Physician Associate/Assistant Impacts on Cancer Diagnosis in Primary Care: A Systematic Review

Background: Symptom recognition and timely referral in primary care are crucial for the early diagnosis of cancer. Physician assistants or associates (PAs) have been introduced in 15 healthcare systems across the world, with numbers increasing to address primary care physician shortages. Little is known about their impact on suspected cancer recognition and referral. This review sought to summarise ndings from international observational studies on PAs’ competence and performance on processes concerned with the quality of recognition and referral of suspected cancer in primary care. Method: A rapid systematic review of international peer-reviewed literature was performed (PROSPERO reference: CRD42019154114). Searches were undertaken on OVID, EMBASE, Web of Science, and CINAHL databases (2009-2019). Studies were eligible if they reported on PA skills, processes and outcomes relevant to suspected cancer recognition and referral. Title and abstract screening was followed by full paper review and data extraction. synthesis of qualitative and quantitative ndings was undertaken on three themes: deployment, competence, and performance. Preliminary ndings were discussed with an expert advisory group to inform interpretation. Results: From 876 references, 15 eligible papers were identied, of which 13 were from the USA. Seven studies reported on general clinical processes in primary care that would support cancer diagnosis, most commonly ordering of diagnostic tests (n=6) and referrals to specialists (n=4). Fewer papers reported on consultation processes, such as examinations or history taking (n=3) Six papers considered PAs’ competence and performance on cancer screening. PAs performed similarly to primary care physicians on rates of diagnostic tests ordered, referrals and patient outcomes (satisfaction, malpractice, emergency visits). No studies reported on the timeliness of cancer diagnosis. Conclusion: This review of peer-reviewed literature combined with advisory group interpretation suggests the introduction of PAs into primary care may maintain the quality of referrals and diagnostic tests needed to support cancer diagnosis. It also highlights the lack of research on several aspects of PAs’ roles, including outcomes of the diagnostic process. them and double screened a proportion ( ≥ 10%). Discrepancies were resolved by discussion between both reviewers and the PI.


Introduction
In health systems such as the UK, most patients with cancer rst present to primary care. 1 The role of primary care is twofold: rst, to conduct investigations in primary care and second, to arrange referrals and tests conducted in secondary care. 2 Research to date has focused on primary care physicians but primary care is changing, with greater input from a range of professionals, such as physician assistants (PAs). [3][4][5] PAs have recently been introduced in the UK as physician associates. Training involves an intense 2-year Masters programme, which covers similar content to medicine (e.g. anatomy, physiology, pharmacology) and provides exposure to clinical environments, including primary care. Masters graduates that pass the national Faculty of Physician Associate exams are then permitted to work as PAs throughout the UK national health service (NHS). Their role is described as complementary to doctors and involves taking medical histories, managing and diagnosing illnesses. 6 (7) PAs in the UK are not currently regulated but following a consultation in 2019 there are plans to introduce it. 7 Responsibilities should also expand when the regulatory framework in England aligns with that of other countries. 8 This will enable them to perform other tasks central to cancer recognition and referral, e.g. ordering x-rays, for which they are trained but not allowed to do under current regulations. Numbers on PAs in England are expected to rise signi cantly following recent health service reforms 4 which include funding for employment of 'additional roles' such as physician associates in newly established primary care networks from 2020/2021. 9 PAs have been working in the USA (as physician assistants) for over 50 years, and 20-30% work in primary care. 10 They have been introduced in several other countries 11 Although PAs' roles vary both within and between health systems, there are opportunities to learn from international experience about the impact of PAs on the quality of cancer referral and recognition in primary care.
It is not clear whether the expansion of PA roles poses a threat or opportunity for the quality of cancer diagnostics in primary care. A systematic review of the international evidence on the role of PAs published in 2013 reported that the quality of evidence was weak with few studies comparing performance with other professionals. 12 This review, however, was undertaken over seven years ago, and since, the volume of studies on PA performance has grown. There have been no systematic reviews examining the impact of PAs on cancer diagnosis.

