Developing targeted client communication messages to pregnant women in Bangladesh: a qualitative study

A qualitative research using phenomenological approach, guided by the Health Belief Model (HBM), was carried out to explore perceptions of ANC and pregnancy complications [32]. The HBM focuses on six constructs: perceived susceptibility and severity of complications; perceived benefits of care, self-efficacy; and barriers and cues to action to attend care. We used the HBM to find out individuals’ perceived knowledge of pregnancy care and practice, to predict why and how women seek care during pregnancy. We developed the interview guidelines using HBM constructs for performing in-depth interviews in order to map the lived experience of pregnancy care (see Additional file 1). The interview guidelines focused on anemia, GDM, and hypertensive disorders in pregnancy. Questions on preferred strategies for communication and participants’ views on text message use to address knowledge gaps were also included in the interview guidelines. We pretested all the interview guidelines and based on the feedback, tools were refined and modified for final data collection. The interview guidelines were developed in English and translated to Bangla by the investigators of the study.

The study was conducted in Matlab South and Matlab North sub-districts in Chandpur District, Bangladesh. The study site mostly consists of rural and riverine delta areas of Bangladesh and is located 80–85 km southeast of Dhaka, the capital of Bangladesh. In Bangladesh, maternal and child health care services are provided by the two separate departments of the Ministry of Health and Family Welfare. Domiciliary community health workers from both departments provide health education on maternal and child health at the village level [33, 34]. At the primary care level, maternal health services are provided at Community Clinics (CC) and Union Health and Family Welfare Centres (UH&FWC) [18]. The UH&FWC cover a population of about 25,000, while CC cover an average of 6000 [33, 34]. The study population included pregnant and postpartum women, husbands, and mothers-in-law of pregnant women. We included husbands and mothers-in-law along with the pregnant and postpartum women as they are the prime household decision-makers of pregnancy-care seeking in this community [35] .

We used purposive sampling to select study participants from the pregnancy register of domiciliary community health workers, based on the inclusion criteria. The following groups were eligible for inclusion in the study: 1) women who were nullipara, primipara or multipara; 2) women with or without antenatal complications during a recent pregnancy; 3) postpartum women who delivered a live baby within 1 month of the interview; 4) husbands of pregnant women; and 5) mothers-in-laws of pregnant women. Husbands and mothers-in-law were not related to the pregnant or postpartum women included in the interviews.

In-depth interviews were chosen as a first step to elicit a vivid picture of the participant’s perspective on and experiences with pregnancy care. In-depth interviews are considered to be a non-threatening method to gather such data. Participants were free to express their true opinions as we assured them that the information shared will not be linked back to the individual participant. Two research investigators with experience in qualitative research conducted all the in-depth interviews. Before the data collection, investigators were well-oriented with the objective of the study as well as the interview guide to foster common understanding of the data collection expectations, processes, and goals. The information collected from the pregnancy register, in terms of inclusion criteria, was verified in the field prior to interviews of the study participants. Research assistants from the study team first contacted eligible participants with a phone call and requested participation in in-person interviews. The two research investigators then visited interested study participants at a time and place convenient and acceptable to them. One female investigator interviewed all the female participants, while the male investigator interviewed all the male participants. We conducted the interviews in the interviewees’ homes or other locations based on their choice. All participants gave informed written consent prior to the interviews and were assigned anonymous study participation numbers before the analysis phase. Each interview took approximately 1 h and was recorded with a digital recorder for transcription. In addition to the in-depth interviews, we also collected background information (see Additional file 2). We conducted the interviews until the data produced little or no new useful information in relation to our study objective. The research team held debriefing sessions at the end of each day of data collection to understand the interpretation of collected data, strengths and weaknesses of interview techniques, and any missed opportunities for further exploration, minimize the confusion among the research investigators, and determine when data saturation was reached. In-depth interviews were conducted from May to June 2017.

We prepared an outline of the purpose and plan for data analysis. Throughout the data collection process, we listened to the tape-recorded in-depth interviews, and read through all the transcribed interviews to identify themes that were discussed in the interviews. All interviews were conducted in the Bengali language, transcribed, and then translated into English. Data were indexed and coded manually, organized into a matrix, and then sub-themes and themes were generated according to the components of the HBM [36]. Table 1 shows an example of the thematic analysis process. Data were compared between and within different types of respondents to strengthen the appropriateness of the findings. Two researchers in the team with experience in conducting qualitative research individually coded the data and finalized the code list by checking for similarities and dissimilarities during data interpretation. Discrepancies were discussed between the two researchers first, and then with a third researcher until consensus was reached. We analyzed the data to understand the views of the different target audiences on the underlying factors that influence their ANC utilization.

Using the results from the in-depth interviews based on HBM constructs, we drafted TCC text messages in English that specifically addressed the knowledge gaps and perceptions that contributed to the low uptake of ANC. We tailored these draft messages to be aligned with specific pregnancy conditions, and the timing of screening for these conditions during ANC to increase client acceptance and enhance the impact on service utilization [37].

