Browse the corpus

This paper by Cooper and colleagues (BMJ 2025;390:e086358, doi:10.1136/bmj-2025-086358, published 3 July 2025) has been amended to remove the mention of a study from Ireland that was not included in the review.

Impact of physician assistants on quality of care: rapid review. OBJECTIVE: To determine the impact of physician assistants, compared with physicians, on quality of care in the context of an ongoing UK policy review. DESIGN: Rapid systematic review. SEARCH STRATEGY: Keyword search of three databases; search and citation tracking of previous systematic reviews. ELIGIBILITY CRITERIA: Empirical studies that quantitatively compared care delivered by physician assistants with care delivered by physicians, including residents, in economically developed countries, published between January 2005 and January 2025. MAIN OUTCOMES OF INTEREST: Measures of outcomes of care, as defined by the Institute of Medicine's definition of quality: safety, effectiveness, patient centredness, timeliness, efficiency, and equity. METHODS: Eligible studies were categorised as primary care, secondary care, physician assistants versus residents in hospitals, diagnosis/performance, and cost effectiveness. Two reviewers independently extracted data on study design, samples, methods, and findings. Each study was assessed using a risk of bias tool. Owing to the heterogeneity of included studies, a narrative synthesis of the main findings was conducted. An assessment of confidence in the body of evidence for each outcome was based on the number and quality of relevant studies and the consistency of results between similar studies. RESULTS: Of 3636 studies screened, 167 studies were eligible and 40 met the inclusion criteria. These consisted mainly of retrospective observational studies of weak quality. Most (31/40) were from the US, and no data from a post-covid-19 context were found. The greatest number of studies with the most consistent results were those that found that physician assistants practised safely and effectively when working under direct supervision and in post-diagnostic care. No difference was found in patient satisfaction between physician assistants and physicians. Although adding physician assistants to medical teams increases access to care, this may reflect the benefits of increased staffing rather than the unique contribution of the physician assistant role. Evidence on cost effectiveness is limited. Patients in the UK are more likely to see a physician assistant if they live in a socioeconomically deprived area. CONCLUSION: The evidence found in this review is limited and does not support the safety or effectiveness of indirect supervision of physician assistants in undifferentiated (pre-diagnosis) settings. National guidance on the supervision and scope of practice for physician assistants can ensure that physician assistants practise safely and effectively. STUDY REGISTRATION: PROSPERO CRD42024614992.

Healthcare systems worldwide face significant workforce challenges.1 In June 2024 the UK’s National Health Service reported 10 745 medical vacancies, equating to around 7% of all medical posts.2 The number of doctors in primary care, who deal with the vast majority of patient encounters,3 is falling.4 These figures are set against a backdrop of population growth, increasing demand, increasing complexity of clinical work owing to multi-morbidity and an ageing population,5 overcrowding,6 and increasing staff attrition rates, which further compound staff shortages.7 Paradoxically, the UK is also experiencing unemployment of doctors, with general practitioners unable to find work,8 and high competition ratios for entry into postgraduate training programmes, meaning that many residents cannot progress directly to their specialty of choice.9 10

hich further compound staff shortages.7 Paradoxically, the UK is also experiencing unemployment of doctors, with general practitioners unable to find work,8 and high competition ratios for entry into postgraduate training programmes, meaning that many residents cannot progress directly to their specialty of choice.9 10 Physician assistants were first introduced in the 1960s in the US in response to medical shortages in certain specialties and mainly rural areas.11 In the US, becoming a physician assistant requires a bachelor’s degree and evidence of a strong foundation in science, previous experience in healthcare, a two to three year accredited master’s degree in physician assistant studies, successful completion of a national licensing examination, and application for a licence to practice. In many states, physician assistants are required to have a formal agreement with a collaborating physician to be able practise.12 Physician assistants perform similar roles to physicians, assessing and examining patients, requesting investigations, prescribing, and making diagnoses and treatment plans.

ce to practice. In many states, physician assistants are required to have a formal agreement with a collaborating physician to be able practise.12 Physician assistants perform similar roles to physicians, assessing and examining patients, requesting investigations, prescribing, and making diagnoses and treatment plans. The first physician assistants graduated from UK pilot programmes in 2007.13 A separate review charting the drivers of the development of the physician assistant role in the UK, including the change of title from assistant to associate, has been published.14 In the UK, becoming a physician assistant requires a bachelor’s degree in a science or a health related discipline, a diploma or master’s degree in physician assistant studies, and successful completion of a national examination.13 No previous clinical experience is required. In 2010 a voluntary register was created, and in December 2024 the General Medical Council became responsible for the future regulation of physician assistants, including setting standards for course providers.15 However, no nationally defined scope of practice or requirement for supervision exists, meaning that local employers make decisions, leading to wide variation in how physician assistants are deployed and supervised.16 In the UK, physician assistants are more likely to be found working in under-resourced areas.17 They are not able to prescribe or request ionising radiation.

uirement for supervision exists, meaning that local employers make decisions, leading to wide variation in how physician assistants are deployed and supervised.16 In the UK, physician assistants are more likely to be found working in under-resourced areas.17 They are not able to prescribe or request ionising radiation. Degrees in physician assistant studies focus on foundational knowledge and clinical skills necessary to assist physicians in patient care.16 Although the physician assistant role is established in several countries, concerns have been raised about the implementation of this role in the UK in six broad areas: scope of practice, patient safety, informed consent, preferential employment conditions, additional workload of physicians supervising physician assistants, and impact on medical training.14 16 18 19 20 Physician assistants see undifferentiated acute presentations in primary care with indirect supervision,21 they have substituted for doctors in hospital on-call rotas,22 and Coroners have issued “action to prevent future deaths reports” in relation to the physician assistant role.23 24 In response, the Academy of Medical Royal Colleges called for a rapid review of the safety, cost effectiveness, and efficiency of physician assistants,25 and the Secretary of State for Health in England and Wales commissioned an independent review of physician and anaesthesia associates in November 2024.26 We therefore decided to do a rapid review to answer the question: “What is the impact of physician assistants on quality of care compared with physicians?” A separate review will explore the impact of physician assistants on resident training.

Rapid reviews are often used to inform health policy. Like systematic reviews, rapid reviews use systematic and explicit methods to appraise, extract, and analyse data, but specific components of the systematic review process are either abbreviated or omitted to provide an evidence synthesis more quickly.27 Rapid reviews can be enhanced with additional steps included to reduce bias, and they have been found to identify the same studies as systematic reviews.28 We followed the updated guidance for rapid reviews by the Cochrane Rapid Reviews Methods Group,27 which includes features enhancing rapid reviews to reduce bias. Our team included an information specialist and experienced systematic reviewers. We used the Institute of Medicine’s definition of quality in healthcare to assess the impact of physician assistants on quality of care.29 We chose this definition because it includes the domains of safety and effectiveness but also includes patient centredness, timeliness, efficiency, and equity, which are also important in determining quality (box 1). We chose to restrict our review to economically developed countries (in North America and Europe, plus Israel, Japan, South Korea, Australia, and Aotearoa New Zealand)30 to ensure that the evidence synthesis would be relevant to the UK and other similar countries developing a physician assistant workforce. Avoiding harm to patients from the care that is intended to help them

We used the Institute of Medicine’s definition of quality in healthcare to assess the impact of physician assistants on quality of care.29 We chose this definition because it includes the domains of safety and effectiveness but also includes patient centredness, timeliness, efficiency, and equity, which are also important in determining quality (box 1). We chose to restrict our review to economically developed countries (in North America and Europe, plus Israel, Japan, South Korea, Australia, and Aotearoa New Zealand)30 to ensure that the evidence synthesis would be relevant to the UK and other similar countries developing a physician assistant workforce. Avoiding harm to patients from the care that is intended to help them Providing services based on scientific knowledge to all who could benefit and refraining from providing services to those not likely to benefit (avoiding underuse and misuse, respectively) Providing care that is respectful of and responsive to individual patient’s preferences, needs, and values and ensuring that patients’ values guide all clinical decisions Reducing waits and sometimes harmful delays for both those who receive and those who give care Avoiding waste, including waste of equipment, supplies, ideas, and energy; includes cost effectiveness Providing care that does not vary in quality because of personal characteristics such as gender, ethnicity, geographical location, and socioeconomic status

Reducing waits and sometimes harmful delays for both those who receive and those who give care Avoiding waste, including waste of equipment, supplies, ideas, and energy; includes cost effectiveness Providing care that does not vary in quality because of personal characteristics such as gender, ethnicity, geographical location, and socioeconomic status Eligible studies were those that quantitatively compared care delivered by physician assistants with care delivered by doctors (known as physicians in the US), including residents. The outcomes were outcomes of care, as defined by the Institute of Medicine’s definition of quality: safety, effectiveness, patient centredness, timeliness, efficiency, and equity. We excluded studies that examined care delivered by nurse practitioners, advanced clinical practitioners, students, anaesthesia associates, surgical care practitioners, radiology assistants, and pharmacists and studies in which physician assistants could not be distinguished from other groups. We also excluded studies comprising descriptive or self-reported processes of care, self-reported competency, or clinical practice in which no objective outcome was measured; studies of physician assistants being added to teams; and studies focusing on educational processes rather than care delivery.

from other groups. We also excluded studies comprising descriptive or self-reported processes of care, self-reported competency, or clinical practice in which no objective outcome was measured; studies of physician assistants being added to teams; and studies focusing on educational processes rather than care delivery. The search strategy was developed with an information specialist (KF). We did a comprehensive search of two major medical electronic databases, Medline and Embase, to January 2025, using the search terms as set out in the supplementary materials. We did a further search using Google Scholar and the search term “impact of physician assistants” limited to the first 200 results. We also hand searched published systematic reviews for eligible studies. Studies were limited to the English language and the past 20 years. After the removal of duplicates, titles and abstracts were screened by a single reviewer, with 20% independently screened by a second reviewer. Single reviewer screening proceeded owing to a good level of agreement being reached (defined as inter-rater reliability of ≥0.8). Potentially eligible studies were then screened by full text. We excluded studies if they did not meet the inclusion criteria or were not available as full text (for example, conference abstracts).

