A package of behavior change interventions and health system strengthening activities delivered to a rural population in East Java, Indonesia, over 18 months will improve women’s diet and iron supplementation in pregnancy, breastfeeding, and complementary feeding practices; promote hygienic food preparation; and promote use of clean drinking water. The interventions will reduce the prevalence of stunting (length-for-age Z score <-2 Z) in children aged 0 to 23 months by 6% (35% in the control group to 28.9% in the intervention group) as assessed in cross-sectional surveys, compared with areas with only the routine integrated village health post services.

The interventions will (1) increase mean length-for-age Z scores, mean weight-for-age Z scores, and mean weight-for-length Z scores at 18 months in the cohort of children followed up longitudinally; (2) decrease the prevalence of low weight-for-age and low weight-for-length in baseline and end-line cross-sectional surveys; (3) decrease the number of children ill with diarrhea, acute respiratory illness, or fever in both cross-sectional and cohort assessments; (4) decrease the prevalence of anemia in pregnant women and children aged <2 years; (5) increase women’s knowledge about appropriate breastfeeding practices; (6) increase median duration of exclusive breastfeeding and the percentage of women exclusively breastfeeding at 6 months and age-appropriate breastfeeding in children aged 0 to 23 months; (7) increase women’s knowledge about appropriate complementary feeding including snacks; (8) decrease the percentage of children aged <2 years receiving snacks; (9) increase the percentage of children consuming ≥4 food groups at ages 6-8, 9-11, and 12-23 months in the end-line cross-sectional assessment and ages 7, 10 and, 17 months in children followed up in the cohort assessment; and (10) increase nutrient intakes, nutrient density, and dietary adequacy for children aged 6-23 months.

The first assessment approach will use repeated cross-sectional surveys of mothers of children aged 0-23 months and pregnant women at baseline before interventions commence and 2 years later at the end line, which is the best approach to assess the impact of the intervention on the stunting prevalence in the study population (population-level impact) [26]. In the cross-sectional surveys, we will collect information on socioeconomic and demographic characteristics, infant feeding practices including intentions of the mother to breastfeed and mother’s self-efficacy with breastfeeding, child morbidity reported by mother/caregiver, contact with the health system and exposure to the interventions, maternal and child anthropometry, and maternal and child hemoglobin. We will assess the impact by comparing the change in the prevalence of key indicators between the treatment groups in children aged <2 years and pregnant women.

The second will be a cohort study that will recruit pregnant women from a subset of study clusters and follow them from late in the third trimester of pregnancy until their child is 18 months old to examine the impact of the interventions on child growth at the individual level. This cohort study will allow us to examine subgroups for any modifying effects on the primary outcome, for example, the modifying effects of different levels of maternal education and household wealth on the impact of the intervention on the study outcomes. We will exclude women who are severely anemic (Hb<7 g/dL) or have chronic diseases such as tuberculosis, women who intend to migrate outside the subdistrict, infants born with a visible congenital defect, and multiple births (twins) from the cohort study.

In the first health system strengthening intervention, Save the Children will launch the implementation of the Baby-Friendly Hospital Initiative (BFHI) [36] in the public sector district and municipal hospitals and birthing centers at the community health centers in the intervention subdistricts. The World Health Organization (WHO) and the United Nations Children’s Emergency Fund (UNICEF) jointly conceived the BFHI in 1991 as a strategy to improve breastfeeding rates worldwide. A hospital or birth center can receive Baby-Friendly designation if they show compliance with the Ten Steps to Successful Breastfeeding [36].

In the second health system strengthening intervention, Save the Children will train village midwives, health workers, and integrated village health post cadres on counseling for appropriate IYCF using the Indonesian Ministry of Health adaptation of the WHO/UNICEF Community Infant and Young Child Feeding Counseling Package [37].

We will employ 4 different behavior change interventions to address program outcomes, as shown in the program impact pathway (Figure 1). We have developed behavior change communications using the Behavior Centered Design theory and employed emotional drivers of behavior change rather than education-focused messaging [33].

The first will be 12 participative and demonstrative behavior change activities called Emo-Demos, developed by the London School of Hygiene and Tropical Medicine, London, United Kingdom, and GAIN [33], which are based on the psychological and environmental determinants of behavior. The Emo-Demos include topics on nutrition during pregnancy (3 activities), breastfeeding (3 activities), care during pregnancy (1 activity), complementary feeding issues (4 activities), and handwashing (1 activity; see Multimedia Appendices 2 and 3 for details of Emo-Demos). The Paramitra Foundation will recruit and train village facilitators to train village midwives and health cadres to deliver Emo-Demos during monthly pregnant women classes and child growth monitoring at the integrated village health post.

