‘Down to the person, the individual patient themselves’: A qualitative study of treatment decision‐making for shoulder pain

Abstract Introduction Many inconsistencies have been identified in the translation of evidence‐based treatment recommendations for musculoskeletal shoulder pain into healthcare services, with little known about factors influencing decision‐making. The objective of this study was to explore the views and experiences of healthcare providers (HCPs) and people living with shoulder pain on treatment decision‐making. Methods Adopting a qualitative design, purposeful sampling was employed to recruit 13 individuals with nonspecific musculoskeletal shoulder pain and 30 HCPs. Data were collected through 1:1 semi‐structured interviews and analysed using an approach informed by Constructivist Grounded Theory. To facilitate analysis, two patient and public involvement (PPI) meetings were conducted. Results Most participants (69%) had shoulder pain of ≥1‐year duration. Biomechanical beliefs about shoulder pain predominated and were heavily influential in decision‐making for both patients and HCPs. Despite a consensus that therapeutic alliance facilitated decision‐making, the extent of collaboration between HCPs and patients in treatment decision‐making was rather limited. In addition to condition‐specific factors, Individual patient characteristics and resources also influenced treatment decisions. Conclusion Findings revealed the complexity of the decision‐making process for both patients and HCPs, exposing substantial gaps between the reported views and experiences of participants and the principles of client‐centred and evidence‐based practice. There is a pressing need to enhance the translation of evidence‐based knowledge into practice in this clinical area. Patient or Public Contribution In line with a consultative approach to collaborative data analysis, a subgroup of participants attended two PPI meetings to provide commentary and feedback on preliminary findings.


| INTRODUCTION
Shoulder pain is the third most common musculoskeletal condition presenting in primary care. It often evolves into a chronic condition, with over half developing persistent pain beyond 6 months. 1 Musculoskeletal shoulder pain describes a spectrum of conditions, including subacromial pain syndrome, shoulder impingement, or rotator cuff disease. 2 Current treatment recommendations commonly include analgesia (e.g., paracetamol and nonsteroidal anti-inflammatory drugs), glucocorticoid injections and exercise therapy as first-line options, with surgery considered a secondary intervention. 3 Many inconsistencies have been identified in the implementation of up-todate treatment recommendations for this population. [4][5][6] Surgery has limited proven clinical benefit for full-thickness rotator cuff tears, 7 and is not recommended for subacromial pain syndrome or rotator cuff disease. 2 This is unsurprising as structural shoulder pathology does not correlate with shoulder pain. 8,9 Nevertheless, shoulder surgery rates continue to increase. 6 A growing number of qualitative studies have explored the experiences of people with shoulder pain and healthcare providers (HCPs) working with this cohort. A recent qualitative synthesis found that individuals with shoulder pain experience emotional, social and functional upheaval, express strong biomechanical beliefs about shoulder pain and are fearful of movement and exercise. 10 A further qualitative synthesis demonstrated a lack of consensus amongst HCPs on how to manage shoulder pain, difficulties implementing research recommendations and challenges in getting patients to 'buy in' to exercise-based treatment. 11 This body of research provides further evidence of inconsistencies in the translation of evidencebased recommendations into healthcare.
Shared decision-making (SDM) is advocated to improve communication of healthcare options, facilitate improved quality of care and better implementation of evidence-based recommendations. 12 A limited body of research on shoulder pain has explored treatment decision-making. Reported patient treatment priorities include a desire to regain movement, understand the problem and be cared for by someone who understands their condition. 13 Patient decisionmaking relating to surgery has also been explored, with failed nonsurgical treatment, pathoanatomical beliefs, limited information on treatment choices/risks and a strong preference for surgery, all increasing the likelihood of pursuing this option. 14,15 While many studies have explored HCP decision-making, relatively few have focused on shoulder pain. One study of physiotherapists (PTs) revealed the influence of both expressed and unexpressed workplace norms, as well as clinical experience on decision-making, sometimes over-ruling research findings. 16 Given the inconsistencies in the translation of evidence to practice and the limited attention to decision-making to date, this study aims to explore the views and experiences of HCPs and people living with musculoskeletal shoulder pain on treatment decision-making.

