Patient involvement in qualitative data analysis in a trial of a patient‐centred intervention: Reconciling lay knowledge and scientific method

Abstract Background We conducted a pilot study of an intervention to facilitate patients’ agenda setting in clinical consultations. The primary aim of the study was to test the feasibility of running the randomized controlled trial. A secondary objective was to assess the extent to which patient and public involvement (PPI) could contribute to the process of qualitative data analysis (QDA). Aims To describe a novel approach to including patient partners in QDA; to illustrate the kinds of contribution that patient partners made to QDA in this context; and to propose a characterization of a process by which patient involvement can contribute to knowledge production. Methods Six patient and public representatives were supported to contribute to data analysis via a range of modalities. During a series of QDA workshops, experienced research staff role‐played consultations and interviews, and provided vignettes. Workshop data and PPI diaries were analysed using thematic discourse analysis. Results We characterized a process of thesis, antithesis and synthesis. This PPI group contributed to the rigour and validity of the study findings by challenging their own and the researchers’ assumptions, and by testing the emerging hypotheses. By training PPI representatives to undertake qualitative data analysis, we transformed our understanding of doctor–patient consultations. Conclusions This research required changes to our usual research practices but was in keeping with the objective of establishing meaningful patient involvement for a future definitive trial. This work was informed by concepts of critical humility, and a process of knowledge production enabled via the construction of a knowledge space.


| INTRODUC TI ON
Enabling patient and public involvement (PPI) in health services research is now a prerequisite. [1][2][3][4] Two of us have theorized PPI as "knowledge spaces," that is spaces where expert knowledge and forms of lay knowledge can interact with each other on an equal basis. 5 In this paper, we trace out how the interaction between different forms of knowledge works in practice drawing on an example of PPI in qualitative data analysis to illustrate our points.
Our analysis of the process has led us to conceptualize this interaction as a dialectical one, operating between lay and academic perspectives.
Previous work has indicated that PPI has a role in qualitative data analysis (QDA). It can correct researchers' misinterpretations and challenge the ways in which findings are reported, 6,7 and add value to the products of research analysis. 8 While qualitative methods to augment clinical trials are mainstream, 9 the role of PPI is often overlooked. 10,11 Where PPI representatives have been involved in QDA alongside clinical trials, the impact of such involvement is often unclear. 12 Examples of PPI in QDA as explicit emancipatory practice, sometimes described as community-based participatory research (CBPR), 13 are prevalent in sociological research and achieved by tailoring how data are presented, engaged with, managed and categorized. Working with disadvantaged women, Jackson 14 provided facilitation alongside simplified instructions and processes, such as encouraging participants to summarize data in lay language. Conducting analysis with people with learning disabilities, Nind 15 recommended tailoring approaches to local contexts, with academics variously taking on the role of "trainer, coach, scaffolder, mentor, partner in dialogue, co-learner, reciprocal learner and practical facilitator." Kaomea 16 has also shown that indigenous participants work with a different set of perspectives and tools when undertaking analysis, which may challenge taken for granted research practices, allowing findings to be challenged and re-examined.
The underpinning philosophies of CBPR and PPI are commensurate, with both valuing meaningful partnership and collaboration to achieve change in the conduct and products of research. 17 We, academic co-authors JF, AG, and NB, also drew on the literature about critical humility and participatory research to guide our conduct. 13,18 Emancipatory practices can be fostered by researchers' "critical humility," or lifelong commitment to self-evaluation and critique, as lifelong learners and reflective practitioners. 18 The political imperative is to subvert or challenge the status quo (here the canon of accepted research methods and practices), with the objective of enabling those who are traditionally the focus of studies (e.g. "patients") to understand their world in order to transform it-a form of "catalytic validity". 19 Although Lather 20 proposed that catalytic validity needs to be consciously designed into the research process in order to democratize knowledge, Baines 21 contends that facilitating meaningful PPI contributions to knowledge would require a shift in the political economy of universities and research funders. Furthermore, the selectivity with which PPI representatives are chosen (reifying some voices to the detriment of others) has also been questioned. 22,23 However, this situation is beginning to change. 24,25 In this study, as the aim was to ensure that patients' concerns were addressed in the consultation, a correspondingly emancipatory approach to the data analysis required patients' perspectives to inform the data analysis. Crucially, analytic judgements about whether the intervention had indeed facilitated the discussions of patients' concerns should be informed by patient perspectives.
Conventionally, these judgements have been made by academic researchers alone. 26

