Contribution of physician assistants/associates to secondary care: a systematic review

Objective To appraise and synthesise research on the impact of physician assistants/associates (PA) in secondary care, specifically acute internal medicine, care of the elderly, emergency medicine, trauma and orthopaedics, and mental health. Design Systematic review. Setting Electronic databases (Medline, Embase, ASSIA, CINAHL, SCOPUS, PsycINFO, Social Policy and Practice, EconLit and Cochrane), reference lists and related articles. Included articles Peer-reviewed articles of any study design, published in English, 1995–2017. Interventions Blinded parallel processes were used to screen abstracts and full text, data extractions and quality assessments against published guidelines. A narrative synthesis was undertaken. Outcome measures Impact on: patients’ experiences and outcomes, service organisation, working practices, other professional groups and costs. Results 5472 references were identified and 161 read in full; 16 were included—emergency medicine (7), trauma and orthopaedics (6), acute internal medicine (2), mental health (1) and care of the elderly (0). All studies were observational, with variable methodological quality. In emergency medicine and in trauma and orthopaedics, when PAs are added to teams, reduced waiting and process times, lower charges, equivalent readmission rate and good acceptability to staff and patients are reported. Analgesia prescribing, operative complications and mortality outcomes were variable. In internal medicine outcomes of care provided by PAs and doctors were equivalent. Conclusions PAs have been deployed to increase the capacity of a team, enabling gains in waiting time, throughput, continuity and medical cover. When PAs were compared with medical staff, reassuringly there was little or no negative effect on health outcomes or cost. The difficulty of attributing cause and effect in complex systems where work is organised in teams is highlighted. Further rigorous evaluation is required to address the complexity of the PA role, reporting on more than one setting, and including comparison between PAs and roles for which they are substituting. PROSPERO registration number CRD42016032895.


Non-narcotic analgesics prescriptions
There were no significant differences among the three providers in the frequency of prescribing non-narcotic analgesics (p=0. 16).

Methodological quality
The studies were of variable methodological quality. The mean score was 80% (SD 19), median 73%, minimum 32%, [38] maximum 100%, [35,39] IQR 73,92. Figure 2 presents a summary of the degree to which the included studies met the criteria of methodological quality and shows that the most important methodological flaws in the included studies were the failure to adjust the analysis for confounding variables, the absence of information to evaluate participants' selection adequacy, and the lack of information about baseline and/or demographic information of the investigated patients or PAs.

Synthesis of findings on the impact of physician associates
We organised our findings by secondary care specialty. Within each specialty, we described the findings within the quality dimensions, [20] presenting the dimension with the largest number of studies within each specialty.

Emergency medicine
The six studies in emergency medicine compared clinical care offered by PAs and physicians of various grades [33,34,35,36] 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   25 Halter et al_PA-SCER_Main text_20170912 Waiting or access outcomes were reported in one Canadian study; [32] the outcomes were leaving without being seen and waiting times. The presence of a PA was reported as significantly reducing the likelihood of a patient leaving without being seen by 44% (the crude rate being 6.5 without and 4.9% with a PA). and the odds of a patient being seen within their benchmark wait time was 1.6 times greater when the PA was involved in the patient's care, with these analyses strengthened by adjustment for hospital, time of patient visit and acuity level. [32] However, the PA was an additional staff resource rather than a substitute in this study, giving extra coverage at the busiest times, alongside also newly appointed nurse practitioners, who increased the odds of being seen on target more than the PAs did, with an odds ratio of 2.1. .
Length of stay was considered in two studies, [31,32] with contradictory results in the comparison against physicians, from different interventions in terms of PAs. Arnopolin and Smithline (2000)[31] reported experienced ED PAs and physicians working solo at different times of day in a satellite unit. This study provided a direct comparison (and control for patient age in the analysis), with a result of a statistically significantly mean longer length of visit (eight minutes) for patients of PAs but also noted that differences in length of visit varied by diagnostic group, with PAs' patients between five and 32 minutes longer. In contrast, Ducharme et al [32] reported that where PAs were an additional staff resource alternating with nurse practitioners, PAs reduced length of stay the by 30% (mean 80 minute reduction).
Cost was considered through total charge (hospital and physician charge) for the visit, [31] with a small but statistically significant decrease per patient reported when patients were treated by a PA, with differences (not statistically significant) by diagnostic groups.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   26  Halter et al_PA-SCER_Main text_20170912 Treatments offered, in terms of analgesia prescribing, were reported in three studies, [33,34,35] with conflicting findings. Secondary analysis of national (USA) ED survey data (1995 to 2004) reported no significant difference by type of provider in frequency of prescribing narcotic and non-narcotic analgesics and in the mean number of prescriptions per visit, but did observe a statistically significantly higher proportion of PAs' cases receiving a prescription compared with those of physicians and nurse practitioners. [33] No adjustment for potential confounders was made. Using the same national survey data but for a subset for long bone fractures, secondary analysis for 1998 to 2003 reported similarly, with those seen by a PA having adjusted odds of 2.05 for receiving opiate analgesia in the ED. This well powered retrospective cohort study of high quality differs from another study of similar quality with somewhat contrasting findings [34]. For patients contacted at an undefined time, on average three days following their ED visit, those attended by an emergency physician had adjusted odds of 3.52 for receiving pain medication while in the ED (29% of their patients) compared to those attended by PAs (10% of their patients), in a prospective cohort study based on patient self-report. [34] Although the period of time for this study is not specified, it first reported in 1998, perhaps suggesting the same decade of data was involved. These three studies did not report the PAs' place in the team or whether they added to, substituted for members of the medical team, nor whether they saw patients as part of a team or solo.
The only study that considered a clinical outcome of care was the oldest study in the review [36], from 1995. PAs were reported to have no statistically significant difference in wound infection rates, in a large sample of patients presenting with lacerations at the ED, compared to other medical staff providers (medical students, residents and attending physicians). [36] However, the authors noted a potential Hawthorne effect as all wounds had been evaluated by an attending physician prior to allocation to one of the medical team members, based on  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   27 Halter et al_PA-SCER_Main text_20170912 their level of training. It was noted that PAs in this study, with nine to 12 years' experience, were classified as experienced (not junior) practitioners.