Once the general topic for each message was agreed upon by the research team, the format of that message was designed based on our review of effective behavior change techniques (BCT) recommended by Abraham and Michie [38]. We applied established theory and behavior change techniques to the development of draft messages to strengthen the potential of achieving specific behavior change [39, 40]. The messages covered six out of 26 BCTs, underpinned by five theories. BCTs included the provision of general information about behavioral risk, information on benefits and consequences of action or inaction of the behavior, encouraging the person to decide to perform, rewarding the effort toward achieving the behavior change, how to perform the desired health behavior, and planning the desired health behavior with a planned outline to perform (Table 2). The theories linked with the six BCTs are the information-motivation-behavioral skills model (IMB), a theory of reasoned action (TRA), theory of planned behavior (TPB), social-cognitive theory (SCogT) and control theory (CT). We matched each component of a given draft message to a specific BCT. We used positively framed messages as opposed to negatively framed messages to engage in healthy behavior [41]. To address the knowledge gaps identified through the HBM constructs, we included information on common pregnancy complications of anemia, hypertension, and diabetes, their consequences, and the benefits of ANC attendance. To nudge the study participants’ decisions towards healthy behavior, messages were personalized using their name [42, 43]. The content of these draft text messages was translated into the Bengali by the research team.

Two focus group discussions were conducted with pregnant women to get their feedback on the draft messages to ensure that the messages were easily understandable and acceptable. Specifically, participants were shown the content of the messages and asked to comment on their understanding of the message and how to improve the specific text and language. Women were selected from community health workers’ pregnancy registers. The results from the focus group discussions were evaluated by the team and the final Bengali text messages were modified based on the results.

This study was approved by the ethics and research committees at the International Center for Diarrheal Disease Research, Bangladesh (protocol number: 16054), and exempt from review by the Regional Ethical Committee in Norway, Southeast region (2017/1018/REK sør-øst C). Individual informed written consent was obtained from all participants.

Of a total of 24 in-depth interviewees (Table 3), ten were pregnant women, five were postpartum women, four were mothers-in-law, and five were husbands. All pregnant women were housewives from 18 to 26 years of age. Among the ten pregnant women, four were pregnant for the first time. At the time of the interview, the mean gestational age was 23 weeks and 4 days (8–36 weeks). Seven pregnant women had less than 10 years of education, and the rest had more than 10 years. All the postpartum women were housewives, from 22 to 32 years of age. Three postpartum women had completed less than 10 years of education and two women, more than 10 years. The five husbands ranged from 25 to 41 years of age; four had less than 5 years of formal education, and one up to 10 years.

The following main findings from our study guided the development of SMS messages for TCC. First, women and families only perceive the need for early ANC when women are sick. Perceived susceptibility to common pregnancy complications, and the knowledge that they may occur without symptoms, was low among study participants. Second, women did not recognize that common pregnancy complications could necessitate care during pregnancy. Third, they did not know that attending ANC could result in fewer complications or earlier detection of complications. Fourth, they were unaware of the appropriate number and timing of ANC visits. Finally, most of the women who received detailed information on when and where to attend follow-up ANC during their first ANC visit did attend those follow-up visits, suggesting that lack of information is a key factor leading to inadequate utilization of ANC and that basic information served as cues-to-action. The evidence did not suggest any specific modifiable factor which could address a majority of women’s self-efficacy and ability to attend ANC.

Given this evidence, and the planned integration of automated data-driven SMS messages into the eRegistry, we designed a series of text messages for women to act as cues-to-action and target women’s knowledge about ANC, specifically on 1) the benefits of attending ANC; 2) women’s susceptibility to complications; and 3) the severity of complications. The messages were designed to match the timing of critical screenings for common conditions such as anemia, hypertension, and GDM. We used clinical data from the eRegistry to further tailor these messages, based on women’s gestational age and common risk factors for the medical conditions targeted (Table 4), in order to make the text messages more individualized and provide added value to the individual woman.

The Bangladeshi government recommends a schedule of four ANC visits for low-risk pregnant women. Reminder messages were designed to be sent 1 week and 1 day prior to the scheduled ANC date; we selected one topic to be highlighted in each of the reminder messages (Table 5). In addition, we created a welcome message to be sent upon enrolment, referral facilitation messages, and facility delivery reminders. The referral facilitation messages were reminders to women that their provider had recorded danger signs or created a referral in the eRegistry.

Given that the messages would be automatically generated within the eRegistry, we were able to additionally tailor each message based on clinical characteristics. Specifically, if data were entered into the eRegistry indicating anemia, hypertension, or gestational diabetes, an algorithm was developed which would take that into consideration and appropriately modify the messages. And, if the risk factors for these conditions were documented in the eRegistry (as shown in Table 4), the text messages would be tailored further.

The initial draft of the TCC messages, based on the HBM and the ANC schedule, and tailored with the behavior change techniques, were translated into Bengali from English and shared with pregnant women in focus group discussions (data not shown). Participants reported that the content of messages was adequate, but some modifications of terminology were needed for ease of understanding. We then reviewed each message and finalized the wording. Each message was adapted to the typical SMS character limits (knowing that for some messages up to three SMS could be sent). In the end, a library of 43 TCC messages was developed based on whether the individual woman had anemia, hypertension, GDM, and associated risk factors in the eRegistry. Examples of the messages are included in Table 6.