A quasi-experimental study design with three arms was conducted. In the first arm mothers and fathers (Mother-Father Intervention - MFI) received breastfeeding education through SMS in addition to standard care; in the second arm only mothers received the breastfeeding education through SMS (Mother’s only intervention - MI) in addition to standard care; and the third arm was the control group (CG) where couples received only standard care.

This research took place in health centers located in Mekelle, Ethiopia. Mekelle has nine public health centers, a tertiary hospital, and three general hospitals. Three public health centers with the highest ANC attendance rates that were at distance from each other were purposively (to get the proposed sample size and to avoid contamination) selected for study participant recruitment. The health centers were randomly (by lottery) assigned to the three intervention arms. Couples receiving services in each health center were assigned to one arm to avoid data contamination. All pregnant mothers in their last trimester and who had one month to their estimated date to give birth in the selected health centers were approached by nurses. First contact was made with mothers and then if they agreed to participate, fathers/partners were contacted by telephone. Couples who did not each have a personal mobile phone; who were not able to read and understand Tigrigna (the official language); were not living together, or where there were issues with the pregnancy or potential issues with breastfeeding were excluded from the trial.

Three community health centers were randomly assigned to the three arms. Couples who received pregnancy services in that health center were by default assigned to one of the arms of the intervention (Fig. 1).

Flowchart of participants in the SMS based breastfeeding education intervention to improve exclusive breastfeeding in Mekelle, Tigray

The theory of planned behavior (TPB) was used as the theoretical underpinning in the development of the SMS text message breastfeeding education approach. TPB is among many social science theories widely used to understand underlying health behaviors and for designing appropriate interventions and has been previously used in the development of breastfeeding interventions [12, 13]. According to the TPB, intention is the most proximal element of behavior. Behavioral intention is affected by attitudes related to the behavior, subjective norm (partner support), and perceived behavioral control (self-efficacy) [26]. The framework generally focuses on the cognitive or modifiable factors and can be used as a suitable theoretical context for designing interventions for behavioral change [26].

The content of the SMS breastfeeding education was developed after conducting an explorative qualitative study through focus groups discussion with fathers and mothers who had a child less than two years of age to inform the intervention. Based on the findings of the qualitative study [27] alignment with international breastfeeding recommendations, and the TPB, the research team developed 16 different weekly messages for fathers and mothers that aligned with prenatal and postnatal milestones and issues (Table 1). The intervention was delivered using a computer-based platform FrontlineSMS software program. The messages were automated “push” messages designed to reduce burden and cost for parents when considering interactive messaging in this setting.

Couples included in the MFI each received a weekly breastfeeding SMS text message. Each parent (mother and father) in this group received four tailored SMS text messages during antenatal care (ANC) (for example “the first milk colostrum is good for baby it will help the baby fight infection”). After delivery, each parent received an additional twelve tailored postnatal (PNC) SMS text messages (for example “encourage your wife to let the baby suckle at her breast to increase milk supply”) through their personal mobile telephones. Each parent in this arm therefore, received a total of 16 weekly SMS text messages over a period of four months. Similarly, the mothers in the MI arm received the four ANC and twelve PNC, weekly SMS text messages through their personal mobile telephones. In addition to the breastfeeding intervention, couples in the MFI and MI received routine ANC and PNC care provided at their respective health centers. Couples in the control group received the ANC and PNC standard care provided at the health center they were attending. The delivery of all messages was tracked, and participants were asked separately about whether they read the messages, showed the messages to others and what they had learned. These data are not presented here.

Where possible, all constructs were measured using tools validated across a range of international contexts, but not necessarily Ethiopia. A recognized process for cultural adaption of tools was used [28]. These tools were first translated into Tigrigna by the primary investigator. The translated questionnaires were then back translated by two public health nutritionists, with differences and comments discussed between the public health nutritionists and the research team. Some modifications were made in the Tigrigna version. Finally, face validity to ensure understanding and language was undertaken with ten mothers and ten fathers, after which some additional wording was changed.

The main outcome of the trial was the proportion of mothers who exclusively breastfed their babies at first, second, and third months. The definition of exclusive breastfeeding was based on the WHO indicator and included infants who were receiving breastmilk only. EBF was assessed using a 24-hour recall, one week recall, and one month recall [29] through phone interview with mothers. The inclusion of one week and one month ensured that any foods and fluids consumed since the last data point were also included.

The secondary outcomes of the intervention comprised breastfeeding attitudes, knowledge, and perceived partner breastfeeding support, for both mothers and fathers. These were collected through face-to-face interview at baseline (during the last trimester of pregnancy) and at the end of the study period (three months post-partum). Breastfeeding self-efficacy was measured for mothers only. Attitudes were measured using the Iowa Infant Feeding Attitude Scale (IIFAS) which has 17 questions and utilizes a Likert scale ranging from 1 = strong disagreement to 5 = strong agreement [30]. Total scores ranged from 17 to 85; with higher scores representing more positive attitudes. Partner breastfeeding support was assessed using the Partner Breastfeeding Influence Scale (PBIS) containing five sub-scales, breastfeeding savvy (learning about breastfeeding and discussing with partner), helping (providing tangible support), appreciation (encouraging and valuing breastfeeding partner), presence (partner’s assistance during breastfeeding), and responsiveness (father’s understanding to the mother’s needs). Each dimension was assessed using a five-point Likert scale from 1 (extremely not supportive) to 5 (extremely supportive). Mean scores were calculated for each breastfeeding support component [31]. A standardized tool from the Food and Agricultural Organization (FAO) of the United Nations (UN) [32] was used to assess breastfeeding knowledge. The tool has 10 open breastfeeding questions, which later coded to “Knows” or “Does not know”.

Maternal breastfeeding self-efficacy was measured using the Breastfeeding Self-efficacy Scale-Short Form (BSES-SF) [33]. The BSES-SF has 14 questions using a five-point Likert scale ranging from 1 (not at all confident) to 5 (always confident). Total scores range from 14 to 70. Higher scores indicated higher self-efficacy. In addition, the sociodemographic, economic, ANC service attendance, birth and infant, and infant characteristics were collected [11].

The sample size for the intervention was calculated using the power calculator and was estimated to be a total of 144 mothers and fathers that is 48 couples in each arm. The sample size was based on estimates of the proportion of EBF in the control group being 0.59 [34], and a 23% expected improvement in EBF in the experimental groups. The sample size was then calculated with the following assumptions α = 0.05, power = 80%, and an expected 10% attrition rate.

Data were entered into IBM SPSS Statistics version 23 (IBM Corp. Released 2015. IBM SPSS Statistics for Windows, Version 23.0. Armonk, NY: IBM Corp). For categorical variables frequency with percent was reported while for continuous variables, either mean or median with standard deviation or interquartile range, respectively, were reported. Normality was tested for each continuous variable using variable inflation factor (VIF).

Chi square test was performed on the baseline characteristics of mothers and fathers. The differences in knowledge, attitude, and self-efficacy after the intervention between the three groups were determined with one-way ANOVA, or Kruskal Wallis test. Binary logistic regression was used to measure the effect of the intervention among the three groups. Baseline variables which were found to be significant (p < 0.05) in the chi square, or one-way ANOVA/Kruskal Wallis test were considered in the final logistic regression as confounders to test the effectiveness of the intervention. The level of significance was set at p-value < 0.05, where the null hypothesis was there was no intervention effect. The risk of stopping exclusive breastfeeding at month 1, month 2, and month 3 was expressed in terms of odds ratio with 95% CI.