Table 2

Main findings of included studies

SpecialtyOutcome measuresFinding(s)Quality scoreKey limitationsStudy details
Emergency medicineLength of visit (LOV)Small but clinically insignificant differences (regression coefficient −8): LOV was 8 min longer when patients were treated by a PA (mean 82 min) than a physician (mean 75 min) (95% CI −10 to −6, p<0.001), although difference ranged from 5 to 32 min dependent on patient condition82%
  • Not randomised

  • Differences by patient condition not explained

  • Limited control for confounders

Arnopolin and Smithline32
Total chargeMean total charge was $159 when patients were treated by a PA and $164 by a physician (95% CI 2 to 14, p=0.013), regression coefficient −8
Emergency medicineLeaving without being seenAbsolute improvement (not controlling for hospital or acuity) from 6.5% to 4.9%; when a PA was on duty, the likelihood that a patient left without being seen was less than half (44% (95% CI 31% to 63%), p<0.01), controlling for hospital and patient acuity73%
  • 2 months’ data

  • Sample size unclear

Ducharme et al 33
Wait time (triage to initial assessment)When a PA was involved in patient care, the odds of the patient being seen within the benchmark wait time was 1.6 times greater than when the PA was not involved (95% CI 1.3 to 2.1), p<0.05, adjusting for hospital, acuity and time of day
LOS in EDWhen a PA was involved in patient care, the LOS in the ED was shorter (mean: 262.4 min vs 182.9 min) than when a PA was not present (30.3%; 95% CI 21.6% to 39%), p<0.01
Emergency medicineProportion of visits in which medications are prescribedSignificant differences were observed between PAs if compared with physicians and to NPs in the proportion of visits in which medication was prescribed: PAs 77.9%, physicians 75.5%, nurse practitioners 75.4% (p=0.001)73%
  • Secondary data analysis

  • No adjustment

  • Treatment outcomes/appropriateness not assessed

Hooker et al 34
Mean number of prescriptions written per visitThere were no significant differences among the three providers in mean number of prescriptions per visit (PA and physician 1.7, nurse practitioner 1.6).
Non-narcotic analgesic prescriptionsThere were no significant differences among the three providers in the frequency of prescribing non-narcotic analgesics (p=0.16).
Narcotic analgesic/NSAID prescription by type of providerThere were no significant differences among the three prescribers in the frequency of narcotic analgesics or NSAIDs recorded (p=0.15 and p=0.06, respectively).
Emergency medicineAnalgesia prescribingEmergency physicians gave some form of ED analgesia to 29% of patients, as compared with 10% of patients seen by PAs (OR 3.58; 95% CI 2.05 to 6.24), adjusting for sex, reported degree of pain and fracture.92%
  • Dependent on patient recall

Kozlowski et al 35
Emergency medicine72-hour revisits to the EDPatients treated only by PAs had significantly lower return rates (6.8%) than for the PA/emergency physician combined group (9.3%) and the emergency physician only group (8.0%), p=0.03.77%
  • No adjustment for significant differences in patient age, admission rate or patient complexity

Pavlik et al 36
Emergency medicineProportion of patients with long bone fracture receiving analgesiaPatients seen by PAs had more than twice the odds of receiving opiates/narcotics (OR 2.05%; 95% CI 1.24 to 3.29) and were more likely to receive other analgesics (OR 1.72%; 95% CI 0.94 to 3.17) compared with those not seen by PAs100%
  • Changes in workload and documentation could have confounded results

Ritsema et al 37
Emergency medicinePatient wound infection rateThere were no significant differences in wound infection rates by practitioner level of training (medical students, 0/60 (0%); all residents, 17/547 (3.1%); physician assistants, 11/305 (3.6%); and attending physicians, 14/251 (5.6%); p=0.14).67%
  • Hawthorne effect

  • Differences in wounds not controlled for

Singer et al 38
Trauma and orthopaedicsTriage time to time seen by orthopaedic service (emergency department) (min)PA presence resulted in a 205 min faster orthopaedic service response time (366 min vs 571 min; p=0.0006).91%
  • Exact cost savings difficult to determine

  • Did not have a way of calculating savings for the time it took for patients to reach the OR from the time of triage

