Objectives The demand for total knee arthroplasty (TKA) is increasing. Differentiating who will derive a clinically meaningful improvement from TKA from others is a key challenge for orthopaedic surgeons. Decision aids can help surgeons select appropriate candidates for surgery, but their uptake has been low. The aim of this study was to explore the barriers and facilitators to decision aid uptake among orthopaedic surgeons.
Design A qualitative study involving face-to-face interviews. Questions were constructed on the Theoretical Domains Framework to systematically explore barriers and facilitators.
Setting One tertiary hospital in Australia.
Participants Twenty orthopaedic surgeons performing TKA.
Outcome measures Beliefs underlying similar interview responses were identified and grouped together as themes describing relevant barriers and facilitators to uptake of decision aids.
Results While prioritising their clinical acumen, surgeons believed a decision aid could enhance communication and patient informed consent. Barriers identified included the perception that one’s patient outcomes were already optimal; a perceived lack of non-operative alternatives for the management of end-stage osteoarthritis, concerns about mandatory cut-offs for patient-centred care and concerns about the medicolegal implications of using a decision aid.
Conclusions Multifaceted implementation interventions are required to ensure that orthopaedic surgeons are ready, willing and able to use a TKA decision aid. Audit/feedback to address current decision-making biases such as overconfidence may enhance readiness to uptake. Policy changes and/or incentives may enhance willingness to uptake. Finally, the design/implementation of effective non-operative treatments may enhance ability to uptake by ensuring that surgeons have the resources they need to carry out decisions.
- qualitative research
- Total Knee Arthroplasty
- decision aids
This is an Open Access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/
Statistics from Altmetric.com
Strengths and limitations of this study
Very few qualitative studies involving orthopaedic surgeons have been published in the literature.
This study took place in one of the largest arthroplasty clinics in Australia. All 20 surgeons performing total knee arthroplasty in this hospital participated in a one-to-one interview.
A theoretical framework was used to systematically explore the barriers and facilitators to uptake of a decision aid by orthopaedic surgeons.
Consistent findings are documented between this single-site study and international surveys of surgeon’ attitudes.
Beliefs and attitudes are distinct from actual behaviour, and therefore the themes elicited in this study do not provide evidence of the actual influences on uptake of a decision aid.
Up to one quarter of total knee arthroplasties (TKA) are performed on inappropriate candidates according to evidence-based criteria1 and a similar proportion experience minimal clinical benefit from surgery.2 The rates of TKA are increasing3; differentiating who will derive a clinically meaningful improvement from TKA from others is a key challenge for orthopaedic surgeons. While the degree of osteoarthritis (OA) severity, pain severity and the impact of pain are key indicators for TKA, surgeons’ consideration of other evidence-based indicators such as psychosocial factors remains varied.4–6 Observations of orthopaedic consultations suggest that other ‘unstated factors’ may also influence clinical judgements such as the surgeons’ beliefs in their own ability to conduct surgery and their ‘instincts’ about the patients’ ability to cope with pain.7
In recent years, there has been a move towards a model of shared decision making in orthopaedics as part of informed consent.8 9 This model implies that surgeons have a duty to inform patients about the benefits and harms of TKA and the likelihood of their occurrence, supporting them to arrive at an informed, shared decision. Evidence suggests that there is often a lack of time during the clinical encounter for patients to consider or discuss all available treatment options and arrive at a decision congruent with their own preferences.10
To help address variations in clinical judgements and promote shared decision making, decision aids may be useful.11 12 Decision aids can be designed to estimate important, patient-specific risks of responding to surgery, based on independent prognostic correlates of post-TKA response such as body mass index, degree of OA severity, preoperative pain, function and mental health.13 Similar aids have been found to predict outcomes in other areas of medicine with superior accuracy to clinical judgements alone.14 15 Decision aids have improved patient knowledge and confidence in decisions16 and have even been found to reduce the rate of surgical procedures.17
Orthopaedic surgeons recognise the need for an aid to support their decision making for TKA and optimise communication with patients.6 18 However, the uptake of decision aids among surgeons has been low.9 A number of factors can influence uptake of a decision aid, and the success of implementation efforts depends on the careful assessment of the barriers to, and facilitators of, uptake.19 The implementation literature advocates the use of theory to ensure the systematic identification of such factors and inform the design of interventions to address them.20 Using theory assists in designing studies that are better able to facilitate behaviour change21 and provides a basis for better understanding the processes underpinning behaviour change.22 To date, studies of decision making among orthopaedic surgeons have been few and atheoretical,6 7 23 and the barriers and facilitators of uptake of decision aids for TKA have not been rigorously explored.9 18
Aim and design
This theoretically informed qualitative study is the first phase of a wider project seeking to design and implement a decision aid into an Australian orthopaedic clinic setting. The aim of this study was to explore the barriers and facilitators to uptake of a TKA decision aid through structured one-to-one interviews with orthopaedic surgeons.
