Refugee and migrants' involvement in participatory spaces in a US practice‐based research network study: Responding to unanticipated priorities

Abstract Background Refugees and migrants face suboptimal involvement in spaces for primary healthcare decision‐making. Given the rising numbers of resettled refugees and migrants in primary care settings in the United States, there is an urgent need for patient‐centred outcome research in practice‐based research networks (PBRNs) with diverse ethnolinguistic communities. This study explored whether researchers, clinicians and patients would achieve consensus on (1) a common set of clinical problems that were applicable across a PBRN and (2) potential clinical interventions to address those problems to inform a patient‐centred outcomes research (PCOR) study in a similar research network. Methods In this qualitative participatory health research study, patients from diverse ethnolinguistic communities and clinicians from seven practices in a US PBRN discussed preferences for PCOR responsive to patients and the clinicians who serve them in language‐discordant settings. Researchers and an advisory panel that included patients and clinicians from each participating practice held regular advisory meetings to monitor progress on project milestones and solve emerging problems. Participants took part in 10 sessions using Participatory Learning in Action and the World Café methods to identify and prioritise their ideas, using questions set for them by the advisory panel. Data were analysed based on principles of qualitative thematic content analysis. Results Participants identified common barriers in language‐discordant healthcare settings, principally patient‐clinician communication barriers and suggestions to overcome these barriers. A key finding was an unanticipated consensus about the need for attention to healthcare processes rather than a clinical research priority. Negotiation with research funders enabled further analysis of potential interventions for care processes to improve communication and shared decision‐making in consultations and the practice as a whole. Conclusion PCOR studies should examine interventions for improving communication between patients from diverse ethnolinguistic communities and primary care staff if the sorts of harms experienced by patients experiencing language‐discordant healthcare are to be reduced or prevented. Flexibility and responsiveness from funders to unanticipated findings are key structural supports for participatory health research in primary care clinical settings with this population and others who experience marginalisation and exclusion. Patient or Public Contribution Patients and clinicians participated in the study both in the formulation of the study question, data collection, analysis and dissemination of these results; consented to their individual participation; and reviewed early drafts of the manuscript.


| INTRODUCTION
The World Health Organisation has developed primary care policies that promote the involvement of patients and communities in health decision-making. These policies emphasise the need for people's involvement as individuals (e.g., in their consultations) and as communities (e.g., in service development, research and policymaking). 1,2 The rationale underpinning these policies is twofold: people have a right and responsibility to be involved and their involvement can lead to a more comprehensive knowledge base to guide the delivery of responsive, person-centred healthcare. Taken together, people's involvement in health decision-making can promote health equity. 3,4 Shared decision-making in consultations, community advisory boards for research, and involvement in mechanisms for policymaking may all be conceptualised as participatory spaces in contemporary primary healthcare. 5 By definition, participatory spaces are shaped by physical, temporal and social dimensions that influence the opportunity for, and enactment of health decision-making. 6 For example, whether brief interactions occur in primary care practices focused on clinical goals; or in community settings with lengthy meetings focused on research or policy goals shapes the scope for involvement in health decision-making. It can, however, be challenging for service providers to adjust to new social identities beyond their caregiving role as they move from the consultation space to those set up for research or service development projects 7 ; and to accept the time required for work in community settings. 8 Thus, participatory spaces in primary healthcare share an implementation problem-typically they are not embedded as routine and normalised practice in primary healthcare. 9 Further, a pattern of exclusion whereby patients and communities who experience marginalisation experience suboptimal involvement may be revealed in each of these participatory spaces. 10,11 Given contemporary global migration patterns, refugees and other migrants in the United States and other resettlement countries bring increasing cultural and linguistic diversity to primary care settings, such that cross-cultural consultations are no longer confined to specialist refugee clinics but are increasingly a feature of mainstream clinical practice even in rural and remote areas. 12,13 There is thus an urgent need to attend to the suboptimal involvement of refugees and migrants in primary healthcare decision-making; to understand their health needs and to take action for equitable health outcomes. Yet, despite the policy and legal imperative, variability in the quality of care for migrants persists. The unavailability in the use of trained interpreters in consultations (in-person or by telephone) leads to inappropriate diagnoses and treatments, medication errors, higher emergency department use, and longer hospitalisations. 14 Despite the mandatory provision of language interpretive services in US primary care settings that care for patients with publicly provided health insurance (Medicare, Medicaid), 15 and further clear evidence for inferior outcomes when providers are linguistically discordant from patients, [16][17][18][19] bilingual clinicians and clinical staff are still the exception rather than the rule, so that US patients with limited English proficiency experience limited access and care quality. 20 Further, in the US, patient-centred outcomes research (PCOR) with refugees and migrants in practice-based networks has been challenging. PCOR focused on various medical conditions including diabetes, asthma and gout found unique relational and communication issues relevant to different individual cultural groups at individual sites-which make involvement with these populations daunting logistically. [21][22][23][24][25][26][27][28][29][30][31][32] This has led to culture-specific approaches to PCOR with such patients, each individualised to the culture in a particular LEMASTER ET AL. | 1597 project. This is valuable for the local setting but means that data and interventions are not easily comparable or generalisable across healthcare systems, thus limiting the scope for systemic change more broadly in primary care services. [21][22][23][24][25][29][30][31]33,34 In the United States, practice-based research networks (PBRNs) are an essential mechanism to deal with these issues because they are designed to involve patients and community members from multiple practices in PCOR. This paper explores the involvement of refugees and migrants from diverse ethnolinguistic populations who do not have a shared language and cultural background with their primary care providers, whose practices participate in a nationwide US PBRN. In line with international evidence, 35 these populations have not been well represented in PBRN activities before this project.
The current study was designed to engage multiple refugees and migrant patient groups in these practices to identify clinical problems that need attention across groups; and to develop a broad-based, well-interfaced network of patient-engaged primary care clinics ready to move forward together to conduct comparative effectiveness studies to address the clinical problem they identified as most important. Although the primary goal of the funded project was to generate an operational guide for future PCOR in PBRNs with diverse ethnolinguistic groups, 36 a secondary goal was to seek consensus across a PBRN on a clinical research project, since consensus is an important indicator of shared decision making given the power asymmetries between patients and practice staff. 37 The overall aim of this paper is to explore patient and clinician perspectives during that process. Specific objectives are to examine whether researchers, clinicians and patients across the emerging network would achieve consensus on (1) a common set of clinical, that is, health problems that were applicable across participating primary care practices, and (2) potential clinical interventions to address those problems to inform a PCOR study in a similar context.
We employ the concept of participatory space as a heuristic device to describe methods, interpret findings and make recommendations.