Trauma and orthopaedics
Four papers reported on PAs working in trauma and orthopaedics. These spanned a 10 year period. Three [37,39,40] focused on an aspect of provision of a hospital trauma service; and one considered planned inpatient care. [38] Two studies described how PAs were substituting for doctors, for residents [40] or GP surgical assistants [38], whilst the others presented service re-organisations of which PAs were a part, seemingly an addition to the pre-existing medical team [37,39] The outcomes assessed were numerous -patient satisfaction, perceptions of other clinical staff, costs, time of various aspects of care, length of stay, operative complications and mortality. The strength of evidence for each outcome is now assessed.
One prospective study reported both patient satisfaction and acceptability of PAs to other clinical staff from surveys of these groups. Positive results were presented from the patient satisfaction survey, although the number of respondents was small and no comparator data were collected. The reports of staff were more mixed, with physician team members being positive and nursing staff more equivocal, expressing concern about the overlap of tasks traditionally considered to be the responsibility of nurses.
Operational measures were addressed in all four of the studies in this specialty, split into a number of outcomes pertaining to time [37][38][39][40] and to cost. [37,38] The evidence of the impact of PAs on access times was equivocal. One study reported how the wait to be seen and the length of treatment by the orthopaedic service in the emergency  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   28  Halter et al_PA-SCER_Main text_20170912   department section of their orthopaedic pathway were significantly shortened when PAs were   substituted directly for doctors, although the authors attributed this to a combination of factors, and   not just to the PAs, including more registered nurse cover, introduction of a family practice resident and other changing practices. [40] Another found the same when PAs were added to the team as part of larger trauma team re-organisation. [37] Median number of weeks to wait for surgical procedures were also reported to be reduced, [38] attributed by the authors to the use of two operating theatres by the surgeon, made possible by the PA preparing and finishing the case.
In terms of time, Althausen et al (2013) [37] reported in detail on operating room times -set up, wound closure to out of theatre, average operating room time -and only noted a minimal (not statistically significant) difference for set up time in a direct comparison study. PAs also released time for supervising physicians and general practitioners (GPs), who had previously acted as surgical assistants [38].
Three high quality studies [37,39,40] reported variably on length of hospital stay, with one showing a significant reduction (three to four hours, a fraction of one day) for all patients when PAs were an addition to either the resident physician team or reorganised trauma panel [39] and two replacement studies finding no difference -when carrying out adjusted analyses of one year against another [40] or when PAs were present or not. [37] Evidence regarding cost was again mixed. Bohm [38] suggests the actual costs of employment were similar to those of the GPs they replaced in the operating room but argue an opportunity cost for others through released time for supervising physicians. However, a nonreplacement model, Althausen [37] reported specific cost savings in the ED and operating room based on time reduction and PA charges (although they noted that only 50% of PA costs were covered through charges). As well as these operational measures, these studies also reported health outcomes, and all reported improvement in these. [37,39,40] One considered the rate of complication from procedures involving physician associates [37] and two reported on mortality. [39,40] In terms of operating room complication rates, these did not differ significantly, but postoperative complications were reported to have decreased and antibiotic use and DVT prophylaxis increased (statistically significantly) for cases with a physician associate present (although it is noted that the tables in this paper presented the findings contradictory to the text and abstract). [37] One study assessing mortality in two, year long periods reported that involvement of PAs in the clinical team had no effect on overall mortality rates [40] while another found that mortality decreased by approximately one per cent with the introduction of PAs to a trauma panel and 1.5% to general surgery residents' teams. [39] However, this could not be directly attributable to the addition of the PA because contemporaneous improvements in efficiency of the trauma service occurred.

Internal (acute) medicine
The only study [41] considering PAs in internal (acute) medicine examined resource use.
This study measured length of stay, direct costs, and outcomes for patients with diagnoses of cerebrovascular accident, pneumonia, acute myocardial infarction discharged alive, congestive heart failure and gastrointestinal haemorrhage. In this controlled comparative replacement (PAs for interns/residents) study no significant differences in length of stay were found between patients admitted by attending physicians to teams with a physician associate or team with an intern/resident, with length of stay considered to be a proxy for severity of illness. Cost in terms of relative value units (RVUs, based on billing information for physician-ordered items, excluding administrative costs outside of the physician's control) was also mostly similar although laboratory RVUs were lower for PAs, that is, they ordered fewer investigations after adjustment for demographics in each diagnostic group. The authors concluded that PAs used resources as effectively as, or more effectively than, residents.

