The Philani MOVIE study: a cluster-randomized controlled trial of a mobile video entertainment-education intervention to promote exclusive breastfeeding in South Africa

This study will achieve the following specific research aims: To

This study will be conducted within the under-resourced settlements of the Western Cape Province in South Africa. In 2010, the infant mortality rate in the Western Cape was reported to be 25 per 1000 live births [43]. A 2014 study of under-resourced communities in the Western Cape, found the EBF rate to be only 6% [44]. Operating within these communities (see Fig. 1), the Philani Maternal Child Health and Nutrition Trust [37] is a non-government organization serving approximately 100 under-resourced neighbourhoods. The Philani Mentor Mother Outreach Program is focused on improving child health outcomes through the deployment of “mentor-mother” CHWs. Mentor-mothers are identified as positive role models within their communities, after they have demonstrated success in raising healthy families despite resource shortages. Once recruited, mentor-mothers complete a standardized training program for delivering in-home health promotion counselling services within their communities. Among other health-promotion tasks, Philani mentor-mothers are trained to monitor growth, counsel parents around proper nutrition and refer sick family members to the nearest clinic when necessary. These mentor-mothers promote a variety of health behaviours, including those related to HIV-prevention, breastfeeding, child nutrition, growth and development [37, 45, 46].

The Philani MOVIE intervention consists of 13 short, (2-5 min.) educational videos created over the course of 10 months in collaboration with community members and local stakeholders in maternal-child health. The health messaging in each video is closely aligned with the most recent WHO recommendations for infant feeding and narrated in English and isiXhosa, the languages spoken most widely among the study participants. The videos use simple syntax, avoiding medical jargon. Table 1 summarizes the video topics included in the Philani MOVIE intervention.

A human-centered design approach was applied during the development of the intervention. Formative feedback was solicited from end-users within the target audience as well as experts in multiple disciplines, including community medicine, academia, public health, media marketing and the performing arts. Scripts were collaboratively edited by content experts, who worked in parallel using Google Drive to allow for reliable version control. Early drafts of videos were then shared via WhatsApp with community members, most of whom did not use email regularly, informing a process of Rapid Iterative Testing and Evaluation [47, 48]. Feedback from community members was gathered via WhatsApp and during face-to-face “bodystorming” (physical brainstorming) sessions [47, 49] in which mentor-mothers tested the intervention through role-play activities involving mock-counseling sessions. Feedback was synthesized on a rolling basis, informing multiple iterations of the content. In this way, the health messages, language, illustrations, characters, color palate, narrative delivery style, logo and soundtrack were collaboratively and iteratively finalized over the 10-month content creation period. Key health and motivational messages were interwoven with animations created by local artists and underscored by the narratives of community mothers and South African celebrity mothers. Both community members and media marketing experts emphasized the importance of representing and celebrating South Africa’s diverse ethnic profile. This feedback shaped production decisions about both the illustrations included in the intervention and the profiles of the mothers featured in the intervention video footage. The trailer for the series can be previewed here https://​youtu.​be/​1sss8ViPKJo [50].

Our study is a cluster-randomized controlled trial with baseline covariate adjustment and stratification [51] (See Fig. 1). The randomization unit is the mentor-mother. Each mentor-mother will enroll several pregnant women over the course of the trial. We conducted analyses of routine program data from these mentor-mothers and found that, among a range of available covariates, the only significant predictor of EBF was the type of neighborhood in which the mentor-mother lived and performed her health promotion duties. These neighborhood types include: informal, formal and mixed neighborhoods. We thus decided to stratify the randomization according to neighborhood type, to ensure balance of intervention versus control assignment by this covariate. Based on the existing literature [11, 52‐59], we also expected that the following covariates would be associated with the primary outcome measure:

To further increase the statistical efficiency of our analysis, we thus decided to measure these covariates at baseline and adjust for them in our primary analyses, in addition to baseline neighborhood type. Randomization of the mentor-mothers was performed, using a computer-generated random allocation sequence, by faculty at Heidelberg University, Germany. The allocations were then transferred by email to Philani for implementation. Cluster randomization was preferred over individual randomization in this trial because each mentor-mother is responsible for counseling pregnant women within her neighborhood, making individual participant randomization logistically impossible without disrupting existing protocols.

This work rests upon the theoretical underpinnings of the Elaboration Likelihood Model (ELM) [78], which suggests that there are two contributing pathways to achieving the changes in attitude that predict a desired behavioral outcome such as EBF. The first “central route” is influenced by the learner’s motivation and ability to cognitively process the information presented (ie: the health benefits of EBF). This can be influenced by factors such as the length of the content and the degree to which the language used is understood by the learner. The second “peripheral route” relies on cues embedded in the method of information delivery that contribute to its relative acceptability to the learner. Positive, peripheral cues, such as the learner’s subjective evaluation of the person delivering the information or the learner’s emotional involvement in the content, lead to peripheral attitude changes which, while less enduring than central attitude changes, can positively influence the learner’s motivation to process the messaging via the central route [78]. Both ELM and a variation on this model, called the extended Elaboration Likelihood Model (eELM) have been used to explain the efficacy of entertainment-education in changing leaner attitudes and behaviors in response to health messages and other types of persuasive messages [16, 19, 79]. According to the eELM model, effectively designed messages, containing positive peripheral cues, can reduce resistance to attitude changes (counter-arguing), when the learner is “transported” (engaged) in the narrative or identifies with its characters. This identification can be empathic or cognitive, through perceived similarity or wishful identification (a desire to emulate the characters featured). This model even suggests that the imagined “pseudo relationships” than can develop between viewers in the general public and on-screen celebrities, can be harnessed as a powerful peripheral cue to enhance attitude change in persuasive messaging [79]. Fig. 3 illustrates the intersecting theories of change on which our intervention is grounded.

This study will benefit from oversight by a data and safety monitoring board (DSMB), which will consist of a senior biostatistician, a senior local health systems researcher and a professor of pediatrics and health policy with expertise in primary care and health outcomes research. The DSMB will meet twice per year. It will evaluate the study progress and provide guidance on study continuation and any potential changes in study conduct in the case of unexpected events or operational difficulties.