Nvivo 11

The open-ended questions were designed in advance. The perspectives of instructors and experts on disease control and prevention concerning the research topic and design were collected through pre-interviews. According to feedback, the interview outline was further revised for formal interviews that proceeded smoothly. Researchers possessing medical literacy, communication skills and enthusiasm for disease control and prevention work are the local people in the participants' areas, pivotal to conducting the interviews smoothly and guaranteeing research accuracy (19 (link)). When collecting data, we concentrated on the oral expressions of the questions. We remained neutral to guarantee that the results were honest reflections of the participants' thoughts. The audio recordings were transcribed into text within a day after the interview. The text was later analyzed following Colaizzi's 7-step analysis method (20 (link)) and coded with the qualitative analysis software NVivo 11.0 (QSR International, Melbourne, Australia). For data entry, the interviewers cleaned and validated the data and provided a clear definition of the categorized framework. Then we coded the data based on the definition (coded twice by two independent coders); internal consistency was also checked. When there were issues, the coders would discuss them until a consensus was reached.

A multi-stage analytical process applying thematic analysis that combined deductive, inductive, and verification techniques to check the reliability and validity of the coding system was conducted using NVivo 11.0 (QSR International, Melbourne, Australia) [28 ]. Firstly, a code book was created a priori by M.V. and W.V.L., which was based on the principles of Appreciative Inquiry and the COM-B model [27 (link),29 ]. Secondly, the transcripts were read as an initial exploration of the data by W.V.L. Thirdly, transcripts were imported into NVivo 11.0 and the code book was used to code the interviews (deductive analysis) in combination with an inductive reflection on the analysis (generation of themes) [31 (link)]. Table 1 gives an overview of the theoretically deduced codes and the inductively produced themes. In terms of reliability and validity, one third of the transcripts (n = 5) was coded by two researchers (M.V., W.V.L.), afterwards codes and results were compared to check if consistent results were reached (in the few cases results were not consistent, the two researchers reached consensus and this was applied to the data). The other two third of the transcripts (n = 10) was further coded by one researcher (W.V.L.). In order to add quotes in the result section of this manuscript, W.V.L. translated the quotes from Dutch to English.

The recorded interview files were converted to text within 24 h following the end of the interview and subsequently analyzed using Colaizzi’s 7-step analysis method [10 (link),17 (link)], and qualitative analysis software NVivo 11.0 (QSR International, Melbourne, Australia) was used to explore the factors affecting the interviewees’ vaccine hesitancy, recommend suggestions, and feasible measures, taking into consideration common themes.
During the research and design stage, the opinions of mentors and disease control professionals on the research topic and study design were obtained by conducting pre-interviews. The survey framework was consequently modified based on the feedback, so that formal interviews could be conducted smoothly. During the data collection stage, we paid attention to the colloquial presentation of the problem and remained neutral to ensure that the interview results truly reflected the interviewees’ ideas. During the data entry stage, the interview data were cleaned and verified, the definition of the classification framework clearly defined, and the data coded to ensure a clear operational definition (coded twice, by two separate researchers, using the NVivo 11.0 software); the internal consistency was also verified. When coding opinions differed, the views were unified after discussion.

We conducted a thematic analysis of the data following steps recommended by Braun and Clarke: becoming familiar with the data, coding data inductively, generating and reviewing themes, and defining and finalizing themes (21 (link)). We repeatedly read the preliminary transcripts, paying attention to the reflective and interpretive meanings therein and identifying the code categories list. Allowing these codes to evolve iteratively, preliminary themes were developed and modified as the data was analyzed. While focusing on recurring themes in the text, we also identified and highlighted cases that contradicted the dominant accounts to draw attention to differences in the experiences of older Chinese women. The key thematic categories were developed from summary statements, memos, and repeated reviews of the data. Data analysis was conducted using NVivo 11.0 (QSR International, Melbourne, Australia).

Thematic analysis is a process of encoding qualitative information under a theme and presenting the repeated information in a systematic way that increases its sensitivity and reliability in order to answer the research question [20 ]. Recently, a number of studies had used thematic analysis method to analysis coroners’ reports [11 (link), 12 (link), 21 (link), 22 (link)] by using inductive and deductive approach. The main difference between inductive and deductive reasoning is that deductive reasoning is based on a hypothesis to test a theory whereas the inductive reasoning generates theory through interpretation of the data.
Our overall approach was similar to those described in Scourfield [11 (link)] in that we took a wider sociological perspective in the assessment of suicide among males. Within this overarching framework, we used deductive process of thematic analysis to identify, analyse, and report repeated patterns (themes) from coroners’ report [23 –25 ]. Nvivo 11.0 was used for data analysis (QSR International Pty Ltd). To protect individual indentity, names, ages and geographical locations have been altered.