Page 3/19
Methods This rapid review aimed to summarise ndings from observational studies reporting on the PAs' performance and competence on processes to support recognition and referral of suspected cancer in primary care. Speci cally, it sought to answer the following research questions: 1. Which tasks relevant to cancer diagnosis in primary care conducted by PAs have been examined in studies conducted in high income country settings?
2. To what extent do these studies indicate whether PAs have the knowledge and skills to support cancer diagnosis in primary care?
3. Informed by research question 1, how does PA performance compare with performance of other clinical professionals on processes (for example investigation, history taking, referral) and outcomes of care relevant to cancer diagnosis?
Through discussion with experts in primary care, cancer diagnosis and the physician associate profession, we sought to contextualise the ndings to primary care in England.
The review was registered with PROSPERO (reference: CRD42019154114) and followed PRISMA guidelines. 13 It sought to provide timely ndings in order to inform the rollout of changes in primary care in England. Therefore, streamlined methods were used which sought to maximise robustness but provide timely ndings. These steps included searching only peer reviewed literature in one language; limiting double screening; focusing only on key elements of quality appraisal tools for appraisal; and integrating a knowledge translation approach into the review. 14

Search strategy
Searches for peer-reviewed papers were undertaken on the following databases: OVID, EMBASE, Web of Science, and CINAHL. The search was conducted in November 2019 with database alerts set up to identify relevant literature published since that date. The search strategy was adapted from Skrobanski et al to include terms related to PAs' potential role in the diagnostic process, informed by the NHS description of typical PA tasks in England (i.e. taking medical histories from patients, performing physical examinations, diagnosing illnesses, performing diagnostic procedures, analysing test results) plus additional duties that PAs may be permitted to undertake once regulation is in place such as ordering x-rays. 6 15 (Supplementary data) Searching was limited to publications in English since 2009. While physician associates were rst introduced in England before 2009, this later date was selected because access to diagnostic tools has changed considerably over the last ten years.
Additional eligible studies were sought by hand-searching reference lists of included studies, consultation with experts and screening of citation alerts since searches were carried out.

Study selection
Eligibility criteria Inclusion criteria -piloted by two researchers on a sample of abstracts -are described in Table 1. Table 1 Review eligibility criteria, highlighting changes made following piloting

Population
Physician associates (PAs) or assistants working in primary care in any high-income country.
CHANGE: The search terms initially included a range of terms for primary care settings. When articles were retrieved, several potentially relevant ones were missing. The search was amended to remove primary care terms. Instead a paper's relevance to primary care in the UK was assessed on all retrieved records, where possible at abstract screening or otherwise at full text review.
Intervention/ Exposure: Actions taken by PAs for patients with any symptom(s) that might be cancer CHANGE: Although initially planned, the search was not limited to papers with mention of possible cancer symptoms.
In practice almost any symptom may be a sign of cancer so it was not feasible to turn this criterion into speci c search terms. However, it was possible to exclude papers on abstract screening or full-text review where the symptom or aims of the study was clearly not relevant to cancer diagnosis, e.g. screening for domestic violence, care for multimorbidity.
Comparator: Any other clinical professional CHANGE: Following the advice of an expert advisor we included studies where PA data were aggregated with nurse practitioners but excluded studies were PA data were aggregated with other professionals. This was in recognition of the fact that, in many settings, nurse practitioner and PA roles may overlap.
Outcomes: Quality of symptom recognition and referral where cancer might be suspected This comprised PA skills, con dence, performance, deployment (activities or decisions undertaken to reach a diagnosis, such as history taking, symptom recognition, referral or investigation, triage and cancer screening referral), satisfaction with care, and adherence to guidance/best practice.
Excluded: Chronic disease management, non-cancer screening or primary prevention. -Editorials, letters or narrative reviews -Systematic reviews though these were rst searched for eligible references.

Screening
All identi ed studies underwent title/abstract and full-text screening. For title and abstract screening, a researcher independently reviewed abstracts of all studies against the inclusion criteria described above. All studies identi ed for inclusion underwent full-text screening. To expedite the review process, multiple reviewers split the screening between them and double screened a proportion (≥ 10%). Discrepancies were resolved by discussion between both reviewers and the PI.

Data extraction and quality appraisal
Data extraction and quality appraisal were undertaken concurrently and split across three reviewers with a proportion (10%) dualextracted by a fourth researcher for quality assurance. A data extraction form was rst piloted then used by four reviewers to extract data on during full text screening on the research question/purpose, study design, setting (clinical and geographical), sample size, sample characteristics, outcomes measured, analysis methods, results and authors' conclusions. Appraisal focused on selected measures of methodological quality and relevance. For methodological quality, researchers considered risk of selection bias based on study descriptions of sampling strategy and response rates; and measurement bias from risks of social desirability or unvalidated measures. Studies with a high risk of bias were still described qualitatively but excluded from synthesis of quantitative results. For relevance, studies were appraised in terms of relevance to (a) UK primary care and (b) cancer diagnosis.
Included studies were narratively synthesised into themes guided by the research questions: 1. PA deployment on tasks to support cancer diagnosis reported in the literature 2. PA competence and skills

PA performance
Quantitative and qualitative data were combined in the narrative synthesis. Quantitative ndings for key outcomes were summarised from studies which compared PA with primary care physician performance and risk of bias was not high.