The second will be a national TV campaign that will have 4 high-quality spots with messages on nutrition during pregnancy (1 spot), breastfeeding (1 spot), and complementary feeding issues (2 spots; see Multimedia Appendices 4 and 5 for a description of these TV spot messages). The TV spots will air on 5 national TV channels, and it will not be possible to contain them to only the intervention subdistricts. Village facilitators will also screen the TV spots using tablets during integrated village health post meetings and with people passing by during street visits.

The third will be the established SMS Bunda intervention, a free, text message service for pregnant women and postnatal mothers that aims to reduce maternal and infant mortality [38]. Midwives encourage women to access the service by registering their mobile number and the expected date of delivery at any time during pregnancy. Initially, in pregnancy, the women receive regular short messages about antenatal care. In the last month of pregnancy, the text message contains information related to the first days after childbirth for mother and baby, including information about early initiation of breastfeeding. In the first month of birth, the messages provide information about the health of mothers and newborns, including exclusive breastfeeding.

In the fourth intervention, GAIN will partner with a company, Nazava, to expand sales of low-cost water filters and promote appropriate water treatment and hygiene practices. Water filters are a low-cost and effective method to purify water that avoids costly fuels (wood or electricity) to boil water. Nazava’s rural sales network in the Malang and Sidoarjo districts as well as trained resellers will market the filters and provide education sessions on water treatment and handwashing.

The primary outcome for the cross-sectional evaluation is the difference in the percentage of stunted (length-for-age<–2 Z scores) children aged 0-23 months between the intervention and the comparison groups at the baseline and end-line surveys. For the cohort evaluation, the primary outcome is the difference in the mean length-for-age Z scores between the intervention and the comparison groups every 3 months from birth to 18 months, as measured in the follow-up assessments.

The secondary outcomes for the cross-sectional evaluation include differences between the 2 treatment groups at the baseline and end-line surveys in the following: (1) the percentage of wasted (weight-for-length<–2 Z scores) children aged 0-23 months and their mean weight-for-length Z scores; (2) the percentage of children aged 0-23 months with low hemoglobin (Hb <11 g/dL) and their mean hemoglobin; (3) the number of events and the mean number of days the children are ill >2 weeks before interview with diarrhea, acute respiratory infections, and fever; (4) the percentage of children aged 0-23 months who are ever breastfed, are breastfed with an hour of birth, are given prelacteal feeds, are exclusively breastfed in the first 6 months of life, continue breastfeeding at 1 and 2 years, and are age appropriately breastfed; (5) the percentage of children aged 6-23 months with minimum dietary diversity, minimum meal frequency, minimum acceptable diet, and frequency of consumption of 7 food groups and iron-rich foods; (6) the percentage of children aged 6-23 months consuming snacks the day before the interview; (7) the mean intake of food energy, protein, carbohydrate, fat, and selected micronutrients from complementary foods and the micronutrient density of the complementary feeding diets compared with desired levels at 6-8, 9-11, and 12-23 months of age; (8) the percentage of children aged 12-23 months at risk of inadequate nutrient intake; (9) the percentage of mothers with appropriate knowledge about IYCF; and (10) the percentage of mothers with high self-efficacy for breastfeeding.

The secondary outcomes for the cohort evaluation include differences between the 2 treatment groups in the following: (1) the percentage of wasted (weight-for-length<–2 Z scores) children and their mean weight-for-length Z scores every 3 months from 3 to 18 months; (2) the percentage of children with low hemoglobin (Hb <11 g/dL) and their mean hemoglobin at 18 months of age; (3) the number of events and the mean number of days the children are ill >2 weeks before interview with diarrhea, acute respiratory infections, and fever from birth to 18 months; (4) the percentage of children who are ever breastfed, are breastfed within 1 hour of birth, are given prelacteal feeds, and are exclusively breastfed each month from birth to 6 months; (5) the median maternal breastfeeding self-efficacy score and the percentage of mothers with high self-efficacy for breastfeeding at 7 days and 3 and 6 months postpartum; (6) the percentage of children with minimum dietary diversity, minimum meal frequency, and minimum acceptable diet at 7, 10, and 17 months of age; (7) the percentage of children consuming snacks within 1 hour of meals and the contribution of snacks to energy intake at 6-8, 9-11, and 16-18 months of age; (8) the mean intake of food energy, protein, carbohydrate, fat, and selected micronutrients from complementary foods and the micronutrient density of the complementary feeding diets compared with desired levels at 6-8, 9-11, and 12-23 months of age; (9) the percentage of women receiving and consuming iron and folic acid supplements during pregnancy; and (10) the percentage of women pregnant in their third trimester meeting minimum dietary diversity and at risk of inadequate nutrient intakes.

