Nvivo version 11

Focus group transcripts were analysed in their original Spanish by researchers at Universidad del Valle and Universidad de Cooperativa del Colombia, sede Villavicencio. English translations were produced, and were provided to researchers at the London School of Hygiene and Tropical Medicine (LSHTM) for independent analysis. Transcripts were analysed by thematic coding according to a pre-defined coding scheme. Coding was performed using the Atlas.ti (version 8.0) (Scientific Software Development Gmbh, Berlin) and NVivo (version 11.0) (QSR International, Melbourne) software packages. There were 7 coding families in total, representing: common mosquito control methods, arbovirus concerns, Zika knowledge, sources of Zika knowledge, changes to control behaviours during the Zika outbreak, ideal mosquito control tools, and views around PPT. Each theme family was divided into a larger number of sub-families. Each interview was independently analysed by at least three individual researchers, and results were compared and discrepancies discussed until a consensus was reached. As the purpose of using the two study sites was to increase diversity and not to directly compare the two cities, comparisons have only been drawn when clear differences between study sites emerged.

All audio-recorded interviews were transcribed verbatim. The analyses were performed using a thematic approach to identify patterns and emerging themes from the quotes based on inductive coding technique within the data and not based on pre-existing assumption or theory [19 (link)]. In this approach, two coders independently reviewed and familiarised themselves with the interview transcripts and later, collaboratively created a codebook of all preliminary codes that emerged from the independent analyses forming a meaningful framework after mapping of similar codes and renaming of the themes. Peer reviews, including counterchecking with other investigators occurred at several stages of the analyses to ensure rigor and reliability of response coding. Using interpretative description method, the responses from participants were embedded within the text to support and build evidence for the proposed themes and sub-themes. All the quotes included in the text were transcribed exactly from the original recordings. In the event of absence of clarity, additional descriptions were incorporated in parentheses to aid in understanding or [sic] was indicated where necessary. All interviews were coded using NVIVO version 11.0 (QSR International Pty Ltd).

Quantitative data were entered into a Research Electronic Data Capture (REDCap) database and summarized with descriptive statistics. We calculated the Pearson correlation coefficient to examine for a relationship between IES score and time since disclosure. We used Grubbs’ test to assess for outliers.
Interviews were recorded, transcribed, and entered into qualitative analysis software (NVivo version 11.0; QSR International, Melbourne, Australia). Two of us (JDG and CGC) reviewed a subset of transcripts to identify preliminary themes and engaged in a consensus-building exercise to arrive at themes for interview coding. We double-coded interviews, adjusting themes as necessary and intermittently assessing interrater consistency. In the event of a Cohen κ coefficient for reliability < 0.8, interviews were reviewed for coding consistency, and consensus was achieved. At the conclusion of data coding, we performed a final consensus exercise to ensure consistency in participants coded for each given qualitative theme. The final Cohen κ was > 0.9. For each theme, we report the frequency count of codes within that theme.

Audio recordings of the focus groups were transcribed and uploaded into NVivo version 11.0 (QSR International Pty Ltd., Doncaster, Victoria, Australia) to aid in qualitative analysis. The moderator and one field note researcher independently read the transcripts and collaborated to develop a codebook. The final codebook reflected the socio-ecological model, interview guide, and themes originating from the discussion content. These themes included: Individual pulse consumption, pulse consumption by family or friends, nutrition information, and policy issues. The grounded theory approach to qualitative analysis was applied to define themes [28 (link)]. There were no unresolved disagreements in coding as the two reviewers collaborated to develop and refine themes after initial coding. Inter-rater reliability between the two coders was high, 99.59, with a kappa value of 0.83 [29 ].

The focus groups were transcribed and thematically analyzed [19 ] inductively in NVivo Version 11.0 (QSR International). Owing to the nature of the focus groups, individual participants were not identified. Thematic analysis is widely used within the social and health sciences [20 (link)] as a tool to examine “repeated patterns of meaning” [18 ] or as a way of identifying and making sense of commonalities within data sets [19 ]. The coding and theme development were inductive, using an iterative process that involved reading and rereading the data sets to establish initial codes that covered the key ideas discussed and then combining similar codes under the themes. Following this process, the themes were reviewed alongside the original data set as a way of checking that the data set was correctly represented and that important data were not missed. The analysis was undertaken by a primary researcher, which was discussed within the research team throughout the analysis and cross-checked by a second researcher.