Impact of employing primary healthcare professionals in emergency department triage on patient flow outcomes: a systematic review and meta-analysis.

OBJECTIVES: To identify, critically appraise and summarise evidence on the impact of employing primary healthcare professionals (PHCPs: family physicians/general practitioners (GPs), nurse practitioners (NP) and nurses with increased authority) in the emergency department (ED) triage, on patient flow outcomes.
METHODS: We searched Medline (Ovid), EMBASE (Ovid), Cochrane Library (Wiley) and CINAHL (EBSCO) (inception to January 2020). Our primary outcome was the time to provider initial assessment (PIA). Secondary outcomes included time to triage, proportion of patients leaving without being seen (LWBS), length of stay (ED LOS), proportion of patients leaving against medical advice (LAMA), number of repeat ED visits and patient satisfaction. Two independent reviewers selected studies, extracted data and assessed study quality using the National Institute for Health and Care Excellence quality assessment tool.
RESULTS: From 23 973 records, 40 comparative studies including 10 randomised controlled trials (RCTs) and 13 pre-post studies were included. PHCP interventions were led by NP (n=14), GP (n=3) or nurses with increased authority (n=23) at triage. In all studies, PHCP-led intervention effectiveness was compared with the traditional nurse-led triage model. Median duration of the interventions was 6 months. Study quality was generally low (confounding bias); 7 RCTs were classified as moderate quality. Most studies reported that PHCP-led triage interventions decreased the PIA (13/14), ED LOS (29/30), proportion of patients LWBS (8/10), time to triage (3/3) and repeat ED visits (5/6), and increased the patient satisfaction (8/10). The proportion of patients LAMA did not differ between groups (3/3). Evidence from RCTs (n=8) as well as other study designs showed a significant decrease in ED LOS favouring the PHCP-led interventions.
CONCLUSIONS: Overall, PHCP-led triage interventions improved ED patient flow metrics. There was a significant decrease in ED LOS irrespective of the study design, favouring the PHCP-led interventions. Evidence from well-designed high-quality RCTs is required prior to widespread implementation. PROSPERO REGISTRATION NUMBER: CRD42020148053. © Author(s) (or their employer(s)) 2022. Re-use permitted under CC BY. Published by BMJ.

The main strength of our systematic review is that our study team engaged and collaborated with patient and public partners during the design, conduct and dissemination phases of the study by following the criteria identified for patient-oriented research which emphasises the active and meaningful engagement of patients as research partners.

This systematic review was conducted using the rigorous Cochrane systematic review methodology and used an a priori registered protocol.

A main limitation of this systematic review is that we did not include non-English language publications.

Introduction

Healthcare systems worldwide experience emergency departments (ED) overcrowding,1–5 which impacts the timely delivery of healthcare,6 7 patient and provider dissatisfaction,8 and other adverse outcomes.9 ED overcrowding is a complex phenomenon and is associated with input (increased patient volume), throughput (ED boarding), and especially output (lack of hospital beds) factors, as well as system-wide influences.10 A large volume of lower acuity patients presenting to ED leads to demand-capacity mismatch and entry block (eg, delays in ED assessment).10 11

Lower acuity ED patients generally include patients: (1) having low acuity triage codes; (2) being discharged quickly or (3) being seen by an alternative primary healthcare provider.12 These alternative primary healthcare providers are typically physicians (family physicians/general practitioners (GPs)), nurse practitioners (NP), nurses with increased authority, or physician assistants who are legally authorised to provide or coordinate healthcare to patients.13 Studies have reported that 8%–62% of all ED presentations are lower acuity.14–16 With ED visits increasing by 20% each year, along with a decrease in operational EDs,17 lower acuity visits may lead to unnecessary diagnostic testing, greater healthcare spending, lost opportunity for continuity of care with primary care physicians, suboptimal care due to hurried management and prolonged ED length of stay (LOS).12 14 18 19 Increased demand for ED services also leads to increased ED wait times and patients choosing to leave ED without being-seen, thus potentially compromising patient safety.17

Worldwide, there is growing interest in interventions and strategies, either to discourage lower acuity ED visits or to reduce the impact of lower acuity visits in the ED by improving patient flow. Studies have investigated the impact of interventions such as public and patient education,14 financial disincentives (higher copayments for lower acuity ED visits),20 increasing after-hours primary care,21 patient redirection to non-ED care alternatives22 and advanced access23 24 to discourage unnecessary ED utilisation. Since EDs have no control over the volume of presenting patients and ED presentations continue to be on the rise,14 17 recommendations have been made to focus on strategies to improve patient flow within the ED.25 Studies have investigated various strategies to improve ED patient flow, including triage related interventions.8 26

