Methods

Design: non‐randomised trial

Timeline: 1 June 1989 to 31 May 1990 (not bank holidays or first 2 weeks of August, February)

Duration: 1 year

Triage: patients categorised by trained nurses based on perceived need for care as either 'primary care' or 'accident and emergency'

Data collection:

Data on process and outcome variables (doctor's use of radiology, haematology, chemical pathology and microbiology investigations, items prescribed), referral and discharge decisions were obtained from hospital records and consultation record forms.

Patient satisfaction and health status were assessed through a simple questionnaire (administered by phone or through post) to assess (1) self reported recovery in 7 to 10 days subsequent to attending ED and (2) health‐seeking behaviour during this period, including re‐attendance at ED or attendance at own GP surgery.

Participants

Intervention group: N = 8 GPs (11 GPs applied, 6 were appointed, 2 left during study and were replaced)

Control group: N = 31 EPs (27 senior house officers, 3 registrars, and 1 senior registrar)

Provider characteristics: none reported

Patients: new ED attendees with 'primary care' suitable problems

Total number of patients: N = 4641; intervention group: n = 1702 patients seen by GPs; control group: n = 2939 patients seen by EPs

Patient characteristics:

Sex: 47.4% female

Age: 41.7% 17 to 30 years

Duration of complaints: 62.2% problems > 24 hours; 20.8% had previously seen a GP

Most common diagnoses: injury and poisoning (44.4%), musculoskeletal diseases (13.7%), non‐specific symptoms and signs (7.0%)

Patient characteristics for control and intervention groups not available.

Setting:

Hospital: one, King’s College Hospital

Country: United Kingdom

Hospital characteristics (1990 figures):

Beds: n/a

Teaching hospital, inner city, "multiethnic, socially deprived"

Yearly attendance: 70,000

Yearly re‐attendance: n/a

Interventions

Intervention: sessional GPs providing care for non‐urgent patients in the ED

Control: regularly scheduled EPs providing care for non‐urgent patients in the ED

Patients referred by GPs were excluded.

Study took place from 1 June 1989 until 31 May 1990 (48 weeks total within 12 months, as bank holidays and the first two weeks of August and February when senior house officers change employment were excluded).

Primary care sessions were established within the ED from 10‐1300 h, 14‐1700 h, and 18‐2100 h each day, except weekends when evening sessions were not available (see Figure 2). 1 physician (either a GP or an EP) was allocated to staff each primary care session according to a weekly rota. All patients triaged as 'primary care suitable' during a particular session were seen by the same physician (a GP or an EP). Medical staff knew patients’ triage status, but patients were unaware of their triage status or the type of physician (GP or EP) they were seeing. Both GPs and EPs were encouraged to use a designated consultation room for primary care sessions and were required to complete a consultation record form for each patient seen. Physicians were unaware how this data would be analysed.

Each week, a random number table was used to select 2 to 3 daytime and 1 evening weekday sessions and 1 daytime weekend session for inclusion in the study (see Figure 2). Hence 8 to 10 sessions, which included a mix of GP and EP assignments, were selected for inclusion each week; this was done for a total of 48 weeks. Physicians were unaware of which sessions were included in the study and what outcomes were being measured. A total of 419 primary care sessions (215 GP‐ and 204 EP‐staffed sessions) were selected by stratified random sampling for inclusion in the study. Primary care sessions staffed by an EP formed the control group.

The study authors noted that there was occasional cross‐over where the allocated physician did not treat primary care patients. This loss of randomisation occurred in both GP‐ and EP‐staffed sessions when the primary care session workload was excessive (to prevent unacceptable wait times) or when EPs were called away to manage urgent patients or to supervise junior physicians in the ED. The frequency and extent with which cross‐over occurred was not reported. To remedy this loss of randomisation, the study authors regrouped patients according to the type of doctor seen and used log‐linear modelling to adjust for confounding factors in their analysis.

