Physician Associates in the UK: a failed experiment?
Profs Trish Greenhalgh and Martin McKee discuss the latest evidence on the use of PAs in the UK
Doctors are highly trained and relatively well-paid. In a health service that is short on both cash and skilled professionals, it is understandable that some have wondered whether some less highly trained and less well-paid people could do some of the more routine aspects of their work. Wouldn’t this reduce overall costs and “free up” some of their time to do more complex work?
This is the reasoning behind the introduction of physician associates (PAs) in the UK (and in a handful of other countries, although differences in training and roles mean that they cannot easily be compared).1 A PA is someone with an undergraduate degree, usually but not always in a science of some sort, who has done an additional two years’ training (compared to doctors, whose basic training is 4-6 years, followed by a similar number of years of postgraduate training). PAs are a great idea in theory. But the experiment appears to have largely failed, for reasons we explain below.
The UK’s PA experiment began in 2006, when just 15 US-trained PAs, recruited by open advertisement, were placed in Scottish hospitals and general practices for short periods, mainly in remote under-doctored settings.2 An independent research team conducted interviews with the PAs, staff who had worked with them and patients who had been managed by them. The results of this pilot exercise were described very briefly and in almost exclusively positive terms in an academic paper. Two PAs who had been placed in intermediate care settings, for example, were described as “confident, flexible and autonomous”, “reported capable of working at level of a staff grade doctor” and “perceived to have a positive impact on patient throughput”. Patients were reported as “satisfied” with care delivered by PAs. Two “minor” safety incidents were documented—one “mix-up” with patient notes and one in which a PA advised a patient to change their drug regimen without consulting their GP (the authors did not report which drugs or anything about the context). Nothing was measured quantitatively. Nobody checked the clinical decisions made by the PAs.
Put another way, 15 self-selecting Americans, who had had longer and more intensive training than was (or is) offered in the UK, were well-received in remote areas of Scotland that had struggled to retain doctors. This led to an assumption that all PAs were safe, were clinically effective and cost-effective, and could be deployed in place of doctors in many, if not all, clinical tasks. Several UK universities began to offer two-year Physician Associate courses, and policymakers created incentives for NHS organisations to hire graduates from these courses instead of (or, sometimes, as well as) doctors in training. This combination of push and pull led to a rapid increase in numbers. As the Financial Times reported recently, the NHS employed only 120 PAs in 2015 but this rose to 3200 by 2023, half of them in primary care.
2023 was also the year when the NHS published its Long Term Workforce Plan,3 against a background of worsening crises. Doctors were on strike (always a sign of a health system in trouble)4 or were emigrating; many posts remained unfilled, and hospitals were spending vast sums on locums hired through extorting agencies.5 The plan was ambitious: it promised major reforms designed to grow the workforce, retain staff, and reform working practices. It seemed authoritative, backed up with detailed modelling and extensively referenced. Yet of the 270 references cited, most related to other policy documents or internal reports. Only six were research studies published in peer-reviewed journals and none of these related to PAs. In this context, the proposal to create a workforce of 10,000 PAs by 2036 seemed plucked out of the sky. Other elements of the Long Term Workforce Plan, such as a proposed expansion of medical schools, had not been thought through.6 Some, such as a proposal to create medical apprenticeships, have already been abandoned.
One of the shortcomings of the Long Term Workforce Plan was lack of clarity on what problem PAs were meant to solve.7 They had initially been called physician assistants, suggesting that they were to assist doctors by relieving them of some basic tasks. But in 2013 they were renamed physician associates. A Faculty of Physician Associates was created within the Royal College of Physicians of London. The government, with enthusiastic support from the General Medical Council (but to the alarm of some doctors), took forward legislation to regulate them.8 Proponents of the PA experiment labeled them as “medical professionals”, ignoring the extensive literature on what constitutes a profession,9 and described their training as “following the medical model”, a term that seemed authoritative but was never actually defined.10 Hospitals and general practices facing difficulties in recruiting doctors began to employ PAs as doctor substitutes, sometimes on the same rotas as doctors.