While the precise role of the NPs, GPs or nurses given increased authority (all referred to as primary healthcare professionals (PHCPs)) in an ED is unclear, they may provide potential benefits to improve ED times and outcomes. Studies have reported the following roles of the PHCPs at ED triage: (1) GP either triaging (seeing and treating, streaming) or supervising triage27–29; (2) NP either alone or working alongside a triage nurse (ordering investigations, streaming, seeing and treating, or assessing patients and discharging/redirecting)17–19 25 30–39; (3) Triage nurse with increased authority given extra capacities outside of their usual scope of practice to order investigations for patients before streaming to the ED MD.40–62 Although, many primary research studies have investigated the impact of PHCPs19 32 36 37 40 63 64 at triage on ED patient flow, to the best of our knowledge, there are no systematic reviews that have summarised evidence from these studies.

The main objective of our systematic review was to identify, critically appraise and summarise evidence on the effectiveness of employing PHCPs at ED triage to improve ED patient flow metrics.

Methods

Using an a priori systematic review protocol developed in collaboration with patient partners, we conducted this review according to guidelines enumerated in the Methodological Expectations of Cochrane Intervention Reviews. Our systematic review is reported using the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guideline.65

Eligibility criteria

We included comparative studies (only English language) of any ED triage intervention that involved a PHCP and was designed to improve ED (adult and paediatric) patient flow metrics. We excluded primary studies involving exclusively emergency physicians (ED MD), such as the triage liaison physician (TLP).26 The primary outcome was the time to provider initial assessment (PIA: time from ED arrival to the time when the patient is first assessed by an ED provider (ED MD, NP or a GP in the ED)). Secondary outcomes were ED LOS (time from ED arrival to disposition), the proportion of patients who left without being seen (LWBS), proportion of patients leaving against medical advice (LAMA), time to triage, number of repeat ED visits and patient satisfaction. The outcome measures were selected a priori in collaboration with the patient partners in the research team. We have reported a more detailed list of the inclusion and exclusion criteria in online supplemental table 1.

Literature search methods for identifying relevant citations

In conjunction with a health librarian (TR), we designed a search strategy for Medline (Ovid) to identify literature relevant to the objective (from inception to June 2018, and later updated in January 2020). Since most of the potentially relevant studies would be performed in the USA, Europe and Commonwealth countries, search results were limited to English language publications. Our Medline search was peer-reviewed by a second librarian (JJ),66 principal investigators (MJ and AA-S) and patient partners (MH and TBe). Once finalised, the Medline search strategy (online supplemental table 2) was adapted for replication in the following databases: EMBASE (Ovid), Cochrane Library (Wiley) and CINAHL (EBSCO). An experienced librarian (NA-Y) searched the included databases up to January 2020. The bibliographic search was supplemented with searching the grey literature (ie, difficult to locate unpublished studies) as listed in online supplemental table 3. We also searched the reference lists of all the included publications for additional relevant studies. We used EndNote (V.X7, Thomson Reuters) for reference management.

Selection of sources of evidence

Two reviewers (RA and (LC or NA-Y)) independently screened the titles and abstracts, and full texts of relevant citations using pilot tested screening forms. Any disagreement on inclusion was resolved through consensus or third party (MJ) adjudication.

Data extraction, data analysis and quality assessment

Standardised data extraction forms were developed to record data from each of included studies after pilot testing. At least two review authors independently extracted baseline characteristics (RA, LC and NA-Y), outcome data (RA, LC) and assessed methodological quality (MJ, RA and LC) on these studies. Disagreements among reviewers were resolved through consensus or third-party adjudication (MJ or AA-S). A meta-analysis of mean differences (MD) in ED times with 95% CIs was planned a priori to derive pooled summary estimates. Heterogeneity among included studies was quantified and tested using I2 (I2 statistic and χ2 statistic, respectively. An I2 value >50% was considered high heterogeneity; we made an a priori methodological decision that heterogeneity indicated by I2 >50% was too high to justify data pooling to generate a summary measure. For studies that did not report any measure of variance we imputed the largest SE from among the included studies. In the event that meta-analysis was not possible, the effect estimates (MD and SE) from included studies reporting data for the primary outcome and ED LOS were depicted in the form of a forest plot for various a priori subgroups (study designs or PHCP interventions). In these cases, where appropriate, the median of the primary study outcome was reported as the average measure.