Consultation and knowledge translation
An expert advisory group was convened for the project: comprising physician associates with experience of working in the UK, GPs, medical educators with PA education programmes and researchers in early diagnosis of cancer. Consultation took place with the group, rstly to identify and include any relevant literature that had not emerged from the database search, and secondly to contribute towards the interpretation and policy and research implications of the results. Themes discussed with the group focused on barriers or opportunities to maximising the contribution of PAs in England and to surface other important perceptions of PA performance and competence not identi ed in the literature.

Description of included papers
The search retrieved 873 unique records, plus three through citation alerts, of which 49 papers were included for full paper review. After excluding those not meeting inclusion criteria, 14 studies from 15 papers remained in the dataset, of which thirteen were from the USA ( Fig. 1). Six of the US-based studies were national. All the other studies were state-wide or regional. Ten studies pertained to primary care settings and ve included both primary and secondary care. (Table 2)  where there was little detail about survey construction or instrument validation. 19 20 Records were obtained from national US health care records systems -Veteran Affairs, 21 MediCare fee-for-service claims 22 , National Practitioner Data Bank claims. 23 . In two studies regional or state-wide records data were used from Kaiser Permanente's electronic medical records in Georgia 24 and, patient records held by local, general practice systems in England. 25 There were two qualitative studies, one of which collected data via focus groups, the other used case studies and collected data through interviews and document analysis.

Sample
All studies included consideration of PAs' competence and performance in relation to other primary care professionals, most commonly primary care physicians. The de nition of primary care physicians varied. In the USA, primary care physicians included general practice, family practice, and internal medicine, 22 and general practice, family medicine, internal medicine, and/or obstetrics/ gynaecology. 18 , whilst in England, primary care physicians were general practitioners only. 25 In ve studies, PA and advanced nurse practitioner (NP) data were aggregated in the paper's main analysis. 16 18 22 24 26 In two studies disaggregated data were available in supplementary data or secondary analyses. 16 22 Sample sizes in surveys were between 50 20 27 and 4891 18 with response rates ranging from 7.7% 20 27 to over 80%. 17 In studies using medical records the smallest sample was 2086. 25 In the largest study there were over 800,000 patient records but only 7% of patients received care from PAs. 21 Of the qualitative studies, Feeley et al conducted focus groups with physicians (n = 56), NP/PAs (n = 47), and patients (n = 103) on colorectal screening. 26 Burrows obtained interview data from 24 PAs, and those that worked with them (17 physicians, 2 medical residents, 2 registered nurses, and 1 family health team administrator) but there was no information on the documentary sources they used. 28 Quality and risk of bias Four quantitative studies had a high risk of bias due to low, unreported or biased response rates. In another quantitative study it was not possible to disaggregate NP from PA performance. These studies were excluded from the quantitative summary of study ndings. Further reporting of bias and other quality/relevance concerns limiting the applicability of the study ndings to this review are in the Appendix.
We also consider the variable relevance of studies to the context of UK primary care in 2020 in Box 1 and the discussion.