All orthopaedic surgeons and registrars performing TKA at one tertiary teaching hospital in Australia were eligible. Eligible candidates received the study details via email from an institutional administration officer and were invited to contact the researchers to arrange an interview at a time and place suitable to them.
The Theoretical Domains Framework (TDF)24 was selected as a comprehensive and validated framework for determining barriers and facilitators related to the implementation of best practice and clinicians’ behaviour change. The TDF was developed by implementation scientists to synthesise existing behaviour change theories into a single framework. A six-stage consensus approach resulted in the identification of 12 theoretical ‘domains’ describing possible mediators of behaviour change.24 A subsequent validation study22 revised the original TDF to 14 domains: knowledge, skills, social/professional role and identity, beliefs about capabilities, optimism, beliefs about consequences, reinforcement, intentions, goals, memory, attention and decision processes, environmental context and resources, social influences, emotions and behavioural regulation. The revised TDF, used in this study, has explained implementation problems and informed the development of theory-informed behaviour change interventions.25–27
Interview questions were developed for each domain of the TDF with the advice of content experts in the TDF (SF), orthopaedic surgery (MD) and decision-making processes (AS), and in consultation with the literature.28 Interviews were prefaced by stating that the researchers planned to develop a decision aid, thus the discussion was centred around a hypothetical decision aid rather than a defined one. In the first part of the interview, questions aimed to elicit current decision-making processes and biases. In the second part, questions aimed to identify beliefs and attitudes towards decision aids and factors that may influence decisions to use one. Table 1 presents the interview schedule.
The interviewer (SB), a female postdoctoral researcher with methodological expertise in qualitative research, had no previous relationship with the participants and no affiliation with the hospital. Data saturation was considered complete when the beliefs and attitudes of all 20 surgeons working in this setting had been elicited. Face-to-face interviews were conducted with 18 participants in a private office; phone interviews were conducted with two participants. Interviews lasted 20–30 min. Nineteen interviews were audio recorded and transcribed verbatim. One participant did not wish the interview to be recorded, therefore hand-written notes were made during the interview. Participant anonymity was ensured at all times. All transcripts were deidentified prior to data analysis. All participants had the opportunity to review the study findings during a presentation at a scheduled surgical meeting. There was consensus agreement with the researchers’ interpretations and no adjustments were made to the study themes.
Adopting an implementation approach,26 three stages of data analysis were conducted. In the first stage, two researchers (SB and EN) independently coded interview transcripts by classifying each interview response or utterance into one of the 14 TDF domains. For example, this response to the first question in the interview schedule: ‘I think 22 per cent is the high end. But there are a lot of different papers that all suggest 10, 15, 20 per cent', was classified into the ‘Knowledge’ domain of the TDF. Definitions for each domain were derived from the literature19 and adapted to the study context. Pilot coding was performed in which the two researchers independently coded two transcripts. Intercoder comparisons resulted in the refinement of domain definitions (see online supplementary file). This process was conducted three times, until the two researchers were confident that all relevant interview responses could be clearly coded into one domain. The two researchers then independently coded all 20 transcripts. Disagreements were discussed, and consensus was reached in each instance. Coded responses were uploaded into a qualitative data sorting software (Codesort29) to facilitate further analysis.