We assessed the included studies using the National Institute for Health and Care Excellence (NICE) quality appraisal tool for quantitative studies of intervention67 as it can be used for multiple study designs. A detailed description is reported under online supplemental appendix methods.

Patient and public involvement

We collaborated with a diverse group of 13 patient partners (self-identified as indigenous, immigrant, white and/or living with disability) during the design phase and the conduct phase of this project, to refine the review question, refine the inclusion criteria, and select patient-important outcomes. Two (TBe and MH) of these patient partners collaborated and supported our grant application to obtain funding for this project. During the conduct phase of this systematic review three patient partners helped refine the search strategy (by identifying missing search terms and suggesting additional search terms in the preliminary search strategy), review to confirm included studies, and in knowledge dissemination (copresented abstract at a conference and co-authoring the manuscript). We have reported the patient partner involvement in this systematic review according to GRIPP2 checklist (short form).68

Results

We identified 23 973 relevant citations from database search, of which 40 met the inclusion criteria17–19 25 27–62 (44 study reports). The study selection process is reported using the PRISMA study flow chart (figure 1).

Figure 1

PRISMA study flow diagram. PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses.

Study characteristics

Included studies were full-length journal articles17–19 25 27–44 46–48 50 51 55 57 58 60 (n=31; 77.5%), abstracts (n=8)45 49 52–54 56 59 62 or thesis (n=1)61 published from 1993 to 2020. More than half were conducted in North America17 18 25 30 31 33–35 38–41 48 50 51 53 54 57 59 61 62 (n=21; 52.5%), and the rest were conducted in Europe19 27–29 36 37 42 44 47 52 56 (n=11; 27.5%), Asia55 58 60 (n=3; 7.5%), Australia32 43 (n=2; 5%), Middle East45 46 (n=2; 5%) or the location was not reported49 (n=1; 2.5%) (table 1). Most studies used a pre–post intervention design17 18 25 30–34 36–39 61 (n=13; 32.5%) and the remaining were characterised as randomised controlled trials (RCT)41 42 44 45 50 51 55–57 59 (n=10; 25%), observational retrospective cohort studies19 28 40 47 53 54 60 62 (n=8, 20%), controlled before and after studies (CBA)27 35 46 52 (n=4; 10%), quasi-randomised trials43 (n=1; 2.5%) observational prospective cohort studies48 49 58 (n=3; 7.5%) or cross-sectional observational studies29 (n=1; 2.5%). The median duration of intervention reported among included studies was 6 months (range: 2.5 days to 17 months). Studies were mostly conducted in urban EDs18 19 25 28 30–33 35–38 40–46 48–51 54 57–59 62 (n=28; 70%), one (2.5%) was conducted in a rural ED and two27 29 (5%) were conducted in a combination of urban and rural facilities; nine17 34 47 52 53 55 56 60 61 (22.5%) studies did not report their setting. We classified the EDs reported in the included studies into paediatric EDs (age <18)18 45 50 53 57 60 (n=6; 15%), mixed EDs seeing both adults and children17 32 48 52 55 58 (n=6; 15%) and adult only EDs30 33 41 43 46 (n=5; 12.5%), depending on the age of the population that the EDs served. More than half of the studies (n=23, 57.5%) did not specifically report the age of the population that their EDs served.19 25 27–29 31 34–40 42 44 47 49 51 54 56 59 61 62 The majority (n=15, 38%) of included studies19 27 28 30 32 33 36–39 43 44 46 55 60 reported enrolling only patients with triage category 4–5 (additional details reported in online supplemental results 1).