| Study design and ethics
This qualitative study was informed by Grounded Theory (GT) methodology, specifically the Constructivist Grounded Theory (CGT) approach, as described by Charmaz. 17 GT methodology is a flexible methodology often used when there is little known about a phenomenon, where a theory is generated from data inductively with the aim of constructing an explanatory theory that reveals a process rooted in the area of inquiry. 17 CGT differs from other GT approaches due to its focus on acknowledging multiple realities, engagement in critical analysis throughout the process and capacity for developing high-level conceptual understanding. 18 Hence, we aimed for a high-level conceptual understanding of treatment decision-making grounded in empirical data. This study adopts the consolidated criteria for reporting qualitative research (COREQ) (File S1). 19  shoulder pain, (see File S2). Given the associated greater morbidity levels and socioeconomic burden, 2 individuals with persistent pain were recruited. Data were analysed in tandem with data collection.
Recruitment was guided by theoretical sampling with efforts made to recruit individuals who could provide relevant data to develop the CGT based on initial analysis. Theoretical sampling was also pursued during the interviews with lines of inquiry being followed in later interviews based on concepts identified during concurrent data analysis of earlier interviews. 21 The final sample size was determined at the point of theoretical saturation, the point in sampling where no new properties emerged and categories appeared to be 'saturated'. 22 Recruitment was affected by the outbreak of the COVID-19 pandemic. In the final month of recruitment (March 2020) several HCP interviews were scheduled with little time between them as the interview team was concerned about the potential impact of the pandemic on future HCP participation. This somewhat limited concurrent data collection and analysis. Following preliminary analysis of these final interviews, it was determined that theoretical saturation was achieved. Subsequently, an additional 1-2 participants across stakeholder groups were recruited for confirmation purposes. 22

| Data collection
Data were collected by C. M. through 1:1 in-depth semi-structured interviews between 6 December 2019 and 26 March 2020, the preferred data collection method when using a CGT approach. 23 The interview guides were developed based on the research objective and covered four key topics for HCPs: clinical experience, diagnostic confidence, treatment decision-making and treatment outcomes, as well as five key topics for patients: personal experience, motivation to seek treatment, treatment expectations, treatment decisionmaking and recovery expectations. To refine these guides, C. M. conducted a pilot interview with one HCP and one individual with shoulder pain, who were not subsequently included in the study. At each of the clinical sites, face-to-face interviews took place in a private room. Those unable to attend face-to-face interviews participated in telephone interviews. Interviews were audiorecorded and fieldnotes were documented. Recordings were transcribed verbatim by a transcription service. Basic demographic information was collected from all participants (File S3). All participants had the opportunity to review transcripts before analysis.

| Data analysis
Data were analysed using an approach informed by CGT. Transcripts were exported to NVivo (Version 12; QSR) and three key analytic techniques of coding, categorization and constant comparison were employed. 17 Data from people with shoulder pain and HCPs were analysed separately and later mapped onto each other. Initially, C. M. coded all transcripts descriptively, later progressing to focused coding, which became more conceptual as the analysis progressed.
Constant comparison of all data and codes across patient and HCP accounts was completed. This facilitated conceptualization of codes into higher-order categories. Four categories were identified and further refined through constant comparison. Drafting analytic memos allowed a record of changes in interpretations of codes and furthered analysis at a conceptual level. Data were analysed by C. M. over a 12-month period from the start of data collection, with regular meetings with K. R. and K. M. to challenge emerging interpretations. 24