Trained field workers will collect measurements of weight, recumbent length in infants, height in mothers, and mid-upper arm circumference in pregnant women using standard methods and equipment [39,40]. We will standardize the measurements of the anthropometry field workers using established methods before and during data collection [39]. In both the cross-sectional and cohort evaluations, 2 field workers will collect duplicate measurements of the weight and length of children aged <2 years. The field workers will take a third measurement if the difference between the first 2 measurements exceeds a predetermined allowable limit (see details below).

We will calculate the Z scores for infant length-for-age and weight-for-length using the 2006 WHO Child Growth Standard [41], and we will define stunting as a low length-for-age or <−2.00 SD from the reference mean and wasting as low weight-for-length or <−2.00 SD. We will collect mid-upper arm circumference from pregnant mothers and define it as low if <23 cm [42].

The field workers will collect all anthropometry data using in-field digital data capture on tablets with routines that immediately calculate Z scores and provide them with a warning about extreme values and the need for remeasurement to check for potential errors (ie, if first and second measurement differed by >0.7 cm for length and by >0.1 kg for weight). We will refer to children with very low weight-for-height (Z score<−3) to their local community health center for assessment and treatment of severe acute malnutrition.

In the cross-sectional assessment, we will ask mothers of children aged <2 years to recall the birth weight and ask them to describe their infant’s birth size using the same question as in the Indonesian Demographic and Health Survey (DHS) [43]. There will be no estimate of the duration of gestation. In the cohort assessment, we will measure birth weight within 72 hours after birth. We will estimate the duration of gestation using the mother’s report of the first date of her last menstrual period.

In the cross-sectional evaluation, we will assess hemoglobin in all women and children. However, in the cohort evaluation we will assess hemoglobin from children at 18 months and mothers during pregnancy. We will collect capillary blood samples and measure hemoglobin using a portable hemoglobinometer (Hemocue). We will refer mothers with a hemoglobin level <12 g/dL and children with a hemoglobin level <11 g/dL to their local community health center for assessment and treatment.

In the cross-sectional and cohort assessments at the scheduled visits (Figure 3), trained interviewers will record maternal recall of symptoms from the preceding 2 weeks of common childhood illnesses (diarrhea, cough, and fever) using the standard Indonesian DHS questions.

In both the cross-sectional and cohort evaluations, we will collect information about IYCF using an interviewer-administered questionnaire on breastfeeding and general complementary feeding practices. In the cross-sectional evaluation, trained interviewers will ask questions about infant feeding practices based on questions from the Indonesian DHS that allow estimation of the standard WHO IYCF indicators [44]. These IYCF questions will include the timing of initiation of breastfeeding; current breastfeeding status; use of bottles for feeding; current use, timing, and introduction of other liquids and solid foods; meal frequency; and consumption of food groups in the previous 24 hours and week before the interview. We will also collect 24-hour dietary recalls (24-HR), which will be used to assess indicators of IYCF (see the Dietary Intake Data section).

In the cohort assessment, we will administer the questions about IYCF at the same time as we collect the anthropometric data, that is, soon after birth, every month in the first 6 months, then every 3 months until they are 18 months of age.

We will estimate indicators of IYCF according to the WHO indicators for assessing IYCF practices [44], including early initiation of breastfeeding, exclusive breastfeeding, predominant and continued breastfeeding, bottle-feeding practices, dietary diversity score, meal frequency, minimal acceptable diet, and timing of initiation of complementary feeding.

Information will be collected about mothers’ intentions to breastfeed their infants when they are pregnant. We will assess the mothers’ self-efficacy with breastfeeding using the breastfeeding self-efficacy short questionnaire developed by Dennis [45], which is a 14-item instrument aimed at measuring breastfeeding confidence, at birth and at 1, 3, and 6 months postpartum.

Quantitative dietary intake data will be collected using 4-pass 24-HR. In the cross-sectional evaluation, we will collect it at baseline and end line from pregnant women and infants aged 6-23 months, whereas in the cohort assessment, we will collect it from mothers during pregnancy (third trimester) and from infants at 7, 10, and 17 months of age. Paper-based data collection was used to collect dietary data in the cross-sectional surveys and for the pregnant women in the cohort study, whereas digital data capture was used to collect the 24-HR from children in the cohort study.

