Gibney 1999.

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	Investigations: blood, X‐ray, any Referrals Prescriptions 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.