| Sampling and recruitment/retention of participants
The clinicians and patients in the project are the samples for the analysis on which this paper is based. Based on principles of purposeful sampling for information-rich participants, 38 all network primary care practices that serve linguistically and culturally diverse patient populations were invited to participate in the project. Those who opted in took part from the inception of the project, including its initial design. In total, seven primary care practices participated from six US states. Practices varied in size (solo to many providers), language preference of patients served and approaches to supporting patients' language needs (in-person vs. phone interpreters vs. bilingual staff). In participating practices, at least 25% of patients were from diverse ethnolinguistic groups.
There were no exclusion criteria for healthcare providers at the participating clinics. Twenty-three clinicians at the seven practices agreed to take part.
Each practice nominated a Site Coordinator and a Lead Clinician (physician) for the project. The Site Coordinator was a practice manager or lead nurse who served as the point of contact between the clinic and the project organising team, as well as with the patients, caregivers or practice staff who took part.
In the first month of the project, to facilitate communication with clinicians and participating patients from different ethnolinguistic groups, Site Coordinators invited at least five patients from each practice who were refugees or migrants and who were bilingual in English to participate in group discussions. Patients were informed that participation was voluntary and that the provision of their medical care would be unaffected by their decision to participate (or not). We did not collect personal health information or ethnolinguistic data from individual participants because the emphasis was on collective experiences and collective knowledge generation at each practice site.
We excluded patients who could not participate for cognitive reasons. Participating patients received usual care from any primary care physician or other clinicians in the practice, including nurse practitioners or physician assistants.
To maximise practice and patient retention, we used published recommendations for minimising losses to follow-up that we have used successfully in prior studies. [39][40][41][42] For practices, we used twice monthly email interaction and/or online meetings. One site withdrew from further participation after session 4 citing their own lack of staff capacity as the reason for doing so. Data provided by that practice before session 5 were included for analysis.
For patients, Site Coordinators collected multiple types of contact information from participating patients/caregivers-telephone (including cellular), mailing address and email addresses for the patients and alternate contacts. We incentivized patient engagement and retention with $30 USD payments for participation in each sitebased discussion. Overall, while not every clinician and patient attended every session, the attendance records at each clinical site showed strong retention of most participants across the sessions. Table 1 describes practices, the number of providers and ethnolinguistic communities served by the practice.

