mhealth-based health promotion intervention to improve use of maternity care services among women in rural southwestern uganda: Iterative development study

Background: Antenatal care (ANC) prevents perinatal morbidity and mortality, but use of these services in Uganda remains low and maternal mortality rates are among the highest in the world. There is growing evidence that mobile health (mHealth) approaches improve timely communication of health-related information and produce positive health behavior change as well as health outcomes. However, there are limited data to guide development of such interventions in settings where ANC attendance and uptake of skilled maternity care are low. Objective: The aim of this study is to develop a novel patient-centered mHealth intervention to encourage and support women to use maternity care services in Mbarara district, southwestern Uganda. Methods: Using an iterative development approach, we conducted formative stakeholder interviews with 30 women and 5 health care providers (HCPs) to identify preferred key ANC topics and characterize the preferred messaging intervention; developed content for SMS text messaging and audio messaging with the help of 4 medical experts based on the identified topics; designed an app prototype through partnership with an mHealth development company; and pilot-tested the prototype and sought user experiences and feedback to refine the intervention through 3 sets of iterative interviews, a focus group discussion, and 5 cognitive interviews. Qualitative data were coded and analyzed using NVivo (version 12.0; QSR International). Results: Of the 75 women who completed interviews during the development of the prototype, 39 (52%) had at least a primary education and 75 (100%) had access to a mobile phone. The formative interviews identified 20 preferred perinatal health topics, ranging from native medicine use to comorbid disorders and danger signs during pregnancy. In all, 6 additional topics were identified by the interviewed HCPs, including birth preparedness, skilled delivery, male partner’s involvement, HCP interaction, immunization, and caring for the baby. Positive audio messaging and SMS text messaging content without authoritative tones was developed as characterized by the interviewed women. The postpilot iterative interviews and focus group discussion revealed a preference for customized messaging, reflecting an individual need to be included and connected. The women preferred short, concise, clear actionable messages that guided, supported, and motivated them to keep alert and seek professional help. Complementary weekly reminders to the women’s significant others were also preferred to encourage continuity or prompt the needed social support for care seeking. Conclusions: We used an iterative approach with diffuse stakeholders to develop a patient-centered audio messaging and SMS text messaging app designed to communicate important targeted health-related information and support rural pregnant women in southwestern Uganda. Involving both HCPs and end users in developing and formulating the mHealth intervention allowed us to tailor the intervention characteristics to the women’s preferences. Future work will address the feasibility, acceptability, and effectiveness of this design approach. We used an iterative design (Figure 1) that comprised the following steps: (1) stakeholder interviews (end-user women and maternal health care providers [HCPs]) to identify key health education topics relevant to the ANC period and characterize women’s preferences for an mHealth-based, social support intervention (formative interviews with predevelopment end users); (2) content development for SMS text messaging and audio messaging with the input of 4 medical experts (2 obstetricians and 2 midwives) for the ANC topics as identified and characterized by the stakeholders; (3) design of an app prototype through partnership with an mHealth development company; and (4) pilot-testing the prototype and obtaining feedback for content refinement through (i) 3 sets of iterative exit interviews (pilot participants), (ii) a focus group discussion (FGD), and (iii) cognitive interviews to further explore user experiences and refine the updated message components and maximize potential impact and sustained use by rural pregnant women. Iterative development of a novel messaging app prototype. We conducted in-depth qualitative interviews with 2 groups of stakeholders: (1) target end users and (2) HCPs. A total of 40 women were invited to complete in-depth interviews, but saturation was attained after 75% (30/40) were interviewed. The women were purposively selected from rural Mbarara district, southwestern Uganda, during the period December 2018 to March 2019. These women were interviewed to identify key health education topics relevant to the ANC period as well as characteristics of a preferred mHealth app. The women were recruited from 10 villages located within 20 km of Mbarara Regional Referral Hospital (MRRH) with the help of existing village health teams (VHTs). VHTs are composed of community-based volunteers who are identified by community members and given basic training on major health programs to mobilize and sensitize communities to use available health services [25-27]. Eligible women for these predevelopment interviews included adults (1) aged ≥18 years (2) who had delivered a baby within the past 3 months, (3) owned or had access to a mobile phone, and (4) were able and willing to give informed consent. The purposeful sample was intended to represent women with differing experiences of pregnancy and ANC and included 15 women who delivered at home and 15 who delivered at a health facility. The interviews were open-ended and organized to cover predesignated core topics. An interview guide was developed and pilot-tested using the constructs of the Healthcare Service Utilization Model as reported elsewhere [26,27] and the Technology Acceptance Model [28,29] (Multimedia Appendix 1). This open-ended approach ensured systematic coverage of specific areas of interest while allowing for unanticipated content to emerge. The interview topics included information and preferred ANC