Outcomes

1. Investigations: laboratory investigations: chemistry, haematology, microbiology; X‐rays; ECGs

2. Prescriptions

3. Referrals to: community or GP; on‐call specialist team; outpatient clinic

4. ED re‐attendance

5. Patient satisfaction, recovery (i.e. health status 7 to 10 days after attending the ED) (questionnaire/survey data)

6. Costs

Notes

Funding: Study authors funded by Lambeth Inner City Partnership and the King's Fund; SETRHA Primary Care Development provided additional funding for conducting the study.

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

High risk

Quote: "General practitioners and accident and emergency medical staff were considered as two groups, and each group was allocated two or three weekday sessions running from 1000 to 1300 and 1400 to 1700, one weekday evening session from 1800 to 2100, and one weekend daytime session for each week during the study period...

...weekly rosters stipulated a named doctor with responsibility for primary care patients for every three hour session" and "a random sample of sessions stratified by time of day and day of week was determined by using a table of random numbers.

...Hence, 8‐10 sessions were sampled each week for a total of 48 weeks. The sample of sessions allocated to accident and emergency staff was the same as those described in the accompanying paper."

See P.1, Col.2, Para.4.

Comment: Primary care sessions selected for inclusion in study were randomly selected using a random number table, however allocation of physicians to selected sessions was not random, but depended on physician availability and scheduling. Also, since nurses performing triage knew if a GP or an EP was seeing the 'non‐urgent' cases, this could affect what type of patients the physician in charge of providing care for the 'non‐urgent' patient group actually saw (i.e. more emergency‐type patients if an EP, and less so if a GP).

Allocation concealment (selection bias)

High risk

Quote: "Patients were unaware of their triage status or the grade and specialty of their doctor". See P.1, Col.5, Para.5

Comment: While patients were unaware of whether they were in the intervention (GP) or control (A&E staff) groups, this did not provide adequate allocation concealment; the type of physician providing care for each primary care session was open and not concealed.

Importantly, triage nurses were not blinded to the grade and speciality of the physician in charge for providing care for 'non‐urgent' patients, which could have affected the triage and therefore also what type of patients the physician actually saw (i.e. more emergency‐type patients if an EP, and less so if a GP).

Incomplete outcome data (attrition bias)
All outcomes

Low risk

Quote: "Not all records were complete" See P.2, Col.2, Para.2

Comment: Unclear whether missing data was predominantly from control or intervention group, or approximately equal across groups. Given binary outcomes and large samples, proportion of missing data probably less than effect size and low risk of bias.

Selective reporting (reporting bias)

Low risk

All outcomes mentioned in methods section were reported.

Other bias

High risk

Quote: "General practitioners worked sessions of only three hours in accident and emergency, compared with senior house officers' and registrars' shifts of up to 11 hours. Duration of shift may affect attitudes to patient care and influence the threshold of initiating referral or investigation." See P.4, Col.2, Para.1

Comments: General practitioners and EPs did not work equal numbers of hours in ED; this imbalance in experience and numbers of patients seen between providers could bias results.

Baseline outcome measures similar

Unclear risk

No baseline measure of outcome reported.

Baseline (provider) characteristics similar

Unclear risk

Quote: in recruiting GPs, "preference was given, firstly to those who had recently completed training (that is, general practitioners registered for similar numbers of years to the accident and emergency doctors) and, secondly, to those with flexible hours of availability". See P.1, Col.2, Para.3

Comment: This does not tell us what the actual provider characteristics were, only what was aimed for in the recruitment process. Also, no data are presented.

Baseline (patient) characteristics similar

High risk

Quote: "Two variables ‐ age and an injury related diagnosis ‐ were found to vary significantly with type of doctor seen. In addition, other variables (such as diagnosis of a mental disorder or a disease of the skin) varied significantly but had small effect sizes." See P.3, Col.2, Para.4, and Table VI.

Knowledge of allocated intervention adequate (Process variables)

Low risk

Unclear if outcomes were assessed blindly, but process variables (laboratory and X‐ray investigations, prescriptions, referrals, admissions) were objective.

Referrals were defined in the primary author's PhD thesis as outpatient, on‐call team and hospital admissions were all counted as referrals.