Principal findings
This systematic review identified a large number of studies of PAs working in secondary care settings, internationally. However, once studies were excluded that ddi not meet the inclusion criteria, only 11 papers remained. Most of the included studies were from the emergency medicine and trauma and orthopaedics specialties, with one from internal medicine. We found no studies in our other specialties of interest -care of the elderly and mental healthwhere other larger groupings of physician associates worked in the UK according to national survey [16] at the time of planning this review. Several of the studies were of high quality, providing comparative data, and some contained statistical adjustments to address confounding; however all findings were observational. While we recognise that trials are rarely feasible in this type of workforce intervention, adjustment for confounding by indication is a serious challenge in this setting, especially when using a limited routine data source, and there was evidence of both residual confounding from imperfect measures of severity [42] and bias from adjusting for co-variates that were not confounders. [43] Quality also varied widely. This is noteworthy considering that this was a relatively recent set of papers. In addition, comparison and synthesis has been limited by the mix in the papers of those who measure outcomes where PAs are an addition to a team (presenting difficulties in attributing the outcomes to PAs as opposed to any other increase in team capacity) and those where PAs substitute for other physicians where the contribution of PAs is actually being measured. Although every paper reported the contribution of PAs in its speciality/subspecialty as overall positive, it is important that the following summary of the main findings of the review is considered in the context of the issues of method and methodological quality.
Results were spread across a number of outcomes, though those related to operational measures were most prevalent. Outcomes reported when employing PAs in emergency medicine were varied. Operational performance results reported were decreased waiting time and reduced length of stay in the emergency department, [32] and an increase in length of visit for those seen by PAs[31] and reduced charges.
[31] Health care outcomes reported were no difference in wound infection rate, [36] and differences which were difficult to interpret, for example an increased prescription rate [33], or increase [35] or decrease in analgesia prescribing. [34] The messages are remarkably similar for trauma and orthopaedics.
Operational measures highlighted no difference to [40] or reduced [37,38,39] waiting times in the emergency, operative and post-operative phases of care; released physician time [38] and reduced cost. [37] Here the evidence on health outcomes was mostly positive -increased adherence to treatment processes such as antibiotic administration [37] and reduced postoperative complications [37] and either no difference [40] or a reduction [39] in mortality. High patient satisfaction and staff acceptability were also reported. [38] The one study in internal (acute) medicine was one of the few using a prospective design, and found few differences in efficiency measures between PAs and residents, although there were lower costs for some conditions. [41] Summarising across the specialties we have reported three studies where PAs were an addition to the team. [32,37,39] In these more patients are reported to have been treated; waiting, ED and operating room times are said to have been shorter and mortality to be lower; assessment of the contribution of PAs as opposed to any increase in team capacity is limited. Six studies which compared outcomes of care by PAs and physicians either when one or the other was providing care or when PAs were substituting overall for physicians [31,34,36,38,40,41] presented mixed results: either no or a very small difference to length of stay, reduced resource used but at equal cost, some time savings to senior physicians, lower analgesia prescribing, no difference in wound infection rate or in acceptability to staff and patients. In the two studies carrying out secondary data analysis we do not know if the PAs were additions or substitutions but both reported higher prescribing by PAs.

Strengths and weaknesses
This review has systematically assessed the body of PA literature most immediately applicable to the current UK secondary care setting. We selected the five specialties in which PAs in the UK were mostly reported to be working [16] and therefore drew together the evidence of most relevance in that context and noted prominent gaps in evidence. However, this excluded evidence from other specialties. We excluded any studies including intensive care data as this overlapped with acute medicine in many abstracts and we could not separately draw this out. We note that this literature appeared to include a greater proportion of studies with stronger study designs, including prospective and randomised designs.
All of the included papers were from North America, with the majority from USA, where health service organisation and the PA role may differ from that in other countries developing the PA role. In the USA PAs can prescribe and are, as a body, more experienced than in countries more recently embracing this role. We planned to carry out meta-analysis as appropriate to the literature included. The diversity of intervention as in initiation of PAs or change to PA practice being measured prevented this, as did identifying the effect of PAs when there were other simultaneous changes, even where a body of literature pertaining to a particular outcome measure, such as length of stay, was included. Although narrative review is more limited in its precision, in following a framework for this, we have aimed to provide a clear rationale for the synthesis and conclusions we draw from it.

Meaning of the study
This evidence is heavily weighted towards process times and patient satisfaction, with much less on health outcomes, although outcomes are crucial to assess safety of practice for all clinicians. Similar findings have been reported in a systematic review of new (non-medical) roles in emergency medicine -reductions in waiting times in emergency departments, high level of patient satisfaction, confidence and acceptance of the roles. [44] Evidence also suggests that the perception of waiting times and satisfaction are correlated. [45] Evidence within the parameters of this review from outside of the USA is very slim, as is evidence from multi-centre studies. The case for PAs in the UK secondary care setting is made on the stability they might offer to medical teams and their broad knowledge in the face of hyper-specialisation. [46] Our recently-acquired knowledge suggests that PAs in England work in teams of multiple medical and other clinical staff grades [47] and that they are seen primarily as a resource where there are significant medical staffing issues. [48] When we place this emerging evidence alongside the exponential growth in training numbers for PAs in England (alongside other UK countries),[49] government support for their professional regulation [50] and increased numbers (in primary care at least) [51] we suggest that this professional group is judged by increasing numbers of employers and workforce planners to be of value, alongside other medical associate professions. [52].
The studies included in this review can be seen as complex interventions in complex systems and yet this has not been considered in the conclusions the authors draw. Well-controlled studies are needed to fill in the gaps in our knowledge about the outcomes of PAs' contribution to the secondary care. High quality substitution evidence from the Netherlands, [53,54] published since we conducted this review, suggests that the role differs in offering greater continuity and PA-managed wards are similarly cost effective to resident-managed wards. More such evidence is required as well as further evaluation from a realist perspective -considering context, mechanisms and outcome -if PAs cannot be separated from service; measurement would utilise the principles of realist complex intervention science [55] or process evaluation to "Clearly describe the intervention and clarify causal assumptions (in relation to how it will be implemented, and the mechanisms through which it will produce change, in a specific context)." [56]

FUNDING AND DISCLAIMER
16604  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47 o n l y 50 ((Efficiency or productivity or economic$ or benefit) adj5 (experience$ or perception$ or view$ or rates or rating or review or audit or impact or influence or effect or outcome or performance or quality)).tw. 34565 51 ((Effectiveness or efficacy or effectivity or capability) adj5 (experience$ or perception$ or view$ or rates or rating or review or audit or impact or influence or effect or outcome or performance or quality)).tw. 35758 52 ((Effectiveness or efficacy or effectivity or capability) adj5 (experience$ or perception$ or view$ or rates or rating or review or audit or impact or influence or effect or outcome or performance or quality)).tw. 35758 53 ((Access$ or responsiveness or timely or timeliness) adj5 (experience$ or perception$ or view$ or rates or rating or review or audit or impact or influence or effect or outcome or performance or quality)).tw. 16251 54 ((Appropriate$ or relevance or relevant) adj5 (experience$ or perception$ or view$ or rates or rating or review or audit or impact or influence or effect or outcome or performance or quality)).tw. 32405 55 ((Cost$ or afford$ value for money or financ$) adj5 (experience$ or perception$ or view$ or rates or rating or review or audit or impact or influence or effect or outcome or performance or quality) General Practitioners: Physicians whose practice is not restricted to a specific field of medicine Geriatrics: The branch of medicine concerned with the physiological and pathological aspects of the aged, including the clinical problems of senescence and senility.