  • Single site with two PAs

Althausen et al 39
Triage time to time of surgery (ER) (min)PA presence resulted in a 360 min improvement in time to surgery (1139 min vs 1499 min; p=0.03).
Operating room complication rates (%)There was no significant difference in the proportion of operating room complications with or without PAs (both 0.65%; p=0.9972).
The use of deep vein thrombosis prophylaxis (%)The use of deep vein thrombosis prophylaxis increased by a mean of 6.73 percentage points (60.69% vs 53.96%; p=0.0084) with PA presence.
Postoperative antibiotic administration (%)Postoperative antibiotic administration increased by 2.88 percentage points with PA presence (94.35% vs 91.47%; p=0.0302).
Postoperative complications (%)There was a 4.67 percentage points decrease in postoperative complications with PA presence (8.16% vs 12.83%; p=0.0034).
Triage time to out of emergency department (min)There was a 176 min decrease in total ER time with PA presence (270 min vs 446 min; p<0.001).
Operating room set-up time (min)There was a marginally improved operating room set-up time by 0.43 min with PA presence (26.6 min vs 24 min; p=0.0034).
Time from wound closure to wheels out (operating room) (min)There was no significant difference for this outcome when the PA was present (7.8 min vs 7.6 min; p=0.5914).
Average operating room time (min)There was no significant difference in the average operating room time when the PA was present (70 min vs 74 min; p=0.44).
Cost savings (emergency department) ($)Based on 50% collection of PA charges and emergency department time savings, per orthopaedic trauma patient seen, PAs saved the hospital $133.53 per patient, resulting in $41 394 in 1 year (310 patients).
Cost savings (operating room) ($)The presence of a PA in the operating room resulted in savings of $3207 based on operating room costs (only set-up time was decreased with presence of the PA).
Hospital length of stay (days)There was no significant difference in the hospital LOS when the PA was present if compared with the presence and the absence of PAs (7.96 days vs 8.57 days; p=0.2662).
Trauma and orthopaedicsPatient satisfaction91.3% of hip patients (total=626, 58.5% response) reported being satisfied or very satisfied and 87.7% of knee patients reported being satisfied or very satisfied with PAs at 1-year follow-up (after surgery)32%
  • Methods are not fully described, for example, no description of data analysis

  • Sample is not described

  • Is this a study about PAs or about the two-room operating model?

  • Patient satisfaction with the surgery at 1 year cannot be attributed to the PA

Bohm et al 40
Perceptions of healthcare providers and patients about PAsPatients: overall patients expressed very positive opinions of PAs who were helpful in providing information and explaining aspects of their care.
Ward nurses: felt that patient care, information flow and patient rounds were enhanced by the PAs; ambiguous as to whether PA tasks fell within the scope of nursing.
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.
CostsThe cost of employing three PAs in 2006 (between $270 000 and $327 000) was found to be similar to the foregone general practitioner (GP) surgical assist fees of $270 226.88.
Time savingsPAs were found to ‘free up’ 204 hours/year (the equivalent of four 50 hours’ work weeks) for their supervising physician (p=not reported). Furthermore, they potentially freed GPs from the operating room to spend more time delivering primary care.
ThroughputIncreased the volume from three to seven primary joint surgeries per day through the use of double rooms in 2006.
Waiting timeMedian wait time for surgery decreased from 44 to 30 weeks.
Trauma and orthopaedicsFracture malunion (maximum angulation criteria) at last clinic visitLikelihood of malunion did not differ significantly if the providers included a PA or not (28% vs 56%, Fisher’s exact test p=0.13) or by number of PAs (p=0.11).82%
  • Unadjusted comparisons

  • Difficult to assess how much of the care was carried out by PAs (analysis is cases with any PA involvement vs cases with no PA involvement)

Garrison et al 41
Trauma and orthopaedics
  • 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.

55%
  • Unable to ascertain which data are descriptive quantitative or gained from qualitative interviews

Hepp et al 42
  • Patient rating of quality of care

All patients responded positively to the PA role; overall rating of PA care of 9.65 of 10.
  • Expected and actual operating room times

Double-room experiment: actual preparation time 39% longer than expected and postsurgery time 37% less than expected (absolute times not given) surgeon time 21% less; 2 hours/day saving
Total new patients seenPreoperative care: 30% increase in numbers of patients seen, noticed in the first year
Trauma and orthopaedicsOverall mortalityThe introduction of PAs to the core trauma panel (group 3 vs group 2) decreased overall mortality (2.80% vs 3.76%, adjusted OR 0.74 (95% CI 0.55 to 0.99), p=0.05). Furthermore, the introduction of PAs to general surgery residents (group 3 vs group 1) decreased overall mortality (2.32% vs 3.82%, adjusted OR 0.6 (95% CI 0.45 to 0.81), p=0.003).100%
  • Not all the covariates which could be significantly associated with outcomes were collected (eg, changes in care).