| Diabetes Intervention for Agenda Trial (DIAT)
We conducted a pilot study of a pre-consultation intervention in which patients were supported by health-care assistants (HCAs) to complete a Web-based intervention to facilitate the production of "their own agenda" for discussion in a consultation with their diabetologists (Diabetes Intervention for Agenda Trial (DIAT); 27,28 ). By combining trial and qualitative methods, we sought to identify how the intervention influenced consultations, and decide whether proceeding to a full trial was appropriate.
Patients were involved at all stages, in order to increase the probability that the intervention would be adopted in practice. 29 The research question was generated from a prioritization exercise, People with diabetes (including PPI representative 1 or "PPI1") suggested that outpatient clinic appointments are pressured times, where health professionals can overlook issues that are worrying patients or where patients can feel inhibited from voicing their concerns. Supported by a PPI Research Fellow (AG), two members of PenPIG (FH-G and JH) who have diabetes joined the research team as funding co-applicants and members of the project management team. They co-wrote the study documentation, assisted with training the HCAs to deliver the pre-consultation intervention, participated in the data analysis and disseminated preliminary findings.
Eight additional people with diabetes, recruited from the NIHR Diabetes Research Network and Diabetes UK, provided feedback on an early iteration of the intervention, and another participated in the trial steering group. With 4 other members of PenPIG, the two coapplicants ("PPI2" and "PPI3") participated in a PPI qualitative data analysis group ("PPI group" with R denoting Researcher).
Seventy-one patients were randomized to either the intervention or usual care. 26 With participants' and diabetologists' consent, we audio-recorded intervention sessions and clinical consultations, to explore how participants generated and utilized the intervention. Thirteen patients randomized to receive the intervention consented to be recorded and twelve had both an intervention session and a consultation recorded, while one had only their consultation recorded. Twelve control patients consented and had a "usual" consultation recorded. Recording consultations facilitated our understanding of the impact of the production of an agenda on the clinical consultation, and its subsequent use in practice when compared with usual care. 24,30 Author A conducted 12 semi-structured interviews with a convenience sample of participating staff and 12 patients (6 in each trial arm) to explore wider organizational factors. Interview questions were based upon the existing literature and PPI perspectives and aimed to capture staff and patient experience of diabetes consultations and the intervention specifically. Data were transcribed verbatim, anonymized and managed using Nvivo software. Preliminary framework analysis 31 facilitated deductive and inductive analysis, enabling the exploration of the role of the intervention and trial processes alongside the elicitation of patient, provider and PPI perspectives.
Qualitative data analysis began with JF and a diabetes specialist nurse (DSN) co-applicant independently familiarizing themselves with the first 20 consultations. 32 In the control arm, we explored the usual consultation style of each participating diabetologist. In the intervention arm, we explored the extent to which diabetologist used the agenda form as a guide for the consultation. A preliminary charting exercise (Table 1) allowed us to identify relationships between the process of agenda identification and agenda use. 23 Thus JF and the DSN co-applicant identified a preliminary typology of consultations (patient ignored, diabetologist led, patient led, diabetologist and patient led).
The aims of this study were to describe a novel approach to involving patient partners in QDA; to illustrate the kinds of contribution that patient partners made to QDA; and to propose a characterization of a process by which patient involvement can contribute to knowledge production.