In the second stage of analysis, one researcher (SB) generated ‘belief statements’ based on the coded interview responses. For example, from the response: ‘I think 22 per cent is the high end. But there are a lot of different papers that all suggest 10, 15, 20 per cent’ classified in the ‘Knowledge’ domain, we generated the belief statement: ‘I am aware of the literature that up to 20 per cent of patients do not have a clinically meaningful improvement from TKA’. Belief statements were worded such that they could describe similar responses from different participants. Belief statements were reviewed by two further researchers (EN and MD), before being interpreted as a likely ‘facilitator’ or ‘barrier’ to surgeon’s uptake of a decision aid. Continuing the example above, the belief statement: ‘I am aware of the literature that up to 20 per cent of patients do not have a clinically meaningful improvement from TKA’ was interpreted as a facilitator to uptake, in that we considered surgeons would be more likely to use a decision aid if they were aware that a substantial proportion of TKA’s resulted in suboptimal outcomes.
In the third stage of analysis, we identified the domains most likely to influence surgeon’s behaviour (ie, using a decision aid or not). This was determined by: (1) frequency of beliefs across transcripts and (2) the perceived strength of beliefs in influencing behaviour. To illustrate, the belief statement: ‘I think that the percentage of my patients who achieve clinically meaningful improvement is higher than that reported in the literature’ appeared in 17/20 transcripts and was considered to have a strong influence on whether a surgeon would use a decision aid or not, implying that behaviour change (ie, using a decision aid) was unlikely if surgeons considered that their patients’ outcomes were already optimised. ‘Knowledge’ was therefore identified as a relevant domain. Where the researchers considered that beliefs within and between domains represented similar barriers/facilitators, these were grouped into themes. We present frequencies of beliefs (see table 2) to provide the reader with a better understanding of the range of interview responses and to assist us in identifying ‘relevant’ domains of the TDF. However, readers should be cognisant that the absence of a belief in a transcript is not the same as a lack of endorsement.
Aggregate data describing the demographics of the sample is provided to protect individual anonymity. The sample comprised of 15 consultant surgeons and 5 registrars. The surgeons’ total experience performing TKA ranged from 6 months to 30 years (mean±SD: 12.9±9.3), and the number of TKAs performed each month ranged from less than 1 to 12 (mean±SD: 5.9±3.0).
Across the 20 interview transcripts, 628 utterances were coded into the 14 domains. Eleven domains of the TDF were identified as relevant: knowledge, behavioural regulation, memory attention and decision processes, beliefs about capabilities, skills, goals, social/professional role and identity, intention, beliefs about consequences, environmental context and reinforcement. Table 2 presents the relevant domains, with specific belief statements supported by example quotes.
The seven themes are described below, illustrated by interview extracts found in table 3 and denoted in the text as ‘Quote’ numbers, for example, Q1.
Knowledge of one’s own patient outcomes
(Relevant TDF domains: Goals, Knowledge, Behavioural regulation and Beliefs about capabilities)
The goal of all participations was to optimise outcomes for their patients (Q1). While almost all participants (n=19) were aware of the literature that up to 20 per cent of patients undergoing TKA have no clinically meaningful improvement from surgery, most believed that this percentage was significantly lower in the patients they operated on (Q2). All participants based this estimation on patient presentations at postsurgical follow-up appointments. However, participants acknowledged the ‘rose-coloured glasses’ they saw their patients through, citing the subtleties of rapport that made patients less likely to report dissatisfaction with surgery (Q3).
It was also recognised that patients with poor post-surgical outcomes may be less likely to attend follow-up appointments, choosing to seek care elsewhere. Participants believed that tracking long-term patient outcomes through an existing joint registry could counteract this (Q4 and Q5).
However, a key problem identified by many participants was how to define clinically meaningful improvement. Seven surgeons emphasised the importance of ‘asking the right question in the right way’ (Q6 and Q7). The discrepancy between surgeons’ perception of clinically meaningful improvement and that of their patients was commonly attributed to mismatched surgeon–patient pre-surgical expectations of TKA. Surgeons reported their expectations of TKA to be a resolution of joint pathology with some corresponding improvement in pain. While disease and pain severity were key considerations in surgical decision making, all participants acknowledged that patient expectations were important (Q8 and Q9).
Reliance on ‘clinical intuition’
(Relevant TDF domains: Memory, attention and decision processes and Skills)
The accurate assessment of patient expectations presented a challenge for many surgeons. While the physical aspects of the clinical assessment were routine skills that all participants believed they had mastered well, many junior and senior surgeons reported difficulties assessing the psychological aspects of the patients’ presentation (Q10).