  • The group 1 period was characterised by a transition from on-call attending surgeons to in-house surgeons and the outcomes may not be homogenous across the study period.

  • Other changes were made, not just individual staff type.

Mains et al 43
Hospital LOSThe introduction of PAs to the core trauma panel (group 3 vs group 2) reduced mean and median hospital LOS (4.32 days vs 4.69 days, p=0.05; and 3.74 days vs 3.88 days, p=0.02, respectively). As well, the introduction of PAs to general surgery residents (group 3 vs group 1) reduced mean and median hospital LOS (4.32 days vs 4.62 days, p=0.05; and 3.74 days vs 3.94 days, p=0.003, respectively).
Trauma and orthopaedicsCollaborative relationshipParticipation during trauma alert calls: PA 100%; resident 51% overall, 88% during on-duty hours. Involvement in minor procedures PA 100% when residents off-duty, 91% overall; resident 95% during on-duty hours, 83% overall.82%
  • Investigators not blinded and all work in the trauma centre investigated

  • No sample size calculation

  • Single site with two PAs

  • Minimal description of data collection method

Oswanski et al 44
Transfer timeAfter controlling for age, gender, race and severity of illness, there was no significant difference in the mean transfer rate overall or for any subpopulation (destination) between years 1998 and 1999.
LOSAfter controlling for age, gender, race and severity of injury, there was no significant difference in the mean LOS overall between years 1998 and 1999.
Mortality rateMortality rate for all patients admitted to the trauma service was 2.2% for both 1998 (8/293) and 1999 (13/479).
Internal medicine30-day all-cause readmissionNo statistically significant difference in odds of readmission between expanded PA (14%) and conventional PA (13.7%) groups (OR 0.95; 95% CI 0.87 to 1.04; p=0.27)91%
  • Non-randomised patient allocation

  • Use of secondary data

  • Readmission to the same hospital only

Capstack et al 45
Inpatient mortalityNo statistically significant difference in odds of mortality between expanded PA (1.3%) and conventional PA (0.99%) groups (OR 0.89; 95% CI 0.66 to 1.19; p=0.42)
Cost of careStatistically significant difference in mean patient charge between expanded PA ($7822) and conventional PA ($7755) groups (3.52% lower; 95% CI 2.66% to 4.39%; p<0.001)
Consultant useNo statistically significant difference in utilisation of consultants between expanded PA (1.3%) and conventional PA (0.99%) groups (OR 1.0; 95% CI 0.94 to 1.07; p=0.90)
Length of stayNo statistically significant difference in length of stay between expanded PA (4.1±3.9 days) and conventional PA (4.3±5.6 days) groups (effect size, 0.99 days shorter; 95% CI 0.97 to 1.01 days; p=0.90)
Internal medicineRelative value units (RVU; ie, costs)(1) Radiology RVUs: there were no statistically significant differences between PAs and residents; (2) total RVUs (excluding pharmacy data): PAs used significantly fewer resources when compared with resident services for pneumonia care (p=0.004), although had a higher mortality rate (% and p value not reported). For all other diagnoses there were no statistically significant differences in total RVUs between PAs and residents; (3) laboratory RVUs: there were statistically significant differences between PAs and residents in laboratory relative value units for stroke (p=0.015), pneumonia (p=0.003) and CHF (p=0.004). In each case, PAs’ RVUs were lower than those of residents.86%
  • RVU figures are not explained

  • Non-random group assignment

  • Single centre

Van Rhee et al 46
Length of stay (LOS)There were no significant differences in LOS between PAs and residents after adjusting for admitting physician effect and other covariates.
Mental health Perceived effect and challenges of delivering psychiatric care with the PA modelParticipants described: improved access to primary care for patients; more timely access to psychiatric appointments and longer appointments; equal team cohesion for the PA or the psychiatrist; decreased wait times and improved access to tertiary care and screening programmes; and implementation challenges of triage hierarchy and patient understanding of the term physician assistant45%
  • Qualitative analysis methods described without detail

  • Short report with overview of themes; no quotations

McCutchen et al 47
  • CHF, congestive heart failure; ED, emergency department; ER, emergency room; GP, general practitioner; NP, nurse practitioner; NSAID, non-steroidal anti-inflammatory drug; PA, physician assistant/associate.