eviewer screening proceeded owing to a good level of agreement being reached (defined as inter-rater reliability of ≥0.8). Potentially eligible studies were then screened by full text. We excluded studies if they did not meet the inclusion criteria or were not available as full text (for example, conference abstracts). We divided included studies into categories by topic. Data extraction was first done by a single reviewer with expertise in that topic and limited to the data relevant to the review question. We piloted the data extraction and summary forms on a small number of included studies and shared them with team members to ensure consistency (see supplementary materials). A second reviewer then independently verified the data extraction for each study. We used a critical appraisal/risk of bias tool applicable to each study’s design to assess the quality of each included study.31 For most studies, this was the checklist for cohort studies (prospective or retrospective). We also applied this checklist to database studies. We rated a cohort study as “good” if the answer to all applicable questions was yes, “moderate” if the answer to all applicable questions was yes and one was rated as unclear, and “weak” if the answer to any applicable question was no or more than one was rated as unclear.

so applied this checklist to database studies. We rated a cohort study as “good” if the answer to all applicable questions was yes, “moderate” if the answer to all applicable questions was yes and one was rated as unclear, and “weak” if the answer to any applicable question was no or more than one was rated as unclear. We did not do meta-analysis owing to the heterogeneity of the included studies in terms of scope, outcomes, and statistical analysis. Instead, we did a narrative synthesis of the main findings, in which we categorised the studies as primary care, secondary care, physician assistants versus residents in hospitals, diagnosis/performance, patient satisfaction, and cost effectiveness. To explore patterns in the data, we further grouped studies by country of origin. An assessment of confidence in the body of evidence for each outcome was based on the number and quality of relevant studies and consistency of results between similar studies. No patients or members of the public were involved in the research, as no funds or time were allocated for patient and public involvement. However, our results have been submitted as evidence to the Leng Review, which includes patient focus groups.

Eligible studies were those that quantitatively compared care delivered by physician assistants with care delivered by doctors (known as physicians in the US), including residents. The outcomes were outcomes of care, as defined by the Institute of Medicine’s definition of quality: safety, effectiveness, patient centredness, timeliness, efficiency, and equity. We excluded studies that examined care delivered by nurse practitioners, advanced clinical practitioners, students, anaesthesia associates, surgical care practitioners, radiology assistants, and pharmacists and studies in which physician assistants could not be distinguished from other groups. We also excluded studies comprising descriptive or self-reported processes of care, self-reported competency, or clinical practice in which no objective outcome was measured; studies of physician assistants being added to teams; and studies focusing on educational processes rather than care delivery.

The search strategy was developed with an information specialist (KF). We did a comprehensive search of two major medical electronic databases, Medline and Embase, to January 2025, using the search terms as set out in the supplementary materials. We did a further search using Google Scholar and the search term “impact of physician assistants” limited to the first 200 results. We also hand searched published systematic reviews for eligible studies. Studies were limited to the English language and the past 20 years.

After the removal of duplicates, titles and abstracts were screened by a single reviewer, with 20% independently screened by a second reviewer. Single reviewer screening proceeded owing to a good level of agreement being reached (defined as inter-rater reliability of ≥0.8). Potentially eligible studies were then screened by full text. We excluded studies if they did not meet the inclusion criteria or were not available as full text (for example, conference abstracts).

We divided included studies into categories by topic. Data extraction was first done by a single reviewer with expertise in that topic and limited to the data relevant to the review question. We piloted the data extraction and summary forms on a small number of included studies and shared them with team members to ensure consistency (see supplementary materials). A second reviewer then independently verified the data extraction for each study. We used a critical appraisal/risk of bias tool applicable to each study’s design to assess the quality of each included study.31 For most studies, this was the checklist for cohort studies (prospective or retrospective). We also applied this checklist to database studies. We rated a cohort study as “good” if the answer to all applicable questions was yes, “moderate” if the answer to all applicable questions was yes and one was rated as unclear, and “weak” if the answer to any applicable question was no or more than one was rated as unclear.

We did not do meta-analysis owing to the heterogeneity of the included studies in terms of scope, outcomes, and statistical analysis. Instead, we did a narrative synthesis of the main findings, in which we categorised the studies as primary care, secondary care, physician assistants versus residents in hospitals, diagnosis/performance, patient satisfaction, and cost effectiveness. To explore patterns in the data, we further grouped studies by country of origin. An assessment of confidence in the body of evidence for each outcome was based on the number and quality of relevant studies and consistency of results between similar studies.

No patients or members of the public were involved in the research, as no funds or time were allocated for patient and public involvement. However, our results have been submitted as evidence to the Leng Review, which includes patient focus groups.

The initial search identified 3639 records. After the removal of duplicates, initial screening by title and abstract resulted in 167 records. Full text screening resulted in 40 studies being included in the review. Figure 1 outlines the search and selection process. Studies were excluded at the full text stage owing to their results not distinguishing physician assistants from other groups (for example, nurse practitioners) (n=61), not meeting the inclusion criteria (for example, not a quantitative empirical study comparing physician assistant care with physician care) (n=62), or data being available only in abstract form (n=4). PRISMA diagram. *Did not meet inclusion criteria: not comparing physician assistant (PA) care with physician care (n=6); comparing different PA staffing models (n=1); reporting impact of PAs being added to medical teams (n=23); describing processes of care (n=18); self-reported competence/outcomes (n=6); systematic review (n=5). NP=nurse practitioner

clusion criteria: not comparing physician assistant (PA) care with physician care (n=6); comparing different PA staffing models (n=1); reporting impact of PAs being added to medical teams (n=23); describing processes of care (n=18); self-reported competence/outcomes (n=6); systematic review (n=5). NP=nurse practitioner The included studies were in the following categories: primary care (6); secondary care (5); physician assistants versus residents in hospitals (14); diagnosis/performance (8); patient satisfaction (3); cost effectiveness (4). The categories were not discrete as several studies had overlapping themes. Of the 40 studies, 32 were conducted in the US, four in the Netherlands, and four in the UK. The tables summarise the included studies, including details of the setting, study design, data analysed, outcome measures, results (including statistical analysis, where present), and quality assessment. The following narrative presents an overview of key findings.

the US, four in the Netherlands, and four in the UK. The tables summarise the included studies, including details of the setting, study design, data analysed, outcome measures, results (including statistical analysis, where present), and quality assessment. The following narrative presents an overview of key findings. Five studies were conducted in the US (table 1).32 33 34 35 36 All were retrospective cohort studies involving hundreds of thousands of patient records. Of these, one study investigated adoption of new chronic disease medications and found that physicians were significantly more likely to prescribe newly approved drugs for chronic diseases than were physician assistants and nurse practitioners.32 Three studies looked at diabetes care and used glycated haemoglobin (HbA1c), systolic blood pressure, and low density lipoprotein cholesterol as surrogates for quality of care.33 34 35 These found no clinically significant differences in outcomes between physician assistants, nurse practitioners, and physicians as the primary care provider. The fifth study compared the practice patterns and quality of care of physician assistants, nurse practitioners, and physicians in community health centres.36 It found no differences in the outcomes studied, apart from that patients were more likely to have smoking cessation and education/counselling services documented if seen by a physician assistant or a nurse practitioner than by a physician. We judged all of these studies to be of weak quality owing to their non-randomised, retrospective design that did not account for potential confounders, such as differences in patients’ characteristics, team care, and visits to other providers.

seen by a physician assistant or a nurse practitioner than by a physician. We judged all of these studies to be of weak quality owing to their non-randomised, retrospective design that did not account for potential confounders, such as differences in patients’ characteristics, team care, and visits to other providers. Summary of results—primary care CI=confidence interval; FDA=Food and Drug Administration; GP=general practitioner; LDL=low density lipoprotein; NP=nurse practitioner; OR=odds ratio; PA=physician assistant; SBP=systolic blood pressure; SD=standard deviation; VHA=Veterans’ Health Administration.

seen by a physician assistant or a nurse practitioner than by a physician. We judged all of these studies to be of weak quality owing to their non-randomised, retrospective design that did not account for potential confounders, such as differences in patients’ characteristics, team care, and visits to other providers. Summary of results—primary care CI=confidence interval; FDA=Food and Drug Administration; GP=general practitioner; LDL=low density lipoprotein; NP=nurse practitioner; OR=odds ratio; PA=physician assistant; SBP=systolic blood pressure; SD=standard deviation; VHA=Veterans’ Health Administration. One study was conducted in the UK and published in 2015 (table 1).37 This was a retrospective study based on a review of consultation notes, linked medical records, and patient satisfaction. It analysed the records of 2086 adults attending for same day/urgent appointments being seen by either a physician assistant or a general practitioner in 12 volunteer practices, of which six employed physician assistants and six did not. The primary outcome of this study was re-attendance within 14 days for the same or a linked problem. Secondary outcomes were related to processes of care: use of diagnostics, referrals, prescriptions, advice given, length of consultation, and cost of consultation. All but one of the practices had guidelines for receptionists to assign patients to physician assistants, defined by the supervising general practitioner. In five practices, physician assistants had longer appointment times or empty slots to give time to consult their supervising general practitioner. Physician assistants and general practitioners saw significantly different patients. Physician assistants’ patients were younger, were more likely to present with minor problems or symptoms, and had fewer chronic diseases, fewer repeat prescriptions, and fewer visits to the practice in the previous three months. No significant difference was found in the primary outcome. After adjustment for clustering at practice level, presenting problem, and patients’ characteristics, no significant differences were found in the secondary outcomes. General practitioners saw three patients for every two seen by physician assistants. Costs per consultation were lower for physician assistants (£28.14 ($38.6; €32.9) versus £34.36 for general practitioners), but lack of data on time spent supervising and signing prescriptions by general practitioners means that the true costs of physician assistants are underestimated to an unknown extent. In a review of the medical records of re-attendances, physician assistants performed significantly better than general practitioners in documenting the consultation. Patient satisfaction was high for both groups.

al practitioners means that the true costs of physician assistants are underestimated to an unknown extent. In a review of the medical records of re-attendances, physician assistants performed significantly better than general practitioners in documenting the consultation. Patient satisfaction was high for both groups. We judged this study to be of weak quality owing to its non-randomised, retrospective design, with general practitioner supervision not accounted for and only seven physician assistants studied.

