Understanding barriers to the introduction of precision medicine in non-small cell lung cancer: a qualitative interview study [version 1; peer review: awaiting peer review]

Background: While treatments targeting genetic mutations and alterations in non-small cell lung cancer (NSCLC) have been available since 2010, the adoption of such examples of precision medicine into clinical practice has historically been slow. This means that patients with NSCLC may not have received life improving and extending treatments which should have been available to them. The purpose of this qualitative interview study was to identify the barriers to the provision of examples of precision medicine for NSCLC. Methods: This study used semi-structured telephone interviews with clinicians, test providers and service commissioners to identify the perceived barriers to providing historical, current, and future examples of precision medicine in NSCLC. Participants were identified through mailing list advertisements and snowball sampling. The qualitative data was analysed using a framework analysis. Results: Interviews were conducted with 11 participants including: five oncologists; three pathologists; two clinical geneticists; and one service commissioner. A total of 17 barriers to the introduction of precision medicine for NSCLC were identified and these were grouped into five themes: the regulation of precision medicine and tests; the commissioning and reimbursement of tests and the testing process; the complexity of the logistics around providing tests; centralisation or localisation of test provision; and opinions about future developments in precision medicine for NSCLC. Conclusions: A number of barriers exist to the introduction of precision medicine in NSCLC. Addressing these barriers may improve access to novel life improving and extending treatments for patients.


Introduction
Advances in the treatment options for people with non-small cell lung cancer (NSCLC) have opened up the possibility of targeting specific mutations in a tumour. The first in class of such medicines to be recommended by the National Institute for Health and Care Excellence (NICE) was gefitinib in 2010 1 . Gefitnib is an EGFR tyrosine kinase inhibitor (TKI). In patients with tumours that showed EGFR mutations, treatment with EGFR TKIs has been shown to extend the length of time a tumour takes to progress to a life-threatening size by two to five months 2,3 . Furthermore, EGFR TKIs may offer greater improvements in quality of life and fewer severe side effects than standard chemotherapy 2,4 .
Other targeted therapies have also been developed and introduced into practice, including crizotinib. For the approximately 5% of patients whose tumours exhibit ALK mutations, crizotinib can be used as a first line treatment 5 . Treatment with crizotinib has been shown to improve patient's progression free survival by over 4 months 6 compared with standard chemotherapy. Treatments targeting a number of other biomarkers including PD-L1 overexpression of ROS1 alterations have also been recommended by NICE and more are currently under evaluation 7,8 These examples of precision medicine for NSCLC offer the potential to improve patients' quality and length of life. There is, however, some evidence that their uptake into practice in NHS England has been slower than anticipated. One factor affecting uptake, was raised during the NICE technology appraisals of the medicines targeting EGFR mutations. The issue raised was that that EGFR mutation testing was not current practice in the NHS before the medicines were available. In the 2010 appraisal of gefitinib, EGFR testing was not widely available in the NHS, although it was predicted that it could be quickly implemented with significant investment 1 . In the 2012 appraisal of erlotinib it was stated during the NICE technology appraisal process that EGFR testing had become best practice. In 2013, the NICE Diagnostic Assessment Programme conducted an evaluation of testing methods for EGFR, which highlighted that there was heterogeneity in how testing was provided 9, 10 . A report published in 2014 by Cancer Research UK estimated that 48% of patients eligible for a targeted treatment for NSCLC were not receiving tests 11 . As a result, 1,429 out of 3,007 patients who could have benefitted from targeted treatments were estimated to be missing out.
There is also evidence of variation in the turnaround time for tests meaning that some patients began treatment before receiving their results 9 . These concerns have been repeated in peer-reviewed published literature and NHS reports 12,13 and a recent survey sponsored by the pharmaceutical company Boehringer-Ingelheim 14 .
When NICE recommends that a medicine should be used in the NHS, as part of the technology appraisal programme process, it becomes a legal requirement that it is made available for all members of the eligible patient group within three-months and paid for by service commissioners 15 . In contrast, when NICE appraises a diagnostic, guidance about its use is produced but there is no legal requirement to fund and provide the test 16 .
As a result there is evidence to support that testing is not immediately available for the entire patient population and complete uptake may require years 11 . Even as testing is made available for all individuals, differences in the quality of testing and the turnaround time may have implications for treatment decisions meaning that the cost-effectiveness of test-treat interventions when used in NHS practice may differ from the cost-effectiveness estimated at the time of appraisal.
Previous research has shown that there have been issues with implementing examples of precision medicine for lung cancer but few have explored why this was the case for these test-treat interventions in particular 17 . One pilot study, conducted to inform a larger qualitative interview study, used face-to-face semi-structured interviews which explored how oncologist's perceptions and work environment affected their use of genomic-targeted medicines in clinical practice in the United States 17 . The published protocol for this study presents the results of a qualitative pilot study 17 . Approximately a third of the ten oncology fellows interviewed in the pilot study were uncertain about guidelines regarding the use of precision medicine as second or third-line treatments for lung cancer while a third of those interviewed were also uncertain regarding how to order testing 17 . Common barriers to performing tests included insufficient tissue samples, the inconvenience of testing and the cost of testing. Facilitators of tests were the ease of testing and deciphering results, as well as patients having health insurance. The cost of treatment was mentioned as a barrier by a smaller number of clinicians. These findings highlight how differences in financing arrangements may impact on the use of precision medicine in oncology.
A number of additional qualitative studies have sought to identify the barriers to precision medicine beyond NSCLC. In a 2013 study based in Canada, which used focus groups to explore the views of physicians about the future role of personalised medicine in health care, eight key relevant concerns about introducing personalised medicine were identified. These eight concerns were: insufficient knowledge; a need for training of physicians; lack of specific guidelines and protocols for using tests; unequal access to testing due to socioeconomic differences; the financial burden of testing on public funds; additional time pressures that precision medicine will put on clinical practice; need for geneticist support after testing 18 . In a qualitative study that used interviews to identify the barriers to introducing a test for the BRCA gene which significantly raises women's risk of breast cancer, clinicians highlighted potential issues with co-ordinating the receiving of test results with the timing of treatment decisions 19 . It was perceived by clinicians that the barriers to introducing BRCA testing could result in "postponement or avoidance of tests, delayed treatment decisions, and proceeding with decisions before test results [were received]".
Despite the number of studies investigating barriers to the uptake of precision medicine in general, there has been a paucity of research focussing on understanding the barriers to implementation of precision medicine for NSCLC into health care systems, in general, and in NHS England, specifically. This study sought to create a typology of the organisational barriers to the introduction of examples of precision medicine in NSCLC. Organisational barriers are those that arise due to the way in which the health care system operates. These organisational barriers are in the control of the health system and so could be removed or reduced by taking actions to improve implementation.