Hospitalization:
The confinement of a patient in a hospital.
Inpatients: Persons admitted to health facilities which provide board and room, for the purpose of observation, care, diagnosis or treatment.   As this review question contained broad terms, these were defined at the outset, as follows: • Physician Associates: trained in a medical model to work in all settings and undertake physical examinations, investigations, diagnosis, treatment, and prescribe within their scope of practice as agreed with their supervising doctor. [1,2] Physician Associates are sometimes described within the term 'mid-level providers' in developed economies: '…..the term mid-level practitioner means an individual practitioner, other than a physician, dentist, veterinarian, or podiatrist, who is licensed, registered, or otherwise permitted by the United States or the jurisdiction in which he/she practices, to dispense a controlled substance in the course of professional practice.
Examples of mid-level practitioners include, but are not limited to, health-care providers such as nurse practitioners, nurse midwives, nurse anaesthetists, clinical nurse specialists and physician assistants who are authorized to dispense controlled substances by the state in which they practice.' [3] While this term is contested as an appropriate umbrella term due to its hierarchical connotations [4,5] and international variation in usage, [6] it appears in the literature regarding Physician Associates.
• Impact: using the broad headings of the components of quality as suggested by Maxwell (1992), [7] augmenting that of Donabedian, [8] that is, effectiveness, efficiency, acceptability, access, equity and relevance; further consolidated in the aspects of quality set out in the NHS Next stage Review (2008)[9]: patient safety, patient experience and effectiveness of care.
• Specialties most frequently employing PAs in England: acute medicine 'Acute medicine is the part of general (internal) medicine concerned with the immediate and early specialist management of adult patients who present to, or from within, hospitals as urgencies or emergencies'. [10] care of the elderly '…geriatric medicine is mainly concerned with people over the age of 75, although many 'geriatric' patients are much older. However, geriatric medicine in the UK is broadly from the age of 65 onwards. Frail older people are those with multiple diseases, that often includes dementia, with reduced functional reserve who tend to present to hospital with 'geriatric syndromes' such as falls, confusion and immobility.' [11] emergency medicine 'Emergency medicine is a field of practice based on the knowledge and skills required for the prevention, diagnosis and management of acute and urgent aspects of illness and injury affecting patients of all age groups with a full spectrum of episodic undifferentiated physical and behavioural disorders; it further encompasses an understanding of the development of

Supplementary file
Study selection 9 State the process for selecting studies (i.e., screening, eligibility, included in systematic review, and, if applicable, included in the meta-analysis).

8-9
Data collection process 10 Describe method of data extraction from reports (e.g., piloted forms, independently, in duplicate) and any processes for obtaining and confirming data from investigators.

10
Data items 11 List and define all variables for which data were sought (e.g., PICOS, funding sources) and any assumptions and simplifications made.

10
Risk of bias in individual studies 12 Describe methods used for assessing risk of bias of individual studies (including specification of whether this was done at the study or outcome level), and how this information is to be used in any data synthesis.

10
Summary measures 13 State the principal summary measures (e.g., risk ratio, difference in means).

10-11
Synthesis of results 14 Describe the methods of handling data and combining results of studies, if done, including measures of consistency (e.g., I 2 ) for each meta-analysis.

Study selection
17 Give numbers of studies screened, assessed for eligibility, and included in the review, with reasons for exclusions at each stage, ideally with a flow diagram.

11+figure1
Study characteristics 18 For each study, present characteristics for which data were extracted (e.g., study size, PICOS, follow-up period) and provide the citations. Table 1 Risk of bias within studies 19 Present data on risk of bias of each study and, if available, any outcome level assessment (see item 12).

Table 2 + Figure2
Results of individual studies 20 For all outcomes considered (benefits or harms), present, for each study: (a) simple summary data for each intervention group (b) effect estimates and confidence intervals, ideally with a forest plot.

Narcotic analgesics/NSAID S prescription by type of provider
There were no significant differences among the three prescribers in the frequency of narcotic analgesics or NSAIDS recorded (p=0. 15 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47 o n l y 32 Halter et al_PA-SCER_Main text_REVISION_submission_Amended_20180321_Clean copy Orthopaedic surgeons: overall the surgeons had very positive opinions of PAs -100% agreement with all survey items: 'a fully trained PA provides surgical assistance equal to an R5 (fifth year of a residency programme)'; 'the presence of PA has improved your job satisfaction'; 'the presence of a PA has safely allowed you to do more surgical volume'; 'the care of your patients in the OR is improved by the assistance of PAs'; 'PAs greatly decrease the amount of "scut work" that you have to do' Operating room nurses: overall OR nurses reported that PAs were valuable team members; improved the care of orthopaedic surgery patients in the operating room; provided surgical assistance superior to family practitioners; and were necessary to run two operating rooms Orthopaedic residents: nearly unanimous that PAs reduced their workload and they generally felt that PAs relieved them of clinical responsibilities so that they could attend teaching.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47 o n l y 33 Halter et al_PA-SCER_Main text_REVISION_submission_Amended_20180321_Clean copy PAs (analysis is cases with any PA involvement versus cases with no PA involvement) Trauma and orthopaedi cs • Perceptions and experiences with the PA • Preoperative care: PA triages, conducts most activities without direct supervision • Operating room: PAs' integration into the OR went well; staff appreciate consistency of the PA; PA acquired skills in a graduated manner -now "preps and closes with patients in OR" • Postoperative care: takes on some of surgical extender role but the role is missed after hours; PA sees 60-70% of all inpatients, freeing up the surgeon; full integration limited by needs for cosignature and verification of orders • Follow-up outpatient care: clinic flow improved • PA is a collaborative member of the team (most mean ratings >4 out of 5.