| Public and patient involvement in data analysis
Following a consultative approach to collaborative data analysis, preliminary findings were presented to a subgroup of participants at two public and patient involvmenet (PPI) meetings in Novemeber 2020 (facilitated by C. M.) for commentray and feedback, in line with a consultative approach to collaborative data analysis. 25 The use of PPI has been advocated to help correct misinterpretations and to challenge the way in which findings are reported. 26,27 For each meeting, 1-2 participants from each group were invited to participate. Meetings were one hour in duration, conducted via Microsoft Teams (V. 1.4.00.11161) and audio-recorded but not transcribed.
Attendees at the PPI meetings were largely in agreement with preliminary findings. One preliminary concept, 'fear of making a mistake', describing decisions to refer onwards, was challenged, with attendees clarifying that this was driven not by fear but as a 'safeguarding' mechanism (PT2), to ensure that nothing was missed.
This refined interpretation was used in the later analysis when documenting beliefs relating to imaging and its role in decisionmaking (Category 4). One person with shoulder pain described their prior expectations of a more hands-on physiotherapy approach and how the detailed information given on treatment options enabled a 'collaborative' (P6) process. Also discussed was the role of trust in being open to exploring other treatment options. The centrality of trust to establishing a therapeutic alliance (TA) and its influence on treatment decision-making was further confirmed by this discussion (Category 1).

| Rigour
The study was conducted in line with guidelines on quality in CGT research. 18 Critical feedback and guidance were provided by K. R., an experienced qualitative researcher with prior experience of GT and critical analyses. A transparent description of how this study was conducted is presented, with recruitment guided by theoretical sampling and saturation. Member-checking enabled participants to confirm the accuracy of data collected. C. M. maintained detailed analytic memos to record changes in the interpretation of codes, while also seeking regular feedback from K. R. and K. M., helping to ensure interpretive rigour. 24 The PPI meetings facilitated additional interpretive analysis and revision, helping to further enhance rigour. 29
Participant demographics and characteristics are presented in File S3. Of note, most participants (69%) had shoulder pain of ≥1-year duration.
Thirty-four interviews were completed face-to-face and the remainder by phone. Interviews ranged in length from 12 to 53 min (average 32).

| Findings
Four categories were identified but the development of a substantive theory integrating these categories was not achieved. The categories do however provide greater conceptual clarity in relation to the factors influencing treatment decision-making for shoulder pain.
Enhancing conceptual clarity without necessarily developing substantive theory is considered a satisfactory outcome for a GT study. 30 HCPs and patients articulated a desire to build and maintain a TA characterized by the trust to support decision-making (Category 1).
Despite HCPs articulating a desire to establish a TA with patients, HCPs appraised patients beyond their shoulders, taking into consideration a wide range of characteristics, such as age and gender

| Limited patient involvement in treatment decision-making
Participants described a wide spectrum of involvement in treatment decision-making, including HCP-led, patient-led and SDM processes.
HCPs largely perceived patients to be passive, preferring their doctor or HCP to take the lead in decision-making.