| Data generation and analysis
We used a blended approach for data generation and coanalysis by combining two participatory approaches that have been effective in qualitative primary care research with refugees and migrants. [43][44][45] We integrated principles from Participatory World Café (PWC) and Participatory Learning in Action (PLA) methods. PWC and PLA are similar group discussion/interaction methods designed to address power asymmetries in participatory spaces between diverse participants. 46 Both methods are designed to bring diverse collaborators from different types of organisations together to generate qualitative data; emphasise 'real-time' coanalysis of emerging themes 47 ; and include prioritisation of the ideas generated in the discussion to guide concrete action/next steps. 44,48 The US-based organising team was responsible for logistical and methodological oversight of the process. Site coordinators and leaders initially came to Kansas City for a 1-day training in which we practised data generation methods to support their consistent use across the seven participating clinical sites. The topic guide for data generation was developed iteratively by the Advisory Board during the project based on initial and emergent areas of investigation (see Supporting Information: Appendix I).  all participants shared all their ideas, participants collectively engaged in a card sort. 48 For this, they rearranged their ideas into emerging themes, so that responses that were more closely related were physically grouped together; and so that those themes which participants collectively perceived to be most impactful to the question were closest to the question on the board. After group members agreed on the rearrangement, they then identified a title for each theme they had created together and labelled the theme on the board. After all data were collected, the author group examined the data iteratively using a thematic content analysis approach. 47 Three coders (J. W. L., C. L., T. M.) independently reviewed the data and its responsiveness to study questions, for example, exploring whether/ how participants would identify a common set of barriers and facilitators applicable across participating primary care practices to conduct PCOR that is responsive to both clinician and patient preferences. The purpose of this analysis was to ensure that the iterative analysis conducted in between data generation sessions was complete and synthesised to understand anticipated and unanticipated emergent findings (described in detail in Section 3). Coders generated memos elucidating these codes during this analysis process, which were reviewed and confirmed by consensus between the three coders.

| PWC method
Final memos were reduced and organised into seven overarching cross-cutting themes, using a similar consensus approach. A fourth researcher who was not involved in the data collection or initial analysis (A. E. M.) reviewed the data, the original list of memos and the final cross-cutting themes to confirm that the data matched the codes.
An early draft of this manuscript was circulated to site leaders to obtain member checks from participants and ensure that the crosscutting themes were consistent with their own perceptions about the project (which they indeed were). The final seven cross-cutting themes are listed in Box 1.
In Section 3, we present key findings in relation to the two objectives of this paper: can researchers, clinicians and patients seek consensus on (1) a common set of clinical problems that were applicable across participating primary care practices and (2) potential clinical interventions to address those problems to inform a research grant application.

| RESULTS
The following themes were identified in our analysis. Please also see Table 1, in which we identify participating practices by letters (A-G).
We refer below to the practices from which comments related to each theme originated.   States and international settings. 16,17,19,[51][52][53]  There was considerable investment in designing an inclusive, participatory research space. The decision to employ PWC and PLA as innovative and interactive material practices to manage and support power asymmetries in these spaces is still relatively rare. 46 More standard methods (interviews, focus groups) dominate the field. 59 The use of PWC and PLA was strongly endorsed by the project participants, 36 which is in line with findings from previous primary care participatory studies with refugees and migrants in the United States and European contexts. 44,60,61 It is important to learn how and whether these methods can be scaled up or normalised in primary care research, and whether their use improves measurable health and healthcare outcomes important to patients and clinicians.
Other innovative methods, such as arts-based, culturally attuned, whole-person methodologies should also be explored to expand the suite of material practices to manage and support power asymmetries in intercultural research spaces. 62

| Methodological critique
The participatory approach in this PBRN-based project was strong and in line with the hallmarks of quality academic-community partnerships with community involvement in developing a research question, governance, data interpretation and dissemination. 63 All participants opted into the project and, thus, we cannot claim representativeness for wider practice or patient populations. Five out of six practices sustained engagement in project activities and dissemination over a duration of 24 months.
The cogeneration and analysis of PWC and PLA data was a strength of the study based on their inherent visual and analytic characteristics that gave voice to patients from diverse ethnolinguistic communities. 61 We believe that the richness of dialogues within and across sites over time deepened our understanding of the issues raised and provided a unique perspective on the field of migrant healthcare. A limitation is that data generated and analysed collectively are not attributable to individual participants and do not generate extensive quotes in the way that other methods such as interviews or focus groups do. The qualitative analysis during, and after, the project ended adhered, however, to a recognised analytic approach. While the former benefited from participants' insights for data interpretation, site participants were not involved in the final retrospective analysis. Finally, patient participants were bilingual in English and their mother tongue, so were not currently experiencing language discordance with their healthcare clinicians; however, all patient participants had at some time in the past experienced language discordance. Data reported herein represent reports from community representatives about extant language discordance experiences in primary care contexts, rather than participants' current personal health care experiences.

| CONCLUSION
Refugee and migrants' meaningful involvement in a participatory research space using innovative and interactive methods was essential to understand their unanticipated priority: to implement interventions to improve their meaningful involvement in other participatory spaces; that is, the clinical consultation space and the space of access. This is exactly the sort of intervention that PCOR studies should undertake if the sorts of harms experienced by patients experiencing language-discordant healthcare are to be Committee.

CONFLICT OF INTEREST STATEMENT
The authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT
Data from the study are summarised in the appendices and available in full upon request from the authors. No material (copyrighted or otherwise) was reproduced from other sources. This is not a clinical trial.

ETHICS STATEMENT
The study was approved by the American Academy of Family Medicine Institutional Review Board.