Methodological quality
The studies were of variable methodological quality. The mean quality score was 79% (SD 0,20), median 82%, minimum 32%, [40] maximum 100%, [37,43] IQR 73,92. Figure 1 presents a summary of the degree to which the included evidence met the criteria of methodological quality and shows that the most important methodological flaws in the included quantitative studies were the failure to adjust the analysis for confounding variables, the absence of information to evaluate participants' selection adequacy, and the lack of information about baseline and/or demographic information of the investigated participants.
Overall, the quality of the included qualitative evidence was low, mainly due to insufficient description of the sampling strategy, data collection and analysis methods. [40,44,47]

Synthesis of findings on the impact of physician associates
We organised our findings by secondary care specialty. Within each specialty, we described the findings within the quality dimensions, [24] presenting the dimension with the largest number of studies within each specialty.

Emergency medicine
The seven studies in emergency medicine variously compared clinical care offered by PAs and physicians of various grades [34,35,36,37] and operational/service measures. [32,33] In only two of these studies was the comparison of PAs and other physicians in a system where  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   38 Halter et al_PA-SCER_Main text_REVISION_submission_Amended_20180321_Clean copy the PAs were described as working 'solo', substituting for physicians at particular times of the day [32] or seeing patients without the input of the attending physician. [36] Waiting or access outcomes were reported in one Canadian study; [33] the outcomes were leaving without being seen and waiting times. The presence of a PA was reported as significantly reducing the likelihood of a patient leaving without being seen by 44% (95% CI 31% to 63%, p < 0.01), the crude rate being 6.5 without and 4.9% with a PA). The odds of a patient being seen within their benchmark wait time was 1.6 times greater (95% CI 1.3 to 2.1, p <0.05) when the PA was involved in the patient's care, with these analyses strengthened by adjustment for hospital, time of patient visit and acuity level. [33] However, the PA was an additional staff resource rather than a substitute in this study, giving extra coverage at the busiest times, alongside also newly appointed nurse practitioners, who increased the odds of being seen on target more than the PAs did, with an odds ratio of 2.1.
Length of stay was considered in two studies, [32,33] with contradictory results in the comparison against physicians, from different interventions in terms of PAs. Arnopolin and Smithline (2000) [32] reported experienced ED PAs and physicians working solo at different times of day in a satellite unit. This study provided a direct comparison (and control for patient age in the analysis), with a result of a statistically significantly mean longer length of visit (eight minutes) for patients of PAs (82 minutes versus the physicians' 75 minutes, 95% CI -10 to -6, p<0.001), but also noted that differences in length of visit varied by diagnostic group, with PAs' patients between five and 32 minutes longer. In contrast, Ducharme et al [33] reported that where PAs were an additional staff resource alternating with nurse practitioners, PAs reduced their length of stay by 30% (mean 80 minute reduction, 183 minutes versus 262 minutes, 95% CI 21.6% to 39%, p < 0.01) . Cost was considered through total charge (hospital and physician charge) for the visit, [32] with a small but statistically significant decrease per patient reported when patients were treated by a PA, with differences (not statistically significant) by diagnostic groups.
Treatments offered, in terms of analgesia prescribing, were reported in three studies, [34,35,37] with conflicting findings. Secondary analysis of national (USA) ED survey data (1995 to 2004) reported no significant difference by type of provider in frequency of prescribing narcotic or non-narcotic analgesics and in the mean number of prescriptions per visit, but did observe a statistically significantly higher proportion of PAs' cases receiving a prescription compared with those of physicians and nurse practitioners (PAs 77.9%, physicians 75.5%, nurse practitioners 75.4%, p=0.001). [34] No adjustment for potential confounders was made. Using the same national survey data but for a subset for long bone fractures, secondary analysis for 1998 to 2003 reported similarly, with those seen by a PA having adjusted odds of 2.05 for receiving opiate analgesia in the ED (95% CI 1.24 to 3.29). [37] This well powered retrospective cohort study of high quality differs from another study of similar quality with somewhat contrasting findings [35] in which for patients contacted at an undefined time (average three days following their ED visit) those attended by an emergency physician had adjusted odds of 3.58 (95% CI 2.05 to 6.24) for receiving pain medication while in the ED (29% of their patients) compared to those attended by PAs (10% of their patients), in a prospective cohort study based on patient self-report. [35] Although the period of time for this study is not specified, it first reported in 1998, perhaps suggesting the same decade of data was involved. These three studies did not report the PAs' place in the team or whether they added to or substituted for members of the medical team, nor whether they saw patients as part of a team or solo.
Two studies considered clinical outcomes of care. One, the oldest study in the review [38], from 1995, reported that in a large sample of patients presenting with lacerations at the ED  40 Halter et al_PA-SCER_Main text_REVISION_submission_Amended_20180321_Clean copy and seen by PAs there was no statistically significant difference in wound infection rates, , compared to other medical staff providers (medical students, residents and attending physicians). [38] However, the authors noted a potential Hawthorne effect as all wounds had been evaluated by an attending physician prior to allocation to one of the medical team members, based on their level of training. It was noted that PAs in this study, with nine to 12 years' experience, were classified as experienced (not junior) practitioners. The other, newer, study [36] used a proxy measure of clinical safety, that is the 72-hour re-attendance (recidivism) rate to the ED for children aged six and younger, and reports that this was significantly lower for those patients treated only by a PA (6.8% versus emergency physician 8.0%, p=0.03), in a large study. However, these rates were unadjusted, and the characteristics of the study population show statistically significantly different mean ages and rate of admission in the patients treated in each group, with PAs seeing the older of the children who were much less likely to be admitted. Although analysis of the recidivism rates by Emergency Severity Index score for patients seen by PAs versus doctors found no statistically significant differences between groups and the authors conclude that PA providers deliver comparable care; the authors themselves consider that it is not known if PAs would have made the same decisions as physicians for the same group of patients. Six papers reported on PAs working in trauma and orthopaedics. These spanned a 14 year period. Four [39,41,43,44] focused on an aspect of provision of a hospital trauma service; and two considered planned inpatient care. [38,42] Three studies described how PAs were substituting for doctors, for residents [44] or surgical assistants [40,42], whilst the others presented service re-organisations of which PAs were a part, seemingly an addition to the pre-existing medical team [39,41,43] The outcomes assessed were numerous -patient satisfaction, perceptions of other clinical staff, costs, time of various aspects of care, patient throughput, length of stay, fracture malunion and operative complications and mortality. The strength of evidence for each outcome is now assessed.