al practitioners means that the true costs of physician assistants are underestimated to an unknown extent. In a review of the medical records of re-attendances, physician assistants performed significantly better than general practitioners in documenting the consultation. Patient satisfaction was high for both groups. We judged this study to be of weak quality owing to its non-randomised, retrospective design, with general practitioner supervision not accounted for and only seven physician assistants studied. Three retrospective cohort studies looking at care in emergency departments were conducted in the US (table 2).38 39 40 These all analysed outcomes such as unplanned re-attendance within 72 hours, patient acuity/diagnosis, and patient flow measures. The first study used data over 12 years from the National Hospital Ambulatory Medical Care Survey.38 This found that physician teams saw older, more severely ill patients and more often saw them overnight. Patients seen by physician assistant or nurse practitioner teams received fewer diagnostics, fewer procedures, and fewer hospital admissions after adjustment for age, severity, and other characteristics of patients. The second study analysed 25 883 patient encounters in a single emergency department.39 Physicians saw older patients, but physicians and physician assistants saw patients with similar acuity scores. Patients seen by physician assistants waited longer to be seen and had a longer lengths of stay. No differences were found in unplanned re-attendances, patient disposition, and computed tomography scans requested. Patient satisfaction was higher for physician assistant care. The third study looked at 10 369 children aged 6 years or younger seen during a 24 month period in a single emergency department and analysed re-attendance rates.40 Physicians saw younger, sicker patients, and no statistically significant difference was found in re-attendance rates by provider. We judged all these studies to be of weak quality because of their non-randomised retrospective design, the confounder of team based care in the emergency department, and no data on whether diagnostics or procedures were appropriate. The second and third study did not adjust for differences in patients’ characteristics. In addition, accepted metrics of efficiency in the emergency department, such as length of stay, have not shown a causal relation with improved patient care.41

ment, and no data on whether diagnostics or procedures were appropriate. The second and third study did not adjust for differences in patients’ characteristics. In addition, accepted metrics of efficiency in the emergency department, such as length of stay, have not shown a causal relation with improved patient care.41 Summary of results—secondary care CI=confidence interval; CT=computed tomography; ED=emergency department; FY=foundation year; LOS=length of stay; NP=nurse practitioner; OR=odds ratio; PA=physician assistant; SD=standard deviation.

ment, and no data on whether diagnostics or procedures were appropriate. The second and third study did not adjust for differences in patients’ characteristics. In addition, accepted metrics of efficiency in the emergency department, such as length of stay, have not shown a causal relation with improved patient care.41 Summary of results—secondary care CI=confidence interval; CT=computed tomography; ED=emergency department; FY=foundation year; LOS=length of stay; NP=nurse practitioner; OR=odds ratio; PA=physician assistant; SD=standard deviation. Two studies were conducted in the UK (table 2).42 43 The first was published in 2020 and compared physician assistants and foundation year 2 doctors in training undertaking consultations in three type 1 emergency departments (24 hour, consultant led emergency departments that treat serious and life threatening conditions) in England.42 It used retrospective chart reviews, a small number of semi-structured interviews with staff and patients, and observations of physician assistants over a 16 week period. The study involved 8816 patient attendances seen by six physician assistants and 40 foundation year 2 doctors. The primary outcome was unplanned re-attendance within seven days as a surrogate of quality of care. Secondary outcomes were length of stay, use of radiography, prescriptions, and referrals. Forty patient records were reviewed for clinical adequacy. The two groups saw significantly different patients. Physician assistants saw younger patients, a higher proportion of patients in minors (less unwell, ambulatory patients), and a lower proportion of patients triaged as urgent/very urgent, and they did not see any patients overnight. Calculation of the primary outcome was limited by a large amount of missing data. This was due to one site where data were not captured on an electronic dataset and were retrieved manually from a random sample of 205 records and then extrapolated for the purposes of the study. No statistically significant difference was found in re-attendance rates after adjustment for patients’ age, sex, and triage score. Additionally, no statistically significant difference was found in prescriptions, admission rates, or discharge summaries completed, but patients seen by a physician assistant were more likely to have radiography and to have a shorter length of stay in the department after adjustment for age, sex, acuity, whether admitted, radiograph taken, and hospital site.

t difference was found in prescriptions, admission rates, or discharge summaries completed, but patients seen by a physician assistant were more likely to have radiography and to have a shorter length of stay in the department after adjustment for age, sex, acuity, whether admitted, radiograph taken, and hospital site. The authors note that they did not take account of differences in staffing levels (for example, day versus night; therapy teams are not available overnight to help to discharge older patients) and no adjustments were made for differences in area of emergency department seen. Of the records that were reviewed for clinical adequacy, three (one physician assistant and two foundation year 2 doctors) were judged to contain errors or omissions that would have altered the patient’s treatment. Of the six patients interviewed who had seen a physician assistant, two thought that they had seen a doctor. We judged this study to be of weak quality owing to its non-randomised retrospective design, the confounder of team based care (including supervision), and no adjustments made for differences in area seen and time of day.

patients interviewed who had seen a physician assistant, two thought that they had seen a doctor. We judged this study to be of weak quality owing to its non-randomised retrospective design, the confounder of team based care (including supervision), and no adjustments made for differences in area seen and time of day. The second UK study was published in 2024.43 This was a retrospective cohort study looking at the outcomes of consultations of physician assistants and foundation year 1 doctors in training in one type 1 emergency department in England. It analysed 7405 records of adult patients who attended between August 2018 and January 2020. The primary outcome was wait time to consultation. Secondary outcomes were length of stay, rates of being left without being seen, and unplanned re-attendances within 72 hours. Physician assistants and foundation year 1 doctors saw significantly different patients. Foundation year 1 doctors mainly worked during normal working hours (weekdays 0800-1600) and mainly saw patients in the urgent treatment centre with a different case mix. Physician assistants mainly worked in majors and the resuscitation room and also worked out of hours. After adjustment for patient’s age, time of day, area of emergency departments seen, and patient’s disposal, no statistically significant difference was found in wait time to consultation, left without being seen rates, or unplanned re-attendances, but patients were in the department for a significantly longer time if seen by a physician assistant. Patients seen by physician assistants were also more likely to be admitted, likely reflecting the different areas they were working in. We judged this study to be of weak quality owing to its non-randomised retrospective design and the confounder of team based care (including supervision) in the emergency department.

. Patients seen by physician assistants were also more likely to be admitted, likely reflecting the different areas they were working in. We judged this study to be of weak quality owing to its non-randomised retrospective design and the confounder of team based care (including supervision) in the emergency department. Fourteen studies compared physician assistants versus residents in hospitals (table 3).44 45 46 47 48 49 50 51 52 53 54 55 56 57 All but two studies were conducted in the US. Summary of results—residents versus physician assistants (PAs) in hospitals CI=confidence interval; CPR=cardiopulmonary resuscitation; FY=foundation year; ICU=intensive care unit; LOS=length of stay; NSAID=non-steroidal anti-inflammatory drug; OR=odds ratio; PA=physician assistant; PGY=postgraduate year; PROM=patient reported outcome measure; SD=standard deviation.

Summary of results—residents versus physician assistants (PAs) in hospitals CI=confidence interval; CPR=cardiopulmonary resuscitation; FY=foundation year; ICU=intensive care unit; LOS=length of stay; NSAID=non-steroidal anti-inflammatory drug; OR=odds ratio; PA=physician assistant; PGY=postgraduate year; PROM=patient reported outcome measure; SD=standard deviation. Seven US studies looked at the impact of using physician assistants compared with residents/fellows as first assistants in different types of low risk surgery (table 3).44 45 46 47 48 49 50 The first found no difference in outcomes, but different types of assistants (physician assistants, residents, and fellows) assisted to varying degrees according to their level of ability.44 The second compared the same experienced physician assistant with rotating orthopaedic fellows and found no difference in patients’ outcomes but longer surgical times for fellows, which reduced over time until no significant difference was seen by the fourth quarter.45 The third study additionally investigated the impact on hospital costs of using physician assistants versus residents as assistants.46 No difference in patients’ outcomes was found, but longer length of surgery involving residents incurred a “hidden” additional cost, although this was neutralised by the fact that physician assistants were more expensive to employ than residents. The fourth study found no difference in outcomes, although significantly more comorbidities were present in patients in the resident/fellow group that were not adjusted for.47 The fifth compared one highly trained physician assistant versus fellows as lead surgeon working with residents in harvesting lungs for transplant and measured surgical injuries to the donor lung.48 Significantly fewer injuries occurred in lungs harvested by the physician assistant compared with fellows, as well as significantly lower rates of pulmonary graft dysfunction grades 2 and 3, and no statistically significant difference was seen in 30 day and one year graft survival rates.

l injuries to the donor lung.48 Significantly fewer injuries occurred in lungs harvested by the physician assistant compared with fellows, as well as significantly lower rates of pulmonary graft dysfunction grades 2 and 3, and no statistically significant difference was seen in 30 day and one year graft survival rates. The sixth study found no difference in complication rates, but operative times were 10 minutes longer on average when a resident participated compared with a physician assistant, surgical fellow, or no assistant.49 The final study found that the presence of a physician assistant, resident, or fellow had no association with length of surgery, complications, and reoperation rates.50 We judged all these studies to be of weak quality owing to their non-randomised retrospective design that did not account for learning curves and other various different confounders, listed in table 3.

