A 2-group parallel arm RCT with hospital-based recruitment and 9 months of follow-up was conducted to test the effect of a text message–based intervention for EBF promotion in participants recruited from the Central Women’s Hospital, Yangon. A total of 353 women were recruited for the study. The intervention group received breastfeeding promotional messages and the control group received pregnancy- and childcare-related messages (except breastfeeding messages) from the time of recruitment until 6 months postdelivery. The study involved conducting a formative (qualitative) study to inform the trial design and to develop text messages, recruiting pregnant women, implementing the intervention, collecting data at baseline and from 1 to 6 months postdelivery by monthly phone calls, and evaluating effectiveness at the end of the study. Figure 1 summarizes the study design of the M528 RCT using the Consolidated Standards of Reporting Trials (CONSORT) template.

We used the Health Belief Model [33] that suggests that a set of beliefs, specifically beliefs about susceptibility, severity, barriers, and benefits, leads to a particular behavior. We postulate that a woman will EBF (a particular behavior) if she believes (perceived benefits) that breastmilk has all the nutrients needed for her baby for 6 months, prevents infection, and is good for the child’s physical and neurological development. She will stop giving water and formula milk while her child is under 6 months of age if she believes that these could cause diarrhea and indigestion (perceived threat: a combination of perceived severity and perceived susceptibility). She will have higher perceived self-efficacy if she receives information on how to overcome barriers (perceived barriers) such as how to manage her family’s influences, solve breast problems, and express breastmilk when returning to work. The Health Belief Model adapted for our study is described in Figure 2; an illustration showing how the text messages relate to the Health Belief Model is described in Textbox 2 (after stage 1: preparatory phase).

The study has been approved by the Myanmar Ethical Review Committee, Department of Medical Research, Ministry of Health and Sports, Myanmar (approval number 7/ethic 2014) and the Department of Health, Ministry of Health and Sports, Myanmar (medical care-2/A-24/2014-659). The Ethical Review Committee, University of Sydney, acknowledged and approved the study based on the approval of the ethics committee in Myanmar. Furthermore, the M528 study trial is registered through Australian New Zealand Clinical Trials Registry [ACTRN12615000063516].

The Central Women’s Hospital, Yangon, was purposely selected as it is the largest tertiary public women’s hospital in Myanmar. The antenatal care clinic operates Monday through Friday and, in 2013, the hospital reported that, on average, 2000 pregnant women visited the clinic each month. The main reason for visits is to have the opportunity to deliver at the hospital free of charge because the majority of mothers attending could not afford private hospitals or clinics. In this setting, there is a higher possibility of recruiting women from a diverse range of socioeconomic backgrounds. The hospital is accredited as a baby friendly hospital initiative and follows guidelines to provide early initiation of breastfeeding. The hospital has a policy to encourage mothers to give colostrum to babies within 1 hour after birth regardless of the type of delivery.

Inclusion criteria were women from 28 to 34 weeks gestation who could access a mobile phone (Android or Java) that could display Myanmar language fonts, who had an uncomplicated singleton pregnancy, who were able to read and write in the Myanmar language, and who lived in an area with mobile network coverage. Exclusion criteria were pregnancy complications, a multiple pregnancy, and known medical conditions including mental illness that might hinder breastfeeding.

Participant eligibility was assessed via a hospital attendance registry (used by hospital staff) and antenatal care records (kept by the participant) in which information such as age, weeks of gestation, address, and phone numbers were recorded. At recruitment, researchers identified potential participants with the help of hospital nurses, explained the study, provided an information statement and consent form written in Burmese, and confirmed eligibility. If a woman agreed to participate, informed consent was obtained, and she was requested to complete the baseline survey questions. The participant’s mobile phone was checked for compatibility with text messages and, if needed, brief training on how to read and send text messages was provided. The recruitment process took 6 weeks (January to February 2015). Out of a total of 450 potentially eligible pregnant women, 353 were confirmed to be eligible and consented to participate in the trial. Each woman received 500 Myanmar kyats (US $0.50) to compensate for sending a text message to inform the research team of her delivery date.

Eligible women were randomized to the intervention or control group according to an allocation sequence generated by a computer program [34] with a block size of 16 to ensure a balanced group size. Each study and group number was kept in an opaque envelope to conceal the allocation sequence and securely stored. At recruitment, research team members sequentially took an envelope for each consenting participant and conducted the baseline interview. After the participant’s information was recorded, the envelope was opened and the group allocation was recorded.