Knowledge of allocated interventions adequate (Patient satisfaction, health status)

Unclear risk

Questionnaires were administered by standardised telephone interview or post within 7 to 10 days of patients' index visit:

"We interviewed the patients again 7‐10 days later by telephone (or sent them a postal questionnaire if they lacked a telephone) about their satisfaction with their assessment and treatment in the department, the extent of their recovery, and the health care they required after attending the department. Responses to questions of satisfaction were recorded on five point Likert scales, ranging from very satisfied to very dissatisfied." See P.1, Col.2, Para.3 (Dale 1996).

Comment: Self reported data and unvalidated questionnaire (as per Dale thesis, no validated questionnaires were available at time of study). Unclear if interviewer was blinded

Blinding of participants and personnel (performance bias)

Low risk

Quote:

"Neither the general practitioners nor the accident and emergency doctors or nurses were informed about the study objectives or whether any particular session was part of the study sample." See P.1, Col.2, Para.4

"Patients were unaware of their triage status or the grade and speciality of their doctor." See P.427, Col.2, Para.5

Comments: All personnel (GPs, EPs, and nurses) were blinded to the study objectives and whether any particular session was part of the study sample, and the patients were unaware which type of doctor they were seen by.

Blinding of outcome assessment (detection bias) (Process variables)

Low risk

Quote: "All doctors...were asked to complete a consultation record form for each patient seen...Doctors remained blind to how data from these forms would be analysed." See P.2, Col.1, Para.3

Comments: Outcomes were objective, and physicians were unaware of what data were being collected for the study. It is unclear if researchers knew which physician saw patients.

Blinding of outcome assessment (detection bias) (Patient satisfaction, health status).

Unclear risk

Unclear if outcome assessors for patient satisfaction and health status were blinded

Adequately protected against contamination

High risk

Quote:

"Although the intention was that all primary care patients would be treated by the allocated doctor, this did not always occur. Firstly, at times when the primary care workload was excessive, other doctors were directed by the nurse performing triage to treat primary care patients to prevent unacceptably long waiting periods from occurring; secondly, registrars in particular were often interrupted from completing primary care sessions by departmental circumstances (such as responding to patients with urgent or life threatening needs or providing advice or supervision to senior house officers). Hence patients were sometimes attended by a non‐allocated doctor, both during sessions originally allocated to a general practitioner and during those allocated to another member of accident and emergency staff." See P.2, Col.1, Para.2

"Since this breakdown of randomisation was not always clearly documented, data for all recorded primary care consultations occurring during the selected sessions were included in the sample, and data on patients were regrouped according to the type of doctor actually seen. The loss of randomisation was allowed for by including confounding factors in the analysis of the data." See P.2, Col.1, Para.2

Methods

Design: non‐randomised trial

Time: March 1996 to September 1996

Duration: 7 months

Triage: patients categorised by receptionists with no formal training into "urgent" and "non‐urgent"

Data collection: Process data were collected from a review of written patient records.

Participants

Intervention group: N = 3 GPs

Control group: N = 8 EPs (1 consultant, 2 registrars, 5 senior house officers)

Provider characteristics: none reported

Patients: all "non‐urgent" and non‐ambulance patients attending the ED; ambulance patients were excluded

Total number of patients: N = 1878; intervention group: n = 771 patients seen by GPs; control group: n = 1107 patients seen by EPs

Patient characteristics: data no longer available

Setting:

Hospital: one, James Connolly Memorial Hospital

Country: Ireland

Hospital characteristics (1996 figures):

Beds: 336, small district hospital, urban/rural mix

Yearly attendance: 25,047

Yearly re‐attendance: 8213

Interventions

Intervention: sessional GPs providing care for non‐urgent patients in the ED

Control: regularly scheduled EPs providing care for non‐urgent patients in the ED (when GP present at the ED)

Patients referred by GPs included.

