencing her ability to choose nutritious foods to feed their children and affecting her ability to manage the available resources to provide food.
Empowering mothers through nutrition education was shown to be effective in addressing poor feeding practices (Kuchenbecker, Reinbott, Mtimuni, Krawinkel, & Jordan, 2017;Reinbott et al., 2016;Waswa, Jordan, Herrmann, Krawinkel, & Keding, 2015) and infant growth (Penny et al., 2005;Shi, Zhang, Wang, Caulfield, & Guyer, 2010;Vazir et al., 2013). As poor feeding practices are considered one of the barriers to optimizing children under two nutritional statuses in this area, efforts to provide mothers nutritional knowledge is required. Therefore, in this study, we delivered nutrition education by combining two approaches: (1) mother nutrition classes with simple and practical messages. In the mother nutrition classes, we promoted dietary diversity through combining seafood with vegetables that were locally available, affordable and part of the daily diet of local residents.
(2) Home visits used cadres as agents to strengthen the

Key Messages
• Nutrition education conducted through nutrition classes combined with regular home visits to promote diverse diets using local food can effectively improve child feeding practices concerning dietary diversity. These approaches provide a great potential to be adopted as part of nutrition programmes in all community health centres in Kendari, SE Sulawesi.
• Using Posyandu's cadres as influencers during the home visits successfully encouraged the participants' to follow the educational intervention.
• Nutrition education in the community should stress more on consuming eggs and nuts as protein sources and the importance of continued breastfeeding to support child growth until 2 years. nutrition messages. Cadres of Posyandu were expected to serve as potential influencers because cadres interact more with the community. We assumed that intervention groups receiving standard care from a Posyandu plus nutrition education would obtain better feeding practices with dietary diversity as the main outcome compared with control groups only receiving Posyandu's standard care. Also, we examined the intervention impact on children's height-for-age z-score (HAZ).

| Study sites
This study took place in Kendari, the capital city of SE Sulawesi in eastern Indonesia. The total population of this city is about 347,496 inhabitants, including about 34,626 children aged 0-4 years (BPS, 2015). The 10 subdistricts in Kendari have varying numbers of Posyandu, depending on the subdistrict's population size. Two subdistricts, West Kendari and Nambo, were purposely selected as the study sites. These sites were selected according to their characteristics, namely, an undernourished children problem, a high level of fishery production and availability and affordability of food in traditional markets to support children's diet diversity. In the study areas, nutrition officers and cadres in 29 Posyandu provided monthly health care for children under five (Kendari Health Office Report, 2017). Because seven Posyandu in two villages were previously exposed to a formative study, only 22 Posyandu were included in the intervention study.

| Study design and participants
This study was conducted using a cluster randomized pretestposttest control group design. Twenty-two Posyandu from 14 villages of Nambo and West Kendari were randomly assigned to an intervention or control group. The participants in intervention groups received standard health and nutrition care plus nutrition education intervention, whereas the participants in control groups only received standard health and nutrition care from the Posyandu. Standard care for children under 2 in all Posyandu included child growth monitoring, nutrition counselling, basic immunizations, Vitamin A supplementation and fortified biscuit for underweight children.
The criteria to participate in this study were mothers or caregivers of children aged 6-17 months listed in the integrated health service post, signed informed written consent and residing in the research area. Exclusion criteria included severe stunting (HAZ < −3SD) or severe underweight status (weight-for-age z-score [WAZ] < −3SD) among children or illiteracy of the mother. A minimum sample size of 126 per group was calculated with 95% confidence, 90% power, 1.5 as the standard deviation, 0.7 increments in child dietary diversity score (CDDS; taken from a related study in Cambodia; Reinbott et al., 2016) and the assumption of 30% loss to follow-up. In total, 266 mother-child pairs from 22 Posyandu working areas were eligible and enrolled in this study.