PA deployment on tasks to support cancer diagnosis
Seven studies reported on general clinical processes in primary care that would support cancer diagnosis. (Table 2) Most commonly, studies reported on ordering of diagnostic tests (n = 6). 16 17 19 22 24 25 Four studies considered referral to other physicians. 16 17 24 28 Two studies reported on physical examinations. 19 17 Only Johnson et al reported on history taking. 19 Two studies explicitly linked care delivery with quality, by examining ordering of guideline-discordant tests or unnecessary referrals 16 or by scoring performance on clinical vignettes. 19 Seven studies reported on tasks related to cancer screening, speci cally breast, (n = 3) 18 20 27 , colorectal (n = 3) 29 , 26 18 ) cervical (n = 2) 30 18 , prostate (n = 1), 21 and endometrial cancer (n = 1). 29 (Table 2) 2. PA competence and skills Four studies across three surveys examined practitioners' knowledge, opinions and understanding of national guideline recommendations on cancer screening. 20 27 30 These studies found that, in common with nurse practitioners, PAs were more likely to recommend breast screening than physicians and both PAs and NPs had knowledge gaps on risk-strati ed screening and referral for genetic counselling in adults at increased risk for colorectal cancer. All these studies, however, had a high risk of bias due to very low response rates.
In Johnson et al's study of performance on history taking, examinations and diagnostic workup, clinicians' responses to two clinical vignettes were scored according to their alignment with national evidence-based and system-speci c recommendations. While no overall difference in PA/NP and physician performance was found, there was a lack of detail on the vignette construction and validation and it was not possible to disaggregate NP and PA performance. 19 Burrows et al's qualitative study in Ontario compared physicians' perceptions of with PAs' XX in family medicine with perceptions from inpatient, outpatient and emergency settings. 28 . It found experienced PAs were often valued as experts, for example: "other consulting services [such as specialist diagnostics] …started to prefer getting consults from the PA because of the PA's understanding of the precise information that the consulting service requires". 28 3. PA performance as measured by processes and patient outcomes Table 3 summarises quantitative ndings in studies with medium or low risk of bias that compared processes and outcomes of care between PAs and primary care physicians. There were no reported differences between PAs and primary care physicians in diagnostic test ordering (3/4 analyses), referrals (3/3 analyses) or screening practices (1/1 analysis). Where differences in care processes were found (1/8 analyses), it was not possible to conclude these differences indicated better or worse quality of care. While PAs had similar rates of guideline-discordant care (PSA screening rates for older veterans with limited life expectancy) to physicians, all practitioners had higher rates than physician trainees. 21 Three studies reported patient outcomes. These comprised re-consultation rates in primary care, 17 25 , satisfaction and malpractice claims, 23 with a brief breakdown of claims due to diagnostic failure or delays in diagnosis, 23 but no studies reported on the timeliness of cancer diagnosis (e.g. stage, survival). There were no reported differences in general patient outcomes (satisfaction, re-consultation rates). While PAs had fewer malpractice payments than physicians, a greater proportion were related to diagnosis. As noted by the authors, it may signal that PAs might be at greater risk of diagnostic error but could also be explained by the presence in the physician group of surgeons and anaesthesiologists -who had malpractice claims related to surgical outcomes. 23 Where it was reported, the pro le of patients seen by PAs differed from that seen by primary care physicians in all but one study.

BOX 1. Applicability of review ndings to current UK context
Preliminary review ndings were presented remotely to the advisory group with comparisons between UK and US contexts ( • Screening lies outside of primary care except for delivery of cervical screening.

Training
• Similar pre-qualifying training: ~2 years intensive core curriculum + national exam. 36 Themes from advisory group discussion to support interpretation of ndings comprised: Implications of 'new' workforce: -PAs in the UK are a young profession. Most PAs entering primary care have little prior clinical experience so will need intensive support. However, freshly graduated, young PAs are often very ready to learn.
-Many new graduates -not just PAs -feel unprepared for General Practice and initially need close clinical supervision. The 'learning curve' in competence may be observed for other professions too. Regulation: -Lack of regulation is recognised as a signi cant barrier to recruiting PAs into primary care. Lack of prescribing rights noted as a particular barrier.
-Some PAs experienced few barriers to their own practice but noted attitudinal barriers from other staff.
-Regulation may in uence standing with doctors & open discussion about prescribing rights.
Other US/UK contextual differences: -The US healthcare system is fragmented between and within states and providers. Variation in PA performance in the UK, therefore may be less variable than in the USA.
-US patients have greater power to 'shop around' than in the UK which affects the clinician-patient power dynamic. Also, patients' perceptions of PAs may be different to GPs which may also result in a different dynamic of PA-patient consultations. This is unexplored in the review.

Summary
This review of peer-reviewed literature combined with advisory group interpretation suggests the introduction of PAs into primary care may maintain the quality of referrals and diagnostic tests needed to support cancer diagnosis. It also highlights the lack of research on several aspects of PAs' roles, including outcomes of the diagnostic process.

