Engaging men in maternal, infant and young child nutrition in rural Tanzania: Outcomes from a cluster randomized control trial and qualitative study

There is growing recognition that engaging men in maternal, infant and young child nutrition (MIYCN) interventions can benefit child health and disrupt harmful gender norms. We conducted a cluster-randomized controlled trial in Tanzania, which engaged men and women in behaviour change via mobile messaging (short message service [SMS]) and traditional interpersonal communication (IPC), separately and in combination. Here, we evaluate intervention effects on individual-level men’s MIYCN knowledge and discuss barriers to male engagement. Eligible clusters were dispensary catchment areas with >3000 residents. Forty clusters were stratified by population size and randomly allocated to the four study arms, with 10 clusters per arm. Data on knowledge and intervention exposure were collected from 1394 men through baseline and endline surveys (March–April 2018 and July–September 2019). A process evaluation conducted partway through the 15–18-month intervention period included focus group discussions and interviews. Data were analysed for key trends and themes using Stata and ATLAS.ti software. Male participants in the short message service + interpersonal communication (SMS + IPC) group reported higher exposure to IPC discussions than IPC-only men (43.8% and 21.9%, respectively). Knowledge scores increased significantly across all three intervention groups, with the greatest impact in the SMS + IPC group. Qualitative findings indicated that the main barriers to male participation were a lack of interest in health/nutrition and perceptions that these topics were a woman’s responsibility. Other challenges included meeting logistics, prioritizing income-earning activities and insufficient efforts to engage men. The use of a combined approach fusing IPC with SMS is promising, yet countering gender norms and encouraging stronger male engagement may require additional strategies. This study draws on data collected during a trial to compare the effectiveness of two nutrition behaviour change interventions, separately and in combination, on MIYCN knowledge and practices. The study design has been described previously (Rothstein et al., 2021). The 2 × 2 factorial, cRCT had four study arms that received the following interventions: (i) interpersonal communication (referred to as ‘IPC arm’); (ii) SMS text messaging (SMS); (iii) a combination of both interventions (SMS + IPC); and (iv) usual care without any special intervention (Control). In this setting, men are encouraged to attend antenatal care visits at public health dispensaries with their pregnant partners, yet usual care does not involve any MIYCN messaging addressed to men. Primary participants were pregnant women and mothers of children <12 months at enrollment; male household members were also enrolled when possible. A cRCT (as opposed to an RCT) was chosen as certain IPC interventions were implemented at the community, as opposed to individual, level. Randomization took place at the level of public health dispensary catchment areas (DCAs); of all 53 public DCAs in the two study districts (ranging in population size from 987 to 21,264), DCAs with <3000 residents (n = 11) were excluded to maximize data collection efficiency. One DCA was also excluded because it contained a large health centre, unrepresentative of normal health service access. From the remaining DCAs, one was randomly selected and excluded to produce 10 DCAs per study arm (40 total DCAs). These DCAs were stratified by population size into >10,000, 4800–10,000 and <4800. Within each stratum, DCAs were randomly allocated to one of four study arms using random number generation by the study PI. Thus, each study arm contained two large‐, three small‐ and five medium‐sized DCAs. All DCAs within each study arm received the same set of interventions, described below, and all were included in the analysis. Data were collected during three periods. Baseline questionnaires were administered following participant recruitment, before intervention implementation (March–April 2018), and endline questionnaires were administered following the intervention (July–September 2019); these data were used for the quantitative impact evaluation. The intervention period ranged from 15 to 18 months depending on the exact timing of recruitment and data collection for each participating household. In addition, partway through the intervention period (April 2019), a qualitative process evaluation, which included focus group discussions (FGDs) and in‐depth interviews (IDIs), was conducted among a subsample of participants to explore factors affecting implementation. The study took place in Mtwara, a rural, agricultural region in southern Tanzania that has historically lacked access to health and nutrition projects. Regional young child nutrition indicators are worse than national averages; for example, only 19.9% of children 6–23 months in Mtwara consume the minimal meal frequency, as compared with 39.9% countrywide, and stunting among children under five is 37.7% in Mtwara versus 34.4% nationally (Ministry of Health, 2016). Two districts, Newala and Tandahimba, were selected in Mtwara as they were most representative of rural Tanzania and had no prior exposure to MIYCN interventions. The sample size (2400 total, with 600 per