Trauma and orthopaedics
Two prospective studies of the addition of PAs to surgical teams, pre-, intra-and postoperatively [40,42] reported both patient satisfaction and acceptability of PAs to other clinical staff from surveys of these groups. Positive results were presented from both studies' patient satisfaction surveys, in large [40] and small [42] response numbers, reporting 91.3% of hip and 87.7% of knee patients being satisfied or very satisfied [40] and an overall rating of PA care of 9.65 out of 10 [42] although no comparator data were collected. The reports of staff were more mixed by staff group in Bohm's study [40] with physician team members being positive (100% agreement with all survey items on the positive contribution of PAs) and nursing staff more equivocal, expressing concern about the overlap of tasks traditionally considered to be the responsibility of nurses; and by impact in different parts of the surgical journey in Hepp's [42] study, where staff ratings were mostly above four out of five, agreeing or strongly agreeing that the PA was a collaborative team member. Staff appreciated continuity and PA advances in skills in the operating room, but did not feel the role could  42 Halter et al_PA-SCER_Main text_REVISION_submission_Amended_20180321_Clean copy offer everything a previous surgical extender did post-operatively, despite being a collaborative team member. [42] Operational measures were addressed in five of the studies in this specialty, split into a number of outcomes pertaining to time [39,40,[42][43][44] and to cost. [39,40] The evidence of the impact of PAs on access times was equivocal. One study reported how the wait to be seen by the orthopaedic service in the emergency department section of their orthopaedic pathway were significantly shortened (366 minutes versus 571 mins; p=0.0006) when PAs were substituted directly for doctors , although the authors attributed this to a combination of factors, and not just to the PAs, including more registered nurse cover, introduction of a family practice resident and other changing practices. [39] Another found the same when PAs were added to the team as part of larger trauma team re-organisation. [39] Median number of weeks to wait for surgical procedures was also reported to be reduced from 44 to 30 weeks, [40] attributed by the authors to the use of two operating theatres by the surgeon, made possible by the PA preparing and finishing the case, similarly to the 30% increased throughput in the number of new patients in the pre-operative stage. [42] In terms of time, two studies [39,42] reported in detail on operating room times-set up, wound closure to out of theatre, average operating room time, and post surgery time.
Althausen et al [39] only noted a minimal (not statistically significant -(26.6 versus 24 mins; p=0.0034) difference for set up time in a direct comparison study, while Hepp [42] describes a 39% reduction in time at this stage. PAs also released time for supervising physicians -204 hours per year (p=not reported) [40] or two hours a day,42] and for general practitioners (GPs) (not quantified), who had previously acted as surgical assistants [40], Three high quality studies [39,43,44] [43] and two replacement studies finding no difference -when carrying out adjusted analyses of one year against another [44] or when PAs were present or not. [39] Evidence regarding cost was again mixed. Bohm [40] suggests the actual costs of employment for three PAs (between $270,000 and $327000) were similar to those of the GPs they replaced ($270226.88) in the operating room but argue an opportunity cost for others through released time for the supervising physicians. However, a non-replacement model, Althausen [39] reported specific cost savings in the ED ($133.53 saving per patient, $41,394 in one year) and operating room ($3,207 saving) based on time reduction and PA charges (taking account that only 50% of PA costs were covered through charges).
As well as these operational measures, these studies also reported health outcomes, and all reported no difference [41] or improvement in these. [39,43,44] Two considered the rate of complication from procedures involving physician associates [37,41] and two reported on mortality. [43,44] In terms of operating room complication rates [39] or the likelihood of fracture malunion if the providers included a PA [41], these did not differ significantly from those of other providers, but postoperative complications were reported to have decreased (8.16 versus 12.83%, p=0.0034) and antibiotic use (94.35 versus 91.47%, p=0.0302) and DVT prophylaxis (60.69 versus 53.96%, p = 0.0084) increased (statistically significantly) for cases with a physician associate present (although it is noted that the tables in this paper presented findings contradictory to the text and abstract). [39] One study assessing mortality in two, year-long periods reported that involvement of PAs in the clinical team had no effect on overall mortality rates [44] while another found that mortality decreased by approximately one per cent with the introduction of PAs to a trauma panel (9.67% versus 12.21%, adjusted  [43] However, this could not be directly attributable to the addition of the PA because contemporaneous improvements in efficiency of the trauma service occurred.

Acute internal medicine
The two studies considering PAs in acute internal medicine both examined resource use and clinical outcomes [45,46] in replacement studies, one prospectively examining the impact of PAs in place of interns/residents [46], the other retrospectively comparing outcomes where PAs made up a greater or lesser proportion of the medical team staff, in place of physicians. [45] Both studies measured length of stay, direct costs, and inpatient mortality for patients with diagnoses of cerebrovascular accident, pneumonia, acute myocardial infarction discharged alive, congestive heart failure and gastrointestinal haemorrhage [46] and those with a principal medical (non-surgical, non-obstetrical) diagnosis code. [45]; the latter study also measuring 30-day all cause readmission Neither study reported any significant differences in length of stay between groups, with length of stay considered to be a proxy for severity of illness. Cost in terms of relative value units (RVUs, based on billing information for physician-ordered items, excluding administrative costs outside of the physician's control) was also mostly similar although laboratory RVUs were lower for PAs, that is, they ordered fewer investigations after adjustment for demographics in each diagnostic group (for stroke, p = .015, pneumonia p = .003 and CHF p = .004). In each case PAs' RVUs were  [46], although the authors reported neither the percentage nor statistical values, and the larger study reported no significant differences in mortality or 30 day all-cause readmission. [45] The authors concluded that PAs used resources as effectively as, or more effectively than, residents [46] at the same time as providing similar clinical quality. [45]