The expansion of PA numbers and the extension of their scope, reflected in their new title, were accompanied by growing concerns. Were PAs really as clinically effective as doctors, even in “low complexity” cases? Were they as cost-effective? Patients were, understandably, confused and many assumed that the person they were seeing was a doctor. Above all, were PAs as safe as policymakers were now assuming? The answer to all these questions is certainly not an unqualified “yes”, and there is growing evidence that it may be a firm “no”.
It's hard to measure whether one health professional is as “clinically effective” or “cost-effective” as another. Let’s say you attend your GP’s surgery for a routine blood pressure check. The healthcare assistant (HCA, someone with minimal training, hired on a low wage to perform very routine tasks) takes your blood pressure very competently. It’s 155/90. They record this on your electronic record. This takes them about 8 minutes, but their hourly rate is cheap. It’s now someone else’s job to become concerned about that blood pressure level and ask you to come in to discuss it.
Alternatively, you might see the PA. They check your blood pressure, explain that 155/90 is just outside the recommended range for someone of your age, and (following a guideline) advise you to start taking your own blood pressure at home and bring the readings in for checking. Or perhaps the PA, knowing that this reading is out of range but unsure of the next step, knocks on the GP’s door and asks their advice. All this might take them 15 minutes (at a higher hourly rate than the HCA, but they’ve achieved a lot more). The GP’s time (90 seconds or so) advising the PA would also need to be factored into the costs.
If you went directly to a GP for your blood pressure check-up, they would follow the same guideline as the PA and give you the same advice, but, on average, they’d do so more quickly (taking perhaps 9 minutes rather than 15). This result would be achieved at a cost marginally more than the cost of the PA (since the PA is both slower and cheaper than the GP), but approximately the same cost if the 90-second input of the advising GP is factored in. However, if your GP was operating at the top of his or her game, they might also notice that you smelt faintly of alcohol even though it was only 11.30 am. In the context of a long-term doctor-patient relationship, they might mention this and ask if things are OK at home. You might then disclose domestic violence—a complex problem that the GP could get started on within the same consultation, thus achieving highly effective and cost-effective care. Less dramatically, your GP might check your blood pressure opportunistically whenever you attended for another reason, which would save you a special trip to either the HCA or the PA.
Unfortunately, the only UK-based study which attempted to compare the cost-effectiveness of PAs and GPs in general practice measured neither the time of the supervising GP (including any delays as the PA and GP found a slot when both were free – with neither hanging around waiting for the other to finish their consultation) nor any of the myriad of subtle things an experienced GP does in a consultation over and above dealing with the problem the patient has made the appointment for.11 That study is often interpreted, erroneously, to mean that PAs can do GPs’ jobs more cheaply than the GPs themselves. In reality, the study showed nothing of the sort.
Here's another example. You twist your ankle while out walking and decide to go and get checked out in the emergency department. You’re in luck – you see the consultant, who quickly and skilfully examines your ankle. She assures you that nothing’s broken and that you don’t need an X-ray. You’re fine with this as she’s clearly an expert. If you’re less lucky, you’ll see a junior doctor or a PA and spend a lot longer in the ED before you depart with reassurance. It may be that the consultant, despite her expensive hourly rate, is the most cost-effective (and safest) of the three. Broadly speaking, the less training someone has had, the longer they take to see patients and the more trigger-happy they are with the prescription pad and the investigation request form. In one study in the emergency department, PAs were 2.7 times as likely to send patients for X-rays as junior doctors, after adjusting for case mix.12
Does this mean that PAs are never effective or cost-effective? Not quite. As we showed in a recent systematic review of the UK evidence published in the British Medical Journal,13 the evidence isn’t that definitive yet. But it’s pointing to the conclusion that if they are effective it is only in very limited circumstances. Furthermore, they are rarely cost-effective when working autonomously, partly because of the double-handling needed to oversee and check their work and partly because of the additional costs of the unnecessary tests they order. The circumstances where they might be effective is when working as part of a team. One study of PAs working on hospital wards, for example, suggested that they were useful and valued when doing “clinically related” tasks (such as finding test results, taking blood or writing discharge summaries) that contributed to the smooth running of the team (one doctor described them as “not the engine, but the oil in the engine”).14 That study set out to measure cost-effectiveness but found that the costs (and hence the value) of PAs could not be meaningfully disaggregated from that of the team as a whole. The question then arises: could these same “oil in the engine” tasks be performed by a cheaper member of staff, such as a clerk, a phlebotomist or a nurse? Inexplicably, such a study, which would involve attending to an entire workforce rather than simply hiring one staff group and trying to measure its impact, has not been done.