Aims and objectives
This study aimed to understand and define the type and extent of barriers experienced by service providers and service commissioners when introducing precision medicine for NSCLC for relevant individuals within specific patient populations.
This study had four objectives to explore the views of stakeholders in the provision of examples of precision medicine for NSCLC to identify: • the types of perceived organisational barriers to introduce examples of precision medicine for NSCLC in NHS England; • the potential impact for NHS patients of the identified different barriers to the provision of licensed test-treat medicines indicated for the treatment of NSCLC; • how the availability of existing licensed test-treat medicines indicated for the treatment of NSCLC has changed over time; • a typology of barriers which may apply to introducing precision medicine beyond NSCLC.

Methods
This study used semi-structured telephone interviews with service provides (clinicians and test providers) and service commissioners to identify the barriers to introducing precision medicine for NSCLC in NHS England. This study was conducted between March 2018 and October 2018. A protocol for this study was published prior to the commencement of recruitment 20 . This study was approved by The University of Manchester Proportionate Review Research Ethics Committee (Reference number: 2017-1885-3619; 25/08/2017). Participants were provided with an information sheet about the project and were asked to return a signed consent form by post or email prior to taking part in the study.
Telephone-based semi-structured interviews were used to collect qualitative data due to the focus of this work on capturing a geographically diverse sample to represent heterogeneity in health care provision. Telephone interviews offer similar advantages to face-to-face interviews while allowing more flexibility in arranging the timing of the interview.