arm) was based on the study's primary outcomes (minimum dietary diversity of women and children), aiming to discern a 10% or greater difference based on a baseline prevalence of 50%, accepting a Type 1 (α) error and power (1 − b) of 80%, a design effect of 1.3 for cluster randomization (with an average cluster size of 60) and a 10% loss to follow‐up. Study participants were recruited through door‐to‐door screening, with a complete enrollment of eligible persons. The husband/male partner of each eligible woman (pregnant woman or mother of a child <12 months) was enrolled if they were available and consented. If the head of the household was a man other than the woman's partner, this person was eligible to serve as the male participant if the partner was not available. Upon study enrollment, male participants were asked to consent to be enrolled in the free Wazazi Nipendeni service (SMS/SMS + IPC arms) or be contacted by a community health worker (CHW; IPC/SMS + IPC arms). For consenting participants in SMS arms, study personnel registered participants’ and other household members’ phones, as allowed, with the Wazazi Nipendeni service immediately after data collection; SMS were expected to begin soon thereafter. For participants assigned to the IPC intervention, activities were initiated within 3 months of baseline. Study households were revisited during endline data collection, and male participants were interviewed if available. In addition, men who were not assessed at baseline but were identified as the partner of a female participant and/or head of household at endline were also eligible to be interviewed at endline; this was done to increase the sample size of men available for analysis. This was the only change to the study design made after trial commencement. Study recruitment and baseline occurred from March to April 2018, and the endline was completed from July to September 2019. Full details on the process evaluation methods are provided elsewhere (Rothstein et al., 2021). Briefly, for IDIs, participants were recruited from villages in SMS and SMS + IPC arms, as the primary objective of IDIs was to understand participants’ experiences of receiving SMS. Men were eligible to participate if they owned a registered phone and had received at least one Wazazi Nipendeni SMS. IDI participants were randomly selected from a list of participants surveyed during the process evaluation. For the FGDs, three villages were identified that had relatively high rates of male participation in IPC meetings (‘active’ villages) and three villages with relatively low male participation rates (‘inactive’ villages), based on monitoring data; this sampling frame was used because the primary objective of the FGDs was to explore barriers to men's participation in the IPC intervention. Men residing within those villages were eligible to participate in FGDs if they had enrolled in the study and provided consent to be visited by a CHW. The IPC intervention consisted of regular group discussions (cluster level) and home visits (individual level) by CHWs to educate participants on MIYCN using the government‐sponsored Mkoba wa Siku 1000 (‘1000 Days Initiative’) programme. During monthly meetings, CHWs played recorded messages and discussed recommended health and nutrition practices using a radio and counselling cards; all messages were targeted to a specific life stage. The original goal was to hold separate women's and men's groups, led by the same CHWs. During implementation, the make‐up of the groups differed across communities based on participation levels: five villages had separate men's groups, 12 villages had mixed‐gender groups and in seven villages, men did not participate at all. Individualized home visits, which happened once every 2 months, were intended to identify and address barriers to recommended practices, reinforce adoption and identify high‐risk women and children. The project drew on existing female and male CHWs, with additional CHWs selected where needed. CHWs underwent 2 weeks of training before implementation, received monthly supervision thereafter and participated in quarterly meetings with all project CHWs to share experiences. They received job aids, such as bicycles and rainboots, as incentives but were not paid. Among households assigned to the SMS and SMS + IPC arms, male and female participants’ mobile phones that were successfully enrolled in the Wazazi Nipendeni service received SMS free of charge with health and nutrition information and life stage‐timed reminders, from pregnancy through the child's fifth year. Registrants could identify themselves as pregnant women, new mothers or supporters (family, friends). Messages covered nutrition, such as breastfeeding and incorporating protein into child diets, and other health topics like child vaccinations. Message frequency varied by month and declined over time, from an average of 4.9 messages/month during pregnancy and the child's first year of life to an average of 1.25 messages/month sent during the child's fifth year. Surveys were conducted in Swahili by a team of trained and experienced data collectors. Data were recorded on tablets and stored on an online server. Upon study enrollment, a survey was completed with the household head or most senior member available to collect sociodemographic, asset, infrastructure and dietary data using questions adapted from Tanzania's Demographic Health