Finally, let’s talk about safety. While it may seem intuitive that recruiting additional staff would make things safer, the empirical evidence does not support this. There is now extensive evidence that adding nursing associates to nursing teams, even when keeping the numbers of nurses the same, leads to worse outcomes.15 The reasons are unclear, but one explanation is that tasks are inappropriately delegated to less well-trained staff.
Deaths linked to decisions by PAs seem to be very rare, but several have occurred in the UK. Emily Chesterton, 30, developed calf pain, palpitations and breathlessness (classic symptoms of a life-threatening thrombosis) four weeks after a bout of COVID-19 (which can, rarely, cause thrombosis). A PA working semi-autonomously in a GP surgery diagnosed a muscle sprain and anxiety. After Emily returned a second time in the same week, complaining of worsening symptoms, they recommended a beta-blocker to slow the heart. Emily died soon afterwards of a thrombosis, having not seen a doctor at all during her final illness.
Pamela Marking, 77, developed severe abdominal pain and brought up blood through her nose; she attended the emergency department and was diagnosed with a “nosebleed” by a PA. She died later at home from internal bleeding. In a Prevention of Future Deaths Report, the coroner expressed concern at the PA’s lack of understanding of the significance of abdominal pain in a patient who was losing blood. She noted that both Mrs Markham and her son had assumed that she had been seen by a qualified doctor, and said that “the term ‘Physician Associate’ is misleading to the public”.
Susan Pollitt, 77, was admitted to hospital with a broken arm and found to have some other conditions including fluid on the abdomen. On the advice of a junior doctor, but without consulting the consultant in charge, a PA inserted a tube to drain off the excess fluid. He ordered that the tube be clamped, and it remained so for several hours. Mrs Pollitt died of septicaemia a few days later. The coroner’s report commented that there had been no clinical indication Mrs Pollitt to have the tube inserted in the first place, that the tube should not have remained clamped, and that the prolonged clamping likely increased the risk of infection.
Ben Peters, an apparently healthy 25-year-old, developed sudden onset chest pains, breathlessness and aching arms (classic symptoms of aortic dissection, a life-threatening emergency). He attended the emergency department and was sent for a chest X-ray, ECG and a blood test for troponin (a test for heart attack), which were all reported as normal. He was then seen by a PA who, after discussing his case briefly with a consultant, discharged him with a diagnosis of panic attack. Ben died later that day, having not been assessed by a doctor.
Safety incidents should never be blamed on one person’s individual failings. We should ask “what system issues have contributed to the incident, and how does the system need to change to prevent a similar incident in the future?”. The coroner’s report on Mrs Pollitt’s death, for example, noted the low staffing levels in the hospital at the time of her death.
There are many additional system issues identifiable from these four cases. One is that all might reasonably been classified as “low complexity” at the stage of initial triage. A young woman with (in her own words) a “muscle strain”, someone with a “nosebleed”, someone with a “broken arm” and an apparently fit young man with chest pain but in whom tests have shown no heart attack.