Principal findings
This systematic review identified a large number of studies of PAs working in secondary care settings, internationally. However, once studies were excluded that did not meet the inclusion criteria, only 16 papers remained. Most of the included studies were from the emergency medicine and trauma and orthopaedics specialties, with two from acute internal medicine and one from mental health. We found no studies in our other specialty of interest -care of the elderly-where another larger grouping of PAs worked in the UK according to a national survey [18] at the time of planning this review. Several of the studies were of high quality, providing comparative data, and some contained statistical adjustments to address confounding; however all findings were observational. While we recognise that trials are rarely feasible in this type of workforce intervention, adjustment for confounding by indication is a serious challenge in this setting, especially when using a limited routine data source, and residual confounding from imperfect measures of severity [48] and bias from adjusting for co-variates that were not confounders[49] were likely. Quality also varied widely. This is noteworthy considering that this was a relatively recent set of papers. In addition, comparison and synthesis has been limited by the mix in the papers of those who measure outcomes where PAs are an addition to a team (presenting difficulties in attributing the outcomes to PAs as opposed to any other increase in team capacity) and those where PAs substitute for other physicians where the contribution of PAs themselves is actually being measured. Although every paper reported the contribution of PAs in its speciality/subspecialty as overall positive, it is important that the following summary of the main findings of the review is considered in the context of the issues of method and methodological quality.
Results were spread across a number of outcomes, though those related to operational measures -waiting times or times taken for treatment, as well as patient satisfaction -were most prevalent. Outcomes reported when employing PAs in emergency medicine were varied. Operational performance results reported were decreased waiting time and reduced length of stay in the emergency department, [33] an increase in length of visit for those seen by PAs [32] and reduced charges. [32] Health care outcomes reported were no difference in 72-hour revisits to the ED [36] or wound infection rate, [39] and differences which were difficult to interpret, for example an increased prescription rate [34], or increase [37] or decrease in analgesia prescribing. [35] The messages are remarkably similar for trauma and orthopaedics. Operational measures highlighted no difference to [44] or reduced [39,40,42,43] waiting times in the emergency, operative and post-operative phases of care; released physician time [40,42] and reduced cost. [39] Here the evidence on health outcomes was mostly positive -increased adherence to treatment processes such as antibiotic administration [39], reduced post-operative complications [39], no difference in fracture malunion [41]and either no difference [44] or a reduction [43] in mortality. High patient satisfaction and staff acceptability, albeit with some caveats, were also reported. [40,42]  The two studies in internal (acute) medicine were of high quality and were among the few replacing physicians with PAs. Both found no differences in clinical outcomes between PAs and residents, or in length of stay, although lower costs were reported. [45,46] In mental health, the one study's qualitative evidence points also to acceptability of the role through team cohesion and improvements in whole system working. [47] Summarising across the specialties we have reported five studies where PAs were an addition to the team. [33,39,42,43,47] In these more patients are reported to have been treated; waiting, ED and operating room times are said to have been shorter and mortality to be lower; however assessment of the contribution of PAs as opposed to any increase in team capacity is limited. Eight studies which compared outcomes of care by PAs and physicians either when one or the other was providing care or when PAs were substituting overall for physicians [32,35,36,38,40,44,45,46] presented mixed results: either no or a very small difference to length of stay, reduced resource used but at equal or reduced cost, some time savings to senior physicians, lower analgesia prescribing, no difference in wound infection rate, inpatient mortality or re-attendance, or in acceptability to staff and patients. In three of the studies we do not know if the PAs were additions or substitutions but two reported higher prescribing by PAs. [34,37] and one no difference in negative outcomes from fracture. [41] Strengths and weaknesses This review has systematically assessed the body of PA literature most immediately applicable to the current UK secondary care setting. We selected the five specialties in which PAs in the UK were mostly reported to be working [18] and therefore drew together the evidence of most relevance in that context and noted prominent gaps in evidence. However, this excluded evidence from other specialties. We excluded any studies including intensive  48 Halter et al_PA-SCER_Main text_REVISION_submission_Amended_20180321_Clean copy care data as this overlapped with acute medicine in many abstracts and we could not separately draw this out, and similarly we excluded studies with medical and surgical specialties combined. We note that this literature appeared to include a greater proportion of studies with stronger study designs, including prospective and randomised designs; in particular we have excluded the recent matched controlled large studies from the Netherlands in which several specialties -some within and some without our inclusion criteria -were studied. [50,51] All of the included papers were from North America, with the majority from the USA, where health service organisation and the PA role may differ from that in other countries developing the PA role. In the USA PAs can prescribe and order ionising radiation, and are, as a body, more experienced than in countries more recently embracing this role.
We planned to carry out meta-analysis as appropriate to the literature included. The diversity of intervention as in initiation of PAs or change to PA practice being measured prevented this, as did identifying the effect of PAs when there were other simultaneous changes, even where a body of literature pertaining to a particular outcome measure, such as length of stay, was included. Although narrative review is more limited in its precision, in following a framework for this, we have aimed to provide a clear rationale for the synthesis and conclusions we draw from it.