Table 1

Characteristics of included studies

Study ID(first author, year)	Country; urban/rural ED; adult/paediatric/mixed ED	Study design	Type ofPHCP	Intervention	Duration of intervention (months)	Study quality
Adamet al, 201445	Saudi Arabia; urban ED; paediatric ED	RCT	Nurse	Nurse triage-plus	0.5	Moderate
Al Abri, 202046	Oman; urban ED; adult ED	CBA	Nurse	Nurse triage-plus	NR	Low
Al Kadhi, 201747	UK; NR; NR	RC study	Nurse	Nurse triage-plus	12	Low
Ashurst, 201448	USA; urban ED; mixed ED	PC study	Nurse	Nurse triage-plus	10	Low
Celona, 201830	USA; urban ED; adult ED	Pre–post	NP	NP Team triage	12	Low
Cheung, 200240	Canada; urban ED; NR	RC study	Nurse	Nurse triage-plus	NR	Low
Day, 201331	USA; urban ED; NR	Pre–post	NP	NP Team triage	1	Low
Demarco, 201049	NR; urban ED; NR	PC study	Nurse	Nurse triage-plus	1	Low
Dixon, 201450	Canada; urban ED; paediatric ED	RCT	Nurse	Nurse triage-plus	12	Low
Edwards, 201132	Australia; urban ED; mixed ED	Pre–post	NP	NP Team triage	17	Low
Fan, 200651	Canada; urban ED; NR	RCT	Nurse	Nurse triage-plus	3	Moderate
Fontanel, 201152	France; NR; mixed ED	CBA	Nurse	Nurse triage-plus	0.08	Low
Gardner, 201833	USA; urban ED; Adult ED	Pre–post	NP	NP Team triage	0.5	Low
Gaucher, 201053	Canada; NR; paediatric ED	RC study	Nurse	Nurse triage-plus	12	Low
Hackman, 201554	USA; urban ED; NR	RC study	Nurse	Nurse triage-plus	17	Low
Hayden, 201417	USA; NR; mixed ED	Pre–post	NP	NP Team triage	2	Low
Ho, 201855	China; NR; mixed ED	RCT	Nurse	Nurse triage-plus	NR	Moderate
Jobé, 201956	France; NR; NR	RCT	Nurse	Nurse triage-plus	NR	Low
Klassen, 199357	Canada; urban ED; paediatric ED	RCT	Nurse	Nurse triage-plus	12	Low
Kool, 200827	Netherlands; both; NR	CBA	GP	GP team triage	12	Low
Lee, 199658	China; urban ED; mixed ED	PC study	Nurse	Nurse triage-plus	3	Low
Lee, 201459	Canada, urban ED; NR	RCT	Nurse	Nurse triage-plus	12	Moderate
Lee, 201641	Canada; urban ED; adult ED	RCT	Nurse	Nurse triage-plus	12	Moderate
Li, 201860	China; NR; paediatric ED	RC study	Nurse	Nurse triage-plus	5	Low
Lijuan, 201761	USA; NR; NR	Pre–post	Nurse	Nurse triage-plus	6	Low
Lindley-Jones, 200042	England; urban ED; NR	RCT	Nurse	Nurse triage-plus	12	Moderate
Love, 201225	USA; urban ED; NR	Pre–post	NP	NP Team triage	0.5	Low
MacKenzie, 201534	USA; NR; NR	Pre–post	NP	NP Team triage	2	Low
Parris, 199743	Australia; urban ED; adult ED	Quasi-RCT	Nurse	Nurse triage-plus	6	Low
Pierce, 201635	USA; urban ED; NR	CBA	NP	NP Team triage	5.5	Low
Rogers, 200436	England; urban ED; NR	Pre–post	NP	NP Team triage	12	Low
Shrimpling, 200237	England; urban ED; NR	Pre–post	NP	NP Team triage	0.75	Low
Sikkenga, 201662	USA; urban ED; NR	RC study	Nurse	Nurse triage-plus	NR	Low
Thurston, 199644	England; urban ED; NR	RCT	Nurse	Nurse triage-plus	2	Moderate
Tsai, 201218	USA; urban ED; paediatric ED	Pre–post	NP	NP Team triage	NR	Low
Tucker, 201538	USA; urban ED; NR	Pre–post	NP	NP Team triage	6	Low
Uthman, 201819	England; urban ED; NR	RC study	NP	NP Team triage	12	Low
van den Bersselaar et al, 201828	Netherlands; urban ED; NR	RC study	GP	GP team triage	11	Low
van Gils-van Rooij, 201829	Netherlands; both; NR	CS study	GP	GP team triage	NA	Low
Zager, 201839	USA; rural ED; NR	Pre–post	NP	NP Team triage	4	Low

CBA, controlled before and after; CS study, cross-sectional observational study; ED, emergency department; GP, general practitioner; NA, not applicable; NP, nurse practitioner; NR, not reported; PHCP, primary healthcare provider; RC study, retrospective cohort study; RCT, randomised controlled trial.

