The study is being conducted in Manna, Shebe-Sombo, and Dedo districts of Jimma Zone, Oromia, Ethiopia. The districts were selected from 21 districts in the zone after confirming the unavailability of interventions aiming to improve maternal and newborn health services in each of them. Jimma Zone is characterized by weak health infrastructure, high health workers to the population ratio (1:329), only 9 health facilities offering comprehensive emergency obstetric and newborn care and a high maternal mortality ratio (412 per 100,000 live births).
A 3-arm, cluster-randomized trial design with a 1:1 allocation ratio was used.
The participants in the intervention were assigned to 3 arms: gain-framed, loss-framed, and control. The detailed description of each arm is presented as follows:
Women in this arm received an automated gain-framed message. The gain-framed SMS text messages, which were tailored based on individual women’s gestational age (GA), expected delivery time, and postnatal period, were sent from the time of enrolment during ANC until the end of the postnatal period to the established list of participants’ phone numbers. Women in this group also received the existing maternal care provided at the health facility.
Women in this arm received automated, loss-framed messages. The loss-framed SMS text messages, which were tailored based on individual women’s GA, expected delivery time, and postnatal period, were sent from the time of enrolment during ANC until the end of the postnatal period to the established list of participants’ phone numbers. Women in this group also received the existing maternal care provided at the health facilities.
Participants in the control group received existing ANC and PNC services provided at health facilities. According to the recent ANC guidelines, pregnant mothers are expected to have 8 ANC and 3 PNC contacts. In addition to these cares, counseling services is recommended on the following topics during ANC visits: (1) lifestyle modification, (2) danger signs and symptoms, (3) birth preparedness and complication readiness, and (4) other issues such as family planning, cord care, bathing, breastfeeding, child immunization, and screening for cervical cancer [
The intervention messages and the messaging system, which automatically dispatches these messages, were developed. These messages were developed based on the concept of message framing, with half of them developed in the form of a gain frame—emphasizing the benefits of maternal and newborn health care practices (eg, choosing to put your new baby to the breast in the first hour he or she is born will increase a baby’s protection from many diseases and illnesses). The other half of the messages were developed as a loss frame—emphasizing loss as a result of not using maternal health services and practicing newborn care (eg, failing to put your new baby to the breast in the first hour he or she is born will lower a baby’s protection from many diseases and illnesses).
The content of the messages was evaluated by mothers and experts before it was used in the intervention. Initially, 120 (60 for each form) English version messages were drafted within the character count of 160, based on the Ethiopian Ministry of Health’s guidelines and the World Health Organization’s recommendations. After discussing with the research team, 32 (16 from each form) messages were discarded. The remaining 88 messages were translated into the Afan Oromo language and used for content evaluation. Mothers evaluated the content of these 88 messages through a card sorting activity, and 10 messages (5 from each form) that were not well understood by mothers were discarded. Following the card sorting activity, 8 experts evaluated the retained 78 messages (39 for each form) using 10 statements. The experts suggested modifications for 14 messages (13 loss-framed and 1 gain-framed), and all of these messages were modified and refined based on the suggestions.
Finally, 39 messages for each message form were retained under 10 thematic areas: nutrition, lifestyle modification, danger signs, birth preparedness, complication readiness, breastfeeding practice, labor, newborn care, immunization, and antenatal and PNC usage. These messages were dispatched using locally developed software that manages the schedule and sends the messages automatically. To facilitate this, a domain, layer 3 VPN, and short code were purchased from Ethio-telecom Company. The software program includes an interface to enter the participant’s name, gestational week, mobile phone number, message type, message category, language option, report generation, and feedback options.
Clusters (Kebeles) with good mobile network coverage in 3 of the selected districts were eligible for randomization. Pregnant women with a GA of 16-20 weeks who either owned a mobile phone themselves or had a family member who owned one, lived in the selected cluster for at least 6 months, provided consent to participate in the intervention, and shared their phone number were eligible for the trial. Pregnant mothers who were unable to read mobile phone messages and had plans to relocate from the study cluster during the follow-up period were excluded from the intervention.
The method suggested by Hooper et al [
Next, the design effect (
To arrive at the final sample size per arm, the sample size for individual randomization is multiplied by both design effects, resulting in (n×
For newborn care practice, the same procedure described earlier was used. First, the individual randomization sample size was calculated to be 118 based on the following assumptions: P1—the proportion of exclusive breast feeding among the control group of 74.8% [
Therefore, the sample calculated for ANC was larger (588) and was the final sample size of the trial to arrive at 21 clusters (7 per arm) and an average of 28 eligible pregnant mothers.