| Nutrition education
Before initiating the intervention, we conducted formative research in two villages in the research area using qualitative methods from March to April 2018. Using in-depth interviews, focus group discussions and observations, we collected information about local child feeding practices. We also interviewed health officers and cadres about child health problems and the complementary feeding programme in their working area. Discussion with the public health office head and nutrition officers generated a recommendation to use the cadres as agents to achieve our intervention target.
We called our educational intervention, 'Gerakan Ibu Cerdas Anak Sehat Bergizi Baik' (GEN ASIK), meaning 'Smart Mother for Healthy Children Movement'. The GEN ASIK nutrition education programme was provided using two approaches: nutrition classes and home visits.
Topics covered in the nutrition classes were taken from the Infant and Child Feeding Counselling Package developed by the UN Children's Fund for Indonesia. Findings from the formative study guided us to provide nutrition messages that fit local needs. For example, considering constant availability and low prices, only two animal sources foods-fish and eggs-were selected to promote in this intervention.
We found that mothers did not give particular food such as eggs for children under 1 year to avoid allergy, so we added food allergy topics as part of our curriculum. Also, we found that mothers had difficulty in determining the adequacy of the amount of food given to their child. Therefore, in advising the amount of food that should be given to children, we used a tablespoon as the standard to substitute plates or bowls. Topics and key messages can be found in Table 1.
The nutrition classes were held June 2018 over four sessions.
Nutrition classes consisted of education lessons with lecture method using pictures and video as media, discussion sessions to enhance participants' understanding and cooking practice. The principal researcher, in cooperation with a medical doctor and nutritionists, provided the lectures during the nutrition classes. Mothers in the intervention group attended four nutrition classes at a community venue for 2.5 to 3 h twice weekly for 2 weeks. Each participant was provided a handbook containing information on complementary feeding and food recipes. The handbook was simple, practical and easy to read because messages were arranged following age group.
At the cooking practices, mothers learned how to prepare meals containing four food groups and home-made snacks with locally available food. The research location was a coastal area where seafoods, for example, anchovies and sea urchins, were affordable. Therefore, we demonstrated how to make anchovy powder, anchovy biscuits and sea urchin cookies (bagea). Participants also practiced making healthy snacks from spinach and tempe. At the end of the cooking sessions, mothers fed their children with the snacks, and we observed together whether the child liked it or not. Continue breastfeeding, seven food groups and the benefit of food for children's health and growth, and food allergy and feeding a sick child.
• Breastfeed your child until 2 years.
• Give your child the best food available at home. • Your child can eat any food that family members eat. • Continue breastfeeding during illness.
• Give fluids and food more frequently during illness.
2 Five principles of feeding for children: Meal diversity • Mixed bubur b or rice with at least the other three groups of foods: -Add fish or eggs in children's meals daily.
-Add nuts such as tempe, green and red beans in your child's meal daily. -Add vegetables such as Moringa leaf, spinach, pumpkin and carrots in your child's meal daily. • Give healthy snacks daily such as fruits (banana, papaya, mango, orange, watermelon and coconut meat), fortified biscuits and home-made snacks. • Avoid giving sweets and salty snacks (processed food) to your child.

Texture
• Gradually increase the texture following child's age: -At 6-7 months, start with thick bubur b .
-At 8-9 months, give your child mushed food/finger food. -At 9-12 months, give your child chopped food/finger food. -At 12 months, give your child family food.

Meal frequency
• For breastfed child: -At 6 months, start with twice daily and gradually increase the frequency following the increase of child's age. -From 9 months, give your child the main meals 3 times daily and healthy snacks 1 or 2 times daily. • For nonbreastfed child: -Give your child main meals 3 times and healthy snack 1 or 2 times.
Responsive feeding • Do not force feed and be patient when feeding your child until finishing all food.
Hygiene • Wash yours and child's hands before meals. 3 Cooking practice 1 • Demonstration of preparing child's meal containing at least 4 food groups from locally available foods • Make nutritious food from food available at home: -Anchovy powder From July to October 2018, home visits were conducted for caregivers in each intervention group. These follow-up visits, were conducted once monthly, aimed to reinforce the messages that had been received by the participants during the nutrition class and established healthy complementary feeding patterns. To monitor the adoption of recommendations for feeding practices, we created a 2-day selfreporting food record form for children's consumption. A monthly health report form was also provided for recording any illness experienced by the children.
A total of 11 cadres were chosen as change agents during the home visits, and each was responsible for one intervention subvillage comprising 10-15 participants. Cadres that were recruited for this study involved women having educational attainment at least senior high school, residing in the area they were responsible for, being active cadres based on the recommendations of the health officer and committed to being change agents during the home visits activity. The 11 cadres involved in this study had previously received training to provide health and nutrition services at the Posyandu and had at least 5 years of experience as volunteers encouraging and mobilizing the community to use the Posyandu.
Before the intervention began, cadres received a 2-day training on complementary feeding practice principles and preparing children's dishes to meet WHO standards using affordable, local food. These cadres were tasked to repeat key messages, encourage mothers to practice the recommendations and collect the food record forms and health report forms from the participants. The principal researcher conducted monthly meetings with cadres at the end of the month to refresh cadres on their tasks and share problems that they found during the home visits. At the meeting time, we determined the participants who needed a joint visit with the cadre and principal researcher the following month.