topics considered useful in supporting women during their pregnancy journey, attitude toward using mHealth technology, performance expectancy, effort expectancy, social influence, facilitating conditions, self-efficacy, anxiety, behavioral intention to use technology, and potential technology engagement or fatigue. Specific data on preferences for messaging, content, frequency, preferred language, length, and timing were also sought. A brief questionnaire at the outset of each interview was administered to collect demographic information (eg, age, occupation, and educational background). In all, 5 key HCPs (2 obstetricians and 3 midwives) were purposively identified from MRRH and another rural maternity health center in Mbarara district. The HCPs included (1) adults aged ≥18 years (2) who were actively engaged in maternity care or policy implementation in Uganda or both, 3) had at least 5 years of experience as HCPs at a busy maternity center, and 4) were able and willing to give informed consent. They were interviewed to explore key ANC health education topics and information to inform the development of the mHealth-based app. All interviews took place at a private location mutually agreed upon by the participant and the interviewer. Each interview lasted 50-70 minutes. Written informed consent was obtained at the outset of each interview session. Qualitative interviews were digitally recorded with the participant’s permission and transcribed verbatim. Using the formative qualitative interviews with the end users, we identified key health education topics along with ANC messages that could be developed to increase ownership, engagement, usability, and acceptability by the intended recipients [28,29]. The HCPs identified additional topics that were a critical part of the health education framework during ANC visits [30]. We used Behavior Change Technique Taxonomy version 1 (BCTTv1) [17,31] because it offered a reliable structure to identify, define, interpret, and characterize key components (active ingredients) of our intervention messages aimed at improving the use of maternity care services. Information was grouped within the BCTTv1 components identified as follows: (1) goal setting (outcomes: improving health-related knowledge and skilled delivery), (2) goal setting (behavior: presenting for ANC and avoiding risky behavior), (3) action planning (planning for scheduled visits, financing, actionable messages, partner involvement, and birth preparedness), (4) feedback on behavior (progress monitoring and app interaction features), (5) prompts or cues (follow-up messages or reminders and information cues, eg, danger signs during pregnancy), (6) credible source of information (systematic content development using experts and continuity of care through regular customized information), (7) instruction on how to perform the behavior, (8) information about health consequences (cautionary social and emotional consequences of, or regrets related to, poor health-seeking behavior), (9) what to do regarding, or where to seek, care or redress (problem solving), (10) review goals (interaction with HCPs and how to review progress), (11) embedded self-monitoring information on progress or preparedness, and (12) active social support for users through regular reminders (self-reminders or through identified social networks). In all, 4 medical experts (2 obstetricians and 2 midwives), different from those interviewed, were engaged to contribute content for the first draft of the SMS text messages and audio messages, identified by the women and the HCPs, to ensure quality and consistency. The content and frequency of these messages were also based on the type of phone, network, preference, and need to ensure effectiveness and usefulness as well as avoid repetition, fatigue, and burdensomeness. During the intervention design, we worked with iStreams, an mHealth app development company in Mbarara town with an existing mHealth platform in Uganda [32]. This local developer designed an initial and novel app prototype that included both SMS text messaging and audio messaging for pilot testing. The design allows training manuals and behavior change communication materials to be integrated within a mobile app. Its unique multimedia design also allows women to listen to messages in their own language or view culturally relevant visuals (such as those that identify danger signs or complications, getting prepared for delivery, expected date of delivery, childbirth checklist, and others). Women were able to register on this platform and be tracked throughout their pregnancy and postpartum periods, receive automatic or scheduled SMS text message reminders, SMS text messages, audio messages, or notifications about upcoming appointments. In addition, the app stores medical information and allows real-time submission of data directly from a mobile phone, allowing managers and supervisors to access up-to-date data on health outcomes. The elements, content, and patterns of the SMS text messaging reminders were customized and the prototype presented as an eBirth platform for SMS text messaging and audio messaging (Figure 2). All messages were developed in English and then translated into the local language, Runyankole, by an experienced translator to ensure that context was maintained. Messages were dispensed in either English or Runyankole as preferred by the recipient. Fixed SMS text messaging data were stored in a secure cloud with iStreams, which is Health Insurance Portability and Accountability Act compliant. The e-Birth app biodata registration form interface. We used the Bendixen approach [33] for designing and developing a user-centered mHealth app, considering 5 overarching goals: (1) ease of use, (2) engagement, (3) education and preparation, (4) motivation and support, and (5) tailoring the system and personalizing the information for end users. Messages were intended to communicate information on the benefits of nutrition, exercise, presenting for ANC, skilled delivery, partner involvement, birth preparedness, monitoring danger signs, and overcoming barriers to access maternity service. The topics of these messages were identified and characterized by