cian assistant, resident, or fellow had no association with length of surgery, complications, and reoperation rates.50 We judged all these studies to be of weak quality owing to their non-randomised retrospective design that did not account for learning curves and other various different confounders, listed in table 3. Five US studies looked at the impact of using physician assistants compared with residents/fellows in general medicine inpatient services (table 3).51 52 53 54 55 The first was a retrospective cohort study analysing 9681 hospital records of patients admitted on weekdays between January 2005 and December 2006.51 It compared hospitalist (a board certified physician specialising in acute inpatient care)-physician assistant teams and attending-resident teams. Inpatient care provided by hospitalist-physician assistant teams was associated with a statistically significantly longer length of stay; however, when adjusted for times when both types of teams could receive admissions, this difference was no longer significant. After adjustment for confounding variables, no statistically significant difference was seen in hospital charges, readmission rates at seven, 14, and 30 days, and inpatient mortality. The second was a retrospective cohort study looking at the care of 95 patients admitted for re-induction chemotherapy for acute myeloid leukaemia between 2008 and 2012.52 Patients were cared for by either physician assistants or house officers, both groups supervised by attending physicians. No statistically significant difference was seen in characteristics of patients admitted to either service. A statistically significantly shorter length of stay, fewer readmissions, and fewer consults per patient were seen for those in the physician assistant service, with no difference in mortality or transfers to intensive care. The third was a retrospective cohort study investigating the care of 5194 consecutive patients admitted to a general medicine service between July 2005 and June 2006.53 It compared hospitalist-physician assistant teams and attending-resident teams. Patients admitted to the hospitalist-physician assistant service were significantly younger, had lower comorbidity scores, and were more likely to be admitted at night. After adjustment for confounders such as age, race, comorbidities, and time of admission, no statistically significant difference was seen in length of stay, inpatient mortality, transfers to intensive care, readmissions, or patient satisfaction.

d lower comorbidity scores, and were more likely to be admitted at night. After adjustment for confounders such as age, race, comorbidities, and time of admission, no statistically significant difference was seen in length of stay, inpatient mortality, transfers to intensive care, readmissions, or patient satisfaction. The fourth was a retrospective cohort study investigating the care of 5346 patients admitted to two different medical intensive care units between January 2004 and January 2007, one run by physician assistants during the day with a fellow and attending physician and the other run by residents with a fellow and an attending physician.54 After adjustment for severity of illness, the physician assistant group had a longer length of stay in intensive care. No statistically significant difference was seen in hospital length of stay, intensive care unit mortality, hospital mortality, or readmission rates. A subsequent matched analysis using a subset of 1249 pairs of patients found no difference in any of the outcomes. The fifth was a before-and-after study between 1996 and 2000 of residents versus physician assistants working with supervising physicians in a general medicine inpatient service.55 Whereas the resident service had access to direct supervision during normal working hours and indirect supervision out of hours, the physician assistant service had access to direct supervision from a hospitalist 24 hours a day, seven days a week. A statistically significant reduction in mortality was seen in the physician assistant service compared with the resident service. No difference was seen in adverse events, readmission rates, and patient satisfaction.

sistant service had access to direct supervision from a hospitalist 24 hours a day, seven days a week. A statistically significant reduction in mortality was seen in the physician assistant service compared with the resident service. No difference was seen in adverse events, readmission rates, and patient satisfaction. All these studies were judged to be of weak quality owing to their non-randomised retrospective design that did not account for various different confounders, listed in table 3. For example, in the first general medicine study,51 when the multivariable analyses were restricted to the subset of hospitalists who served as attending physicians on both teams, the increase in length of stay associated with hospitalist-physician assistant teams was no longer significant, suggesting that it is the supervisor rather than the physician assistant or resident who most influences patient management. In the fourth study,54 residents covered the physician assistant intensive care unit overnight; and in the fifth study,55 the authors conjectured that onsite availability 24/7 of a hospitalist could have contributed to the reduction in mortality, which is in keeping with other studies showing that hospitalists improve quality of patient care.

esidents covered the physician assistant intensive care unit overnight; and in the fifth study,55 the authors conjectured that onsite availability 24/7 of a hospitalist could have contributed to the reduction in mortality, which is in keeping with other studies showing that hospitalists improve quality of patient care. Two multicentre studies were conducted in the Netherlands (table 3). The first analysed 2307 records of adult inpatients on participating medical and surgical wards to look for non-adherence to prescribing guidelines.56 Wards staffed by physician assistants at least 51% of the time during normal working hours were compared with wards staffed by residents, both groups being supervised by physicians. No statistically significant differences in outcomes were found between the two groups. The second study used the same dataset to investigate quality of care.57 The primary outcome was length of stay, and secondary outcomes consisted of 11 quality indicators (for example, in-hospital mortality, unplanned transfers to intensive care, cardiopulmonary resuscitation, pressure ulcer development after admission) and patient experience. No statistically significant differences were found in length of stay or any of the quality indicators between the two groups. Patients’ experiences of care were all rated statistically significantly higher on the wards that involved physician assistants. We judged these studies to be of weak quality owing to their non-randomised retrospective design and various confounders not accounted for (see table 3).

ty indicators between the two groups. Patients’ experiences of care were all rated statistically significantly higher on the wards that involved physician assistants. We judged these studies to be of weak quality owing to their non-randomised retrospective design and various confounders not accounted for (see table 3). Eight studies compared the diagnosis or management of physician assistants versus physicians (table 4).58 59 60 61 63 64 65 66 Seven of these were published in the US. Summary of results—diagnosis/performance CI=confidence interval; GP=general practitioner; NNB=number needed to biopsy; NP=nurse practitioner; OR=odds ratio; PA=physician assistant.

Eight studies compared the diagnosis or management of physician assistants versus physicians (table 4).58 59 60 61 63 64 65 66 Seven of these were published in the US. Summary of results—diagnosis/performance CI=confidence interval; GP=general practitioner; NNB=number needed to biopsy; NP=nurse practitioner; OR=odds ratio; PA=physician assistant. Four studies investigated prescribing practices (table 4).58 59 60 61 The first used national databases to analyse the prescribing quality of physician assistants, nurse practitioners, and physicians in outpatients/ambulatory care between 2006 and 2012.58 Overall mean performance across all indicators was 58.7%. Statistically significant differences were found in three of 13 quality indicators (see table 4). The second study used a large national database to investigate differences in opioid prescribing between physician assistants, nurse practitioners, and generalist physicians between 2013 and 2016.59 It found that physician assistants and nurse practitioners made up a disproportionately high number of the prescribers after adjustment for potential confounders, such as practice setting. The third study investigated opioid overprescribing in primary care and found that most nurse practitioners and physician assistants prescribed opioids in a pattern similar to that of physicians, but the nurse practitioner and physician assistant groups had more outliers who prescribed high frequency, high dose opioids than did physicians.60 The fourth study used clotrimazole-betamethasone prescribing as a surrogate of quality of dermatology care, using a database of 301 million outpatient visits for common inflammatory or fungal skin conditions.61 Most visits were to primary care physicians (44.7%) and dermatologists (38.8%). A physician assistant was the sole provider for 0.9% of visits. Each visit was analysed according to whether a physician, physician assistant, or both were involved in the visit. Direct supervision was defined as a visit in which patients were seen by both a physician assistant and a physician. In multivariate logistic regression analyses, prescription rates were as follows: physician assistants in primary care without direct supervision 16.9%, physician assistants in primary care with direct supervision 8.3%, primary care physicians 4.9%, dermatology physician assistants without direct supervision 3.8%, dermatology physician assistants with direct supervision 1.1%, dermatologists 0.2%, and other physicians 1.7%.

primary care without direct supervision 16.9%, physician assistants in primary care with direct supervision 8.3%, primary care physicians 4.9%, dermatology physician assistants without direct supervision 3.8%, dermatology physician assistants with direct supervision 1.1%, dermatologists 0.2%, and other physicians 1.7%. We judged the first three studies to be of weak quality owing to their non-randomised retrospective design and various confounders not accounted for (see table 4).58 59 60 The fourth study assumed that dermatologists provided gold standard care,61 as clotrimazole-betamethasone overuse by primary care physicians has been observed and may represent clinical uncertainty or unfamiliarity with studies examining the therapeutic efficacy of combination versus monotherapy agents.62 The study included multivariate logistic regression analyses to deal with potential confounders; although whether statistically significant differences in patients’ characteristics existed between groups is unknown, the database was vast, and it is unique in measuring the effect of direct supervision.

therapy agents.62 The study included multivariate logistic regression analyses to deal with potential confounders; although whether statistically significant differences in patients’ characteristics existed between groups is unknown, the database was vast, and it is unique in measuring the effect of direct supervision. One study analysed 597 consecutive patients undergoing routine, average risk screening colonoscopy at a single centre between July 2015 and June 2016 (table 4).63 It compared performance of physician assistants, gastroenterology fellows, and gastroenterologists. Physician assistants performed better than fellows in mean intubation time and had higher caecal intubation rates than did gastroenterologists (98.8% v 94.8%; P=0.04). Adenoma detection rates were similar in all groups. Physician assistants with at least 15 years’ experience had shorter mean intubation times than did the two attending gastroenterologists with similar experience (15.6 v 7.5 min; P=0.002). The authors speculated that this could be because the physician assistants kept more up to date with latest endoscopic techniques. We judged this study to be of weak quality owing to its non-randomised retrospective design and no information on how patients were prospectively triaged to different clinicians. Patients found to have poor bowel preparation were excluded from the retrospective analysis.

re up to date with latest endoscopic techniques. We judged this study to be of weak quality owing to its non-randomised retrospective design and no information on how patients were prospectively triaged to different clinicians. Patients found to have poor bowel preparation were excluded from the retrospective analysis. Three studies investigated patient management by physician assistants versus physicians.64 65 66 The first was a US study published in 2018 that investigated the accuracy of skin cancer diagnosis by physician assistants versus dermatologists.64 It analysed 33 647 cancer screening records and found that the number needed to biopsy to diagnose one skin cancer was statistically significantly higher for physician assistants compared with dermatologists. Screenings performed by dermatologists were also more likely to result in a diagnosis of melanoma in situ, but no difference was seen between physician assistants and dermatologists in diagnosing invasive melanomas and other skin cancers, which tend to be more obvious. We judged this study to be of weak quality for its non-randomised retrospective design and not accounting for differences in patient’s characteristics (see table 4). The second was a US study published in 2016 that analysed 178 035 malpractice claims during a 10 year period.65 It found that physicians had higher rates of malpractice claims (range 11.2-19 per 1000 clinicians) than did physician assistants (range 1.4-2.4 per 1000 clinicians). The three most common claims for all groups were diagnosis related (32.2%), surgery related (26.0%), and treatment related (19.8%). Physician assistants were significantly more likely to have diagnosis related and treatment related malpractice allegations than were physicians (P<0.001). We judged this study to be of weak quality because of potential confounders—not all malpractice and adverse events are reported or claims made; and in the US, a plaintiff may hold the physician, as a supervisor, accountable for the actions of his or her employees.