Because the intervention could not be blinded, we minimized bias by ensuring both groups received messages and by not explaining in the consent and information sheets how the text messages differed between groups. The information statement provided only general information that each participant could be allocated randomly into a group; would receive messages containing information on pregnancy, childcare, and breastfeeding practices; and that the frequency of messages could vary from 1 to 3 times per week. Research staff and interviewers (via telephone) were blinded to each participant’s group status although they might guess the group based on participants’ responses. To minimize bias, we also blinded research staff to the trial aims and hypotheses.

Stata version 13.0 (StataCorp LLC) was used for sample size estimation and data analysis. Sample size was calculated with the assumptions of 80% power, 5% 2-sided alpha, and 13% expected loss at follow-up. The EBF rate of Myanmar children at 6 months of age was 15% [4], and we hypothesize that EBF in the intervention group will be increased 2-fold. The estimated sample size was 312 (156 per group) and the final sample size requirement was 353 participants to allow for 13% loss at follow-up.

Baseline data were collected from participants during recruitment. Follow-up (outcome) data were collected by monthly phone calls at 1 to 6 months postdelivery. Process evaluation (evaluation survey and in-depth interviews) was undertaken at completion of the study. Quantitative data were collected on Android tablets using the Dimagi CommCare app [35] to reduce errors and facilitate data processing. All questionnaires were pretested prior to transferring to tablets. Full details of the baseline, follow-up, and process evaluation measurements and timing of each measurement are shown in Table 1.

Baseline questionnaires comprised instruments which have been validated and tested for reliability and were adapted from the 2011 Myanmar Multiple Indicator Cluster Survey [4] 2015, Myanmar census questionnaires [36], the 2011 Demographic and Health Survey, Bangladesh [7], and a 2012 study from Shanghai, China [37]. At baseline, participants were asked to provide information on their socioeconomic status, previous pregnancies and breastfeeding history, knowledge and sources of breastfeeding information, confidence to breastfeed and perceived self-efficacy to breastfeed, intention to breastfeed, and intended breastfeeding patterns.

We will examine the relationship between breastfeeding duration and self-efficacy using self-report Likert scales (no confidence= 1, somewhat confident= 2, sometimes confident= 3, confident= 4, very confident= 5) [38,39]. With 5 levels of responses to 33 items, total scores range from 33 to 165 and will be categorized as low self-efficacy (1 to 82) and high self-efficacy (83 to 165).

Follow-up comprised a postdelivery survey, feeding follow-up assessments, morbidity, breastfeeding self-efficacy, and social desirability assessments. Questionnaires have been adopted from the Global Strategy for Infant and Young Child Feeding [6] indicators for assessing infant and young child feeding recommendations by WHO [38-40], breastfeeding efficacy scales [41,42], a short version of the Marlowe-Crowne Social Desirability Scale [43], and unpublished questionnaires being used in ongoing projects in Bangladesh (MJD and ML) and Indonesia (MJD).

The postdelivery form was completed when the child was 10 days old and was used to assess his or her postdelivery status, including the date, place, and type of delivery and the child’s birth weight, sex, and perinatal outcomes. The feeding follow-up form was completed every month and asked if the child was breastfed and if other liquids or foods were given in the last 24 hours and over the preceding month since the last contact (or birth for the first follow-up). Each participant completed 6 assessments of their feeding practices, including detailed information about the different types of liquids and foods (Textbox 1).

Child morbidity status and self-efficacy were assessed when the child was 1, 3, and 5 months of age using items from a validated instrument [7,41,42]. Child morbidity questions were used to record whether the child has had any signs or symptoms such as fever, cough, cold, running nose, rapid breathing, difficulty in breathing, chest indrawing, and diarrhea and dysentery. If one of these was present, we further explored the duration of illness. Breastfeeding self-efficacy was collected at 1, 3, and 5 months postdelivery to ascertain the mothers’ confidence to breastfeed, the relationship between self-efficacy and breastfeeding duration, and whether text messages improved breastfeeding self-efficacy. Studies report that women with a higher perceived self-efficacy for breastfeeding tend to initiate breastfeeding and persist even through challenges, and we hope that women in our study will increase their confidence or self-efficacy to continue EBF because of the knowledge gained from the text messages they received [41,42,44,45].