For each target group and data collection period, 24-HRs will be proportionately collected on all days of the week to account for any day of the week effects on dietary intakes. In subsamples of pregnant women and infants aged 12-23 months, we will collect repeated 24-HRs on nonconsecutive days (n=40 per target group per data collection period) to adjust nutrient intake distributions for intrasubject variability [46] when estimating the percentage at risk of inadequate nutrient intakes.

We will collect all 24-HRs in the participants’ homes by trained interviewers using standardized methods [46]. Briefly, in the first pass, each respondent will be asked to recall all foods and beverages consumed by herself (or her child) during the preceding day. In the second pass, for each food and beverage listed, we will record details about the time of consumption, the food and beverage (brand names, ingredients, and cooking methods), and any food items added (eg, sugar added to tea). In the third pass, we will ask the respondents to estimate the amount they (or their child) consumed, after probing for leftovers. Portion sizes will be quantified by weighing the indicated amounts of real foods, photographs, food models, or by the number of standard units consumed. We will develop a standard portion size manual to ensure that all interviewers use consistent methods. We will record recipes, including the amounts of raw food used in the recipe, the cooking method, and the total amounts cooked and consumed. In the fourth pass, the interviewers will review the dietary data for completeness, and they will ask whether the previous day’s diet represented a typical day and, if not, how it differed from usual, and for infants, whether they were breastfed on the previous day.

Dietary data will be collected using either a paper-based questionnaire or an in-field digital data capture on tablets for multiple-pass 24-HR that is under development. The digital data capture system saves data processing time and reduces errors by ensuring that probes and food portion size estimations are standardized. At the end of each data collection day, all dietary data will be checked for completeness by the team supervisor. Paper-based questionnaires will be entered into computer files with special-purpose data entry screens, and digitally captured questionnaires will be downloaded directly to a central server. The 24-HR records from both sources will be reviewed by trained nutritionists to monitor data quality and to select the appropriate food composition for foods not linked to an existing food composition item. We will monitor dietary data quality by applying a plausible energy range for children (by age group and breastfeeding status) to identify energy under- and over-reporting.

We estimated the targeted sample size for the study using the following assumptions: (1) in the 51 subdistricts in the Malang and Sidoarjo districts, there is a total population of approximately 4.4 million, with an average of approximately 86,000 per subdistrict; (2) assuming a crude birth rate of 17 per 1000 population per annum, 5% refuse, and 10% data losses, each subdistrict cluster would yield an average of 620 live births over 6 months; (3) an expected prevalence of stunting for children aged 0-23 months of 35% in the control clusters (based on unpublished survey data from Brebes, West Java, and Indonesian DHS data); (4) 80% power and 5% 2-sided alpha, and 1:1 ratio of interventions (subdistricts with and without the Baduta program) allocated; (5) design effect of 1.5 (lower than reported for the 2011 Indonesia DHS survey because of larger sized clusters); and (6) expected percentage point difference in the prevalence of stunting (0-23 months) between the intervention groups of 6% (35.0% in the control to 28.9% in the intervention group). (This is a relative reduction of 17%, which is lower than the 29% reduction in stunting reported in a recent meta-analysis [47] of education interventions to improve complementary feeding.)

A standard formula using established methods [48] for comparison of proportions to calculate the number of subjects required per intervention arm to test H₀: π₀=π₁ is as follows:

In the cross-sectional evaluation, the sampling design will follow a 3-stage cluster sampling procedure. In the first stage, in each of the 12 subdistricts selected for the study, we will randomly select 10 villages using the probability proportionate to size sampling method. In the second stage, we will select 2 villages from each village using simple random sampling. In the third stage, in each of the selected subvillages, we will list all households and conduct a mini census of children aged <2 years and pregnant women. Using this framework, we will use simple random sampling to select 12 children aged <2 years and their mothers and 2 pregnant women.

In the cohort evaluation, we will construct a sampling frame of all pregnant women in their third trimester from all the selected villages using the client lists of midwives and health volunteers. We will systematically contact the pregnant women on the list across all 6 subdistricts to obtain the required sample.

We have described the approach and criteria for selecting clusters in the section above on Cluster Selection and Design of Randomization. We will seek Gatekeeper consent [26] from the heads of the 2 study districts, the selected subdistricts, and the sampled villages.

In both the cross-sectional and cohort evaluations, we will recruit either women with children aged <2 years or pregnant women from the village-level sampling frames by contacting them at home. Trained field staff will explain in detail the background and objectives of the study to the women and provide a written information sheet to all women contacted. Written informed consent was obtained from women who agreed to participate in the study.