malpractice allegations than were physicians (P<0.001). We judged this study to be of weak quality because of potential confounders—not all malpractice and adverse events are reported or claims made; and in the US, a plaintiff may hold the physician, as a supervisor, accountable for the actions of his or her employees. One UK study, published in 2016, compared the performance of physician assistants and general practitioners by using video recordings of 62 primary care consultations in volunteer practices in England.66 Five general practitioners and four physician assistants participated. The consultations involved adults presenting for same day/urgent appointments. Quality of consultations was assessed by experienced general practitioners masked to the role of the clinician, using the Leicester Assessment Package. Statistically significant differences were found in the number and nature of presenting complaints in each group (see table 4). General practitioners performed better in all domains of the consultation than did physician assistants and were statistically significantly better at problem solving and patient management. No consultation was deemed unacceptable. We judged this study to be of weak quality because of several limitations, including a small number of volunteer participants, differences in patients’ characteristics between the groups, and the low the inter-rater reliability of the Leicester Assessment Package (κ=0.602, 95% confidence interval 0.428 to 0.777).

able. We judged this study to be of weak quality because of several limitations, including a small number of volunteer participants, differences in patients’ characteristics between the groups, and the low the inter-rater reliability of the Leicester Assessment Package (κ=0.602, 95% confidence interval 0.428 to 0.777). Three studies focused on patient satisfaction, comparing physician assistants and physicians (table 5).67 68 69 Of these, two were from the US and one was from the Netherlands. The first,67 published in 2005, analysed 146 880 randomly sampled completed surveys by Medicare beneficiaries in the US between 2000 and 2001. It found that patients were generally satisfied with their care and no difference were apparent between physician assistants, nurse practitioners, and physicians. This study was limited by the low percentage of respondents (2.8%) who identified a physician assistant or nurse practitioner as their primary provider, which was lower than expected and could be explained by the fact that many physician assistants and nurse practitioners work in physicians’ offices as part of a healthcare team. The second study analysed 12 386 surveys completed after dermatology outpatient visits at one US institution between April 2019 and December 2021.68 It found that patient satisfaction was consistently high for physician assistants, residents, and dermatologists throughout the study period. No statistically significant differences were observed between dermatologists and physician assistants. Scores were statistically significantly lower for residents, but the effect size was small (Cohen’s d=0.29). The third study surveyed patients who had received care from either a general practitioner or a physician assistant out of hours in primary care in the Netherlands between July and August 2014.69 A total of 214 patients completed the survey (27% response rate). Patients were highly satisfied with their care and seemed to be just as satisfied with care they received from physician assistants as from general practitioners. We judged all these studies to be of weak quality owing to their unknown or low response rates and potential confounders such as non-responder bias or patients being unable to identify their provider as a physician assistant.

as satisfied with care they received from physician assistants as from general practitioners. We judged all these studies to be of weak quality owing to their unknown or low response rates and potential confounders such as non-responder bias or patients being unable to identify their provider as a physician assistant. Summary of results—patient satisfaction CI=confidence interval; GP=general practitioner; NP=nurse practitioner; PA=physician assistant.

as satisfied with care they received from physician assistants as from general practitioners. We judged all these studies to be of weak quality owing to their unknown or low response rates and potential confounders such as non-responder bias or patients being unable to identify their provider as a physician assistant. Summary of results—patient satisfaction CI=confidence interval; GP=general practitioner; NP=nurse practitioner; PA=physician assistant. Four studies looked at cost effectiveness (table 6).70 71 72 73 Of these, three studies were conducted in the US and one in the Netherlands. The first used US Veterans’ Affairs data between 2012 and 2013 to analyse the records of 47 236 patients with medically complex diabetes.70 Case mix adjusted total care costs were 7% lower for patients of physician assistants, driven by less use of pharmacy and outpatient services by physician assistants than by physicians. Although the study adjusted for differences in the case mix of patients assigned to different provider types, unmeasured confounders not adjusted for included team based care and the smaller average caseload of physician assistants compared with physicians, which could allow them more time for patient care activities. The second study,71 published in 2020, analysed the potential economic impact of using physician assistants to provide routine postoperative care instead of surgeons, on the basis of data provided by 16 physician assistants and six surgeons. Owing to their lower salaries, physician assistants were less expensive than surgeons in providing routine postoperative care and could potentially save surgeons time that they could use for other activities. The third study,72 published in 2020, compared healthcare utilisation and costs among patients with diabetes cared for by physician assistants, nurse practitioners, and physicians as their primary providers. It used the same Veterans’ Affairs database as the second study in this category,70 but it analysed 368 481 records of patients with diabetes. Its findings and limitations were similar—case mix adjusted total care costs were 6% lower for patients seen be physician assistants, driven by lower inpatient, outpatient, and pharmacy costs. We judged all these studies to be of weak quality owing to their retrospective design, various confounders not accounted for (see table 6), and focus on costs rather than cost effectiveness.

al care costs were 6% lower for patients seen be physician assistants, driven by lower inpatient, outpatient, and pharmacy costs. We judged all these studies to be of weak quality owing to their retrospective design, various confounders not accounted for (see table 6), and focus on costs rather than cost effectiveness. Summary of results—cost effectiveness CI=confidence interval; ED=emergency department; GP=general practitioner; NP=nurse practitioner; OR=odds ratio; PA=physician assistant; QALY=quality adjusted life years. One study,73 using the same dataset as two previous studies from the Netherlands,56 57 measured costs and patients’ quality of life scores. After adjustment for differences in medical specialty, hospital type, diagnosis, comorbidities, type of admission, and discharge destination, it found that the mean total costs per patient and quality of life measures did not differ significantly between the two groups. We judged this study to be of weak quality owing to its retrospective design and various confounders; for example, the comparison was wards staffed by physician assistants at least 51% of the time during normal working hours working with residents versus wards staffed by residents, with both groups being supervised by physicians.

Five studies were conducted in the US (table 1).32 33 34 35 36 All were retrospective cohort studies involving hundreds of thousands of patient records. Of these, one study investigated adoption of new chronic disease medications and found that physicians were significantly more likely to prescribe newly approved drugs for chronic diseases than were physician assistants and nurse practitioners.32 Three studies looked at diabetes care and used glycated haemoglobin (HbA1c), systolic blood pressure, and low density lipoprotein cholesterol as surrogates for quality of care.33 34 35 These found no clinically significant differences in outcomes between physician assistants, nurse practitioners, and physicians as the primary care provider. The fifth study compared the practice patterns and quality of care of physician assistants, nurse practitioners, and physicians in community health centres.36 It found no differences in the outcomes studied, apart from that patients were more likely to have smoking cessation and education/counselling services documented if seen by a physician assistant or a nurse practitioner than by a physician. We judged all of these studies to be of weak quality owing to their non-randomised, retrospective design that did not account for potential confounders, such as differences in patients’ characteristics, team care, and visits to other providers. Summary of results—primary care

Five studies were conducted in the US (table 1).32 33 34 35 36 All were retrospective cohort studies involving hundreds of thousands of patient records. Of these, one study investigated adoption of new chronic disease medications and found that physicians were significantly more likely to prescribe newly approved drugs for chronic diseases than were physician assistants and nurse practitioners.32 Three studies looked at diabetes care and used glycated haemoglobin (HbA1c), systolic blood pressure, and low density lipoprotein cholesterol as surrogates for quality of care.33 34 35 These found no clinically significant differences in outcomes between physician assistants, nurse practitioners, and physicians as the primary care provider. The fifth study compared the practice patterns and quality of care of physician assistants, nurse practitioners, and physicians in community health centres.36 It found no differences in the outcomes studied, apart from that patients were more likely to have smoking cessation and education/counselling services documented if seen by a physician assistant or a nurse practitioner than by a physician. We judged all of these studies to be of weak quality owing to their non-randomised, retrospective design that did not account for potential confounders, such as differences in patients’ characteristics, team care, and visits to other providers. Summary of results—primary care CI=confidence interval; FDA=Food and Drug Administration; GP=general practitioner; LDL=low density lipoprotein; NP=nurse practitioner; OR=odds ratio; PA=physician assistant; SBP=systolic blood pressure; SD=standard deviation; VHA=Veterans’ Health Administration.

Five studies were conducted in the US (table 1).32 33 34 35 36 All were retrospective cohort studies involving hundreds of thousands of patient records. Of these, one study investigated adoption of new chronic disease medications and found that physicians were significantly more likely to prescribe newly approved drugs for chronic diseases than were physician assistants and nurse practitioners.32 Three studies looked at diabetes care and used glycated haemoglobin (HbA1c), systolic blood pressure, and low density lipoprotein cholesterol as surrogates for quality of care.33 34 35 These found no clinically significant differences in outcomes between physician assistants, nurse practitioners, and physicians as the primary care provider. The fifth study compared the practice patterns and quality of care of physician assistants, nurse practitioners, and physicians in community health centres.36 It found no differences in the outcomes studied, apart from that patients were more likely to have smoking cessation and education/counselling services documented if seen by a physician assistant or a nurse practitioner than by a physician. We judged all of these studies to be of weak quality owing to their non-randomised, retrospective design that did not account for potential confounders, such as differences in patients’ characteristics, team care, and visits to other providers. Summary of results—primary care CI=confidence interval; FDA=Food and Drug Administration; GP=general practitioner; LDL=low density lipoprotein; NP=nurse practitioner; OR=odds ratio; PA=physician assistant; SBP=systolic blood pressure; SD=standard deviation; VHA=Veterans’ Health Administration. One study was conducted in the UK and published in 2015 (table 1).37 This was a retrospective study based on a review of consultation notes, linked medical records, and patient satisfaction. It analysed the records of 2086 adults attending for same day/urgent appointments being seen by either a physician assistant or a general practitioner in 12 volunteer practices, of which six employed physician assistants and six did not. The primary outcome of this study was re-attendance within 14 days for the same or a linked problem. Secondary outcomes were related to processes of care: use of diagnostics, referrals, prescriptions, advice given, length of consultation, and cost of consultation. All but one of the practices had guidelines for receptionists to assign patients to physician assistants, defined by the supervising general practitioner. In five practices, physician assistants had longer appointment times or empty slots to give time to consult their supervising general practitioner. Physician assistants and general practitioners saw significantly different patients. Physician assistants’ patients were younger, were more likely to present with minor problems or symptoms, and had fewer chronic diseases, fewer repeat prescriptions, and fewer visits to the practice in the previous three months. No significant difference was found in the primary outcome. After adjustment for clustering at practice level, presenting problem, and patients’ characteristics, no significant differences were found in the secondary outcomes. General practitioners saw three patients for every two seen by physician assistants. Costs per consultation were lower for physician assistants (£28.14 ($38.6; €32.9) versus £34.36 for general practitioners), but lack of data on time spent supervising and signing prescriptions by general practitioners means that the true costs of physician assistants are underestimated to an unknown extent. In a review of the medical records of re-attendances, physician assistants performed significantly better than general practitioners in documenting the consultation. Patient satisfaction was high for both groups.