Meaning of the study
This evidence is heavily weighted towards process times and patient satisfaction, with much less on health outcomes, although outcomes are crucial to assess safety of practice for all clinicians. Similar findings have been reported in a systematic review of new (non-medical)  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   49 Halter et al_PA-SCER_Main text_REVISION_submission_Amended_20180321_Clean copy roles in emergency medicine -reductions in waiting times in emergency departments, high level of patient satisfaction, confidence and acceptance of the roles. [52] Evidence also suggests that the perception of waiting times and satisfaction are correlated. [53] Evidence from outside of the USA is very slim, as is evidence from multi-centre studies. The implications of this for policy can be seen in two ways.
Firstly, the limitations to evidence could be considered a cause for some concern, particularly in light of exponential growth in training numbers for PAs in England (alongside other UK countries), [54] government support for increased numbers (in primary care at least) [10] and for recent consultation on the introduction of statutory regulation for PAs, alongside judgment by employers and workforce planners of the role's value, alongside other medical associate professions. [55,56] Numbers of PAs are also rising rapidly in the US. [4] That said, the evidence presented is this review is positive and likely supportive of the direction of travel in policy. In addition, the case for PAs in the UK secondary care setting is made on the stability they might offer to medical teams and their broad knowledge in the face of hyperspecialisation[57] and recently-acquired knowledge -although not covered in this review due to its inclusion of PAs from across multiple specialties -suggests that PAs in England work in teams of multiple medical and other clinical staff grades [58] and that they are seen primarily as a resource where there are significant medical staffing issues. [59] High quality, multi-centre matched controlled substitution evidence from the Netherlands, [50,51] reassuringly also offers similar evidence to that included in our review regarding no difference in a large number of inpatient and post-discharge clinical outcomes, alongside an increase in patient satisfaction. The study found no difference in total healthcare costs or Quality Adjusted Life Years, despite lower personnel costs. The authors conclude that PA substitution appeared safe. The studies included in this review can be seen as complex interventions in complex systems and yet this has not been considered in the conclusions the  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   50 Halter et al_PA-SCER_Main text_REVISION_submission_Amended_20180321_Clean copy authors draw. Well-controlled studies are needed to fill in the gaps in our knowledge about the outcomes of PAs' contribution to the secondary care. More such evidence is required as well as further evaluation from a realist perspective -considering context, mechanisms and outcome -if PAs cannot be separated from service; measurement would utilise the principles of realist complex intervention science [60] or process evaluation to "Clearly describe the intervention and clarify causal assumptions (in relation to how it will be implemented, and the mechanisms through which it will produce change, in a specific context)."[61]

FUNDING AND DISCLAIMER
1989. By exploding this term, we do include MeSH terms found below it in the MeSH hierarchy as follows: benchmarking.

Psychiatry:
The medical science that deals with the origin, diagnosis, prevention, and treatment of mental disorders.

Traumatology:
The medical specialty which deals with wounds and injuries as well as resulting disability and disorders from physical traumas.
Treatment Outcome: Evaluation undertaken to assess the results or consequences of management and procedures used in combating disease in order to determine the efficacy, effectiveness, safety, and practicability of these interventions in individual cases or series. Year introduced: 1992 As this review question contained broad terms, these were defined at the outset, as follows: x Physician Associates: trained in a medical model to work in all settings and undertake physical examinations, investigations, diagnosis, treatment, and prescribe within their scope of practice as agreed with their supervising doctor. [1,2] Physician Associates are sometimes described within WKH WHUP µPLG-OHYHO SURYLGHUV ¶ LQ GHYHORSHG HFRQRPLHV µ« WKH WHUP PLG-level practitioner means an individual practitioner, other than a physician, dentist, veterinarian, or podiatrist, who is licensed, registered, or otherwise permitted by the United States or the jurisdiction in which he/she practices, to dispense a controlled substance in the course of professional practice.
Examples of mid-level practitioners include, but are not limited to, health-care providers such as nurse practitioners, nurse midwives, nurse anaesthetists, clinical nurse specialists and physician assistants who are authorized to dispense controlled substances by the state in which they SUDFWLFH ¶ >3] While this term is contested as an appropriate umbrella term due to its hierarchical connotations [4,5] and international variation in usage, [6] it appears in the literature regarding Physician Associates.
x Impact: using the broad headings of the components of quality as suggested by Maxwell (1992), [7] augmenting that of Donabedian, [8] that is, effectiveness, efficiency, acceptability, access, equity and relevance; further consolidated in the aspects of quality set out in the NHS Next stage Review (2008)[9]: patient safety, patient experience and effectiveness of care.

+ reference
Eligibility criteria 6 Specify study characteristics (e.g., PICOS, length of follow-up) and report characteristics (e.g., years considered, language, publication status) used as criteria for eligibility, giving rationale.

9
Information sources 7 Describe all information sources (e.g., databases with dates of coverage, contact with study authors to identify additional studies) in the search and date last searched. 8 Search 8 Present full electronic search strategy for at least one database, including any limits used, such that it could be repeated.

Supplementary file
Study selection 9 State the process for selecting studies (i.e., screening, eligibility, included in systematic review, and, if applicable, included in the meta-analysis).

8-10
Data collection process 10 Describe method of data extraction from reports (e.g., piloted forms, independently, in duplicate) and any processes for obtaining and confirming data from investigators.

10
Data items 11 List and define all variables for which data were sought (e.g., PICOS, funding sources) and any assumptions and simplifications made.

10
Risk of bias in individual studies 12 Describe methods used for assessing risk of bias of individual studies (including specification of whether this was done at the study or outcome level), and how this information is to be used in any data synthesis.

10
Summary measures 13 State the principal summary measures (e.g., risk ratio, difference in means).

10-11
Synthesis of results 14 Describe the methods of handling data and combining results of studies, if done, including measures of consistency (e.g., I 2 ) for each meta-analysis.

10
Additional analyses 16 Describe methods of additional analyses (e.g., sensitivity or subgroup analyses, meta-regression), if done, indicating which were pre-specified. n/a

Study selection
17 Give numbers of studies screened, assessed for eligibility, and included in the review, with reasons for exclusions at each stage, ideally with a flow diagram.

11+figure1
Study characteristics 18 For each study, present characteristics for which data were extracted (e.g., study size, PICOS, follow-up period) and provide the citations. Table 1 Risk of bias within studies 19 Present data on risk of bias of each study and, if available, any outcome level assessment (see item 12).

Table 2 + Figure2
Results of individual studies 20 For all outcomes considered (benefits or harms), present, for each study: (a) simple summary data for each intervention group (b) effect estimates and confidence intervals, ideally with a forest plot. n/a DISCUSSION Summary of evidence 24 Summarize the main findings including the strength of evidence for each main outcome; consider their relevance to key groups (e.g., healthcare providers, users, and policy makers).