| Comparison with existing literature
Participants in this study generally reported limited patient involvement in treatment decision-making. An SDM approach is the preferred model for treatment decision-making, shown to improve patient outcomes and reduce healthcare costs. 31,32 SDM is described variably as a joint process where clinicians and patients make decisions together, 12 or as a continuum, where the extent to which the patient-and clinician-lead decision-making varies depending on patient preference and the context. 33 Although numerous studies indicate that most patients seek to share decision-making, 34,35 few patients within the current study described a shared approach. In some cases, HCPs associated a more passive approach with older patients, echoing the findings of previous studies and potentially reflecting ageist views of older adults as being less able to participate in decision-making. 36 We found little evidence of patients being encouraged to adopt a more active role, 37 or of HCPs effectively communicating research evidence in a 'clear, understandable and non-misleading manner' to support SDM. 38 As shoulder pain is a multidimensional condition associated with a multitude of nonbiological factors, 39 a biopsychosocial approach to its management is recommended. 40 The use of a biopsychosocial approach for back pain has spurred the use of interventions, such as cognitive behavioural therapy (CBT) 41 and pain neuroscience education. 42 Despite a biopsychosocial approach being recommended for shoulder pain, we found that biomechanical belief predominated as has been previously reported. 10 decision-making arising from such beliefs is inconsistent with the best available research evidence. 8,9 Patients' preferences are just as important as research and clinical expertize in treatment decision-making. 43 However, we found in most cases that HCP beliefs and clinical expertize far outweighed research evidence or patient preference. Some HCPs described the use of their 'expert' title as a 'lever' to foster compliance, irrespective of whether the proposed treatment reflected up-to-date recommendations. The way in which information is provided to patients heavily influences treatment decision-making in relation to surgery. 44 We found little evidence that HCPs consistently provide patients with balanced up-to-date information on the associated risks and/or benefits of different treatment options. Furthermore, our findings highlight the disparity between the private and public health systems in relation to expedient access to treatment, inappropriate referral and access to treatment incongruous with evidence-based treatment hierarchy. Such inappropriate specialist referrals and underutilization of appropriate treatment for rotator cuff disorders have been described elsewhere. 45 Both patients and HCPs emphasized the importance of TA in treatment decision-making. The 'active' components of a TA are 'empathy, congruence and unconditional positive regard'. 46 Additionally, a TA can facilitate patient-centred care. 47 Although HCPs acknowledged the importance of a TA, descriptions of various patient factors influencing decision-making are incongruent with the tenets of a TA. Some HCPs expressed negative expectations in relation to the engagement and recovery of women, those with chronic or widespread pain, as well as those involved in medico-legal claims.
While some of these factors have been identified as negative prognostic indicators for shoulder pain, [48][49][50][51] it is important that HCPs do not perpetuate or reinforce stereotypes in clinical practice.
Adopting a subjective approach can negatively impact the quality, consistency and accuracy of decision-making. 52 While TA was valued by HCPs, our findings revealed it is not being fully realized in practice, with negative consequences for patient involvement and the potential for clinical biases to adversely influence treatment decision-making.

| Strengths and limitations
To our knowledge, this is the first qualitative study to explore the views of both people with shoulder pain and HCPs in relation to treatment decision-making. Strengths of the study include the recruitment strategy involving recruitment at four clinical sites, within both private and public healthcare settings, including a large number and broad range of highly experienced HCPs and patients who had experiences of a variety of treatments, with attention to both theoretical sampling and saturation. Data were subjected to prolonged analysis and PPI consultation during analysis, thereby contributing to the study rigour. 29 Although each stakeholder contributed to discussions during the PPI sessions, formal training may have further enhanced their participation, confidence and contributions. 53 A further limitation of this study is that data collection and analysis were completed by a PT, thereby potentially influencing the responses collected and the analysis as such.
Reflexivity, facilitated through fieldnotes and research team discussions, aimed to minimize this potential impact. While the interviewers' background may have influenced participants' responses, the data collected included wide-ranging experiences of physiotherapy, both positive and negative. Furthermore, the involvement of participants in the interpretation of data collected and the refinement of identified categories was a further strategy to minimize the impact of the research team on the results and conclusions drawn. People with ≥6 weeks' history of musculoskeletal shoulder pain were eligible to participate. However, most participants (69%) had shoulder pain of ≥1-year duration, with only one participant reporting symptoms of ≤3 months. Although people with shoulder pain discussed decisionmaking across the entire period from symptom onset, the experience of those whose symptoms resolve within 3 months is not well represented in this study.

| Implications for clinical practice and future research
The findings of this study can be used to support an SDM approach and evidence-based treatment decision-making in shoulder pain. It is MAXWELL ET AL. | 1115 recommended that future research on this topic should adopt a PPI approach from conception to dissemination as this would likely facilitate improved stakeholder engagement and enhanced application of findings in clinical practice. 54 Future exploration of factors influencing patient's treatment decision-making in the early phase after shoulder pain onset is warranted.