The study design was informed by critical elements drawn from frameworks on intervention development (17) (26) (27). Specifically, the intervention was developed using a sequential mixed methods study design (28). This consisted of:
- exploring the evidence base through an umbrella review, and identifying key international standards on the appropriate use of antibiotics;
- undertaking detailed formative research through a) a qualitative study to explore the most appropriate mechanisms through which to embed the intervention within the existing health system and community infrastructure, and to understand patterns of knowledge, attitudes and practice regarding antibiotics and antibiotic resistance; and b) a household survey – which drew on the qualitative findings - to quantify knowledge, and reported attitudes and practice regarding antibiotics and antibiotic resistance within the target population;
- drawing on appropriate theories regarding change mechanisms and experience of implementing community engagement interventions to co-produce the intervention processes and materials with key stakeholders at policy, health system and community level.
Detailed findings from the umbrella review and from the investigation into knowledge, attitudes and practices regarding antibiotics and antibiotic resistance are reported elsewhere. This paper focuses on reporting on the process of intervention development as well as the most appropriate mechanisms through which to embed the CDA within the existing health system and community infrastructures of rural Bangladesh.
Study setting
The study was conducted in one upazila (sub-district) of Comilla, a peri-urban district about 100km south-east of the capital, Dhaka, with a population of 5.4 million. The district has around 410 functional CCs. Comilla was selected in consultation with the Ministry of Health and Family Welfare (MOHFW), based on the fact that members of the research team had previously worked with the MOHFW to enable the community health care providers in CCs in Comilla to deliver a basic package of essential care. The upazila in which we conducted the study was selected purposively, due to ease of access for researchers. The five CCs included in the study were selected purposively based on prior information regarding the functionality of the CG and the CSG. There is one CG and three CSGs per community clinic. Each group has 17 members. The members of the groups are clearly specified within policy, and they are supposed to represent a broad spectrum of the population. They encompass males and females, different ages, different socio-economic groups, and different professions. The members have been selected based on the information that is provided from the upazilla health complex regarding the categories of people to be included. The key responsibilities of these groups centre around managing the community clinic regarding issues such as opening and closing times, medicine supply, resolving problems related to electricity and other infrastructural issues. We selected two CCs with highly active groups, two with moderately active groups, and one with a relatively inactive group in order for us to better understand variation in the potential to embed the community engagement approach within the existing infrastructure of the CCs.
Qualitative study methods
A formative qualitative study was conducted in order to: a. inform intervention design, by exploring potential key issues and implementation strategies, including the most appropriate mechanisms through which to embed the intervention within the existing health system and community infrastructure; and b. to understand the accessibility of health services, and patterns of knowledge, attitudes and practice regarding antibiotics and antibiotic resistance in order to inform the design of the household survey.
One interview was conducted with the Union Health and Family Planning Officer (UNFPO). Interviews were conducted with each of the five CHCPs who work within the five CCs. Ten focus group discussions (FGDs) each with 6-8 participants were held with community members. Participants were purposively sampled from the CG and CSG, two per community clinic catchment area, one male and one female. By selecting community members from these pre-existing groups, which are representative of a broad spectrum of the population, the variation of the sample was maximised in terms of gender, age, education, employment and socio-economic status and, therefore afforded us with a wide range of opinion within the limited resources and time available to us.
Interview and focus group discussion guides were developed collaboratively by the research team. Reviews of key international guidance on antibiotic stewardship, peer-reviewed literature, and existing knowledge of the health system and cultural context informed the design of the guides. Translations from English to Bengali were undertaken by the Bangladesh-based research team, who discussed the most appropriate terminology to convey critical concepts. Guides were pre-tested with two respondents per guide in a different CC catchment area from the five in which the study was conducted and minor adjustments made after the pre-testing.
All interviews and FGDs were held in private rooms within CCs. Interviews were conducted and focus group discussions facilitated by male and female researchers experienced in qualitative data collection (PB, DB, FF, SH). They were audio-recorded and detailed notes were taken. Interviews and focus group discussions were transcribed by two Bengali speaking research assistants and then translated into English by bilingual researchers. Transcriptions and translations were checked by FF, PB and SH. Data was managed using NVivo 11. Analysis of the data was undertaken using a framework approach (9) using the following steps: Familiarization - key themes were identified during a meticulous review of the transcripts; Thematic framework construction - themes deriving from the study objectives and other key issues that emerged from the data were identified and used to assemble a coding/thematic framework – this process was undertaken by two researchers with the coding frame being developed through a review of a sub-sample of transcripts; Indexing - the data were coded according to the thematic framework by target group and re-organized into sections under each theme - transcripts were coded by one researcher and the coded transcripts independently reviewed by another, with any disagreements discussed; Interpretation - each thematic area was compared between respondent groups, similarities and associations between themes were identified and findings were interpreted.