Strengths and limitations
This review, the rst of its kind to focus on cancer diagnosis, provides timely insights into the contribution of PAs in an important sphere of activity at a time of rapid expansion of the physician associate profession in England. It also addresses some of the limitations of the last major systematic review in 2013 examining the contribution of PAs to primary care, which reported the quality of evidence was weak with few studies comparing performance with other professionals. 12 There are important limitations, however.
Most (13/15) studies came from USA, which limit the transferability of ndings to other healthcare systems. In particular, in the US the role of primary care professionals in cancer diagnosis may be different; they are not always required for referral to specialists but they are often central in organising cancer screening (a task led by cancer screening hubs in England). Studies undertaken in other countries (Netherlands, Israel, Germany) were identi ed but excluded because PAs were not deployed in primary care settings. However, eligible studies from the UK and Canada -where access to specialist care is normally via a family physician 37 -provided corroborative and complementary insights to those from USA. Moreover, US-based studies have relevance internationally for two other key reasons. Firstly, the drivers for the introduction of PAs have been experienced globally, i.e. shortages in primary care providers amid increasing patient demand, and shifts to multidisciplinary models of primary care teams to provide care. 3 Secondly, they give some indications of how PAs that are regulated and integrated into the healthcare system might perform on processes such as ordering of ionizing radiation that are not currently permitted in the UK.
None of the studies sought speci cally to investigate the impact of PAs on cancer diagnosis. Some excluded cases with 'red ag' symptoms which might exclude cases where cancer was suspected. However, red-ag symptoms are present in only a minority of cancer diagnoses, and UK guidance speci cally recommends investigation of a wide range of symptoms. 35 Five studies presented only aggregated data for NPs and PAs. Numbers of PAs may be smaller than NPs, so there is a risk that ndings are driven by NPs rather than PAs. This aggregation, therefore, may miss important differences in care. Where sub-analyses had disaggregated data, PAs data was often more similar to primary care physicians than NPs. Comparison with existing literature Our principal nding -that in most studies PAs performed similarly to physicians -is largely in line with ndings from other studies. 12 In the UK a suite of studies examining the impact of PAs in primary care at micro, meso and macro levels in 2014 reported PAs were acceptable, effective and e cient in complementing the work of GPs. 32 25 40 At this time, however, there were just 25 PAs working in primary care, with around half trained outside of the UK, which may limit the transferability of this study to a context where most PAs have been trained in the UK and their presence is the norm, not the exception. As others have noted, this nding does not mean that PAs and physicians deliver equivalent care in general. Indeed, in common with other studies, the pro le of patients seen by PAs often differed from those of primary care physicians, and generally seemed to be healthier. 40 The ndings may indicate, however, that there are circumstances in which the additional clinical acumen amongst primary care physicians gained by more training and experience may not be required. 41 In common with the wider literature, this review also highlighted that PAs' deployment varied between (and within) settings. 32 ; 2842 Limits on their role, due to lack of regulation and prescribing rights, is understood as a signi cant barrier in the UK. 43 However, aside from regulation, there are other barriers to delegation. In particular, there is evidence of some resistance and hostility from other health care professionals where there is perceived role overlap or competition for training opportunities. 42 This resistance appears to lessen when there is greater understanding of the role. 43 For PA skills to be utilised appropriately, the whole primary care team need to be clear about and accept the role of PAs in their setting. This role clarity is also required by non-clinicians also to ensure that patients are triaged to the most appropriate clinician. 4445 Role clarity does not mean uniformity; evidence from our review and advisory group stakeholders (Box 1) suggested that the exibility and adaptiveness of the PA role in general can enable PAs to develop in different ways as required by their particular healthcare system.

Conclusions and implications for research, policy and practice
This review suggests that the expansion of PAs working in primary care may maintain the quality of care needed to support cancer diagnosis. This is important, given concerns that PAs might provide poorer quality of care. 16192428 It is also important to guide deployment of PAs in contexts like the UK, where, following regulation, their roles could be expanded to cover tasks like ordering of ionizing radiation. The review also highlights important gaps in the evidence base, particularly how primary care workforce changes may impact on the timeliness of cancer diagnosis. For research to explore the impact of new professions on the timeliness of diagnosis, amendments research and monitoring are needed to collect data on consultations with a range of professionals other than physicians.
Although we discovered no adverse outcomes from the introduction of PAs, it is clear that PAs need to be actively integrated into their working environments. Integration of PAs may require strategies for the whole practice. For example, support for clinical supervisors could enable them to maximise safe delegation to PAs. Support to primary care leaders could promote PAs' integration into wider team, through clarifying respective clinical roles.
The context of primary care has altered signi cantly since the studies in this review were conducted. International guidance on the role of primary care in cancer acknowledges the planned structural shift away from a model of the lone practitioner, but provides no insight into the potential role of PAs. 3 In addition, routes to cancer diagnosis have been affected in unforeseen ways by the Covid pandemic. These include, but are not limited to, patients' reluctance to consult primary care when they experience possible cancer symptoms, a switch to remote instead of face-to-face consultations which may decrease clinicians' capacity to spot subtle symptoms, and a delay in diagnostics referrals due to lockdown backlogs and decreased capacity in order to maintain social distancing. 46 Further studies should examine the impact of emerging professions such as PAs on timely cancer diagnosis in this new context of primary care.

Declarations
Ethics approval and consent to participate: not applicable, ethical approval was not required for this study Consent for publication: not applicable Availability of data and materials: all data are to be found in the publications references in the article Competing interests: The authors declare that they have no competing interests  This is a list of supplementary les associated with this preprint. Click to download. SupplementarydataDecember2020.docx