Conducted March to September 1996 (7 months). This study was designed by the same author‐group as Murphy 1996. 3 GPs were hired by the hospital to work on a sessional basis. The frequency and duration of GP sessions in the ED were not reported. As in the Murphy 1996 study, non‐urgent patients were allocated to either a GP or an EP in alternating (but not random or consecutive) order according to time of registration. Triage status did not factor into the order in which patients were seen, as only two triage categories were used: "urgent" and "non‐urgent". As in Murphy 1996, the control group comprised non‐urgent patients seen by EPs when a GP was on‐site.

Outcomes

1. Investigations: blood, X‐ray, any

2. Referrals

3. Prescriptions

4. Admissions

Notes

Funding: not reported

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

High risk

Quote: Allocation of patients "to either GP or A&E staff was the same as our previous study (Murphy 1996) and was performed according to time of registration." See P.1, Col.2, Para.5.

Comment: Sequence generation was non‐random; patients were seen in temporal order, and allocation to provider was not necessarily consecutive, depending on the length of previous consultations.

Allocation concealment (selection bias)

High risk

Quote: "An unstructured receptionist‐based triage system divides all non‐ambulance patients into two categories: 'urgent' and 'non‐urgent'." See P.1, Col.2, Para.3.

Comment: Patient allocation occurred as individuals entered the study (by attending the ED). It is unclear how physician allocation to primary care sessions was performed.

It is not specified whether nurses performing triage were blinded; nurses' knowledge of whether a GP or an EP was working could have affected triage and the type of patients that physician working in primary care sessions saw (i.e. more emergency‐type patients if an EP, and less so if a GP).

Incomplete outcome data (attrition bias)
All outcomes

Unclear risk

Not specified in the paper

Selective reporting (reporting bias)

High risk

All outcomes mentioned in the text were reported in the results, however the study was designed and carried out by same author‐group as Murphy 1996, and fewer outcomes are reported without explanation.

Other bias

Unclear risk

It is probable that GPs and EPs did not work equal numbers of hours in the ED; this imbalance between providers in experience and numbers of patients seen could bias the results.

Baseline outcome measures similar

Unclear risk

No baseline measure of outcome reported.

Baseline (provider) characteristics similar

Unclear risk

No provider characteristics reported.

Baseline (patient) characteristics similar

Unclear risk

Quote: "There were no differences in age, sex, socio‐economic status, registration with a GP or type of presenting complaint between patients seen by a GP or usual A&E staff." See P.1, Col.2, Para.6.

Comment: No data on patient characteristics were reported, hence we cannot corroborate that the patient groups seen by GPs or EPs were comparable in terms of duration of complaints, diagnoses, etc.

Knowledge of allocated intervention adequate (Process variables)

Low risk

The outcomes are objective.

Knowledge of allocated interventions adequate (Patient satisfaction, health status)

Unclear risk

Not specified in the paper

Blinding of participants and personnel (performance bias)

Unclear risk

Not specified in the paper

Blinding of outcome assessment (detection bias) (Process variables)

Unclear risk

Unclear if outcomes were assessed blindly, but process variables (laboratory and X‐ray investigations, prescriptions, admissions) were objective.

A definition of what constituted referrals in the study was not provided; if only some types of referrals (e.g. to on‐call physicians) were counted, this would not objectively account for the total referrals made (e.g. to non‐physician health professionals) by both intervention and control groups.

Blinding of outcome assessment (detection bias) (Patient satisfaction, health status).

Unclear risk

Not specified in the paper

Adequately protected against contamination

High risk

Quote: "Study enrolment only occurred when both GPs and usual A&E staff were on duty together." See P.1, Col.2, Para.5.

Comments: General practitioners and EPs worked simultaneously in primary care sessions, and overlap and contamination between groups was possible.

Methods

Design: pragmatic randomised trial

Time: first participant enrolled February 2014

Duration: not described

Triage: participants triaged by trained nurses using the Australasian Triage Scale

Data collection: baseline data collected from all consenting participants during enrolment. Pain score reduction reported by the participant, all other outcomes collected from the ED patient information system and electronic health record.