The majority (82.5%) of included studies were of low methodological quality and the remaining seven (17.5%) included studies (RCTs)41 42 44 45 51 55 59 were of a moderate methodological quality (table 1 and online supplemental table 4).

We categorised the triage interventions involving PHCP reported by the included studies, in comparison to the traditional (nurse-led) triage model (figure 2), as follows: (1) GP team-triage27–29 (n=3, 7%): where GP was involved in the ED triage (triaging or supervising triage) either seeing and treating low-acuity patients or streaming moderate to high-acuity patients to the ED MD; (2) NP team-triage17–19 25 30–39 (n=14, 35%): where the NP was located at the ED triage area working alongside a triage nurse, either ordering investigations at triage before streaming to ED MD, seeing and treating low-acuity patients, directing low-acuity patients to a GP located within ED for treatment, or assessing patients and discharging/redirecting with a same day appointment with a GP at an adjoining GP clinic; (3) Nurse triage-plus40–62 (n=23, 58%): triage nurse with increased authority (extra capacities outside of their usual scope of practice) to order investigations for patients before streaming to the ED MD. The traditional ED care model with an ED nurse-led triage, followed by the ED MD assessment was considered standard of care and the comparator in all the included studies (figure 2).

Figure 2

Various models for PHCP involvement in triage of emergency department patients. ED, emergency department; GP, general practitioner; MD, mean difference; NP, nurse practitioner; PHCP, primary healthcare professional.

Provider initial assessment

Fourteen studies17 18 25 27 29 30 33–36 38 42 46 55 (35%) reported the effect of PHCP triage interventions on PIA in comparison to a traditional nurse-led triage. Using a forest plot, we depicted the effectiveness of the PHCP triage interventions on PIA subgrouped by study design (figure 3). Two RCTs42 55 (of moderate quality), reported a non-significant small decrease in PIA in the PHCP triage intervention (nurse triage-plus) group compared with the traditional nurse-led triage model (mean difference (MD) −0.36 min (95% CI −4.53 to 3.81); two studies; I2: 39%; p=0.20; moderate quality). Three CBA studies27 35 46 (low quality) reported a decrease in PIA in the PHCP triage intervention group (median (range) = −18 min (95% CI −2.3 to −31)).

Figure 3

Effectiveness of primary healthcare professional (PHCP) interventions on time to provider initial assessment (in minutes) subgrouped by study design. The horizontal black lines represent 95% CIs and the red dots in the middle represents point estimates (mean difference). CBA, controlled before and after; RCT, randomised controlled trial.

All eight pre–post studies17 18 25 30 33 34 36 38 (low quality) reported a significant decrease in PIA (median (range)=−24.65 min (95% CI −3 to −50)) in the PHCP triage intervention group, compared with the traditional nurse-led triage model. Exploration of heterogeneity among pre–post studies (I2: 100%) revealed four studies25 30 33 34 that contributed to all the observed heterogeneity; however, we were unable to identify specific reasons for heterogeneity. A sensitivity analysis without these four studies showed a significant mean decrease of PIA by 26 min favouring the PHCP triage intervention group (NP team triage). One cross-sectional observational study29 (low quality) failed to identify a difference in PIA in the PHCP intervention group (4.4 min). The results for PIA sub-grouped by various PHCP interventions is reported under online supplemental appendix results 2. We have depicted the effectiveness of each of the three models of PHCP triage interventions on PIA separately using a forest plot (online supplemental figure 1).

Three studies25 32 36 37 reported greater percentage of patients seen within benchmark times in the NP team triage intervention groups compared with the traditional nurse-led triage model (online supplemental figure 2). A fourth study32 reported that greater percentage of patients (all ATS categories) were seen within benchmark times in the NP team triage group compared with the traditional nurse-led triage group (data not shown).

ED LOS

ED LOS was reported by thirty studies (75%).17 18 29 31 33–36 38–41 43–52 54–60 62 Using a forest plot, we have depicted the effectiveness of the PHCP triage interventions on ED LOS sub-grouped by study design (figure 4). Eight RCTs41 44 45 50 51 55 57 59 (six42 45 46 52 56 60 of moderate quality and two50 57 of low quality), reported a significant decrease in ED LOS (MD −15.31 min (95% CI −18.35 to −12.27); eight studies; I2: 0%; p<0.00001) in the PHCP triage intervention (nurse triage-plus) group compared with the traditional nurse-led triage model. The CBA studies35 46 52 (low quality) reported a significant decrease in ED LOS (mean difference −63.17 min (95% CI −101.93 to −24.40); three studies; I2: 51%; p=0.001) in the PHCP triage intervention group (two nurse-triage plus and one NP team triage) compared with the traditional nurse-led triage model and the three retrospective cohorts40 54 60 (low quality) also reported a significant decrease in the ED LOS (MD −13.96 min (95% CI −19.31 to −8.61); three studies; I2: 37%; p<0.00001) in the PHCP triage intervention group (nurse triage-plus), compared with the traditional nurse-led triage model.