Sample frame
The sampling frame aimed to identify stakeholders with experience of introducing precision medicine for NSCLC. The relevant stakeholders where identified by examining the NICE care pathway for patients with NSCLC. The relevant stakeholders were drawn from three groups: clinicians; test providers, for example pathologists and geneticists; and service commissioners which may include individuals who are members of care commissioning groups or those involved in commissioning at the national level through NHS England. The principle service providers of interest were oncologists and respiratory physicians specialising in lung cancer but also geneticists and pathologists who are key in providing examples of precision medicine for NSCLC. Examples include EGFR, ALK and PD-L1 testing for medicines such as erlotinib, ceritinib and pembrolizumab.
While factors linked to demand for precision medicine by patients, such as uptake of testing or treatment or adherence to medicines treatment may also impede the implementation of precision medicine, the focus of these interviews was to identify potential barriers focussing on the perspective of the health care system (supply-side capacity constraints). This focus was taken because these capacity constraints in the supply of testing or treatments are within the control of the health system. For this reason, patients were not interviewed in this study.
Clinicians and test providers with over seven years of NHS experience were targeted as such individuals were more likely to have direct experience of the introduction of EGFR and ALK testing and treatment as they were working in clinical practice. Clinicians and test providers were recruited via the British Thoracic Oncology Group (BTOG) [15] and the Royal College of Pathologists (RCPath) [16]. Details about the study and an invitation to participate were circulated via the BTOG mailing list which currently has 2083 members and the RCPath list that has over 11,000 members.
The targeted service commissioner sample comprised hospital, regional and national level individuals involved with service commissioning and funding decisions. Examples of service commissioners may involve members of care commissioning groups, hospital finance staff and decision makers involved with national organisations, such as NICE. As service commissioners were likely to come from a range of organisations, there was no universal sampling frame available to reach them. Service commissioners were therefore recruited using existing links and collaborations within the supervisory team to identify an initial sample. As in recruitment for the clinician sample, geographical diversity was sought through purposive sampling and service commissioners were required to have been in a relevant position when EGFR and ALK mutation based testing and subsequent treatment were introduced.

Sample technique
Purposive sampling was used to gain a diverse sample in terms of the setting and geographical location of testing and treatment 21 . These characteristics were deemed likely to be important in the context of introducing examples of precision medicine as experiences may vary depending on the size and nature of hospitals. For example, mutation testing services may be more readily available in larger teaching hospitals with established links to laboratories. For smaller, general hospitals there may be a greater logistical challenge in sending samples for testing and receiving results in a timely manner.

Sample size
Calculating a target sample size based on defined rules is not relevant in the design of qualitative studies 22 . The target sample size in qualitative studies is informed by the aim of the analysis. In this study, interviews were used to identify the breadth of experiences, thoughts, or opinions on a given subject. This study therefore started with an approximate target sample of ten clinicians or test providers, and ten service commissioners. Inductive thematic saturation was then used that means sampling continued iteratively until no new themes arose from the collected data that were analysed alongside data collection 23 .

Recruitment process
Information regarding the study was sent to clinicians and test providers using mailing lists, with contact details of the principal investigator provided for those interested in taking part. The individuals that expressed an interest in taking part were then subsequently sent more detailed information about the study. Service commissioners were directly sent an email including information about the study and the contact details of the principal investigator. Snowball sampling was used for both samples whereby participants were asked if they knew any other individuals who meet the inclusion criteria who may have been interested in taking part in the study 24 .
Clinicians and service commissioners who were interested in taking part in the study were asked in the mailing list adverts to email or phone a named individual (SW) to express an interest in taking part. The researcher then emailed the potential participant a participant information sheet. After receiving an information sheet, potential participants were given at least 24 hours to consider taking part in the study. If they agreed to take part they were asked to complete a written consent form and to return a copy to the researchers by post or email.

Data collection
Semi-structured interview schedules were created for the two study samples. The interview schedule was piloted with two clinicians before study recruitment began. The interview schedule for service providers (see extended data 25 ), and service commissioners (see extended data 26 ) were, informed by a systematic review of previous economic evaluations of precision medicine (including health technology assessments) 27 and consultation with two expert clinical advisors who are lung oncologists. The core questions for each of the two interview schedules were similar. There were slight variations in the way some of the questions were asked depending on the particular role of the interviewee. For example, clinicians were asked primarily about their experience offering treatments to patients while for geneticists and pathologists the focus was on offering testing.
All telephone interviews were conducted by one researcher (SW) at The University of Manchester and were digitally-recorded. Phone calls and recording took place in an enclosed office. The recording device and memory card containing the interview recordings were stored in a locked draw in a secure university office. The recordings were saved onto an encrypted university computer and the files password protected. The lead researcher (SW) transcribed the first three recordings from the interviews. The remaining recordings were transcribed verbatim by a contracted transcription company called 1 st Class Secretarial Services 28 . Recording were transferred using a secure, encrypted connection. Recorded interviews were deleted from recording devices after they were stored on a computer and anonymised and then destroyed completely at the end of the study. Interview transcripts will be stored for 10 years.