| Collecting data at baseline and endline
Baseline data were collected in May 2018, and endline data were collected 6 months later. In each subvillage, mother-child pairs, selected from the health service monitoring book, were invited to health service posts. They were asked to provide their child growth monitoring books to verify their child's age. Before collecting any data, written informed consent was obtained from the mothers.
A total of 10 assistants, who obtained a bachelor's degree in public health promotion or nutrition, were recruited for the study.
They were trained concerning anthropometric measurement and the interview techniques for 24-h recall. When collecting data, the principal researcher supervised the research assistants in performing their tasks and provided feedback for the next data collection.
Data were collected including information about mother and child characteristics, sociodemographics and household backgrounds. Participants' economic status was determined based on monthly income (average cash earned monthly, not including household valuable assets, savings or aid). The original measure was collapsed in three income groups: Indonesian rupiah (IDR) less than 2,170,000; IDR 2,170,000-3,000,000; IDR more than 3,000,000. Families in the top category received income below monthly regional minimum wage in the research area. The Indonesian government sets the monthly regional wage in SE Sulawesi at IDR 2,170,000 or equivalent to 155.24 USD monthly (BPS, 2019). Household food security was assessed using the US Household Food Security Survey Module (Melgar-Quinonez et al., 2006). (1) grains, roots and tubers; (2) legumes and nuts; (3) milk; (4) flesh food (meat and fish); (5) eggs; (6) Vitamin A-rich fruits and vegetables; and (7) other fruits and vegetables. The DDS assesses how many food groups are included in the food consumed by children during the past day, with scores ranging from 0 to 7. MDD is defined as consuming foods from four or more food groups daily. For breastfeeding children, MMF is defined as being fed twice or more daily for children 6-8 months or at least 3 times daily for children aged 9-23 months.
For nonbreastfeeding children, MMF is defined as being fed four times or more daily. MAD is defined as receiving the MAD apart from breast milk. For breastfeeding children, MAD is achieved when they reach both MDD and MMF. For nonbreastfeeding children, MAD is achieved when they receive two or more milk feedings daily, achieve at least the MDD not including milk feeding and meet the MMF. All these indicators were measured using 24-h recall (WHO/UNICEF, 2010).
Anthropometric parameters were measured following standardized procedures. Children's weight was measured using an electronic weighing scale with 0.01-kg accuracy, and the supine length was measured using a horizontal board with a head and sliding foot piece with 0.1-cm accuracy. Mother's weight was measured using an electronic weighing scale that was precise to 0.1 kg, and height was measured to the nearest 0.1 cm using a microtoise. All measurements were taken twice, and all equipment were calibrated every measurement day.
Children's height and weight were converted to HAZ and WAZ using the WHO Anthro Software (WHO, 2010). A cut-off score of <−2 standard deviations for HAZ and WAZ was used to classify stunting and underweight status, respectively.

| Statistical analysis
Descriptive analysis was performed to provide general information concerning the characteristic of the study population. Differences in subject characteristics and observed variables between intervention and the control groups were tested using the independent t test for continuous variables, the Mann-Whitney test for ordinal variables or nonnormal data distribution and the χ 2 test for nominal variables.
Independent t test was used to compare score changes on dietary diversity and children anthropometry between control groups and intervention groups. Effect size was calculated for the mean score change between control and intervention group. Difference in score changes was divided by the control group standard deviation to determine effect size. The effect of the intervention on DDS as the main outcome was estimated by mixed effects models adjusting clustering within subvillages, child's age, parent's education, parent's occupation, number of children and income. We also used mixed effects models to measure other outcomes (HAZ, WAZ, length and weight) accounting for clustering within subvillages and baseline characteristics (parent's education, parent's occupation, child's age, child's sex, mother's height, number of children, birth weight and birth length). SPSS, Version 18.0 for Windows, was used for data analysis (SPSS Inc., Chicago, IL, USA).  Figure 1 displays the trial profile. At baseline, 266 mother-child pairs were eligible to participate (134 pairs in the intervention and 132 pairs in the control groups). At endline, 242 mother-child pairs were available for final measurement (126 in the intervention and 116 in the control groups). Twenty-four participants were unavailable because of migration or not being at home at follow-up visit time. Table 2 shows the characteristics of the study participants. No significant differences were found regarding any of the main characteristics between intervention and control groups.  Table 3 presents child feeding practices at baseline and endline. No significant differences were observed for child feeding practices in intervention and control groups at baseline, except for continued breastfeeding (70.6% in the intervention vs. 56.0% in the control group, P = 0.018).