We also assessed social desirability status, which is the tendency respondents have to answer questions in a way that is viewed favorably by others (such as research team members). This can interfere with the interpretation of average tendencies as well as individual differences [43]. By measuring social desirability status, we were able to assess whether follow-up data were influenced by bias. We used the validated and reliable short version of the Marlowe-Crowne Social Desirability Scale, which is based on a subset of 13 items adapted from Reynolds and Gerbasi [43].

Before the start of the trial, research team members received training by the principal investigator on how to collect data by tablet, conduct recruitment, implement group allocation, conduct follow-up phone calls, and monitor times to call participants. The schedule of enrollment, interventions, and assessments is shown in Table 2.

Implementation was organized in 2 phases, preparation and intervention service delivery.

In designing the control group, we followed the recommendations made in a review article on the design of control groups [54]. To minimize intervention bias, we sent nonbreastfeeding-related text messages with mainly pregnancy- and childcare-related content but at a lower frequency (once per week). Messages were developed after reviewing the literature about pregnancy and childcare education, including WHO online resources [55,56] and information and communication materials about infant feeding produced by the Department of Health [50] and Save the Children, Myanmar [51]. Messages included the importance of following recommended pregnancy care such as taking iron and folic acid regularly, following a balanced diet, getting regular exercise, making regular health center visits, and identifying danger signs before delivery. Examples of message topics after delivery include following recommended baby immunization schedules, how to practice proper infant care, and how to protect the baby from mosquitoes to avoid malaria. Each control participant received an estimated 38 messages (1 per week) over 9 months, which were informed by findings from the formative study [16].

The primary outcome is the rate of EBF at 1 to 6 months of the infants’ age measured at monthly intervals after delivery. Secondary outcomes are median duration of EBF and rates of early initiation of breastfeeding (within 1 hour after birth); predominant breastfeeding; current breastfeeding; bottle feeding; and early introduction of solid, semisolid, or soft foods at 1 to 6 months of infant’s age (measured at monthly intervals) [38-40]. We used WHO infant and young child feeding definitions for this study, and definitions of these outcomes are described in Multimedia Appendix 1 [38].

An evaluation of the use of mobile phones to promote breastfeeding practices was conducted with participants from the intervention group. Information on text delivery success, user experiences (user-friendliness), acceptability (trust), comprehension, new information learned, and feedback from received messages was assessed by trained callers during phone calls at 5 to 6 months after delivery. We measured acceptability (proportion of participants who trust the messages), comprehension (proportion of participants who can describe the last message received), and new information learned (proportion of participants who indicate they learned new information about breastfeeding). These instruments have been widely used in mHealth evaluations in other countries [20,57-60] and took about 10 minutes to administer.

At completion, a qualitative study was conducted with intervention group participants. Based on information from follow-up calls, participants were split into 3 subgroups: EBF for 6 months, predominant and other types of breastfeeding (excluding EBF), and breastfed for less than 1 month only. In-depth interviews were held with approximately 25 women (7 to 9 women from each group) using semistructured guidelines. We explored user experiences in receiving SMS text messages, perceptions about the number of messages received, message delivery success, acceptance of the service delivery model, and effect of messages on breastfeeding practices. We expected that this number of participants would allow us to reach saturation point [60]. There were variations in the numbers of participants we could interview based on breastfeeding outcome, participant’s availability, logistic reasons, and feasibility of the study.

Quantitative data were collected via tablets and automatically submitted to the Dimagi server from which we could monitor data and generate reports. Stata software version 13 (StataCorp LLC) was used for data analyses. The intervention and control group outcomes will be compared using intention-to-treat principles [61]. Descriptive analyses were used to summarize participant baseline characteristics and breastfeeding practices. Likert scales were analyzed using Mann-Whitney U tests or cumulative odds ratios and proportions between groups using chi-square tests. Duration of EBF was compared by survival analyses with log-rank tests for between-group comparisons. If differences in baseline characteristics of the groups are identified, multivariable analyses will be undertaken. We report results using the CONSORT guidelines for reporting of randomized trials [62]. As follow-up data collection had to rely on the mothers’ self-reporting alone, we adjusted the potential social desirability bias [43] in our analyses to confirm the accuracy of self-reported outcome measures.