Three retrospective cohort studies looking at care in emergency departments were conducted in the US (table 2).38 39 40 These all analysed outcomes such as unplanned re-attendance within 72 hours, patient acuity/diagnosis, and patient flow measures. The first study used data over 12 years from the National Hospital Ambulatory Medical Care Survey.38 This found that physician teams saw older, more severely ill patients and more often saw them overnight. Patients seen by physician assistant or nurse practitioner teams received fewer diagnostics, fewer procedures, and fewer hospital admissions after adjustment for age, severity, and other characteristics of patients. The second study analysed 25 883 patient encounters in a single emergency department.39 Physicians saw older patients, but physicians and physician assistants saw patients with similar acuity scores. Patients seen by physician assistants waited longer to be seen and had a longer lengths of stay. No differences were found in unplanned re-attendances, patient disposition, and computed tomography scans requested. Patient satisfaction was higher for physician assistant care. The third study looked at 10 369 children aged 6 years or younger seen during a 24 month period in a single emergency department and analysed re-attendance rates.40 Physicians saw younger, sicker patients, and no statistically significant difference was found in re-attendance rates by provider. We judged all these studies to be of weak quality because of their non-randomised retrospective design, the confounder of team based care in the emergency department, and no data on whether diagnostics or procedures were appropriate. The second and third study did not adjust for differences in patients’ characteristics. In addition, accepted metrics of efficiency in the emergency department, such as length of stay, have not shown a causal relation with improved patient care.41

Fourteen studies compared physician assistants versus residents in hospitals (table 3).44 45 46 47 48 49 50 51 52 53 54 55 56 57 All but two studies were conducted in the US. Summary of results—residents versus physician assistants (PAs) in hospitals CI=confidence interval; CPR=cardiopulmonary resuscitation; FY=foundation year; ICU=intensive care unit; LOS=length of stay; NSAID=non-steroidal anti-inflammatory drug; OR=odds ratio; PA=physician assistant; PGY=postgraduate year; PROM=patient reported outcome measure; SD=standard deviation.

Seven US studies looked at the impact of using physician assistants compared with residents/fellows as first assistants in different types of low risk surgery (table 3).44 45 46 47 48 49 50 The first found no difference in outcomes, but different types of assistants (physician assistants, residents, and fellows) assisted to varying degrees according to their level of ability.44 The second compared the same experienced physician assistant with rotating orthopaedic fellows and found no difference in patients’ outcomes but longer surgical times for fellows, which reduced over time until no significant difference was seen by the fourth quarter.45 The third study additionally investigated the impact on hospital costs of using physician assistants versus residents as assistants.46 No difference in patients’ outcomes was found, but longer length of surgery involving residents incurred a “hidden” additional cost, although this was neutralised by the fact that physician assistants were more expensive to employ than residents. The fourth study found no difference in outcomes, although significantly more comorbidities were present in patients in the resident/fellow group that were not adjusted for.47 The fifth compared one highly trained physician assistant versus fellows as lead surgeon working with residents in harvesting lungs for transplant and measured surgical injuries to the donor lung.48 Significantly fewer injuries occurred in lungs harvested by the physician assistant compared with fellows, as well as significantly lower rates of pulmonary graft dysfunction grades 2 and 3, and no statistically significant difference was seen in 30 day and one year graft survival rates.

Five US studies looked at the impact of using physician assistants compared with residents/fellows in general medicine inpatient services (table 3).51 52 53 54 55 The first was a retrospective cohort study analysing 9681 hospital records of patients admitted on weekdays between January 2005 and December 2006.51 It compared hospitalist (a board certified physician specialising in acute inpatient care)-physician assistant teams and attending-resident teams. Inpatient care provided by hospitalist-physician assistant teams was associated with a statistically significantly longer length of stay; however, when adjusted for times when both types of teams could receive admissions, this difference was no longer significant. After adjustment for confounding variables, no statistically significant difference was seen in hospital charges, readmission rates at seven, 14, and 30 days, and inpatient mortality. The second was a retrospective cohort study looking at the care of 95 patients admitted for re-induction chemotherapy for acute myeloid leukaemia between 2008 and 2012.52 Patients were cared for by either physician assistants or house officers, both groups supervised by attending physicians. No statistically significant difference was seen in characteristics of patients admitted to either service. A statistically significantly shorter length of stay, fewer readmissions, and fewer consults per patient were seen for those in the physician assistant service, with no difference in mortality or transfers to intensive care. The third was a retrospective cohort study investigating the care of 5194 consecutive patients admitted to a general medicine service between July 2005 and June 2006.53 It compared hospitalist-physician assistant teams and attending-resident teams. Patients admitted to the hospitalist-physician assistant service were significantly younger, had lower comorbidity scores, and were more likely to be admitted at night. After adjustment for confounders such as age, race, comorbidities, and time of admission, no statistically significant difference was seen in length of stay, inpatient mortality, transfers to intensive care, readmissions, or patient satisfaction.

Eight studies compared the diagnosis or management of physician assistants versus physicians (table 4).58 59 60 61 63 64 65 66 Seven of these were published in the US. Summary of results—diagnosis/performance CI=confidence interval; GP=general practitioner; NNB=number needed to biopsy; NP=nurse practitioner; OR=odds ratio; PA=physician assistant.

Four studies investigated prescribing practices (table 4).58 59 60 61 The first used national databases to analyse the prescribing quality of physician assistants, nurse practitioners, and physicians in outpatients/ambulatory care between 2006 and 2012.58 Overall mean performance across all indicators was 58.7%. Statistically significant differences were found in three of 13 quality indicators (see table 4). The second study used a large national database to investigate differences in opioid prescribing between physician assistants, nurse practitioners, and generalist physicians between 2013 and 2016.59 It found that physician assistants and nurse practitioners made up a disproportionately high number of the prescribers after adjustment for potential confounders, such as practice setting. The third study investigated opioid overprescribing in primary care and found that most nurse practitioners and physician assistants prescribed opioids in a pattern similar to that of physicians, but the nurse practitioner and physician assistant groups had more outliers who prescribed high frequency, high dose opioids than did physicians.60 The fourth study used clotrimazole-betamethasone prescribing as a surrogate of quality of dermatology care, using a database of 301 million outpatient visits for common inflammatory or fungal skin conditions.61 Most visits were to primary care physicians (44.7%) and dermatologists (38.8%). A physician assistant was the sole provider for 0.9% of visits. Each visit was analysed according to whether a physician, physician assistant, or both were involved in the visit. Direct supervision was defined as a visit in which patients were seen by both a physician assistant and a physician. In multivariate logistic regression analyses, prescription rates were as follows: physician assistants in primary care without direct supervision 16.9%, physician assistants in primary care with direct supervision 8.3%, primary care physicians 4.9%, dermatology physician assistants without direct supervision 3.8%, dermatology physician assistants with direct supervision 1.1%, dermatologists 0.2%, and other physicians 1.7%.

One study analysed 597 consecutive patients undergoing routine, average risk screening colonoscopy at a single centre between July 2015 and June 2016 (table 4).63 It compared performance of physician assistants, gastroenterology fellows, and gastroenterologists. Physician assistants performed better than fellows in mean intubation time and had higher caecal intubation rates than did gastroenterologists (98.8% v 94.8%; P=0.04). Adenoma detection rates were similar in all groups. Physician assistants with at least 15 years’ experience had shorter mean intubation times than did the two attending gastroenterologists with similar experience (15.6 v 7.5 min; P=0.002). The authors speculated that this could be because the physician assistants kept more up to date with latest endoscopic techniques. We judged this study to be of weak quality owing to its non-randomised retrospective design and no information on how patients were prospectively triaged to different clinicians. Patients found to have poor bowel preparation were excluded from the retrospective analysis.

Three studies investigated patient management by physician assistants versus physicians.64 65 66 The first was a US study published in 2018 that investigated the accuracy of skin cancer diagnosis by physician assistants versus dermatologists.64 It analysed 33 647 cancer screening records and found that the number needed to biopsy to diagnose one skin cancer was statistically significantly higher for physician assistants compared with dermatologists. Screenings performed by dermatologists were also more likely to result in a diagnosis of melanoma in situ, but no difference was seen between physician assistants and dermatologists in diagnosing invasive melanomas and other skin cancers, which tend to be more obvious. We judged this study to be of weak quality for its non-randomised retrospective design and not accounting for differences in patient’s characteristics (see table 4). The second was a US study published in 2016 that analysed 178 035 malpractice claims during a 10 year period.65 It found that physicians had higher rates of malpractice claims (range 11.2-19 per 1000 clinicians) than did physician assistants (range 1.4-2.4 per 1000 clinicians). The three most common claims for all groups were diagnosis related (32.2%), surgery related (26.0%), and treatment related (19.8%). Physician assistants were significantly more likely to have diagnosis related and treatment related malpractice allegations than were physicians (P<0.001). We judged this study to be of weak quality because of potential confounders—not all malpractice and adverse events are reported or claims made; and in the US, a plaintiff may hold the physician, as a supervisor, accountable for the actions of his or her employees.