| Training
JF, AG, and NB trained a group of six PPI representatives (the PPI group, including co-applicants PPI2 and PPI3), in QDA techniques (such as paying attention to nuances in consultations, and constructing patterns within and across consultations 32  No initial discussion of agenda form. D begins by identifying P on his records and checking which medication P is taking. P is slightly confused, and D says: 'I suppose one thing this illustrates is really useful if you can bring a copy of your prescription with you.' P says that he recently had a heart attack, but D cannot find a record of relevant blood tests, and D asks P if he is 'quite sure' of when he was in hospital. D says that a recent HbA1c result 'wasn't very good' and P admits that he has been struggling with selfmonitoring, and has been having high results, which he attributes to stress about his sisters illness. D then refers to the agenda form: "Okay now you have got this list of questions here for the study and the first thing you ask is my blood sugar levels are always high I find the emotions play a big part in raising these levels. Can I increase the dose of Exenatide? The answer to that last question is no, you can't." There is a lengthy discussion re: Exenatide, with D proposing other medications to control HbA1c, and P suggesting that D has not got his dose correct and AG therefore role-played the anonymized consultations from the charting exercise, making sure that they captured the audible nuances of the audiotape, and provided vignettes which summarized the "cases." Role-play provides insight into social interaction and can provide a stimulus for discussion. 33 Vignettes are stories about individuals, situations and structures, which allow actions to be explored in context and provide a less personal means of exploring sensitive topics. 34 In Workshop 1 (3 hours), the researchers introduced the study to the PPI group members who were unfamiliar with the project and the aims of the workshops. They described data collection and preliminary analysis to date. Large data matrices were used to demonstrate how the researchers had conducted their preliminary coding, and participants were given instructions as to how they might produce their own codes about "what is going on" and "whose agenda is dominant" from the data that they would be given. The researchers then provided the PPI group with a vignette (Patient 1- Figure 1) and asked them to read the transcript and observe a role-play of a consultation in the usual care arm of the trial.
PPI participants were then provided with the researchers' preliminary typology, as a stimulus to develop their own codes. The researchers had produced a form to guide participants in conducting their analysis, but this was felt by the group to be too constraining.
Instead, "codes" were identified from line-by-line discussion of the transcript. These analytical discussions were captured by the researchers using a flip chart. The skilled PPI facilitator summarized PPI discussions and probed further for clarification. Notes of these discussions informed revisions to the preliminary coding frame, which, in turn, were fed back to the PPI group and shaped their subsequent analysis.
In Workshop 2 (6 hours), role-play was used to portray a consultation in the intervention arm of the trial (Patient 2- Figure 1) and a subsequent qualitative interview in which the participant discussed his experience, alongside a contextual vignette. The PPI group read and discussed transcripts from both the consultation and interview, and were asked to critique and develop the typology. In Workshop 3 (2 hours), the PPI group was provided with their revised typology, with the codes that they had identified, supported with descriptors from the previous discussions. They were given 4 further vignettes from a purposive sample of the consultation data. In small groups, participants led the analytical discussions, testing the codes and descriptors against the data, in the same way that teams of qualitative researchers conduct analytical discussions.
Thus, over the course of three workshops, and using a hands-on approach, the researchers and the PPI group went through the key stages of framework analysis, consisting of familiarization with the data, coding, developing an analytical framework, applying the framework, charting and interpreting the data. 31,33

| Data collection
The 2 PPI co-applicants (PPI2 and PPI3) kept diaries and included information which they felt comfortable sharing with the wider research team. 35 Repeat semi-structured interviews (2 qualitative interviews, one in each year of the project, with each of the two co- West-Preston (13/NW/0123). The 2 PPI co-applicants provided written consent for their diaries and interviews to be analysed. As the group's PPI analysis was a PPI activity to shape research, NHS REC approval was not required.

Patient 1:
This 28-year-old woman was seen in outpatients at Hospital 1. She was randomized to the "usual-care" arm of the study and saw her diabetologist as usual. She has type 1 diabetes and was concerned about "hypos' and panic attacks. This consultation was 19 minutes. On the baseline questionnaire, she indicated that she did not have an eating plan. At 3 months post-consultation, she repeated that she did not have an eating plan and that she ate high-fat foods and chocolate. She stated that she had not received dietary advice in the previous 2 years. At six months post-consultation, she indicated that she had still not changed her treatment.