The idea that patients can be straightforwardly and proactively divided into “complex” (to be seen by doctors) and “non-complex” (safe to be seen by PAs) is dangerously incorrect. Rather, some life-threatening illnesses begin in ways that are indistinguishable from simple, self-limiting conditions. An early cancerous breast lump may feel just like a benign lumpy breast. An ectopic pregnancy (which, if untreated, can kill both the foetus and the mother) may first show itself with some mild “spotting” (a common symptom in normal pregnancies). Sepsis (blood poisoning) can start with symptoms that mimic flu. The rash of meningococcal disease looks similar (but not identical) to that of many benign viral illnesses.
Doctors acquire their clinical knowledge in two broad ways. First, from detailed training in the human body, how it works, why it goes wrong and how particular treatments work—in other words, anatomy, physiology, pathology and pharmacology. Second, from “scripts”—the accumulated stories of the hundreds of patients they have seen personally, and also those that they have discussed with other doctors. In their two years of training, PAs can only cover the basic medical sciences extremely superficially, something confirmed by inspection of their curricula. An undergraduate degree in, say, physics, is no substitute. And they also lack exposure to the scripts. With two years exposure focused on patients who have been deemed “low complexity” means that it could take them a lifetime to accumulate sufficient complex cases to hone their clinical instincts sufficiently to practise safely.
All this points to the conclusion that PAs should never be seeing what are known as “undifferentiated” patients—that is, patients whose problems have not yet been sorted by severity or risk. While undifferentiated patients in general practice and in the “minors” queue in the emergency department are, statistically speaking, likely to have mundane and easily diagnosable problems, it is also the case that some patients with as-yet undiagnosed serious and even life-threatening conditions will be in the same queue.
Another system problem, illustrated by Mrs Pollitt’s case, is the danger of “taskification”—uncoupling a medical procedure from the knowledge and professional attitudes needed to perform that task safely and appropriately (and know when not to perform it). Anecdotal examples of unnecessary procedures performed by PAs are not hard to find on a Google search, though in many of these, key medical details are lacking or impossible to verify. In our systematic review, we found no studies of the clinical indications and safety of interventional procedures performed by PAs (mainly because the literature relates largely to PAs working to a narrower scope of practice and under much closer supervision than is now often the case in the UK).
The safety incidents described above also raise the question of what we mean by “supervision”. In each of the above cases, a doctor would have been technically overseeing the PA’s decisions. In reality, the doctor must either repeat the history-taking and examination themselves or take the PA’s word for some or all of the assessment. If the former, the arrangement cannot possibly be cost-effective. If the latter, they become at least partly accountable if anything goes wrong.
In November 2024, Wes Streeting, the Secretary of State for Health and Social Care, commissioned public health doctor Gillian Leng to conduct a review of PAs and the related group anaesthetic associates (AAs). Specifically, she was asked whether their roles are safe and whether they are effective “as members of a multidisciplinary team, across all tasks, roles and settings?”. She has been asked to look comprehensively at the roles and settings in which PAs and AAs work, including internationally, and covering a range of potential activities.
These are very important questions and our review of research in the British Medical Journal sought answers to some of them. As we have noted above, despite searching assiduously and systematically, we have been unable to find evidence that either PAs or AAs are as safe and effective as doctors in the settings in which they are deployed. We found no evidence that they are cost effective and have strong grounds to believe that, given current salary scales, there is no model of care that would enable them to be cost effective undertaking their current roles. We are also aware, as described above, of serious concerns about their safety and, while any things can go wrong with any type of health worker, we cannot ignore the warnings issued by coroners who do not make such reports unless they have evidence of real concern about the wider context.
The Leng Review has been asked to look at international evidence. This is welcome, given how often health policy in the UK ignores lessons from elsewhere. However, this must be undertaken with particular caution here given the evidence that these roles are so very different in the very few countries where they exist in more than tiny numbers.