Participants

Intervention group: N = 9 emergency NPs

Control group: N = 17 emergency medicine registrars

Years of postgraduate training (minimum): 3 years

Patients: all patients presenting to the ED with "pain" and allocated to the "fast‐track" zone

Total number of patients: intervention: 130; control: 128

Patient characteristics:

Sex: intervention: 47% female; control: 39% female

Age (median): intervention: 33 years; control: 30 years

Pain score (median): intervention: 5; control: 5

Setting:

Hospital: one, adult tertiary ED

Country: Australia

ED characteristics (2013 figures):

Major urban teaching hospital

Yearly attendance: 65,000

Interventions

Intervention: People presenting with pain, who were triaged to fast‐track area (Australasian Triage Scale 2 to 5), were randomly assigned to receive either standard ED medical care or emergency NP care.

Control: Care was provided by medical officers with assistance from registered nurses, if required.

Outcomes

Primary outcomes: pain score reduction and time to analgesia

Secondary outcomes: waiting time, number of patients who did not wait, length of stay in ED, re‐presentations with 48 hours

Integrity of the intervention measured through clinicians' use of evidence‐based guidelines for management of knee, ankle, and burns injury. (Outcomes as per the published protocol.)

Notes

Funding: National Health and Medical Research Council postgraduate scholarship through Queensland University of Technology, Australia (principal investigator)

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Low risk

"Randomization was performed with an allocation sequence of four and generated by computer random number generator and then transcribed into opaque sequentially numbered sealed envelopes" (p.775)

Allocation concealment (selection bias)

Unclear risk

"Each envelope contained a card with the allocation group recorded and treatment pack. Allocation adhered strictly to the generated sequence and was maintained. Both participants and treating staff were aware of treatment allocation." (p.775)

Incomplete outcome data (attrition bias)
All outcomes

Low risk

No participants lost to follow‐up; 2 participants allocated to intervention excluded from analysis as consent forms not signed (0.02%).

Selective reporting (reporting bias)

Low risk

All outcomes specified in the protocol published (primary and secondary outcomes reported separately).

Other bias

Low risk

No other risk detected.

Baseline outcome measures similar

Low risk

Clinical research assistants used an examination cubicle to recruit and consent patients and collect baseline demographic information.

Baseline (provider) characteristics similar

Unclear risk

Not described

Baseline (patient) characteristics similar

Low risk

Little or no differences between groups (Table 1)

Knowledge of allocated intervention adequate (Process variables)

Low risk

Most outcomes are objective.

Knowledge of allocated interventions adequate (Patient satisfaction, health status)

Low risk

Not applicable, not outcomes for this study

Blinding of participants and personnel (performance bias)

Unclear risk

Not enough information to ascertain risk of bias

Blinding of outcome assessment (detection bias) (Process variables)

Low risk

Primary investigators were blinded to treatment allocation for data analyses. Most outcomes were objective.

Blinding of outcome assessment (detection bias) (Patient satisfaction, health status).

Low risk

Not applicable, not outcomes for this study

Adequately protected against contamination

Unclear risk

Not enough information to ascertain risk of bias. Both medical officers and NPs worked in fast‐track area at overlapping times.

Methods

Design: non‐randomised study

Time: August 1993 to October 1994

Duration: 15 months

Triage: Patients triaged by trained nurses based on physiological criteria as (1) life‐threatening, (2) urgent, (3) semi‐urgent, or (4) delay acceptable.

Data collection:

Process information (investigations, referrals, prescriptions, etc.) was collected from hospital records.

The numbers of patients re‐attending the ED within 1 month of the index visit was determined using the hospital's mainframe computer.

Patient satisfaction was assessed immediately by a blinded interviewer using the consultation satisfaction questionnaire. Health status was determined 1 month after the initial consultation by means of a simple questionnaire (4 questions) completed by telephone or letter.

Marginal (materials and disposables) and total (marginal plus all staff) costs were determined in conjunction with the hospital's finance department and X‐ray and laboratory staff. Costs were calculated for the following: full blood counts; measurements of blood urea and plasma electrolyte concentrations, plasma glucose concentration, and serum amylase activity; sequential multiple analysis with computer (SMAC); and chest, limb, skull, spine, and abdominal radiographs. Based on the hospital admission profile, an estimate of the average cost per admission was also obtained.