Patient 2:
This 57-year-old man with type 2 diabetes was seen in outpatients in Hospital 2. He was randomized to the "treatment arm" of the study, so before his consultation he was seen by a health-care assistant, used the Web-based intervention and produced an agenda form. The consultation was 35 minutes. He was the only study participant who did not give consent for his intervention session to be recorded, and he implied that this was because of his concern about sexual problems. However, he was the only person in the treatment arm, to have his consultation with the Diabetologist recorded and participate in an explanatory indepth interview.

| Data analysis
Thematic discourse analysis was developed over the course of 3 workshops using all sources of data, paying attention to micro-level discourse and the construction of the meaning of involvement to illustrate the cumulative process of knowledge production in a knowledge space 31,32 by Authors JF, AG, and NB, with further clarification from FH-G and JH. The results of each workshop are presented in turn, reflecting our analysis at both a workshop and cumulative level, and our objective of assessing the contribution of PPI to the process of QDA.

| Workshop 1: Hypothesis testing
By providing the PPI group with a preliminary typology, the academic co-authors were demonstrating an analytical process without suggesting that the PPI group should validate those suggestions. 37 In terms of the critical humility of the researchers, and recognizing that our preliminary analyses were led by our own assumptions, we saw this as part of a conversation in which we were opening ourselves up to their experiential knowledge and interpretations as one would with other research and clinical partners. 38 The PPI group subsequently developed the preliminary typology by proposing the terms "patient identified" (rather than "patient led") and "diabetologist facilitated" (rather than "diabetologist led") He fed in at the important places with his stuff.

| Workshop 2: Challenging
Here, we explore one of the coding discussions ("patient identified and diabetologist facilitated"), to demonstrate that the PPI group were able to develop a refined typology of the consultations which challenged the assumptions of the project team, as well as their own views. The group were discussing how to code a particular consulta-  Members of the PPI group were given a further opportunity to amend the analytical typology, and it was concluded that the data fitted the categories earlier identified by the PPI representatives.

| Workshop 3: Re-challenge
Finally, the group were able to produce a highly nuanced synthetic ac- In this final example, the PPI group identified that diabetes consultations required care that was tailored to the needs of the patient and contextual, suggesting that an agenda form was helpful in some situations but not others-a potential challenge to the premise of the feasibility study. Thus, the PPI contribution and wider analysis of the DIAT feasibility study suggested that the study protocol would need further development before a definitive randomized controlled trial would be warranted. By equipping the PPI group with analytical techniques, they were able to engage with the data and crucially, in this clinical trial of a patient-centred intervention, make the findings clinically relevant.