| Child feeding practices
Children's dietary diversity was developed with a total score of 7. Between the 2 time points, mean DDS increased from 2.37 to 3.87 in the intervention group (Supporting information) and from 2.35 to 3.32 in the control group. The intervention group gained more points in diversity score than did the control group (1.49 ± 1.36 vs. 0.96 ± 1.31, P = 0.002). At endline, the percentage of children achieving two indicators, MDD and MAD, was significantly higher in the intervention compared with the control group (MDD: 69.8 vs. 48.3%; MAD: 57.1 vs. 22.4%). In terms of MMF, no significant change was found in both groups because it was high since baseline.
We also observed a decrease in the percentage of mothers who continued breastfeeding in both groups.
The types of food consumed by children were grouped in seven food categories. After implementing the intervention, more children in the intervention than those in the control group consumed flesh foods (81.0 vs. 65.5%), eggs (45.2 vs. 27.6%), legumes and nuts (28.6 vs. 7.8%) and Vitamin A-rich vegetables and fruits (73.0 vs. 56.9%).
Moreover, more children in the intervention group than those in the control group reported consuming healthy snacks (45.2 vs. 20.7%).
However, the percentage of children receiving sweets and salty snacks remained high in both groups (45.2 in the intervention and 65.5% in the control group). In terms of handwashing practice before preparing meals and feeding child, no significant difference was found in both groups (94.4 vs. 87.9%, P = 0.072; 92.9 vs. 88.8%, P = 0.272).
Further, the intervention effect on children's DDSs as the main outcome was analysed using a mixed model ( Table 5). The result showed a positive significant intervention effect in which children in the intervention group showed a larger DDS than children in the control group (Beta = 0.34, P = 0.038, 95% confidence interval: 0.02 to 0.66). At baseline, the anthropometry characteristics were similar in both groups. After implementing the intervention, increment in weight was significantly greater in the intervention than in the control group (1.19 ± 0.54 vs. 1.04 ± 0.53, P = 0.040), whereas increment in length did not significantly differ between the two groups (5.48 ± 1.21 vs. 5.27 ± 1.54, P = 0.261). From baseline to endline, a decreasing trend in mean HAZ scores was shown either in the intervention (−1.13 ± 1.00 to −1.52 ± 0.83) or control groups (−1.33 ± 1.08 to −1.73 ± 0.91). Change in mean HAZ scores did not have significant differences between the two groups (−0.38 ± 0.53 vs. −0.39 ± 0.64,

| Children's nutritional status
From baseline to endline, stunting prevalence was observed to increase from 28.4% to 36.2% in the control group, whereas it remained stable (19.8%) in the intervention group. Moreover, both intervention and control groups showed a reduced prevalence of underweight (14.3% to 7.1% vs. 20.7% to 16.4%, respectively). At endline, stunting and underweight prevalence were significantly lower in the intervention than the control group (P = 0.004 and P = 0.025).
Using linear mixed model (Table 5)