Three studies focused on patient satisfaction, comparing physician assistants and physicians (table 5).67 68 69 Of these, two were from the US and one was from the Netherlands. The first,67 published in 2005, analysed 146 880 randomly sampled completed surveys by Medicare beneficiaries in the US between 2000 and 2001. It found that patients were generally satisfied with their care and no difference were apparent between physician assistants, nurse practitioners, and physicians. This study was limited by the low percentage of respondents (2.8%) who identified a physician assistant or nurse practitioner as their primary provider, which was lower than expected and could be explained by the fact that many physician assistants and nurse practitioners work in physicians’ offices as part of a healthcare team. The second study analysed 12 386 surveys completed after dermatology outpatient visits at one US institution between April 2019 and December 2021.68 It found that patient satisfaction was consistently high for physician assistants, residents, and dermatologists throughout the study period. No statistically significant differences were observed between dermatologists and physician assistants. Scores were statistically significantly lower for residents, but the effect size was small (Cohen’s d=0.29). The third study surveyed patients who had received care from either a general practitioner or a physician assistant out of hours in primary care in the Netherlands between July and August 2014.69 A total of 214 patients completed the survey (27% response rate). Patients were highly satisfied with their care and seemed to be just as satisfied with care they received from physician assistants as from general practitioners. We judged all these studies to be of weak quality owing to their unknown or low response rates and potential confounders such as non-responder bias or patients being unable to identify their provider as a physician assistant.

Four studies looked at cost effectiveness (table 6).70 71 72 73 Of these, three studies were conducted in the US and one in the Netherlands. The first used US Veterans’ Affairs data between 2012 and 2013 to analyse the records of 47 236 patients with medically complex diabetes.70 Case mix adjusted total care costs were 7% lower for patients of physician assistants, driven by less use of pharmacy and outpatient services by physician assistants than by physicians. Although the study adjusted for differences in the case mix of patients assigned to different provider types, unmeasured confounders not adjusted for included team based care and the smaller average caseload of physician assistants compared with physicians, which could allow them more time for patient care activities. The second study,71 published in 2020, analysed the potential economic impact of using physician assistants to provide routine postoperative care instead of surgeons, on the basis of data provided by 16 physician assistants and six surgeons. Owing to their lower salaries, physician assistants were less expensive than surgeons in providing routine postoperative care and could potentially save surgeons time that they could use for other activities. The third study,72 published in 2020, compared healthcare utilisation and costs among patients with diabetes cared for by physician assistants, nurse practitioners, and physicians as their primary providers. It used the same Veterans’ Affairs database as the second study in this category,70 but it analysed 368 481 records of patients with diabetes. Its findings and limitations were similar—case mix adjusted total care costs were 6% lower for patients seen be physician assistants, driven by lower inpatient, outpatient, and pharmacy costs. We judged all these studies to be of weak quality owing to their retrospective design, various confounders not accounted for (see table 6), and focus on costs rather than cost effectiveness.

The review question was: “What is the impact of physician assistants on quality of care compared with physicians?” We compared care delivered by physician assistants with care delivered by physicians, including residents, in economically developed countries, using the Institute of Medicine’s definition of quality in healthcare.29 The review found mainly retrospective observational studies of weak quality that varied as to whether statistical adjustments were made for confounders, as well as the statistics used to present findings. The results of studies were also spread across a range of different outcomes and settings, which makes synthesis difficult. The weak nature of the evidence is an important finding in itself. The findings are discussed below in light of this limited evidence. “Safe” is defined as avoiding harm to patients from the care that is intended to help them.29 The greatest number of studies with the most consistent results in this review were those which found that physician assistants practised safely when working under direct supervision and in post-diagnostic care.33 34 35 44 45 46 47 48 49 50 51 52 53 54 55 56 57 63 This is consistent with past studies which have found that physician assistants perform competently within the framework of their delegated responsibilities.74

at physician assistants practised safely when working under direct supervision and in post-diagnostic care.33 34 35 44 45 46 47 48 49 50 51 52 53 54 55 56 57 63 This is consistent with past studies which have found that physician assistants perform competently within the framework of their delegated responsibilities.74 None of the studies in the review was designed to measure harm. This is difficult to do, especially in primary care where a significant proportion of consultations are for minor ailments and the number of physician assistants in the UK is relatively small.75 Evidence of harm is more likely to be found in organisational safety reporting systems or by hand searching coroners’ reports and litigation records. The most common approaches for measuring harm include reporting by staff, analysis of existing databases, reviewing patient records manually or using automation, and asking professionals or patients to recall errors.76 These methods are potentially biased and time consuming, and numbers need to be large to look for relatively rare events such as death.

suring harm include reporting by staff, analysis of existing databases, reviewing patient records manually or using automation, and asking professionals or patients to recall errors.76 These methods are potentially biased and time consuming, and numbers need to be large to look for relatively rare events such as death. Most studies in this review involved retrospective review of patient records, which has limitations. For example, of the four studies comparing physician assistant versus physician performance in the UK, three consisted of retrospective record reviews and used outcomes such as unplanned re-attendance to the same provider as surrogates for safety.37 42 43 This is unreliable, as UK patients may return to another provider if they had a problem with the first one, especially in urban areas with multiple providers. Retrospective record reviews may be limited in detecting diagnostic error, the most common and most dangerous of medical mistakes.77 If the documented diagnosis is wrong but the treatment is appropriate to the wrong diagnosis, then potential patient harm due to a wrong diagnosis will not be detected. This is also a limitation of indirect supervision.

be limited in detecting diagnostic error, the most common and most dangerous of medical mistakes.77 If the documented diagnosis is wrong but the treatment is appropriate to the wrong diagnosis, then potential patient harm due to a wrong diagnosis will not be detected. This is also a limitation of indirect supervision. UK general practitioners performed better in all domains of the consultation compared with physician assistants but were significantly better at problem solving and patient management.66 Physician assistants were significantly more likely to have diagnosis related and treatment related malpractice allegations than physicians.65 One systematic review of the impact of physician assistants in the emergency department found that physician assistants were rated highly by their colleagues in patient education, history taking, and physical examination but lower in diagnosis and management. The willingness of physicians to be treated by a physician assistant decreased as the severity of the clinical scenario increased (44.3% falling to 0.8%),78 implying that physicians did not feel safe being treated by a physician assistant when the scenario was deemed to be outside the scope of their training.78

gement. The willingness of physicians to be treated by a physician assistant decreased as the severity of the clinical scenario increased (44.3% falling to 0.8%),78 implying that physicians did not feel safe being treated by a physician assistant when the scenario was deemed to be outside the scope of their training.78 “Effective” is defined as providing services based on scientific knowledge to all who could benefit and refraining from providing services to those not likely to benefit.29 Physician assistants have been found to compare favourably with physicians in post-diagnostic care outcomes—for example, in studies of diabetes33 34 35—and in general medical inpatient care when compared with residents.51 52 53 54 55 56 57 Physician assistants were also more likely to provide smoking cessation advice and health education.36 37 Evidence from studies measuring processes of care in undifferentiated (not yet diagnosed) settings is limited. These studies had mixed results and did not assess the appropriateness of investigations and treatments.38 39 42 58 59 60 The clotrimoxazole-betamethasone prescribing study showed that physician assistants were more likely than physicians to prescribe this drug ineffectively, although safely.61

ed) settings is limited. These studies had mixed results and did not assess the appropriateness of investigations and treatments.38 39 42 58 59 60 The clotrimoxazole-betamethasone prescribing study showed that physician assistants were more likely than physicians to prescribe this drug ineffectively, although safely.61 “Patient centred” is defined as providing care that is respectful of and responsive to individual patients’ preferences, needs, and values and ensuring that patients’ values guide all clinical decisions.29 This review found that patients’ satisfaction, pertaining to communication (attention, information sharing, customised care, and respect) and time spent with the patient was consistently high, and no difference was apparent between physician assistants and physicians, although patients may not always know they are seeing a physician assistant rather than a physician.42 69

ing to communication (attention, information sharing, customised care, and respect) and time spent with the patient was consistently high, and no difference was apparent between physician assistants and physicians, although patients may not always know they are seeing a physician assistant rather than a physician.42 69 “Timely” is defined as reducing waits and sometimes harmful delays for both people who receive care and those who give care.29 Studies in emergency departments that focused on commonly used patient flow indicators had mixed results.38 39 40 42 43 However, several studies have shown that adding physician assistants to medical teams increases access to care. This has been found in settings as varied as emergency medicine,79 surgery,80 outpatient clinics,81 and nursing homes.82 For example, a study investigating the impact of regular visits by a physician assistant specialising in geriatric medicine reduced the number of annual hospital visits by 38%.83 Although adding physician assistants to medical teams seems to increase access to care, this may reflect the benefits of increased staffing rather than the unique contribution of the physician assistant role.

a physician assistant specialising in geriatric medicine reduced the number of annual hospital visits by 38%.83 Although adding physician assistants to medical teams seems to increase access to care, this may reflect the benefits of increased staffing rather than the unique contribution of the physician assistant role. “Efficient” is defined as avoiding waste (for example, unnecessary tests and referrals) and also includes cost effectiveness.29 A previous study found that both physician assistants and nurse practitioners were statistically significantly more likely to refer unnecessarily than were physicians84; however, none of the studies in this review was designed to assess waste, apart from one study which found that physician assistants did statistically significantly more biopsies than dermatologists to diagnose skin cancer.64 In this review, three of the four studies assessing cost effectiveness were from the US.70 71 72 They analysed costs, not necessarily cost effectiveness. Caution is needed in applying evidence from a very different healthcare system with higher costs, lower efficiency, and worse outcomes compared with the UK.85 In the single study from the Netherlands,73 total hospital costs and measures of patients’ quality of life did not differ between physician assistant-physician care and resident-physician care. Physician assistants, however, are cheaper to train than residents.

wer efficiency, and worse outcomes compared with the UK.85 In the single study from the Netherlands,73 total hospital costs and measures of patients’ quality of life did not differ between physician assistant-physician care and resident-physician care. Physician assistants, however, are cheaper to train than residents. “Equitable” is defined as providing care that does not vary in quality because of personal characteristics such as gender, ethnicity, geographical location, and socioeconomic status.29 Physician assistants in UK are more likely to be found working in under-resourced areas.21 86 This review also found that patients are more likely to see a physician assistant if they live in a socioeconomically deprived area in the UK.37 Although the physician assistant role increases patients’ access to care, the distribution of physician assistants risks maintaining or establishing new inequities in healthcare.