| D ISCUSS I ON
There have been calls both to standardize the practice of involving participants in clinical trials 40 and the reporting of involvement in qualitative analyses. 41 We have responded by demonstrating the process which was undertaken by the PPI group, facilitated by experienced qualitative researchers (JF and NB) and skilled PPI facilitation (AG). By applying our in-depth knowledge of qualitative research theory and PPI practices, we were able to improve the inclusivity and understanding of this PPI group, 42 by developing an approach more akin to the emancipatory and participatory practices described in the sociological literature. [14][15][16] To facilitate catalytic validity, defined as the degree to which research moves those it studies to enable them to better understand the world and transform it, 19 the academic researchers reflected on their own practices and attempted to foster mutually beneficial and non-paternalistic partnerships with members of the PPI group. 18 This required the deconstruction of established qualitative practices, to enable the PPI representatives to participate in the interpretation of data and knowledge production. This experience led us to reflect that the emancipatory or ethical imperative of PPI is not at odds with the scientific methods employed in clinical trials (of testing and refining hypotheses).
In trying to understand and conceptualize the process described above, both for ourselves and others, we came to the conclusion that what had taken place could be best understood as a Previous research has identified how non-research-based professionals can feel isolated by the process of having specific tasks within a trial, which can inhibit their sense of agency. 43 Clinicians have suggested that lack of time allowed by planned research activities has limited their ability to adequately recruit patients to clinical trial targets, 44,45 while others have detailed the emotional labour of managing patients' (mis)understanding of randomization processes. 46 Even though the PPI activities in this study were crafted to achieve longitudinal inclusion, this did not mitigate the partial view of the research that PPI collaborators expressed early in the research process. We thus considered the experience as akin to the boundaries of inclusion described by some clinicians participating in trials. However, by making changes to our usual research practices, the contribution to QDA specifically was perceived by PPI collaborators was a particularly valuable experience, because it went some way towards valuing their knowledge and redistributing power, which enabled them as "knowledgeable actors" within a knowledge space. 5 In turn, this enabled us to develop a far more nuanced understanding of diabetology consultations and identify the limitations of our intervention within current healthcare provision. 27,28 In 1808, Fichte 47 identified that forming a thesis involves setting out or laying down an argument. The thesis is then challenged by its antithesis. This challenge is then itself challenged, and a process of reconciliation is sought in a synthesis where both views are accommodated or surpassed. In an ongoing process, this synthesis becomes the starting point for a new cycle of knowledge creation. 48 For Fichte, this process was not the preserve of academics, but rather antithesis was the process by which others held the elite in check. 49 In 1940, Popper 50 contended that this dialectic triad (of thesis, antithesis and synthesis) surpassed the traditional scientific method of trial and error, as it contained an imperative to reconcile or further explore different perspectives, which were absent from other methods of scientific enquiry.
A thesis is formed when an idea or hypothesis is outlined, and we began our analytical process by orientating the PPI group to our preliminary typology in response to the overall research question which the pilot trial sought to answer. The PPI group subsequently developed the preliminary typology by proposing their own terms.
An antithesis is the challenge to a thesis by the presentation of contrasting or opposing views, and as the PPI group discussed the data with each other, they began to challenge the researchers', each other's, and indeed their own interpretations. With such a wealth of experience, we did not expect the PPI group to have a unified voice, and here, we demonstrate how they wove individual stories into storylines which, as qualitative researchers, we typically think of as "themes." A synthesis occurs when the challenge produced by an antithesis is itself challenged. It becomes the next thesis to inform the analytical process, which the PPI group engaged with by testing out their own hypotheses. In the final analysis workshop, members of the group demonstrated their ability to analyse and synthesize the findings from the qualitative data.
This approach may have utility in other areas of health services research, and clinical trials more specifically, for example when assumptions about the acceptability or effectiveness of trial interventions warrants further exploration to identify core components and their optimal alignment. 29 By detailing our approach, we are able to suggest a template for how other researchers might facilitate PPI in QDA (Table 2). We are aware that this is one illustrative example of a broader dialectical process. Here, the PPI group was provided with a typology to initiate the dialectical process, although this initial "thesis" could have been generated by patients, with academics providing the "antithesis." Similarly, the dialectic process could have continued, but we were constrained by the timeframe of the funded project.
A strength of this approach is that it facilitates meaningful patient and public involvement which we feel both augments the validity of the research findings and provides a replicable process that could be employed in other clinical trials or qualitative studies. A limitation might concern the individuals who were engaged in this study.

| CON CLUS ION
We successfully trained and collaborated with PPI representatives to undertake qualitative data analysis which transformed our understanding of doctor-patient consultations, 24  We contend that the driver for actively involving patients and members of the public in clinical trials should not merely be its requirement for successful funding. 38 Rather, the imperative should be to develop more holistic understandings of complex problems via a change in the way that knowledge is produced, 48 Dingwall et al 47 proposed that qualitative research can augment the findings of clinical trials and address the humanitarian issues of equity and effectiveness. We propose that by undertaking QDA with PPI representatives using a dialectic process, we collectively produced more valid and clinically relevant findings.