Survey and Malaria Indicatory Survey and an International Food Policy Research Institute (IFPRI) study of the same SMS programme in Tanzania (IFPRI, 2020; Ministry of Health, 2016). Food insecurity was assessed through USAID's Household Food Insecurity Access Scale (Coates et al., 2007). A separate survey was administered to the male participant to assess intervention exposure, IYCF knowledge (through 18 multiple‐choice questions) and mobile phone usage; these questions were developed by the research team. These same questions were administered to men at endline, along with a module exploring men's perceptions of the intervention(s). FGDs were facilitated by two local interns familiar with Mkoba wa Siku 1000 and loosely structured by a discussion guide. FGDs explored men's level of interest in child health and nutrition topics, barriers to participation in IPC activities and suggestions for how to better engage men. Those who had attended meetings and/or interacted with a CHW during the study period were asked to describe their reactions to the content and impressions of the CHW. FGDs were held in a neutral community location and audio‐recorded. Issues surrounding male participation in IPC activities were explored through IDIs to ascertain individuals’ perspectives outside of a group context and in more villages. IDIs were conducted by data collectors using semistructured guides that included questions similar to those of the FGDs, took place in a private location and were audio‐recorded. From the cRCT data, our outcome of interest was individual‐level men's nutrition and health knowledge; this was a secondary outcome of the study, for which the primary outcomes were maternal and child dietary diversity. From the qualitative data, we sought to understand barriers and facilitators to engaging men in behaviour change communication on nutrition topics. Quantitative data were analysed using Stata SE15 (StataCorp). Percentages and mean/standard deviation were generated for categorical and continuous variables, respectively. Baseline characteristics of households and men were compared across arms using one‐way analysis of variance for continuous variables and χ 2 tests for categorical variables. Variables with significant differences (p ≤ 0.05) across arms were adjusted for in multivariate models assessing programme effects. Exposure to the IPC intervention was measured by participation in a group meeting and receipt of a home visit by a CHW for nutrition counselling over the previous 12 months and frequency of these activities over prior 12‐ and 3‐month periods. Exposure to the SMS intervention was assessed through Wazazi Nipendeni enrollment, ever receiving a Wazazi Nipendeni SMS or SMS with health and nutrition information, and the number of SMS received in the prior 30 days. Analysis of impact measures only considered observations with data from both baseline and endline. Men's nutrition knowledge scores were based on 14 questions after four of the original survey questions were excluded from analysis due to uncertain validity. The remaining questions were categorized into three areas—breastfeeding, child feeding and micronutrients, and hygiene (Supporting Information: Table 1). Participants received one point for each correct response, resulting in three area scores, and a total score was calculated as a sum of these scores (range: 0–14); these scores were then normalized (to 0–1 each) for further analysis. The effect size for men's knowledge was estimated using mixed‐effects linear regression models. Using baseline and endline data, changes in knowledge scores within a given study arm were examined and compared with the control group change in the intervention period. In mixed‐effects models, the random‐effect component included indicators to control for clustering at individual and village levels. The fixed component included the study arm, time of survey (baseline = 0; endline = 1) and an arm–time interaction term. Clustering at the village level was adjusted for; the adjusted models included four variables that differed across groups at baseline (religion, drinking water source, type of toilet facility and garden ownership). The treatment effect and associated p value were derived using contrasts of average marginal effect. Audio recordings from the FGDs were used to complement the detailed notes taken during each discussion, which were then translated into English. IDI recordings were transcribed verbatim by native Swahili speakers and simultaneously translated into English. Deductive and inductive codes were then applied to the transcripts using ATLAS.ti qualitative management software (Scientific Software Development) by the first author. Following procedures for content analysis developed by Hsieh and Shannon (2005), codes were organized into categories, subthemes and then broader themes (Hsieh & Shannon, 2005). The first author sought feedback on initial interpretations from the data collectors, strengthening the trustworthiness of the findings. Ethical approval was obtained from Tanzania's National Institute for Medical Research (NIMR/HQ/R.8a/Vol.IX/2684). Written informed consent was obtained from all male and female participants before enrolment, with separate consent for audio recording where relevant. The cRCT was registered on ClinicalTrials.gov ({"type":"clinical-trial","attrs":{"text":"NCT03297190","term_id":"NCT03297190"}}NCT03297190).