both the women (end users) and the HCPs during the predevelopment stage, and the content was developed by health experts. Scheduled SMS text messaging reminders were incorporated as part of the intervention as a stimulus, prompt, or cue to take action. Through the VHTs, we screened and enrolled 30 pregnant women (3 successive iterations of 10 pregnant women) from communities residing within 20 km of MRRH who had not presented for ANC by the beginning of their third trimester (determined by their last menstrual period) to test and assess preliminary feasibility, acceptability, and usability of the novel app through postuse qualitative interviews. An iterative approach of interviewing 10 women in each of the 3 groups was also considered sufficient to obtain rich, specific, and purposefully focused information from participants who had been exposed to the intervention [34]. These novel messages and their content were also tested to ensure ownership, relevance, consistency, and expectations among these pregnant women. After enrollment, messages were sent to these 30 women (audio messages or SMS text messages or both), depending on the participant’s choice. Women with access to a mobile phone in their household were registered, and they received the current version of the messages through the eBirth app prototype (Figure 2) for at least 3 months, a period of time chosen to include a minimum of 3 ANC visits and delivery. Messages were sent in a specific sequence, depending on the month of pregnancy, to cover appropriate topics identified in the formative interviews. In the case of some women, SMS text messaging reminders were incorporated and sent to their significant others or social supporters to help remind them and support them on their upcoming ANC visit and maternity journey [23]. However, the social supporters were not given any prior recommendations or instructions guiding them on how to respond to the SMS text messaging reminders because the intervention was designed to provide social support by building on already existing supportive relationships of the study participants. To test for frequency preferences, we first sent out messages daily for 2 weeks, then weekly at the chosen times, and then twice a week alternating between 8 AM and 8 PM. A message delivery log was monitored on the app platform. We followed the enrolled women through delivery. Upon completion of the 3-month message delivery period, we interviewed the women using semistructured questionnaires to obtain feedback on content, preferred terminologies, language, and ease of use in obtaining the needed support. We assessed phone use and responsiveness by how often the women read the SMS text messages, received calls, texted back or texted at all, confirmation of receipt, phone calls made, and the times when they missed calls or did not read the SMS text messages sent to them. The women were interviewed on technology acceptance, performance and effort expectations, whether they preferred SMS text messaging or audio messaging as a medium of information delivery, their attitude toward SMS text messaging or audio messaging technology, other preferred terminologies, content scheduling, facilitators, technology engagement, convenience, social influence, facilitating factors, anxiety, need for help using the app (self-efficacy), behavioral intention to use, and preliminary feasibility (network challenges, phone ownership, battery life, resources, frequency, and timing). The app prototype was modified based on feedback from each iterative round of the pilot interviews. After the third modification, of the 30 participants, 10 (33%) who had had similar exposures to the intervention were randomly recruited from the pilot to constitute an FGD aimed at further refining the relevant message components and helping to limit or prioritize the number of topics included in the messaging app as recommended [35]. Finally, 5 cognitive interviews with a new set of women were conducted to further refine the updated messages and maximize potential impact and sustained use by rural pregnant women. We described demographic and clinical data for all qualitative, iterative, and FGD interview participants using standard descriptive statistics. Qualitative analysis began with repeated review of the initial transcripts to identify relevant topics of ANC care, as well as characterization of a preferred mHealth app. Qualitative data were coded with the aid of the data management software, NVivo (version 12.0; QSR International). Coded data were iteratively reviewed and sorted to identify repeated themes (topics) arising from the data. Themes were generated using inductive content analysis [36]. The suggested content consisted of descriptive labels that defined and specified each theme (topic) meaning, along with illustrative quotes taken from the qualitative interviews. Themes were harmonized to be inclusive throughout all the development stages. Coding was guided by questions about attitude, perceived importance, usefulness, responsiveness, preliminary feasibility, experience with the messaging app, and suggested changes. Negative, positive, and neutral perceptions as well as attitudes were also identified and coded. Data analysis was performed jointly by ECA, GRM, and JN. Both JN and ECA double-coded 5 sampled transcripts, yielding a Cohen κ of 0.796. Together with GRM, we resolved disagreements until we were satisfied with the consistency in our coding to generate a codebook. We aimed at ensuring consistency in coding. For the iterative interviews, data for frequency, timing, and frequency of messaging during the iterative testing were described using Stata software (version 12.0; StataCorp). All personnel involved in the project had relevant training in human subject research ethics. The study was reviewed and approved by the institutional ethics review committees of Mbarara University of Science and Technology and the Uganda National Council for Science and Technology, Kampala, Uganda. All consenting participants gave written informed consent before study enrollment; in the case of those who could not write, a thumbprint was obtained on the consent form as approved by the ethics committees.