In this light, it is perhaps regrettable that the Leng Review is not charged with asking a more fundamental question. If the PA role did not already exist, would there be any reason to create it? PAs came about as a new role in the NHS without any clear assessment of need. Had this been done, it is likely that the result would have been very different. Doctors do often need assistance, but not usually with their clinical work. Rather, they seek solutions that reduce the time they spend entering data onto computer systems, arranging tests and referrals, and a myriad of administrative tasks. This is the role that has been created, much less controversially, in France. Instead, the UK now seems to be faced with a solution in search of a problem. A wider remit would have allowed the Leng Review to ask what is needed to make the work of doctors more efficient, cost-effective and fulfilling (and thus encourage more of them to stay in the NHS)?
The current situation is one that satisfies no-one. It is increasingly being seen as a serious policy failure. The question then becomes how to resolve it while doing the least damage. PAs and AAs have already invested considerable personal and financial resources in their career. They have been let down by those in authority, who now have a responsibility to support them through any future transition to other roles. Hopefully, the Leng Review can inform this process.
1. Showstark M, Smith J, Honda T. Understanding the scope of practice of physician associate/physician associate comparable professions using the World Health Organization global competency and outcomes framework for universal health coverage. Hum Resour Health 2023;21(1):50. doi: 10.1186/s12960-023-00828-2 [published Online First: 20230623]
2. Farmer J, Currie M, Hyman J, et al. Evaluation of physician assistants in National Health Service Scotland. Scott Med J 2011;56(3):130-4. doi: 10.1258/smj.2011.011109
3. NHS England. NHS Long Term Workforce Plan. London: NHS England. Accessed 29th February 2024 at https://www.england.nhs.uk/publication/nhs-long-term-workforce-plan/ 2023.
4. Weil LG, Nun GB, McKee M. Recent physician strike in Israel: a health system under stress? Isr J Health Policy Res 2013;2(1):33. doi: 10.1186/2045-4015-2-33 [published Online First: 20130815]
5. Weil LG, Nun GB, McKee M. Recent physician strike in Israel: a health system under stress? Israel journal of health policy research 2013;2:1-11.
6. Geary U, McKee M, Petty-Saphon K. Mind the implementation gap: a systems analysis of the NHS Long Term Workforce Plan to increase the number of doctors trained in the UK raises many questions. Br Med Bull 2024;150(1):1-10. doi: 10.1093/bmb/ldae002
7. Vaughan L, McKee M. Sorting out scope of practice of physician associates in the UK: the clock is ticking. Bmj 2024;387:q2445. doi: 10.1136/bmj.q2445 [published Online First: 20241114]
8. Rimmer A. Physician associates: Doctors raise alarm over legislation to allow GMC regulation. BMJ: British Medical Journal (Online) 2024;384:q156.
9. Freidson E. Theory and the Professions. Ind LJ 1988;64:423.
10. Salisbury H. Helen Salisbury: Training in the medical model. Bmj 2023;383:2793. doi: 10.1136/bmj.p2793 [published Online First: 20231128]
11. Drennan VM, Halter M, Joly L, et al. Physician associates and GPs in primary care: a comparison. British Journal of General Practice 2015;65(634):e344-e50.
12. Halter M, Drennan V, Wang C, et al. Comparing physician associates and foundation year two doctors-in-training undertaking emergency medicine consultations in England: a mixed-methods study of processes and outcomes. BMJ Open 2020;10(9):e037557. doi: 10.1136/bmjopen-2020-037557 [published Online First: 20200901]
13. Greenhalgh T, McKee M. Physician Associates and Anaesthetic Associates in the UK: A rapid systematic review of recent UK-based research to inform the Leng review. BMJ 2025;388:e084613.
14. Drennan VM, Halter M, Wheeler C, et al. The role of physician associates in secondary care: the PA-SCER mixed-methods study. Health services and delivery research 2019;7(19):1-158.
15. Greenley R, McKee M. How will expansion of physician associates affect patient safety? Bmj 2024;386:q1377. doi: 10.1136/bmj.q1377 [published Online First: 20240705]
Really helpful review. Thankyou.
Very interesting read, thanks