Participants

Intervention group: N = 5 GPs

Age (median): 32 years

Years since registration (median): 7 years

Control group: N = 13 EPs (1 consultant, 2 registrars, 10 senior house officers)

Age (median): 26 years

Patients: new ED attendees triaged as "semi‐urgent" or "delay acceptable"

Total number of patients: N = 4684; intervention group: n = 2303 patients seen by GPs; control group: n = 2381 patients seen by EPs

Patient characteristics:

Sex: 41.4% female

Age: median 28 to 34 years

Years since registration (median): 6 months

Duration of complaints: 44% problems > 24 hours; 92.6% registered with GPs (unclear how many saw GP prior to attending)

Most common diagnoses: musculoskeletal (50.9%), skin complaints (19.0%), and neurological (8.8%) 

Setting:

Hospital: one, St James' Hospital

Country: Ireland

Hospital characteristics (1992 figures):

Beds: 490, catchment 219,300 people

Major teaching hospital

Yearly attendance: 40,159

Yearly re‐attendance: 7589

Interventions

Intervention: sessional GPs providing care for non‐urgent patients at hospital ED

Control: regularly scheduled EPs providing care for non‐urgent patients when GP present in department

Patients referred by GPs (20%) were excluded.

The study took place between August 1993 and October 1994 (15 months). 3 GPs were hired to work two 4‐hour shifts each week alongside EPs. During these primary care shifts, non‐urgent patients were allocated to either the GP or EP according to registration time. The control group comprised non‐urgent patients seen by EPs when a GP was on‐site. The allocation of patients was predictable but not necessarily consecutive, as the order in which patients were allocated depended on the length of consultations. In addition to temporal ordering, patients were also ordered by triage category: triage category 3 patients were seen prior to category 4.

The GPs and EPs in this study had access to all of the same ED facilities, and patients were unaware what type of physician was treating them.

Outcomes

1. Investigations: blood, X‐ray, any

2. Referrals

3. Prescription

4. Disposal to: community, hospital, outpatient clinic

5. Admissions

6. Re‐attendance within 1 month; 2 years

7. Patient satisfaction

8. Health status

Notes

Funding: Department of Health through the General Practice Unit of the Eastern Health Board

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

High risk

Quote: "Randomisation of patients to the general practitioner or accident and emergency staff depended on time of registration. Once patients were registered their charts were divided according to triage category on to four separate shelves and then placed in line by strict temporal order. Doctors took the first chart on the triage 3 shelf and continued doing so until the shelf was empty. They then moved to the triage 4 shelf." See P.2, Col.1, Para.3

Comment: Sequence generation was non‐random; patients were seen in temporal order, and allocation to provider was not necessarily consecutive, depending on the length of previous consultations. Although a research nurse was employed to ensure adherence to the temporal order, this open allocation method could be problematic if the triage information recorded on chart influences physician's choice to accept or reject a patient (by waiting for the other physician to take the top chart). For example, GPs investigated fewer semi‐urgent (triage 3) and more delay‐acceptable (triage 4) patients than EPs. See P.3, Table 1:

• GPs saw 1516 and EPs saw 1837 triage 3 patients.

• GPs saw 787 and EPs saw 544 triage 4 patients.

Allocation concealment (selection bias)

Unclear risk

Quote: "General practitioners...were dressed similarly to the usual staff and patients were unaware that they were being seen by a general practitioner" See P.2, Col.1, Para.2‐3

Comment: Patient allocation occurred as individuals entered the study (by attending the ED) and was carried out by a study researcher and enforced by the triage nursing team. It is unclear whether the same person conducted both steps of the randomisation process. Physicians were not blinded to the triage category of the patients being seen, however patients were probably unaware of the type of physician treating them.

It is unclear how physician allocation to primary care sessions was performed.