s review also found that patients are more likely to see a physician assistant if they live in a socioeconomically deprived area in the UK.37 Although the physician assistant role increases patients’ access to care, the distribution of physician assistants risks maintaining or establishing new inequities in healthcare. The strengths of this review are that it explores the available evidence across all domains of quality in healthcare, and its findings are directly relevant to the debate on regulation and deployment of physician assistants in the UK. The review was conducted according to Cochrane rapid review guidance, with systematic search strategies, dual screening, and risk of bias assessments. The limitations are that we found only four UK studies that were relevant to the review question. The vast majority of included studies were of weak quality and retrospectively analysed routinely collected data, not always using strategies to deal with confounding factors. Nearly all studies were from the US, and we found no data from a post-covid-19 context. However, the weak nature of the evidence is an important finding and establishes the need for further research.

ity and retrospectively analysed routinely collected data, not always using strategies to deal with confounding factors. Nearly all studies were from the US, and we found no data from a post-covid-19 context. However, the weak nature of the evidence is an important finding and establishes the need for further research. Many arguments for introducing physician assistants into the NHS hinged on the assumption that physician assistants could be indirectly supervised by a named clinician (for example, in primary care), thereby increasing the efficiency of a service. However, the evidence does not support the safety or effectiveness of indirect supervision of physician assistants in undifferentiated (not yet diagnosed) settings. Rather, studies show that physician assistants perform safely and effectively under direct supervision, working in post-diagnostic care, or doing procedures for which they are highly trained, working as part of a medical team. The inability of UK physician assistants to prescribe or request ionising radiation limits the transferability of these findings. Reports in the UK that physician assistants are seeing undifferentiated patients without adequate supervision, acting as senior decision makers, and supervising residents have caused concern.87 88 89 90 91 National guidance on the supervision and scope of practice for physician assistants can ensure that they practice safely and effectively.

physician assistants are seeing undifferentiated patients without adequate supervision, acting as senior decision makers, and supervising residents have caused concern.87 88 89 90 91 National guidance on the supervision and scope of practice for physician assistants can ensure that they practice safely and effectively. Very few well designed studies on the impact of physician assistants on the domains of quality of healthcare have been conducted in the UK. Future research could attempt to answer the questions in box 2 in the UK context. In what scopes of practice do physician assistants perform most safely and effectively? What is the impact of direct supervision versus indirect supervision of physician assistants on patient diagnosis and management outcomes? What is the impact of physician assistants on resident training? How best could physician assistants be deployed in a way that positively affects resident training and patient care? What is the impact (in terms of cognitive load, clinical efficiency, and wellbeing) on physicians of supervising physician assistants? What are public perceptions of the physician assistant role? What is the impact of physician assistant patterns of employment on equity of access to physician care? In what defined settings or scope are physician assistants cost effective?

What is the impact (in terms of cognitive load, clinical efficiency, and wellbeing) on physicians of supervising physician assistants? What are public perceptions of the physician assistant role? What is the impact of physician assistant patterns of employment on equity of access to physician care? In what defined settings or scope are physician assistants cost effective? The evidence to inform how physician assistants should be deployed effectively and safely in the UK is limited. However, the findings from this review are consistent with previous studies and can inform UK practice. A legitimate role exists for physician assistants working alongside physicians in well defined roles under supervision. However, indirect or unsupervised management by physician assistants of undifferentiated symptoms and disease may risk patients’ safety.

“Safe” is defined as avoiding harm to patients from the care that is intended to help them.29 The greatest number of studies with the most consistent results in this review were those which found that physician assistants practised safely when working under direct supervision and in post-diagnostic care.33 34 35 44 45 46 47 48 49 50 51 52 53 54 55 56 57 63 This is consistent with past studies which have found that physician assistants perform competently within the framework of their delegated responsibilities.74 None of the studies in the review was designed to measure harm. This is difficult to do, especially in primary care where a significant proportion of consultations are for minor ailments and the number of physician assistants in the UK is relatively small.75 Evidence of harm is more likely to be found in organisational safety reporting systems or by hand searching coroners’ reports and litigation records. The most common approaches for measuring harm include reporting by staff, analysis of existing databases, reviewing patient records manually or using automation, and asking professionals or patients to recall errors.76 These methods are potentially biased and time consuming, and numbers need to be large to look for relatively rare events such as death.

“Effective” is defined as providing services based on scientific knowledge to all who could benefit and refraining from providing services to those not likely to benefit.29 Physician assistants have been found to compare favourably with physicians in post-diagnostic care outcomes—for example, in studies of diabetes33 34 35—and in general medical inpatient care when compared with residents.51 52 53 54 55 56 57 Physician assistants were also more likely to provide smoking cessation advice and health education.36 37 Evidence from studies measuring processes of care in undifferentiated (not yet diagnosed) settings is limited. These studies had mixed results and did not assess the appropriateness of investigations and treatments.38 39 42 58 59 60 The clotrimoxazole-betamethasone prescribing study showed that physician assistants were more likely than physicians to prescribe this drug ineffectively, although safely.61

“Patient centred” is defined as providing care that is respectful of and responsive to individual patients’ preferences, needs, and values and ensuring that patients’ values guide all clinical decisions.29 This review found that patients’ satisfaction, pertaining to communication (attention, information sharing, customised care, and respect) and time spent with the patient was consistently high, and no difference was apparent between physician assistants and physicians, although patients may not always know they are seeing a physician assistant rather than a physician.42 69

“Timely” is defined as reducing waits and sometimes harmful delays for both people who receive care and those who give care.29 Studies in emergency departments that focused on commonly used patient flow indicators had mixed results.38 39 40 42 43 However, several studies have shown that adding physician assistants to medical teams increases access to care. This has been found in settings as varied as emergency medicine,79 surgery,80 outpatient clinics,81 and nursing homes.82 For example, a study investigating the impact of regular visits by a physician assistant specialising in geriatric medicine reduced the number of annual hospital visits by 38%.83 Although adding physician assistants to medical teams seems to increase access to care, this may reflect the benefits of increased staffing rather than the unique contribution of the physician assistant role.

“Efficient” is defined as avoiding waste (for example, unnecessary tests and referrals) and also includes cost effectiveness.29 A previous study found that both physician assistants and nurse practitioners were statistically significantly more likely to refer unnecessarily than were physicians84; however, none of the studies in this review was designed to assess waste, apart from one study which found that physician assistants did statistically significantly more biopsies than dermatologists to diagnose skin cancer.64 In this review, three of the four studies assessing cost effectiveness were from the US.70 71 72 They analysed costs, not necessarily cost effectiveness. Caution is needed in applying evidence from a very different healthcare system with higher costs, lower efficiency, and worse outcomes compared with the UK.85 In the single study from the Netherlands,73 total hospital costs and measures of patients’ quality of life did not differ between physician assistant-physician care and resident-physician care. Physician assistants, however, are cheaper to train than residents.

“Equitable” is defined as providing care that does not vary in quality because of personal characteristics such as gender, ethnicity, geographical location, and socioeconomic status.29 Physician assistants in UK are more likely to be found working in under-resourced areas.21 86 This review also found that patients are more likely to see a physician assistant if they live in a socioeconomically deprived area in the UK.37 Although the physician assistant role increases patients’ access to care, the distribution of physician assistants risks maintaining or establishing new inequities in healthcare.

The strengths of this review are that it explores the available evidence across all domains of quality in healthcare, and its findings are directly relevant to the debate on regulation and deployment of physician assistants in the UK. The review was conducted according to Cochrane rapid review guidance, with systematic search strategies, dual screening, and risk of bias assessments. The limitations are that we found only four UK studies that were relevant to the review question. The vast majority of included studies were of weak quality and retrospectively analysed routinely collected data, not always using strategies to deal with confounding factors. Nearly all studies were from the US, and we found no data from a post-covid-19 context. However, the weak nature of the evidence is an important finding and establishes the need for further research.

Many arguments for introducing physician assistants into the NHS hinged on the assumption that physician assistants could be indirectly supervised by a named clinician (for example, in primary care), thereby increasing the efficiency of a service. However, the evidence does not support the safety or effectiveness of indirect supervision of physician assistants in undifferentiated (not yet diagnosed) settings. Rather, studies show that physician assistants perform safely and effectively under direct supervision, working in post-diagnostic care, or doing procedures for which they are highly trained, working as part of a medical team. The inability of UK physician assistants to prescribe or request ionising radiation limits the transferability of these findings. Reports in the UK that physician assistants are seeing undifferentiated patients without adequate supervision, acting as senior decision makers, and supervising residents have caused concern.87 88 89 90 91 National guidance on the supervision and scope of practice for physician assistants can ensure that they practice safely and effectively.

Very few well designed studies on the impact of physician assistants on the domains of quality of healthcare have been conducted in the UK. Future research could attempt to answer the questions in box 2 in the UK context. In what scopes of practice do physician assistants perform most safely and effectively? What is the impact of direct supervision versus indirect supervision of physician assistants on patient diagnosis and management outcomes? What is the impact of physician assistants on resident training? How best could physician assistants be deployed in a way that positively affects resident training and patient care? What is the impact (in terms of cognitive load, clinical efficiency, and wellbeing) on physicians of supervising physician assistants? What are public perceptions of the physician assistant role? What is the impact of physician assistant patterns of employment on equity of access to physician care? In what defined settings or scope are physician assistants cost effective?

The evidence to inform how physician assistants should be deployed effectively and safely in the UK is limited. However, the findings from this review are consistent with previous studies and can inform UK practice. A legitimate role exists for physician assistants working alongside physicians in well defined roles under supervision. However, indirect or unsupervised management by physician assistants of undifferentiated symptoms and disease may risk patients’ safety.

Physician assistants (PAs) were introduced in the US in response to medical shortages in certain specialities and regions The first PAs graduated from UK pilot programmes in 2007 Concerns have been raised about the implementation of the PA role in the UK, particularly as “doctor substitutes”

The greatest number of studies with the most consistent results were those that found PAs to practise safely and effectively when working under direct supervision and in post-diagnostic care The evidence is limited and does not support the safety or effectiveness of indirect supervision of PAs in undifferentiated (not yet diagnosed) settings