Following the randomization of the clusters, lists of pregnant women in the randomized kebeles were obtained from the health extension worker’s registration book. Midwives then identified eligible pregnant women using 2 strategies: the last menstruation period (LMP) and the fundal height measurement technique (in the case of an unknown LMP) to determine the GA of mothers. The data collector provided a detailed explanation of the inclusion criteria and overall intervention procedure to eligible mothers by using the information sheet. Eligible pregnant mothers were enrolled in the intervention after obtaining informed consent regarding their willingness to give their phone number for the intervention and their approval to receive text messages to their own or their family’s phone during the intervention period. The recruitment was done after randomly selecting 28 participants from a list of eligible pregnant mothers in each cluster. After recruitment, information, such as the participant’s name, mobile phone number, and GA, was gathered during the baseline assessment. This information was then entered into the software to dispatch the messages according to the GA of those assigned to the intervention arm.
A total of 6 primary health care units, 2 from each district, were randomly selected, and kebeles in these primary health care units were used as units of randomization for the trial. A simple randomization with a 1:1 allocation was used to randomize the clusters (kebeles). After listing 30 kebeles that fulfilled the eligibility criteria, 21 of them were randomly selected for randomization. To assign the clusters to the intervention groups (gain-framed group, loss-framed group, and control group), the following steps were followed. First, the 21 clusters were alphabetically listed and labeled 1 to 21. Next, random numbers were generated using Excel 2019 (Microsoft Corp), and the generated values were copied and pasted as values next to the list of clusters. The list was then sorted in ascending order based on the generated random number. Finally, the first 7 clusters in the sorted list were assigned to the gain-framed group, the next 7 to the loss-framed group, and the last 7 to the control group. After the randomization, eligible pregnant women with a GA of 16-20 weeks were identified from the selected cluster by midwives. The midwives used 2 strategies: the LMP and the fundal height measurement technique in the case of an unknown LMP to determine the GA of mothers (
The allocation process was done by a person who was not involved in the implementation of the trial, and it was masked from the researcher involved in the implementation of the intervention. The interviewers who collected the data were blind to the intervention assignment. It is not possible to blind participants due to the nature of the intervention.
The primary and secondary outcomes are provided in
Maternal health service usage
Newborn care practice
Knowledge about maternal health services
Knowledge of danger signs (during pregnancy, labor and delivery, and the postnatal period)
Knowledge of newborn care practice
Knowledge about newborn danger signs
Attitude toward maternal health services
Attitude toward newborn care
Self-efficacy toward maternal health service usage
Self-efficacy toward newborn care practice
Iron supplementation
Maternal morbidity
Neonatal morbidity
Operational definitions are as follows:
Maternal health service usage:
when mothers receive at least 8 ANC follow-up visits, give birth in a health facility, and visit a health facility for PNC follow-up contact from health professionals within 7 days of delivery.
Newborn care practice:
this refers to the care provided to the baby from birth to 28 days of age by the mothers, including early initiation of breastfeeding, exclusive breastfeeding, feeding colostrum, thermal care, hygienic cord care, and seeking care for an ill newborn. If a mother feeds colostrum, initiates breastfeeding within 1 hour after birth, feeds a newborn only breast milk, practices thermal and hygienic cord care, and seeks care for her ill newborn, she is considered to be practicing newborn care, and if she misses one of them, she is considered to be not practicing newborn care.
Knowledge about maternal health services:
participants were asked to mention the recommended newborn care practice using 16 questions. All the responses were counted, and the higher score indicates a higher knowledge of newborn care practice.
Knowledge of maternal danger signs:
this was measured by 7 items related to maternal danger signs during pregnancy, labor & delivery, and the postnatal period such that yes=1 for correct statements and no=0 for incorrect statements. The scores of correct answers were summed, and a higher score was interpreted as a higher level of knowledge of maternal danger signs.
Knowledge of newborn danger signs:
this was measured by 12 items related to neonatal danger signs, such that yes=1 for correct statements and no=0 for incorrect statements. The scores of correct answers were summed, and a higher score was interpreted as a higher level of knowledge of newborn danger signs.
Knowledge of newborn care practice:
participants were asked to mention the recommended newborn care practice using 7 questions. All the responses were counted, and the higher score indicates a higher knowledge of newborn care practice.
Attitude toward maternal health services:
this refers to mothers’ feelings toward maternal care practice. Attitude toward maternal care practice was measured by 10 items on a 5-point Likert scale ranging from 1=strongly disagree to 5=strongly agree. The cumulative scores were used in the analysis once negatively phrased items were reverse-coded. A higher composite score indicated a more favorable attitude toward maternal care practice.
Attitude toward newborn care practice:
this refers to mothers’ feelings toward newborn care practices. It was measured by 16 items, which are designed in a 5-point Likert scale format ranging from 1=strongly disagree to 5=strongly agree. The sum score was used in the analysis after reverse coding the negatively worded items. The higher composite score was interpreted as a favorable attitude toward newborn care practice.
Self-efficacy to practice maternal health service usage:
self-efficacy in practicing maternal care is an individual belief in mothers’ ability to perform maternal care practices. It was measured by 10 items designed on a 5-point Likert scale ranging from 1=strongly disagree to 5=strongly agree. Once negatively worded items were reverse-coded, the sum score was used in the analysis. A score with a high value indicated greater self-efficacy in practicing maternal care.