For process evaluation, we used Stata software for survey data analysis. We conducted descriptive analyses to summarize participant baseline characteristics and their experiences and feedback about receiving text messages. Text message delivery success was measured by the frequencies of messages received, and proportions were compared with chi-square tests. For the qualitative study, we used thematic analysis. All digitally recorded interviews were transcribed verbatim in Burmese in Microsoft Word and were reviewed to check for accuracy and translated to English. A list of thematic codes was developed by MPH, which was independently reviewed and verified by AA. The data were then be manually coded by MPH for emerging themes, which were again verified by other investigators and the most relevant themes are summarized in a document.

We hope to achieve our outcomes as SMS text messaging is increasingly being applied to improve reproductive, maternal, neonatal, and child health with growing evidence of its effectiveness [20-28,31,32,59,60]. This study is important as our intervention will leverage the power of mobile phone technology at an appropriate time following the opening up of the telecommunications market in Myanmar. Incorporating feedback from the community and mobile service providers prior to the study was a key element in the study design. We could envisage similar larger scale interventions implemented within the health system with women recruited from other public and private hospitals or community settings in the future. We also acknowledge that a number of factors may interfere with the success of the intervention. We anticipate the possibility of mobile coverage variation across geographic areas as the launch of mobile operators only started in 2014. Another barrier is that people often change phone numbers in order to receive better network and promotional prices leading to difficulties in sending text messages and making follow-up phone calls. Successfully upscaling the intervention would have the potential to improve child nutrition and consequently to reduce neonatal and infant morbidity and mortality where causes are associated with suboptimal breastfeeding practices. With the endorsement of Sustainable Development Goals, nutrition is central to 12 of the 17 development goals [63] and the role of breastfeeding is essential in achieving these goals. We also anticipate that continued growth in the coverage of the mobile network could attract the Ministry of Health and Sports and policy makers to implement mHealth-based behavior change interventions in national programs.

A key strength of our study is the use of an RCT design to assess the impact of text messages on improving breastfeeding practices. Although the trial cannot be fully blinded, research team members who recruit participants and conduct follow-up phone calls are blinded to group allocation. By sending messages with different content to both control and intervention groups, we reduce the possibility of indirect effects of participants receiving health messages on the targeted behaviors. Both groups receive pregnancy and child healthcare-related text messages, but only the intervention group receives messages about breastfeeding. The intervention is feasible to implement because of the significantly reduced prices of subscriber identity module (SIM) cards, increased availability of cheap smart phones, and improved coverage of a 3G network. In addition, female literacy is high with 94% of urban women and 84% of rural women able to read text messages [36]. The timing of the study may allow new investments for mHealth-based EBF promotion interventions.

Limitations include recruitment from only 1 hospital, which will reduce the generalizability of results. To date, only major cities in Myanmar have reliable access to the mobile network but this is changing rapidly as the network is expanded, offering opportunities to adapt the intervention for rural populations. The hospital selected for recruitment, which has a patient population of women from diverse ethnic and socioeconomic backgrounds, is the largest public hospital providing free quality delivery care service. Women come from all over the country, including from rural and urban slum areas. It is likely that the participants have similar feeding practices to women from other areas as feeding practices only differ slightly between states and regions [4]. We acknowledge that the intervention group has more frequent SMS text messages than the control group and it could be a limitation of the study. However, we assume that the difference in SMS text message frequency might not constitute an intervention. Several studies including studies conducted in Australia [22], Iran [64], and Korea [65] delivered text messaging service once a week to the intervention groups and findings reported that these interventions were effective.

Another limitation is that there may be technological challenges when participants are not familiar with the use of mobile phones. However, we have compensated for this by providing training. As with other mobile phone programs, we anticipate a possible low response rate because participants may not answer calls, may switch off their phone, or may change their phone number. Solutions to these problems include recording alternate contact numbers during recruitment, calling repeatedly if the phone is powered off, or trying to reach the participant based on the suggested time to call.

This is the first Myanmar RCT to test the effectiveness of mobile text messaging (mHealth) in promoting EBF practices. Our results will help determine whether text messaging is an effective and feasible method for promoting appropriate feeding practices and will inform further research to assess how this model could be replicated in the broader community.