Background
The health benefits of exclusive breastfeeding (EBF), especially in low- and middle-income countries (LMIC), have been well documented. These include protection against infectious diseases, improved measures of intelligence and decreased risk of becoming overweight and developing diabetes. Globally, EBF could save the lives of an estimated 823 000 children under the age of five each year. Yet, despite World Health Organization recommendations that children should be exclusively breastfed for the first six months of life, EBF rates hover at only 37% in LMIC.
South Africa has one of the lowest rates of EBF in the world. Estimates of EBF rates under six months in South Africa range from 8% to the most generous estimate, 32% reported in the South Africa Demographic and Health Survey 2016. The health problems that result are largely due to unsafe formula feeding and the widespread practice of introducing solid foods early (often before three months) into the infant's diet. In South Africa, more than 70% of infants are given solid foods before reaching six months, the recommended age for initiating complementary feeding. Unsafe formula feeding and early introduction of solid foods can lead to diarrheal disease from pathogens entering the gut, as well as nutrient deficiencies when the foods introduced displace breastmilk without providing adequate nutrient density. The limited availability of reliable national data on infant feeding in South Africa complicates the design and evaluation of infant feeding interventions. However, systematic reviews of the literature suggest that large-scale interventions focused on educating mothers can increase the prevalence of EBF and decrease infant mortality.
Narrative, entertainment-education (E-E) approaches to health education have emerged as potentially powerful strategies for promoting positive health behavior change. E-E involves the delivery of health messages by embedding them in an entertainment media framework. A substantial body of research suggests that E-E is an effective way of influencing beliefs, attitudes and behaviors. The characteristics of particularly impactful E-E media include: a) appealing narratives, b) high-quality production, c) persuasive messages that are unobtrusive and d) high potential for involvement or identification with the presenters or characters portrayed. While E-E can take many forms, a video-based approach readily supports the optimized integration of these characteristics.
Video-based content, optimized for mobile devices, may also facilitate broad dissemination of health education, especially as the global penetration of mobile technology increases. Recent advances in mobile health (mHealth) messaging have already begun to facilitate the delivery of health messages at scale and researchers are increasingly recognizing the potential for mobile phones and tablets to play an important role in health education interventions in LMIC. South Africa is an example of a LMIC at the forefront of mHealth initiatives, due to its rapidly expanding penetration of mobile phones. National, maternal-child mHealth initiatives built around free, text-based messaging have attracted global attention and been well received. General "tech-savviness" in South Africa continues to grow, with 37% of adults reporting "smartphone" ownership and 42% reporting daily internet usage in 2016. Active experimentation with "WhatsApp", a widely used mobile communication tool that supports the transfer of videos, is also underway in many sectors, including health.
Even for South Africans who do not yet have access to mobile phones, a feasible dissemination pathway lies in the delivery of educational videos by community health workers (CHWs) who bring teaching tablets to their in-home counseling sessions. Numerous interventions that support women in their homes have demonstrated efficacy in improving breastfeeding rates. The Philani Maternal Child Health and Nutrition Trust is an example of a successful community based organization, employing "mentor-mothers", CHWs who offer in-home health promotion counseling to pregnant women and mothers within their communities. Data from both case studies and quantitative research suggest additional improvements in breastfeeding rates when synergistic interventions - for example, mass media interventions - are used to supplement successful community-based ones. Developing interventions that boost the efficacy of successful community-based programs represents a critical step towards ending preventable child and maternal deaths by 2030 (as called for by the World Health Organization (WHO) and UNICEF. Prior research in South Africa suggests that such synergistic interventions are even more effective when they are created in close collaboration with community-based programs, actively involving local stakeholders in the content creation process.
Aims
This study will achieve the following specific research aims: To establish the
1. Effectiveness of the Philani MOVIE intervention for improving exclusive breastfeeding practices among study participants
2. Effectiveness of the Philani MOVIE intervention for improving other infant feeding practices, and maternal knowledge among the study participants
Methods
A cluster-randomized controlled trial will be used to measure the effects of the Philani MOVIE intervention on infant feeding behaviors. Eighty-four mentor mothers will be randomized 1:1 into intervention and control arms, stratified by neighborhood type. Mentor-mothers in the control arm will provide the usual Philani Intervention Model (PIM) perinatal in-home counseling. Mentor-mothers in the intervention arm will provide standard care plus the MOVIE intervention (PIM+M). Based on sample size calculations described below, at least 1008 pregnant women will be enrolled in the study and mother-child pairs will be followed until five months of age. Cluster randomization will be done by a third party who is not employed by Philani, nor an in-country member of the research team.