Self-efficacy to practice maternal and newborn care:
this refers to an individual belief in one’s ability to perform maternal and newborn care practice. It was measured by 9 items, which are designed in a 5-point Likert scale format ranging from 1=strongly disagree to 5=strongly agree. The sum score was used in the analysis after reverse coding the negatively worded items. A score with a high value indicates higher self-efficacy toward practicing newborn care.
The baseline survey was conducted in March 2023, while the end-line survey was conducted in February, 2024. Trained interviewers collected data for both the baseline and end-line surveys through face-to-face interviews using a structured questionnaire. The questionnaire addressed various domains, including sociodemographic factors, socioeconomic factors, obstetric history, maternal health service usage, newborn care practice, knowledge about maternal health services, knowledge about maternal danger signs, knowledge about newborn danger signs, knowledge about newborn care practices, attitude toward maternal health services, attitude toward newborn care practice, self-efficacy toward maternal health services, self-efficacy toward newborn care practice, iron supplementation, and maternal and neonatal morbidity. Participants who were lost to follow-up were recorded, along with their reasons for discontinuation and the outcome of their pregnancy. The data were checked for inconsistencies and errors during the data collection period by dedicated supervisors and submitted to the central data management unit.
A total of 10 data collectors and 2 supervisors were recruited during each data collection period. A 2-day training was given for data collectors and supervisors on the questionnaire content, measurements, ethical issues, and how to use the ODK app (during end-line assessment), and they were kept uninformed about cluster allocation.
The collected data during the baseline and end-line was checked for completeness at the time of data collection by supervisors, and questionnaires with missing items were communicated with data collectors for correction. Double entry was done for baseline data to identify and manage mistakes during the data entry, while the end-line data were retrieved from KoBo Toolbox. Data cleaning was done to identify the duplicate records and missing and outlier values; to verify that the values were in the proper format and expected range; and to correct the errors. Only the study team has access to the intervention data sets. Information about clusters and participants was not shared with any third party, both during and after the intervention. Moreover, all personal identifiers were removed from the database, and paper-based filled questionnaires were stored in a locked cabinet. Data were stored in a database on a password-protected computer, and only the research team could access it.
Double-data entry was accomplished using Epi-data (version 4.6, released in 2020; the EpiData Association) and exported to the SPSS (version 25; IBM Corp) for analysis. The report of this study will be made according to the CONSORT-EHEALTH (Consolidated Standards of Reporting Trials of Electronic and Mobile HEalth Applications and onLine TeleHealth) standards for reporting randomized controlled trials. Exploratory subgroup analysis was done to identify whether the effects of intervention vary with sociodemographic characteristics, socioeconomic characteristics, obstetric history, and psychological factors. To assess the characteristics of the study population through the lenses of different variables, selected data were tabulated by study arms; this helped us to identify variations and similarities among study arms. The summary statistics used for categorical variables were frequencies and percentages. For continuous variables, the mean and standard deviation will be presented.
To observe the effectiveness of mobile phone messaging-based message framing interventions for improving maternal and newborn health service usage, we will use an intention-to-treat approach where all randomized participants will be included in the statistical analysis, regardless of whether they received the intervention or not. A generalized linear mixed model will be used to adjust for both the clustering effect and the participant-level intracluster correction. The outcome variables are assumed to have a multivariable normal distribution. The odds ratio will be used to measure the effectiveness of the intervention on the outcome variable with a 95% CI.
Ethical approval was received from the Jimma University Institutional Review Board (ref no JUIH/IRB/358/23). Participating in this trial was completely voluntary, and women can opt out of receiving the messages by sending “stop” to the domain. Informed consent was obtained for taking part in the intervention, baseline, and end-line assessments. The interviewer read the contents of the informed consent form and outlined the purpose of the study, the institutions and researchers involved, the expectations of women, and the risks and benefits associated with the trial using the local language. Women also get an explanation about their rights as participants and have their questions answered before enrollment.
Women’s phone numbers were kept encrypted, and no one could access them without a server technician and researcher. Codes were assigned to each enrolled woman to ensure confidentiality; mobile numbers were entered into the dashboard anonymously. Data were shared with the research team after removing participants’ names. No compensation was provided to participants for taking part in this trial.
Monthly monitoring of intervention implementation was conducted by collecting data through face-to-face interviews using a predefined questionnaire. Participants were asked the number of maternal and newborn information they have received through SMS, the number of messages read, the service they have used following the received message, and any pregnancy outcomes. They were also asked about any barriers that they have faced to receive the messages and the usage of services after receiving messages. Moreover, the functionality and stability of the messaging system were monitored continuously for the duration of the project. Data about the SMSs being sent and delivered were also automatically recorded by the messaging system and checked weekly.