Data analysis
The objectives of data analysis were to: identify potential barriers; and create a typology of barriers that may prevent patients' access to precision treatments for NSCLC. The qualitative data were analysed using framework analysis. 29 .
In the initial data familiarisation stage, one researcher (SW) transcribed and read the first three interviews in order to gain an in-depth understanding of the initial themes emerging from the data. The initial key themes identified during the data familiarisation stage, alongside evidence from previous research formed an initial thematic framework against which the selection of data was sorted and collected 30 . As semi-structured interviews were used for this study, many of the themes originated in the questions contained in the interview schedule. As new themes were identified in the data, they were added to the framework.
Each transcript, facilitated using NVivo software (version 10) 31 , was indexed against these identified themes, with sections from the text which support different themes annotated for later retrieval.
In the context of this study, the identified themes were the range of barriers which occur in providing and accessing examples of precision medicine for NSCLC and views about which barriers were most significant in restricting the provision of precision medicines.

Results
The section reports the results from 11 interviews with participants including; five clinicians, three pathologists, two clinical geneticists, and one service commissioner. All of the clinicians were consultant oncologists. The pathologists comprised two consultant histopathologists and a biomedical scientist. The geneticists were clinical scientists and the service commissioner was a consultant in clinical genetics. Participants were based in a range of hospitals including city-based teaching hospitals and city-based general hospitals serving town and rural communities. The interviews took place between March 2018 and October 2018 and each interview lasted a mean of 23 minutes (range: 10 minutes to 39 minutes). Data saturation was achieved in the clinician sample but it was not possible to recruit sufficient numbers of test providers or service commissioner to achieve data saturation and as such results from this sample can only be viewed as indicative.
A total of 17 barriers to introducing precision medicine for NSCLC were identified. These 17 barriers were grouped into five key broad themes (see Table 1): the managed entry of precision medicine for NSCLC; the commissioning and reimbursement of precision medicine for NSCLC and specifically Funding mechanisms and service commissioning The most commonly discussed barrier across service providers was a lack of clarity in the commissioning process for precision medicine, with a particular focus on the availability of funding for testing. Due to the low availability of biomarker testing before and at the time of NICE approval for a precision medicine, it is common for the manufacturer of the medicine to provide initial ('pump-priming') funding to the NHS to help conduct testing. The process, while loss-leading for the manufacturer, allows clinicians to start using targeted medicines when testing does not exist in the health system. However, many participants highlighted the problems which occurred when companies withdrew funding for EGFR and ALK mutation testing before NHS funding for tests had been established: Geographical inequity in access to testing arising from difficulties in receiving reimbursement for testing was a common theme in the interviews. While some service providers struggled to provide testing due to a lack of available funding, others used different strategies to ensure a continuity of test provision. Some clinicians believed that hospitals with bigger reputations

Tests conducted in different locations
For hospitals which do not conduct testing in-house, the samples may have to be sent to different hospitals for different tests. This when combined with the sequential testing of samples may further add delays to receiving results.

Providing Testing Locally or a Through A Centralised Service
Testing is centralised Centralised testing may mean longer turnaround times for tests as they struggle to meet the demand from multiple hospitals Testing is localised Localised testing may mean that testing is of poorer quality as pathologists do not get to see as many samples so have less opportunity for learning

Barriers to Introducing Future Barriers to Precision Medicine
Less accurate IHC screening The IHC screening stain for ROS1 mutations has a lower specificity than that of ALK and so more tests will have to be sent for more costly and complex FISH testing could provide more funds for testing. In some cases research funding was used to fund testing until NHS funding was made available: For many interviewees, it was not clear who should take responsibility for the cost of testing given who accrues the potential benefits. This confusion in responsibility was because precision medicine involves a complex intervention comprising a test-treat strategy and involves processes using many different specialities including pathology, genomics, and oncology. As such while specialities such as pathology and genomics may be covering the cost of testing, the benefits are realised almost entirely in oncology:

"we are paying huge amounts still for sending tests for PD-L1 testing which is actually not our test. In a way it is a test for oncologists" [TP3, Histopathologist, City based teaching hospital]
If testing helps to prevent patients from receiving an intervention that will not benefit them, as was the case when EGFR testing for gefitinib was introduced, the expense paid for by pathology or genomics may even save money for oncology services. The complexity of the testing pathway and the lack of clarity about test commissioning appeared to create conflict between the stakeholders in the provision of the precision medicine: "historically these sorts of  Potential barriers in the quality and speed of service provided by pathology laboratories may occur due to an increasing pressure on services due to the volume of samples processed and the increased workload involved in processing each sample. For example, pathologists first conduct haematoxylin and eosin stains that help to highlight abnormal cell structures to inform a cancer diagnosis. The pathologist then determine the type of lung cancer present by looking at the cells and potentially conducting additional immunohistochemistry tests. If the cancer is identified as NSCLC, the pathologist will then prepare part of the sample to send for EGFR mutation testing and potentially conduct a number of IHC tests on the remaining parts of the sample for other markers such as ALK or PD-L1.
An additional barrier presented in providing PD-L1 testing is that there is some subjectivity involved in the interpretation and reporting of the results of IHC stains: Despite the increased workload, pathologists suggested that they struggled to find sufficient resources to support their services, therefore creating another barrier to returning test results promptly and allowing treatment to be started quickly. This problem did not seem to occur in genomics laboratories. Providing testing locally or a through a centralised service

"if there's a little bit of tumour in a big resection slice like that might be a resection slice and you might have thousands of tumour cells in that. And then they're asking you to say, well, is it one per cent of the tumour positive or less than one per cent? So if you have a little bit of staining in the tumour but it's less
The size and location of the laboratory conducting ALK IHC or FISH testing was identified as a potential barrier to the timely use of precision medicine. However, there were significant differences in interviewee attitudes to whether testing should be conducted more locally or should be conducted in fewer, centralised laboratories. In addition, opinions on the issue of centralisation or localisation appeared to depend on the nature of the test and role of the service provider. For example, it was generally considered that as EGFR testing required specialist knowledge this test would need to be conducted in larger genomic medicine centres: There was substantially more disagreement about where ALK testing, when compared with EGFR testing, should be conducted. While ALK testing was originally conducted using centralised FISH (fluorescence in situ hybridisation)-based genomic testing, the move towards immunohistochemical screening for the mutation increased calls for the analysis to be moved into local pathology labs. Many participants believed that moving testing "in-house", often in the same hospital that the patient was being treated at, would mean cheaper testing and a shorter turnaround time for tests leading to better outcomes for patients. This was seen to be particularly true of PD-L1 testing, another immunohistochemical-based test, which is currently only conducted by a small number of centralised pathology labs. At the time of the interviews, NGS-based testing was used only in research but not in clinical practice due to its high cost. However, as more mutations become relevant in clinical decision-making, the balance between the cost of NGS, and associated interpretation of the result, and the cost and interpretation of a number of individual tests may become more favourable. NGS-based testing also has the advantage of requiring less tissue from patients and may reduce the number of repeat biopsies that are required 33 . However, as highlighted by one pathologist, NGS-based testing would not be able to guide treatment of PD-L1 targeting therapies as currently only immunohistochemical-based tests are available for PD-L1 overexpression.
Importantly, patients' tumour profiles can change over time, therefore, additional tests may be required regardless of how advanced the original test methods were. Such changes include the addition of resistance mutations such as EGFR T790M 34 and ALK G1269A 35 following treatment with EGFR TKIs and ALK inhibitors respectively. To identify these potential mutations in patients who stop responding to treatment, additional testing is required which in the past has required an additional invasive biopsy to be conducted. Recently circulating tumour DNA (ctDNA) testing has been introduced by some laboratories which allows such mutations to be identified from a blood sample rather than a tumour sample 36 . Some participants suggested benefits to using blood-based tests instead of tumour samples for initial identification of mutations:

Discussion
This qualitative interview study identified 17 barriers that may impede the introduction of examples of precision medicine for NSCLC. These barriers were grouped into five themes: the managed entry of precision medicine for NSCLC; the commissioning and reimbursement of precision medicine for NSCLC and specifically the test component of precision medicine; the complexity of the logistics around providing tests; opinions about whether test provision should be localised or centralised; and opinions about future developments, including potential barriers to their introduction, in precision medicine for NSCLC.
The existence of these barriers may explain the slow adoption of the test-treat interventions into the clinical practice for treating patients with NSCLC. Therefore, these barriers can also be identified as capacity constraints (bottlenecks) in the health care system (supply-side). The development of this typology of barriers, and associated supply-side capacity constraints, to the introduction of examples of precision medicine for NSCLC may aid in predicting the potential challenges that may be faced in introducing future examples of precision medicine in NSCLC, specifically, and other disease areas, generally. By being aware of the barriers and taking action to address them before the introduction of a new test-treat intervention, it may be possible to increase the speed of implementation of precision medicine into clinical practice and ensure that all patients receive appropriate, high quality testing and treatment.
Some of the specific barriers raised in this qualitative study were consistent with barriers mentioned in published reports and papers authored by institutions involved in the provision of examples of precision medicine. For example, a recent report by Cancer Research UK highlighted the strain on pathology laboratories in the NHS 37 . In particular, this report highlighted increasing workloads for cytopathologists who diagnose cancer while the number of new professionals working in those roles was growing at a slower rate. The report suggested that while molecular pathologists who are involved in biomarker tests were currently able to manage their workloads, this was partly due to the current underuse of these tests to guide treatment with precision medicine.
The potential barriers associated with localised testing have been recognised more broadly across pathology testing and there is currently a national strategy to create a national network of pathology laboratory hubs 38 . It has been argued that "consolidating pathology services allows for the most consistent, clinically appropriate turnaround times, ensuring the right test is available at the right time" 39 . Furthermore, it has been suggested that centralising testing could save the NHS £200million by 2020-2021 through economies of scale. The potential to benefit from economies of scale when centralising testing has also been highlighted by Buckell et al. (2015) who suggest potential efficiency savings of 13% or up to £390 million per year 40 .
There were some limitations to this qualitative study. It transpired during interviewing that one of the targeted samples in this study, the service commissioners responsible for financing for tests and targeted treatments were not identifiable as an individual. This specific role is not part of NHS service commissioning processes. One service commissioner was interviewed, but their primary commissioning role was in the provision of germline genetic and genomic tests. The individual only had second-hand knowledge of the commissioning of the companion diagnostic tests required for precision medicine in NSCLC. It became apparent through this participant, and those working in other specialities, that in essence the role of service commissioner for such test-treat diagnostics does not exist and that commissioning arrangements were often laboriously discovered following conflict between hospital managers and test providers.
There were also difficulties in interviewing a sufficient number of participants in the test provider sample. Due to a national policy of restructuring NHS genetics and genomics services during the recruitment phase for this interview, it was difficult to recruit clinical geneticists who make up a sub-section of the test provider sample due to their excessive workload. It was possible that due to the small number of these participants who were interviewed that saturation was not reached in this subsample. It was therefore possible that other barriers to precision medicine in NSCLC exist which were not identified in this study. In addition, it was difficult to identify participants who had experienced the introduction of EGFR mutation testing and treatment with gefitinib. At the time of recruitment it had been over 8 years since this example of precision medicine had been approved by NICE and few participants were working in an NHS role at this time. As such, the coverage of the barriers to the introduction of EGFR mutation and targeted treatment appeared, in this study, to be less extensive than for ALK or PD-L1 testing. This extensive length of time also created a greater risk of recall bias when compared to ALK testing (approved four years prior to the study), and PD-L1 testing (approved two years prior to the study but still being implemented). However, the extent and range of barriers that were identified in this study was sufficient for generating a typology of capacity constraints.

Conclusion
This qualitative study, that used semi-structured telephone interviews, has described and generated a typology for barriers that exist to the introduction of examples of precision medicine in NSCLC into the NHS in England. These barriers may result in patients not receiving access to potential beneficial testing and treatment, healthcare providers offering testing at a financial loss, and poorer quality testing leading to worse outcomes for patients. While some progress in addressing these barriers has been made, some have been faced in the introduction of all example of precision medicine currently available in the NHS.

Data availability
Underlying data The interview transcripts created as part of this study contain a large amount of personal identifying information including names, job titles, workplaces, and references to colleagues and as such have not been provided alongside this manuscript. In addition, this data is not available to reviewers. Consent was provided for the inclusion of anonymised quotes in this manuscript by participants.