Only a few participants were aware of any validated tools to assess pre-surgical patient predictive factors. One participant had prior knowledge of a decision aid but had not used it. None of the other participants were aware of any decision aid for TKA. Participants relied on their ‘clinical intuition’ for patients who were less likely to do well. A ‘gut-feeling’ for patients was developed with experience over time (Q11).
The role of aids in supporting clinical decision making
(Relevant TDF domains: Intention and Reinforcement)
Fifteen of the participants believed an aid would support decision making, like a ‘barometer’ or weather app to forecast outcomes, particularly in patients they were unsure about (Q12, Q13). However, most participants (n=13) believed an aid would be insensitive to nuances at an individual level and could therefore not replace their clinical acumen (Q14).
All participants expected to be provided with evidence that a decision aid had been rigorously validated and shown to have high specificity and sensitivity before considering using it. Participants were more likely to trust this evidence if it came from their own institution (Q15). However, half of the participants reported that evidence supporting the validity of a decision aid would not be sufficient to convince them to use it. Instead they would need to see how the tool correlated with their own clinical decision making (Q16).
Implications of a decision aid for patient–surgeon communication and shared decision making
(Relevant TDF domains: Beliefs about consequences and Memory attention and decision processes)
Nine participants perceived that a decision aid would give them an evidence-based approach for saying ‘no’ to patient, particularly for those participants who reported difficulty declining for surgery (Q17). Many participants (n=10) believed an aid could be an important component of patient informed consent, providing patients with objective data on their likely risk outcomes, enabling them to have the appropriate expectations to weigh up the risk and benefits of surgery for themselves. In this way, a decision aid was seen as a valuable support to shared decision making (Q18).
When asked how they would feel about operating if a decision aid predicted a patient had a 50 per cent chance of not responding to TKA, surgeons were divided in their responses. While 11 believed that decisions to proceed would need to be shared with the patient (Q19), eight reported they would not consider surgery unless there was a greater than 80% per cent chance the patient would respond (Q20).
Ethical and legal concerns about decision aids
(Relevant TDF domain: Beliefs about consequences)
Eight participants had medicolegal concerns about documenting a specific risk value in patient records, with a few believing that such information would have to be deliberately withheld from patients in the case it fell into the ‘wrong’ hands (Q21). While some participants believed it would be unethical not to use a decision aid if it had been shown to improve patient outcomes, others were concerned about the ethical implications of a tool if imposed cut-offs were used to deny patients’ surgery (Q22 and Q23).
Available resources and organisational culture as barriers to uptake
(Relevant TDF domains: Environmental context and resources, Beliefs about consequences and Social/professional role and identity)
Almost all participants expressed concerns about making an aid compulsory and imposing mandatory cut-off levels. While many recognised that implementing the tool in this way would have the potential to improve the use of valuable health resources and save costs (Q24), several participants commented that mandatory cut-offs would only be possible if an effective, non-operative alternative was made available for patients that were denied surgery. The existing lack of effective alternative to surgery was seen as a key barrier (Q25).
There was widespread agreement among surgeons (n=17) that the logistics of agreeing on a cut-off value for acceptable risk, and the threat to surgeons’ professional identity as a patient-centred practitioner, were insurmountable barriers to imposing mandatory cut-offs (Q26 and Q27).
Format and content of a decision aid
(Relevant TDF domain: Environmental context and resources)
Finally, physically integrating an aid into clinical practice was not seen as a key barrier to implementation. An electronic or online format was seen as the most likely way an aid could be implemented, particularly given the strict time constraints placed on outpatient clinics (Q28).
Most believed that an aid would be best used within the patient–surgeon consultation (Q29 and Q30), while a couple suggested that an aid could be designed for patients to use on their own or with a support network to save time in the clinical consultation (Q31 and Q32).
Three domains did not appear to have a salient influence on the target behaviour. These were: Optimism, Emotion and Social influences. Optimism was closely connected to the domain reinforcement; that is, surgeons reported being neither optimistic nor pessimistic about the utility of an aid until they had seen evidence of its effectiveness. The Emotion domain did not appear relevant as described by one surgeon: ‘I am not worried about the implications of a tool. TKA is wellness surgery, in the worst case scenario they don’t get a new knee’. The Social influence of patients did not appear relevant as surgeons believed patients were accustomed to filling out questionnaires. While surgeons were more likely to trust an aid if it had been developed by their peers, they preferred to test the tool ‘in their own hands’ and therefore the social influences of other surgeons using an aid appeared limited.