Figure 4

Effectiveness of primary healthcare professional (PHCP) interventions on ED Los (in minutes) subgrouped by study design. The horizontal black lines represent 95% CIs and the red dots in the middle represents point estimates (mean difference). ED, emergency department; RCT, randomised controlled trial.

Among eight pre–post studies17 18 31 33 34 36 38 39 (low quality), all reported a decrease (five were significant) in ED LOS (median (range) = −28 min (95% CI −16.65 to −102) favouring the PHCP triage intervention group (NP team triage). Exploration of heterogeneity among pre–post studies (I2: 99%) revealed four studies18 31 33 39 that contributed to all the observed heterogeneity; however, we were unable to identify specific reasons for heterogeneity. A sensitivity analysis without these four studies showed a significant mean decrease of ED LOS by 17 min favouring the PHCP triage intervention group (NP team triage). One quasi-RCT,43 and one cross-sectional observational study29 reported no significant differences in ED LOS between comparison groups. Among the three prospective observational cohorts,48 49 58 one reported significant decrease in ED LOS whereas other two reported a non-significant decrease in ED LOS favouring PHCP intervention group. The ED LOS subgrouped by various PHCP interventions is reported under online supplemental appendix results 3. We have depicted the effectiveness of each of the three models of PHCP triage interventions on ED LOS separately, using a forest plot (online supplemental figure 3).

Other outcomes

Ten studies17–19 25 30 33 34 38 60 61 reported data for percentage of patients LWBS (table 2). Eight studies reported a reduction in percentage of patient LWBS in the NP team triage intervention group, except one17 (five18 19 25 33 34 reported statistically significant decrease). Two60 61 studies reported a non-significant decrease in percentage of patients LWBS in nurse triage-plus intervention group. The median effect of all estimates is a reduction in LWBS of −2.31% (IQR: −0.39 to –3.77). Three pre–post studies17 34 38 reported the effect of NP team triage intervention on percentage of patients discharged as LAMA (table 3). One study showed a non-significant decrease favouring the intervention group and another two showed a non-significant increase in the percentage of patients LAMA.

Table 2

Leave without being seen outcome data reported by included studies

Study ID(first author, year)	Triage intervention	Study design	Intervention (%)	Comparator (%)	Percentage difference	Reported statistical significance
Celona, 201830	NP team triage	Pre–post	4.7	3.3	1.4	NR
Love, 201225	NP team triage	Pre–post	0.93	3.39	−2.46	Significant
MacKenzie 201534	NP team triage	Pre–post	0.7333	2.96	−2.2267	Significant
Gardner, 201733	NP team triage	Pre–post	2.2	4.6	−2.4	Significant
Hayden, 201417	NP team triage	Pre–post	5.8	5.4	0.4	NS
Tsai, 201218	NP team triage	Pre–post	3	9.7	−6.7	Significant
Tucker, 201538	NP team triage	Pre–post	1.3	5.07	−3.77	NR
Uthman, 201819	NP team triage	Retrospective cohort	2.2	3.9	−1.7	Significant
Li, 201860	Nurse triage-plus	Retrospective cohort	0.7	6.9	−6.2	NS
Lijuan, 201761	Nurse triage-plus	Pre–post	7.13	7.52	−0.39	NS

NP, nurse practitioner; NR, not reported; NS, not significant.

Table 3

Leave against medical advice outcome data reported by included studies

Study ID(first author, year)	Triage intervention	Study design	Intervention (%)	Comparator (%)	Percentage difference	Reported statistical significance
MacKenzie, 201534	NP team triage	Pre–post	0.22	0.33	−0.11	NS
Tucker, 201538	NP team triage	Pre–post	1.41	1.29	0.12	NS
Hayden, 201417	NP team triage	Pre–post	1.4	0.06	1.34	NR

NP, nurse practitioner; NR, not reported; NS, not significant.