Incomplete outcome data (attrition bias)
All outcomes

Low risk

Quote: "The hospital's computer could not locate 83 (2%) of the 4684 patients enrolled in the study. Thirty three had been seen by the general practitioners and fifty by the usual accident and emergency staff." See P.4, Col.2, Para.4

Comment: There were similar numbers of missing records across the 2 groups, and a relatively small portion of data was missing, hence probably low risk of bias.

Selective reporting (reporting bias)

Low risk

All outcomes mentioned in the text were reported in the results.

Other bias

Unclear risk

Quote: Each GP "worked two four hour sessions a week, managing non‐emergency patients". See P.2, Col.1, Para.2.

General practitioners and EPs did not work equal numbers of hours in the ED; this imbalance between providers in experience and numbers of patients seen could bias the results.

Baseline outcome measures similar

Unclear risk

No baseline measure of outcome reported.

Baseline (provider) characteristics similar

High risk

The median age and time since registration were not equal between GPs and EPs. The median age of the 5 GPs employed during the project was 32 years, compared with 26 years for EPs. Similarly, the median time since full registration was 7 years for GPs and 6 months for EPs. See P.3, Col.2, Para.3

This difference in experience between the groups could bias the study outcomes.

Baseline (patient) characteristics similar

High risk

Quote:

"There were significant differences (in presenting complaints)....between (triage 3) patients seen by the general practitioners and those seen by the usual accident and emergency staff". See P.4, Table 3.

"There were no differences between triage 4 patients seen by general practitioners and those seen by the usual accident and emergency staff". See P.3, Col.2, Para.5

Comment: High risk of bias because patient diagnoses in control and intervention groups were not equal.

Knowledge of allocated intervention adequate (Process variables)

Low risk

Unclear if outcomes were assessed blindly, but process variables (laboratory and X‐ray investigations, prescriptions, referrals, admissions) were objective.

(Referrals were "when a second doctor was formally requested to review a patient and did so". P.2, Col.2, Para.2)

Knowledge of allocated interventions adequate (Patient satisfaction, health status)

Unclear risk

Quote:

"Patient satisfaction was assessed immediately by a blinded interviewer using the consultation satisfaction questionnaire." See P.2, Col.2, Para.4

"Health status was determined after one month by means of a simple questionnaire completed by telephone or letter".

Patient satisfaction was assessed blindly. Unclear if health status was assessed blindly. See P.2, Col.2, Para.4

Comment: Self reported data, and unclear if questionnaires were validated or if health status was assessed blindly.

Blinding of participants and personnel (performance bias)

Unclear risk

Quote: "General practitioners...had access to the same facilities as the usual medical staff. They were dressed similarly to the usual staff and patients were unaware that they were being seen by a general practitioner".

Comment: Patients were unaware of which type of physician they were seeing.

It is unclear whether medical practitioners were aware of the study objectives. Knowledge of study objectives may have affected performance (e.g. consciously choosing to order fewer investigations or make more referrals to the community rather than to a second doctor).

Blinding of outcome assessment (detection bias) (Process variables)

Unclear risk

It is unclear if outcomes were assessed blindly, but most process measures were objective items such as the number of investigations ordered, prescriptions given, and admissions made.

Referrals were only counted in the study if "a second doctor was formally requested to review a patient and did so" (See P.2, Col.2, Para.1). Hence any referrals to community or non‐physician healthcare providers (e.g. community nurses, social workers, mental health professionals) were excluded, and detection bias could have been introduced if physicians were aware of the study definition or outcome; we therefore judged the risk of bias as unclear.

Blinding of outcome assessment (detection bias) (Patient satisfaction, health status).

Unclear risk

Quotes:

"Patient satisfaction was assessed immediately by a blinded interviewer using the consultation satisfaction questionnaire." See P.2, Col.2, Para.4

"Health status was determined after one month by means of a simple questionnaire completed by telephone or letter". See P.2, Col.2, Para.4

Comment: Satisfaction assessment was blinded, but it is unclear if health status assessments were blinded.

Adequately protected against contamination

Unclear risk

Unclear. General practitioners and EPs worked simultaneously in primary care sessions, and overlap and contamination between groups was possible. See P.2, Col.2, Para.2, 4‐6