An organization not responsible for the intervention will recruit and manage a data collection team. For the cross-sectional evaluation, we will recruit 10 field coordinators, 130 interviewers, and 20 nurses or midwives to collect blood samples and to take anthropometric measurements. We will divide the staff into 10 field teams, with each team consisting of 1 field coordinator, 1 assistant field coordinator, 8 enumerators for interviews of mothers and caregivers of children aged <2 years, 2 interviewers for pregnant women, 1 interviewer for dietary data, and 2 anthropometry field workers who also collected the finger-prick blood sample for hemoglobin assessment using Hemocue. We will train the cross-sectional evaluation team for 1 week on an overview of the Baduta study, the use of the CommCare data capture app, household listing and data collection procedures, explanations of study instruments (listing forms and questionnaires), quality controls for data collection, and a field plan. The training will include a 1-day tryout to practice the household listing, interviews of mothers of children, and pregnant women using the CommCare data capture app.

For the cohort evaluation, we will recruit 2 field coordinators and 8 interviewers, all of whom will be graduates from a local nutrition academy. We will divide this staff into 2 field teams, 1 team for each district. We will train the cohort evaluation team for 1 week on an overview of the Baduta study, use of the CommCare data capture app, the schedule of cohort data collection, explanations of study instruments, anthropometry, dietary and hemoglobin assessment, and tryout of all the cohort data collection procedures. The 2-day training for blood collection will cover standard phlebotomy methods, use of Hemocue for hemoglobin assessment and its calibration, and a practice session in the field. The 2-day training for anthropometric measurements will be competency based and will use standardization procedures to check the performance (precision and accuracy) of the anthropometry field workers.

For the cross-sectional evaluation, we will conduct a household listing in each subcluster (hamlet) before the interviews to obtain a list of all children aged <2 years and pregnant women living in the subcluster. We will identify households by applying a sticker that has an identification number for each house listed. We will interview women at home but invite and accompany them and their selected children (aged <2 years) to a central location in the village to collect blood samples and record anthropometric measurements. We will use the same field methods for both the baseline and end-line surveys. We will conduct the end-line survey 2 years after the baseline survey.

For the cohort evaluation, the data collectors will visit each woman a total of 15 times from late in the third trimester of pregnancy until the child is 18 months old (Figure 3).

For the cross-sectional evaluation, we will use the house-listing form (paper based) to record the identification number and collect household location and composition, including details about each member such as date of birth and gender. We will use a registration module (part of the CommCare app) to record the name and address of the potential respondents selected based on the house-listing form. This module will then direct interviewers to the appropriate questionnaire modules for children aged <2 years or pregnant women. The respondent will be either the mother and caregiver of the child aged <2 years or the pregnant women. The questionnaires for children aged <2 years (CommCare) will record the family sociodemographic characteristics, the history of pregnancy and delivery, the history of antenatal care services, exposure to the interventions (end line only), breastfeeding practices, use of integrated village health posts, morbidity, knowledge about breastfeeding, self-efficacy for breastfeeding, socioeconomic status, food security, and use of mobile phones.

Similarly, the questionnaires for pregnant women (CommCare) will record family sociodemographic characteristics, the history of pregnancy and delivery, exposure to interventions, knowledge about breastfeeding, knowledge about food to eat during pregnancy, intention to breastfeeding, socioeconomic status, and food security. We will use a special-purpose module (CommCare) to record the anthropometric and hemoglobin measurements. We will capture dietary intake data from children aged <2 years and pregnant women using the multiple-pass 24-HR (paper-based).

For the cohort evaluation, in pregnancy and until the child is 18 months old, the field staff will collect information about the location and composition of the household, pregnancy knowledge, breastfeeding intention, infant feeding knowledge, household food security, birth outcome and delivery services, antenatal care, IYCF practices, utilization of integrated village health post services and immunizations, exposure to Baduta interventions, child morbidity, maternal and child dietary intake, maternal and child anthropometry, and hemoglobin assessment (Figure 3). We will capture dietary intake data from the children and pregnant women using the multiple-pass 24-HR (paper based for the women and electronic data capture for the children).

We will capture data electronically on Android tablets in the field using a special-purpose CommCare app developed for the study. The interviewers will record the information on structured, error-detecting forms on tablets or mobile devices and dispatch it directly to a server for cleaning and merging. The app will navigate the interviewers through the different questionnaires for women with a child aged <2 years or pregnant women or the different visit forms in the cohort evaluation. We will generate and distribute routine field team performance reports based on the data captured to help supervisors manage field activities.