Intervention
The Philani MOVIE intervention consists of 13 short, (2-5 min.), videos conveying current WHO infant feeding recommendations, the health rationale for these recommendations, historical information on global breastfeeding trends and contemporary challenges facing breastfeeding mothers. The intervention uses a narrative, entertainment-education approach, incorporating the personal stories of six South African mothers, three celebrities and three community mothers, alongside animations, illustrations and voiceovers by local South African voice talent.
Sample size calculation
The sample size calculation in this trial was based on the primary outcomes: exclusive breastfeeding in the first month and in the fifth month of life. Calculations were performed using standard methods for cluster-randomized controlled trials with baseline covariate adjustment and stratification. The randomization units were the mentor-mothers. An intra-cluster correlation of 0.1 was assumed for the two primary outcomes. Routine program data describing the performance of the 84 mentor-mothers participating in this trial was used to inform the power calculation. Based on this data, and a range of other data sources on breastfeeding in South Africa the investigators assumed that 40% of mothers exclusively breastfed their infants in the first month of life and 10% of mothers exclusively breastfed their infants in the fifth month of life. Assuming a correlation between the baseline measurements and the primary outcome of 0.30 and an average enrollment of 12 pregnant women per mentor-mother over the course of this study, the investigators estimated that the trial would have 80% power to detect at a 5% significance level a 13-percentage point increase in the primary outcome in the first month of life and a 9-percentage point increase in the primary outcome in the fifth month of life. This minimal detectable difference met the investigators' condition for policy relevance (an improvement of more than 15 percentage points). The total sample size was thus set for outcome assessment at 840 pregnant women plus 20% (to allow for loss to follow-up), i.e. 1008 pregnant women.
Data collection and training
Data will be collected via two tablet-based questionnaires, one in the first month post-delivery and the second in the fifth month post-delivery. Both questionnaires will be completed directly on the tablets by the participants and will include visual and audio prompts to overcome literacy barriers. Each of these surveys will be followed within one week by a telephone survey, administered by a professional phone survey research company, and will be used to verify the responses recorded on the tablet-based surveys and gather additional data on specific complementary foods introduced and on maternal knowledge.
Mentor-mothers in both the intervention and control groups will undergo training on basic research procedures. In total, six groups of \~15 mentor-mothers will attend a one-day training workshop over a one-week period, followed by a half-day refresher training workshop within three weeks of the first training. The workshops will include training on informed consent procedures, ethical principles, and the use of tablet devices to collect data on infant feeding practices. Mentor-mothers will be trained to open the questionnaires on the tablet and administer a practice questionnaire with the participant to ensure that the participant is comfortable and competent in working with the tablet. Once the mentor-mother is confident in the participant's ability to work with the tablet, they will hand it to the participants to complete the questionnaires independently, without input from their mentor-mother.
Mentor-mothers in both intervention and control groups will be afforded the opportunity to troubleshoot any concerns related to the use of the tablets or the research procedures for the duration of the study.
Mentor-mothers in the intervention group will undergo separate training sessions aimed at familiarizing them with the content of the Philani MOVIE intervention as well as allowing them to practice integrating the intervention into their existing perinatal home-visiting program. These training sessions will include role play activities, followed by feedback from peers and training facilitators.
Data Analysis
The primary analysis will be based on intention-to-treat at the unit of the individual newborn/individual mother, with standard errors adjusted for clustering at the level of the unit of randomization, i.e. the mentor-mothers. The investigators will adjust for baseline neighborhood type in addition to six additional baseline covariates detailed in the study design description. Mixed-effects generalized linear models will be used for the primary analysis, using a fixed effect for the assignment to treatment vs. control arm and a random effect for mentor-mother. In particular, the investigators will use modified Poisson models (generalized linear models with Poisson distribution and a logarithmic link function), for the binary outcomes, i.e., the primary outcomes as well as secondary outcomes, 3-13. Modified Poisson models were chosen because they generate estimates of risk ratios (and thus avoid the well-known interpretational difficulties associated with odds ratios) and because they converge more easily than alternative models that generate risk ratios (such as binomial models). For the continuous secondary outcomes (14 and 15), generalized linear models will be used with Gaussian distribution and identity link function.