Survey methods
The aim of the survey was to quantify knowledge and reported attitudes and practices regarding antibiotic use and antibiotic resistance, in order to inform the focus of the key issues to be addressed within the intervention. The survey tool was informed in part by the findings of the qualitative study.
We attach our questionnaire as supplementary material. We developed our questionnaire after reviewing findings from the qualitative study and conducting a rapid literature review of relevant studies (29-31). Our questionnaire contained 85 questions for females: 42 in relation to themselves, 19 in relation to their children and 24 in relation to their husbands. We chose to focus on females, as time and resources prevented us from also surveying males and because our piloting demonstrated that females were able to respond to questions regarding themselves, their husbands, and their children on this topic. Prior to surveying we pilot tested the questionnaire twice to check it was understandable, feasible and acceptable for the respondents and interviewers, and adapted as necessary. The testing took place in Comilla district, but outside the study area, to ensure similarity in context. Women in five households participated in the pilot. Trained data collectors conducted the survey. We recruited four data collectors and one supervisor who were provided with two and half days training.
To obtain rapid responses to our survey we used a non-probability cluster sampling approach slightly modified from the WHO approach used in their Expanded Programme on Immunization (EPI) vaccination coverage surveys (32) which is known to generally achieve its aims in terms of providing reasonable estimates (33, 34), but it does have significant limitations and risks of bias (20). However, as the primary aim of the survey was to provide rapid, cost-effective information to inform the development of the intervention the compromise was felt justified.
We aimed to survey a total of 245 women, as this would allow us to estimate outcome percentages and their 95% confidence intervals with an absolute margin of error of ± 10% (suitable for our purposes), assuming an outcome of 50% (the least precisely estimable outcome percentage) and a design effect of 2.5 due to the clustered sampling.
We analysed the data using the R version 3.4.2 (35) and where necessary the “Survey” (36, 37) package. We first described the sample’s characteristics in terms of common socio-demographic variables. We then produced estimates of outcomes as either percentages (for categorical variables) or means (for continuous variables) with their associated 95% confidence intervals, adjusted for the clustered sampling design.
Structured approach to developing intervention processes and tools
Interactions among the study team
The intervention was co-produced through a structured process of engagement with the wider research team, as well as with key stakeholders at policy, health system and community levels. A document outlining key issues, implementation strategy, and intervention tools was developed and updated throughout the intervention development phase.
When preliminary results from the formative research phase were available, a workshop was conducted, which brought together the wider study team to review preliminary findings, discuss implications for intervention development and refine key issues. It also served to identify knowledge gaps and means of eliciting required information. Throughout the intervention development phase, a small intervention development working group had weekly calls to review progress, discuss emerging findings from the formative research and recommendations from stakeholders, and make executive decisions. The wider study team was kept informed and provided feedback during monthly team calls.
Co-production of intervention with key stakeholders
First, a one-day a workshop was conducted with mid-level policy makers and practitioners with an interest in antibiotic resistance and community engagement in Bangladesh. In this workshop, participants refined tailored messages regarding antibiotic resistance and adapted them to the local context. Participants also provided feedback on aspects of the intervention design, particularly in relation to mechanisms to embed the intervention within the existing health system infrastructure. Second, a half-day workshop was conducted with representatives from the villages in the study area. This was used to validate key issues and obtain feedback on the proposed intervention design. Results, insights and recommendations from each workshop were summarised in a comprehensive workshop report. The Bangladesh-based study team continued to engage informally with key stakeholders throughout the intervention development period.
Development of intervention materials and pre-testing
Following the stakeholder workshops, a local artist developed visual materials illustrating the intervention’s tailored messages in an iterative process of drafting images and refining them based on feedback and suggestions from the study team. The images were used to develop a flipchart and a leaflet to support information sharing and stimulate discussion among Community Dialogue participants. The images were pre-tested in two focus group discussions with community members from a community in Dhaka (one with females, one with males). Discussions focused on establishing whether the drawings were understood as intended, whether they were culturally appropriate and whether community members liked their design. A range of non-visual tools to support sensitisation, training, community dialogue sessions, supervision, monitoring and evaluation were developed by the study team. All intervention materials were developed in English and subsequently translated into Bengali.