Six studies27 28 31 38 39 53 reported the impact of PHCP interventions on the number of repeat ED visits, and the majority of them reported a decrease in the number of repeat ED visits after PHCP intervention (online supplemental appendix results 4). Ten studies17 27 33 38 41 48 49 51 59 61 reported the effect of PHCP intervention on patient satisfaction, and the majority of them reported an increase in patient satisfaction (online supplemental appendix results 5). Three studies34 36 55 reported the impact of PHCP interventions on the time to triage, and all of them reported a decrease in the time to triage after PHCP intervention (online supplemental appendix results 6).

Discussion

Main findings

This systematic review has summarised the best available evidence from 40 unique comparative studies on the effectiveness of the PHCP triage interventions to improve ED patient flow metrics and mitigate the negative impacts of ED overcrowding. The findings in this systematic review shows that the PHCP-led triage interventions significantly decrease the ED LOS and lead to improvements in key ED patient flow metrics such as PIA, proportion of patients who LWBS, triage time, ED visits and patient satisfaction.

Although this systematic review highlights the positive impact of three unique PHCP triage models on key ED patient flow metrics, it is important to note that the most comprehensive evidence (data for the primary review outcome and all of the secondary outcomes) was available mainly for the nurse triage-plus and NP team triage models, with the least evidence available for the GP team triage model.

Comparison with other reviews

To the best of our knowledge, this is the first review to investigate specific triage interventions involving NPs and GPs. Previous work had focused specifically on the impact of TLP26 or triage nurse ordering69 on ED patient flow metrics. Rowe et al26 investigated the impact of TLP’s and reported reductions in ED LOS and PIA. However, the interventions mainly involved emergency physicians. Previously, Jennings et al70 published a systematic review on the impact of emergency NP services in the ED and narratively concluded that although not enough data were available for meta-analysis, NPs within ED may have a positive impact on waiting times, patient satisfaction, and quality of care. Again, this review did not focus on NP at triage. A recent Cochrane review investigated the role of primary care professionals (emergency NP and GP) in the ED71 and concluded that due to limited evidence and suspected bias in allocations of ED patients it was unclear if hiring primary care professionals would decrease PIA, ED LOS and other ED metrics. It is important to note, however, that this Cochrane review did not investigate the role of primary care professionals at ED triage.

ED LOS and PIA

ED wait times for care delivery is a key performance indicator in many ED settings and our systematic review findings indicate that the PHCP-led triage intervention consistently decreases ED wait times (PIA) and ED LOS. In this review, pre–post studies contributed to the majority of the evidence for effectiveness of PIA (NP team triage). Although heterogeneous and of low quality, the results indicate important potential for the role of NPs in the triage process to reduce ED wait times, improve patient satisfaction and other key ED metrics. A significant decrease in ED LOS was observed with the RCTs (median: −16.8 min) although this was comparatively smaller than the significant decrease observed with the CBA (median: −64 min) or the pre–post studies (median: −28 min). As the minimal clinically important difference in ED LOS is generally accepted to be approximately 30 min (clinically significant), the PHCP-led triage interventions could potentially have a positive impact on ED LOS, if implemented.

Type of PHCP

One may argue that similar results could be seen with ED MD at triage. Although true, the cost of adding an NP could be far less than adding an ED MD.38 In our review, we found only three studies reporting evidence on the role of GP’s in ED triage, with one CBA study27 reporting statistically significant decrease in PIA, and a cross-sectional study29 reporting an increase in PIA when triaged and treated by a GP. The increase in PIA, however, was reported to be due to an increase in the number of self-referrals in order to be seen by the GP involved in triage and treatment of low-acuity patients.29 In the reported GP team triage interventions, an ED and a GP clinic were co-located and had a joint common entrance, with the GP assistant (supervised by GP) and/or a GP being responsible for the triage of patients (for both ED and GP clinic) and for the treatment of low-acuity patients. The third study28 reported that GP team triage and x-ray requests at the joint triage reduced the annual ED patient visits. High-quality studies investigating the effectiveness of GPs at ED triage would be valuable.

In our review, the evidence on effectiveness of nurse triage-plus model came mostly from moderate quality studies (RCT or CBA) and showed significant decrease in PIA, ED LOS, and an improved patient satisfaction. Many factors such as patient acuity, EMS traffic/volume and referral patterns often dictate the degree of ED crowding and each ED has their own ‘signature’. For example, in settings where most patients present with ambulatory, single system problems, a nurse triage with extra skills might be effective. Conversely, a nurse triage-plus intervention may be less effective when faced with the challenges of an ED setting with high volumes of trauma, EMS traffic and high acuity patients.