| DISCUSSION
In this study, we promoted the importance of applying five principles of child feeding (meal diversity, meal quantity, meal frequency, meal consistency and hygiene). Meal diversity was the main focus in which mothers were encouraged to add fish and eggs as protein sources from animals, nuts, vegetables and fruits to their children's daily diets to reach a minimum of four of seven food groups, as recommended by the WHO. In this present study, information exposure through mothers' classes, strengthened by cadres' repeated home visits, improved the mothers' awareness of the importance of appropriate complementary feeding and subsequently increased the quality of children's diet. Simple and practical messages with explanations of the benefits of each food for children encouraged mothers to adopt new knowledge, which was reflected in the more diverse diet of children.
A research on maternal nutrition knowledge and the demand for micronutrient-rich foods in Indonesia found that mothers with nutrition knowledge allocated larger budget to buy micronutrient-rich foods than did mothers without nutrition knowledge (Block, 2004). In this study, the intervention group exhibited higher achievement in diversity scores than those of the control groups, indicating the quality of children's diets increased in the intervention group. Moreover, a significant improvement was observed regarding two indicators of feeding practices, MDD and MAD, as shown in the intervention group. More children in the intervention group achieved MDD and MAD than those in the control group at endline. For MMF indicator, the proportion of children achieving in both groups already had been high since the baseline survey and did not change much afterward.
Our nutrition education provided new knowledge and information that overcame mother's incorrect feeding practice. Caregivers in the intervention group were more aware of the dangers of excessively consuming unhealthy snacks than at baseline, which could be seen in the decreased percentage of children consuming sweet and salty snacks, along with increased intake of healthy snacks. Our 24-h recall showed between the two main meals, mothers in the intervention group provided their children with snacks such as local fruits, fortified biscuits and local cakes, either self-made or purchased, mostly made from flour, coconut, bananas and eggs.
The success of several nutrition education interventions targeting improved complementary feeding practices has been well documented. Related nutrition education intervention programmes in China and Kenya, using similar approaches to those in this study (combining group training sessions, cooking demonstrations and home visits), were demonstrated to improve children's dietary diversity (Shi et al., 2010;Waswa et al., 2015). Likewise, interventions in Zimbabwe, Cambodia and Malawi showed improved quality of children's diets (Kuchenbecker et al., 2017;Paul et al., 2012;Reinbott et al., 2016). In Indonesia, a study in East Nusa Tenggara Province reported that educational intervention concerning mothers of children aged 9-16 months could improve children's nutrient intake. In this study,  (Fahmida et al., 2015).

P value n (%) Mean (±SD) n (%) Mean (±SD)
In this study, significant improvement of children's DDS in the intervention group was shown. Notably, mean CDDS remained below WHO minimum standard. For two recommended food groups, eggs and nuts, the percentage of children consuming these items remained lower than expected (<50%). Moreover, although the importance of continued breastfeeding was included as one of the nutrition messages and stressed in repeated home visits by cadres, the significant decrease was shown in the percentage of those still breastfeeding children. The main reason that prompted mothers to stop breastfeeding was low breast milk production (60%). Other reasons were pregnancy, the child wanted to stop and being busier. In this study, to avoid dissemination of complicated messages to participants as our main concern was improving the more diverse diet on children, we only inserted short explanations about the benefits of continued breastfeeding for children and mothers to support The main strength of our study was the use of Posyandu cadres as change agent to reach the target population. Cadres resided in the same villages as those of the study participants, and they knew the participants in their working area personally. This simplified the process of making repeated visits to participants' homes. In our study, we observed that cadres were a trusted source of information for mothers and they could be a connecting bridge between health services and the community. The strategy of enlisting the assistance of one cadre per subvillage as an influencer successfully encouraged participants. Participation was high in that all participants in the intervention group attended all nutrition class sessions and cooperated in the follow-up visits. Less than 10% of participants were lost to follow-up, and most participants were in the control subvillages.
Another strength of our study was its randomized controlled design, its trained team and its use of standardized evaluation procedures.
Some limitations were observed in our study. Due to the limited number of nutrition officers from the local health department, they were not directly involved in implementing educational interventions.
Nutrition classes were delivered by a study team and may have affected nutrition class sustainability. For future interventions, involving nutrition officers and midwives in the nutrition education sessions would be important as they are the primarily responsible for health and nutrition care in their working area. Furthermore, in training the cadres, we focused on providing knowledge and skills concerning appropriate complementary feeding and gave minor concern to communication skills as an influencer during home visits. For future interventions, cadre training should include communication skills in the curriculum. Another limitation was our sample size calculation did not take clustering into account. We calculated the design effect (DE) by assuming the intercluster variation to be low at 0.01 (Dreyhaupt, Mayer, Keis, Öchsner, & Muche, 2017;Shi et al., 2010).