Study design
The study applied a parallel mixed-methods design for a prospective comparative cohort, nested within the BabyGel pilot cluster randomised trial.
Study setting
The study setting was that of the BabyGel pilot trial (reported elsewhere), i.e. homes of pregnant women from villages in Mbale District in rural eastern Uganda. The district has 912 villages, each village had approximately 130 households and about five people lived in each household, 56% of whom are women of reproductive age, as documented in the Mbale District health office records. In this study area, only 6% of women aged 15–49 years had completed education at secondary level or higher, compared to the national average of 33%. Literacy was less than 68% [18].
Selection of villages
We considered all BabyGel pilot trial villages in which there was ongoing recruitment of pregnant women within the strict 8 weeks’ recruitment period of the BabyGel pilot trial. However, this particular study was introduced in the sixth week and we included only six BabyGel trial villages. We recruited participants before the end of the eighth week. To attain the required sample size for this study, we included seven more villages that were neighbouring the BabyGel pilot trial villages. These non-trial villages were purposively selected based on the availability of pregnant women meeting the original inclusion criteria for the BabyGel pilot trial.
Study population
Participants included eligible pregnant women prior to recruitment into the BabyGel pilot cluster randomised trial. As in the pilot trial, the confirmed pregnant women had a gestation of 34 weeks or more and were able to speak or understand English or Lumasaba (the local language). We excluded women who had already been recruited into the pilot trial before this nested study.
Recruitment of participants
We recruited a total of 30 pregnant women in their homes from 13 villages in the last 3 weeks of BabyGel pilot trial recruitment (Fig. 1). Most of the women (70%, 21/30) were recruited from BabyGel pilot trial villages, comprising 20% (21/103) of the BabyGel recruits. The other nine women were recruited from non-trial villages.
The research midwife visited homes in each village with the aid of the village health team member/worker (VHW) and screened women for eligibility into the BabyGel pilot trial. The 30 recruited women gave initial signed informed consent (on a standard one-page information consent form) for participation both in the BabyGel pilot trial and to undergo a specific form of consent for the BabyGel pilot trial. These women were told during the informed consent process that they would be followed up 48 h (or 2 days) later to test their comprehension of the informed consent information and to identify their preference.
The three different consent models were administered to eligible pregnant women systematically over the last 3 weeks of recruitment into the BabyGel pilot cluster randomised trial. In each week, before receiving informed consent for entering the trial, study information was presented to each woman according to that week’s consent model. In the first week, the standard researcher-read participant information sheet was used, followed by a slide-show message in the second week and a video message in the last week. as illustrated in Fig. 2a. The standard researcher-read consent method was used to recruit the rest of the participants into the BabyGel pilot trial before the start of the consent nested study.
Each participant could ask questions or discuss issues at any point during the information presentation, which the research assistant responded to. At the end, a copy of the participant information sheet was given to each woman. Those women who agreed to participate in the trial completed a consent form with a signature or thumbprint and also signed by an impartial witness.
Researcher-read consent
This is the standard consent model commonly used to provide information to the BabyGel study participants. The research assistants read the study information from the ethically approved participant information sheet (Additional file 1) to each participant in either English or Lumasaba (Fig. 2b).
Slide-show message consent
In the second week, instead of research assistants reading out the participant information sheet as in week 1, they presented the study information to each participant orally using the approved 21 slides (Additional file 2) presented on a flip chart (in the style of an A4 ring-bound calendar) with text in bullet points and pictorial illustrations (Fig. 2c). The slide show was in either English or Lumasaba according to the participant’s preference. During the presentation, the participant concurrently read the pre-specified text and watched the pictorials on the reverse of the flip chart. After the presentation, the woman was allowed to see certain slides again if she wished.
Video-message consent
In the third week, the research assistant showed the approved study video (Additional file 3) to the eligible women with or without her family members present. The woman watched the video in either English or Lumasaba on a 5-in touchscreen Samsung S4 smartphone as presented by the researchers (Fig. 2d). At the end of the video, the woman was allowed to watch it again if she wished.
Information consent models
The participant information sheets, the slides in the flip charts and the videos were developed in a rigorous systematic stepwise process (Additional file 4). The lead investigator trained the research assistants on the presentation of the three consent methods using role playing, peer reviews and critiques before participants were enrolled using any of the methods.
Data collection
Three research assistants (a research midwife and two qualitative researchers) visited each woman in her home 48 h after giving informed consent. Qualitative and quantitative were data collected using Open-Data Collection Kit (ODK) forms installed on Samsung smartphones. These were synced over a wireless internet connection to the online database on the University of Liverpool’s server once the research team returned each evening to the research office in the Sanyu Africa Research Institute.
Quantitative data collection
The primary outcome was understanding and recalling information about the BabyGel pilot trial 48 h after being recruited. Data were collected electronically into the modified version of the Quality of Informed Consent (QuIC) case report form on the ODK system operated smartphones (Additional file 5). The form or data tool was designed to measure objective and subjective comprehension of the study information [5]. The research midwife administered the questionnaire face to face. At the start of the evaluation, the woman was instructed to respond to each question or rank her understanding of something as best as she could without consulting the participant information sheet. Correct and incorrect answers were assigned a QuIC score of 100 points and 0 points, respectively [5], and the answers were entered into the ODK electronic data capture system on the smartphone.
After answering questions on comprehension, the woman was shown the other two consent models, which were not used at the time of her recruitment. It was explained to the participant what the two alternative consent models entailed and she was not necessarily taken through the informed consent process again. Each woman was asked to rank the consent models in terms of preference on a 5-point scale of 1 (“I don’t like it all”) to 5 (“I like this most”). Lastly, the final item asked the participant to state her preferred model, giving reasons for her choice. The quantitative data were entered directly into the ODK form on the smartphone.
Qualitative data collection
Through semi-structured interviews, two qualitative researchers (MRN and JK) explored aspects of the comprehension of the informed consent messages and the preferences relating to the informed consent process [19]. All responses were audio-recorded and the qualitative researchers entered notes directly into the ODK software on the smartphone.
Data analysis
All analyses of quantitative variables were conducted using StatsDirect 3 to generate descriptive statistics. Statistical significance was set at the conventional 5% level.
Audio-recorded responses were transcribed and translated verbatim by the study’s main qualitative researcher (MRN). The transcripts were analysed using the inductive thematic approach to qualitative analysis, by describing and analysing patterns within the data [20]. The qualitative researchers (MRN and JK) familiarised themselves with the data by repeatedly reading the transcripts before coding them in detail to ensure all relevant codes were included in the initial pool of codes. The pool of codes was arranged into potential themes based on occurrence, significance and overlap. Together with the lead investigator (JD), the qualitative researchers (MRN and JK) reviewed the themes in relation to the generated codes and the entire dataset. We then agreed on the themes and defined a coding schedule, which was used to code all interviews again to ensure reliability and consistency of coding. QSR NVivo 10 (qualitative analysis software) was used to organise codes and themes.