It would be generally expected that the addition of any qualified staff in the ED, including addition of NP in triage, would tend to make efficiency of the ED operations better. Although we did not assess staff satisfaction in our review, it would be intuitive to think that the addition of NP in the ED triage may also help improve ED staff satisfaction. Many government-funded EDs are cash-constrained and often cannot add additional resource without strong justification and/or reducing funding elsewhere. While addition of NPs, TLPs or GPs at triage may help, there is still lack of published comparative effectiveness and economic evaluation research to produce a clear cost-effectiveness recommendation. While comparative effectiveness research may prove logistically difficult in the ED and outcome measurements need to be granular and robust (eg, including intended and unintended consequences), these studies are critical to developing recommendations.

Overall, the evidence synthesised by our review indicates that the PHCP-led triage interventions significantly decrease PIA or ED LOS compared with the traditional nurse-led triage model. The studies in this review demonstrate promise to improve ED patient flow metrics by either seeing and treating non-urgent patients in the triage area, starting investigations at triage for moderate to low acuity patients, or assessing and making decision to re-direct very low-acuity patients to an adjoining GP clinic with same day appointments. All of these could mitigate ED overcrowding. Since ED wait times are multi-factorial it cannot be expected that one solution will solve such a complex problem. Each ED will need an individualised approach. Moreover, while calling for improved research quality, we believe comparative effectiveness studies with health economic outcomes are required to fully weigh the costs and benefits associated with any intervention.

Strengths and limitations

We acknowledge the following limitations in interpreting the results of this systematic review. All systematic reviews are susceptible to publication and selection bias. Selection bias was minimised by using a comprehensive, peer-reviewed search strategy developed by an experienced information specialist. Selection bias was also addressed by using two independent reviewers and third-party adjudication. We evaluated the quality of each included study using the NICE Quality Appraisal Tool67; that is tailored to quantitative studies investigating public health interventions. A few included studies reported the effectiveness of GP team triage intervention on the review outcomes, thus limiting conclusions on GP-led triage interventions. Most of the included studies were of pre–post intervention design providing low quality evidence. Even the included RCTs were only of moderate quality, thus evidence from high-quality studies is lacking, limiting the confidence that can be placed on the results. Nevertheless, irrespective of the study design, we observed a significant decrease in ED LOS favouring PHCP-led triage intervention. Although we used a comprehensive search strategy, for the sake of feasibility we did not consider non-English language studies and the possibility of missing some of the other language studies remains. Despite the compressive search strategy, publication bias is likely since many operational studies never reach publication and many of those would be negative. We also encountered issues with missing data in some of the included studies and resorted to imputation techniques as we were unable to obtain data from study authors. The included studies in this systematic review did not focus on clinical outcomes, such as delayed or missed diagnosis, but it would be important for future studies to quantify relevant clinical outcomes. As included studies were conducted in various countries, health systems and societal contexts, the results from one may not be compatible with evidence from other jurisdictions.

Notwithstanding the above concerns, we believe this review has many strengths, including the rigorous Cochrane systematic review methodology employed and the use of an a priori registered protocol. In addition, our study team included patient partners who collaborated with the investigators during the design, conduct and dissemination phases of the study. Following the criteria identified for patient-oriented research which emphasises the active and meaningful engagement of patients as research partners, twelve diverse group of patient partners from three Canadian provinces (Manitoba, Alberta and Quebec) were engaged from the design stage and throughout the research process around decisions and in knowledge dissemination.

Conclusions

PHCP-led triage interventions could be an effective strategy to improve ED patient flow overall by decreasing ED LOS, PIA, time to triage or ED visits, and by improving patient satisfaction. While these triage interventions may work in specific settings, each ED is unique, and policy would have to be evaluated specific to that facility and system. High quality methods are also necessary to further support PHCPs role in ED triage, and it is important for future studies to focus on cost efficiency or incremental value for money as these are critical real-world issues. Additionally, future research could focus on generating high quality evidence on the effectiveness of GP triage intervention. The acceptability of a PHCP-led interventions in an ED could also be formally ascertained in future studies as experience and beliefs of ED staff may play a role in the success or failure of the policy to implement PHCPs in triage. Finally, the research gap involving rural EDs needs to be addressed.