These findings have implications for the future design and implementation of decision aids into surgical clinical practice. It is possible that current decision-making biases may be key barriers to uptake. Surgical decision making involves the consideration of the risks versus benefits of surgery.30 In this study, participants expressed confidence in their ability to successfully remove diseased tissue and correct joint deformity, and the likelihood of some corresponding improvement in pain severity was perceived to be high. At the same time, participants perceived that the risk their patients would not respond to TKA was low. One potential explanation for this is an apparent lack of clarity around the definition of ‘clinically meaningful improvement’ and ‘non-response’. The literature suggests that patients’ perceptions of ‘clinically meaningful improvement’ depend on their satisfaction with improvements in pain and function, closely linked to their expectations of surgery.31 In contrast, the surgeons in this study believed that a resolution in joint pathology and some corresponding improvement pain could be considered ‘meaningful improvement’. This lack of clarity is important, as surgeons who believe that their patient outcomes are already optimised may be less motivated to use a decision aid.
Another explanation may be that surgeons in this sample exhibited an overconfidence bias. Indeed, most participants believed their own patient outcomes were better than that reported in the literature. While it is possible that this belief is accurate and outcomes among this sample of surgeons are indeed above average, similar observations were made in a survey study involving 700 orthopaedic surgeons in Europe and North America.32 In that study, 83 per cent of surgeons surveyed considered themselves to be above-average diagnosticians, 74 per cent believed they were above average surgeons and 25 per cent believed they were in the top 5 per cent of surgeons.32 This overconfidence bias may lead to a confirmation bias in which surgeons only notice things that agree with their point of view and are less attentive to alternative viewpoints due to cognitive dissonance.33 A confirmation bias can have consequences for the uptake of a decision aid as surgeons may be less likely to consider evidence that contradicts their clinical experience.
Surgical decision making also involves weighing up of the risks versus benefits of non-operative interventions.30 The participants highlighted a lack of effective non-operative interventions for end-stage knee OA. A study among Dutch orthopaedic surgeons similarly documented a lack of confidence in the efficacy of non-surgical treatments associated with a decreased referral rate.34 Indeed, there is currently a dearth of evidence-based, non-operative interventions for patients with end-stage knee OA either as an alternative to surgery or as a definitive management when the patient is deemed unsuitable for TKA. In the absence of a treatment alternative, surgery with all its risks and costs is often the default intervention. This is an important barrier to uptake as evidence suggests surgeons will not use an aid if they do not have the resources to carry out the decisions28 or if the aid does not support their own view about effective treatments.35
The participants in this study believed a decision aid could promote shared decision making and enhance informed consent for TKA. Similar attitudes towards decision aids were identified in a previous review9 and a survey of the wider population of orthopaedic surgeons.18 It may be that surgeons can be motivated to uptake a decision aid, because it is the ethical thing to do.36 However, our study identified concerns about the ethical implications of imposing mandatory cut-offs that would determine eligibility for TKA. These concerns may reflect the participants’ desire to defend professional discretion and autonomy. We take the opportunity to emphasise the function of a decision aid is to support, not replace, clinical decision making. However, concerns were also expressed about the legal implications of using a decision aid. This finding contrasts with a study involving US orthopaedic surgeons who believed decision aids may reduce litigation and malpractice insurance premiums by enhancing shared decision making.37 It may be that these concerns are unique to the medicolegal context of Australia. Certainly, surgeons will be less likely to be uptake a tool if they perceive it makes them vulnerable to medicolegal action.
The Ready, Willing and Able model posits that these three preconditions must be satisfied before a decision aid will be adopted.38 ‘Ready’ refers to the perceived benefit of changing the status quo and adopting an aid to support decision making. ‘Willing’ refers to the perceived legitimacy of a decision aid and a willingness to overcome objections and concerns. ‘Able’ refers to the ability to adopt the decision aid given available resources.39 Implications for enhancing the readiness, willingness and ability of Australian orthopaedic surgeons to uptake a TKA decision aid are described below.
To enhance readiness to uptake, current decision-making biases may need to be addressed. Information to counter these biases could be provided through audit and feedback. Audit and feedback is defined as a ‘summary of the clinical performance of healthcare provider(s) over a specified period of time’ (Ivers et al, p5).40 It may be particularly effective when clinicians’ ability to accurately self-assess is limited.41 This assumes that clinicians are motivated to improve care but lack intention to change current behaviour because they are unaware of their suboptimal performance. While effect sizes may depend on the format and context in which feedback is provided, findings from a Cochrane review suggest that audit and feedback can yield important improvements in professional practice.40 Clarity and consensus on the definition of ‘clinically meaningful improvement’ and ‘non-response’ to TKA is likely to influence the success of an audit and feedback intervention in this context. Future research is needed to explore how audit and feedback is best delivered in this context including the immediacy and frequency of feedback, and the potential to incorporate other approaches such as the use of surgeon incentives.42
To enhance willingness to uptake, an audit and feedback intervention before and after using a decision aid may be necessary for orthopaedic surgeons to gain confidence in its legitimacy. The findings suggest that validating a decision aid in large multicentre trials may not be sufficient to encourage uptake, as surgeons believed such trials do not account for variations in case load and years of experience. Surgeons perceived themselves as scientists and ‘artists’, believing that surgical success depended, in part, on the ‘hands’ that performed it. This suggests that orthopaedic surgeons may need to be given the opportunity to validate a decision aid ‘in their own hands’.
Addressing concerns about the medicolegal implications of a decision aid may also be necessary to enhance willingness to uptake. This may require the consultation of multiple stakeholders including policy makers, lawyers and institutional administration before the implementation of a decision aid. Legislative changes in the USA have recognised decision aids as a higher standard of informed consent,8 and in the UK, incentives to use decision aids are being developed.9 There is evidence that mandating the use of a decision aid as a requirement for gaining informed consent results in significantly improved patient outcomes including a reduction in the rate of TKA by 38 per cent in a 6-month period.43 Future studies are required to explore if policy changes and/or incentives might influence willingness to uptake a decisions aid in the Australian context.
To enhance ability to uptake, it appears important to address the dearth of non-operative alternatives to surgery for end-stage knee OA so that surgeons have the resources to carry out decisions. Future research to design and implement effective, evidence-based, non-operative interventions should be a priority. Targeting orthopaedic surgeons’ beliefs about the efficacy of such interventions through education strategies will also be important. The study findings suggest that ability to uptake may be enhanced if a decision aid is packaged in an electronic or online format that is quick and easily accessible. Focus groups involving surgeons and patients to gain feedback on prototypes of a decision aid during the design phase is important to ensure that the format and the content are accessible to a range of end users.
Qualitative studies involving orthopaedic surgeons are rare. The strength of this study is the 100 per cent participation rate by orthopaedic surgeons in one tertiary hospital setting. While the sampling strategy means the generalisability of these findings to other contexts may be limited, we have documented significant similarities with international studies. We acknowledge that while beliefs, attitudes and intentions can predict behaviours with a degree of accuracy, they are distinct from actual behaviour.44 Thus, the themes elicited in this study do not provide evidence of the actual influences on uptake of a decision aid.
Using a theoretical framework to systematically explore barriers and facilitators to uptake, this study will inform the design and implementation of future TKA decision aids.13 The findings suggest that a multifaceted approach will be required to ensure that orthopaedic surgeons are ready, willing and able to use a decision aid that can reduce the suffering and economic burden of ‘failed’ TKA. Research exploring patients’ beliefs and attitudes towards a TKA decision aid is required prior to implementation.
We acknowledge Pascal Buenzli for assistance with the data management software and inter-rater reliability calculation. We extend our thanks to all the surgeons for kindly donating their time to participate in this study.
Contributors SB: led the study design, data collection, data analysis and manuscript preparation. EN contributed to the data analysis. AS, PC and MD contributed to the study design, data analysis and manuscript preparation. SF contributed to the study design and manuscript preparation.
Funding This study was funded by the University of Melbourne FBE/MDHS Collaboration Seed Funding Grant and by the National Health & Medical Research Council (APP1122983). A/Prof Dowsey holds an National Health & Medical Research Council Career Development Fellowship (APP1122526).
Competing interests None declared.
Ethics approval This study was carried out in compliance with the Helsinki Declaration. Approval was given by St Vincent’s Hospital Melbourne, Australia (QA 019 /_16).
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Requests regarding the datasets used in this study can be directed to the first author. Full interview transcripts will not be shared to protect the anonymity of the study participants.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.