Article Text
PDF
Abstract
Objectives Patient flow and crowding are two major issues in ED service improvement. A substantial amount of literature exists on the interventions to improve patient flow and crowding, making it difficult for policymakers, managers and clinicians to be familiar with all the available literature and identify which interventions are supported by the evidence. This umbrella review provides a comprehensive analysis of the evidence from existing quantitative systematic reviews on the interventions that improve patient flow in EDs.
Methods An umbrella review of systematic reviews published between 2000 and 2017 was undertaken. Included studies were systematic reviews and meta-analyses of quantitative primary studies assessing an intervention that aimed to improve ED throughput.
Results The search strategy yielded 623 articles of which 13 were included in the umbrella review. The publication dates of the systematic reviews ranged from 2006 to 2016. The 13 systematic reviews evaluated 26 interventions: full capacity protocols, computerised provider order entry, scribes, streaming, fast track and triage. Interventions with similar characteristics were grouped together to produce the following categories: diagnostic services, assessment/short stay units, nurse-directed interventions, physician-directed interventions, administrative/organisational and miscellaneous. The statistical evidence from 14 primary randomised controlled trials (RCTs) was evaluated to determine if correlation or clustering of observations was considered. Only the fast track intervention had moderate evidence to support its use but the RCTs that assessed the intervention did not use statistical tests that considered correlation.
Conclusions Overall, the evidence supporting the interventions to improve patient flow is weak. Only the fast track intervention had moderate evidence to support its use but correlation/clustering was not taken into consideration in the RCTs examining the intervention. Failure to consider the correlation of the data in the primary studies could result in erroneous conclusions of effectiveness.
- emergency department
- crowding
- emergency care systems
- efficiency
Statistics from Altmetric.com
Key messages
What is already known on this subject
Patient flow is a major issue in ED service improvement.
An extensive volume of literature exists on the interventions to improve patient flow.
An umbrella review provides a comprehensive analysis of the evidence from existing systematic reviews on the interventions that improve ED patient flow.
What this study adds
The evidence supporting the interventions to improve patient flow is weak.
Only the fast track intervention had moderate evidence to support its use but clustering of data was not taken into consideration in the randomised controlled trials examining the intervention.
Failure to consider the clustering of data may produce misleading conclusions regarding the effectiveness of the intervention.
Introduction
Patient flow and crowding are two major issues in ED service improvement. Although previously published literature have used these terms interchangeably, in order to suggest better quality improvement measures, it may be necessary to distinguish between the two terms. In 2006, Asplin advocated for a shift in focus from ED crowding to patient flow.1 In Asplin’s view, measuring crowding may be unproductive and suggested a shift from crowding to flow measurements, recognising that measuring patient flow may be more achievable and useful to improve ED care.1
Consensus definitions and measures of ED patient flow and crowding do not yet exist. For this review, patient flow may be described in terms of the progressive movement of patients through care processes from arrival until the patient physically leaves the ED, with movement referring to the conversion of an input into an output.2 3 ED crowding may be described in terms of an imbalance between the demand and capacity to provide care.4
Hwang et al further simplify crowding measurements, categorising it as flow and non-flow, where non-flow leads to crowding.5 Asplin suggested that the ‘fundamental metric of patient flow is throughput’, which may be measured using ED throughput time, that is, time from patient arrival to exit in the ED.1 In terms of metrics, it may be inferred from Hwang et al that patient flow may be measured using time-intervals, while non-flow (crowding) be measured by using numerical counts.5
ED quality indicators from Hospital Episodes Statistics UK and the National Ambulatory Medical Care Survey in the USA include measures such as time to treatment, time to initial assessment, total time in the ED.6 7 This is consistent with Asplin’s measure of ED throughput time and suggestions by Hwang et al to use time intervals to measure patient flow.
Although this review attempts to separately consider patient flow and non-flow (crowding), a close relationship does exist between the two. A crowded ED may result in poor patient flow because of the demand for care. In other words, the number of patients exceeds the capacity to match that demand and consequently this will lead to a downstream effect on the progressive movement of patients, thus hindering patient flow.8 In an ED with poor patient flow, patients may not move through the processes of care at an adequate rate, which eventually may result in ED crowding.8 Thus, it is possible that identifying factors that optimise patient flow may also address crowding.
A substantial amount of literature exists on the interventions to improve patient flow and crowding. An initial quick search in Medline for studies exploring ED patient flow, identified 266 primary studies, 18 systematic reviews and 11 other review types. Reviews assessed specific interventions, making it difficult for policymakers, managers and clinicians to be familiar with all the available literature and identify which interventions are supported by the evidence. Hence, to improve the ED in a holistic manner, policymakers, managers and clinicians may have to familiarise themselves with all the available literature. This may prove to be a difficult task for managers and clinicians.
A comprehensive review of the literature should assist in identifying and assessing the evidence base, and subsequently choosing effective interventions to improve ED patient flow. One method to accomplish this is to compile the evidence from existing systematic reviews. The Cochrane Collaboration describes this as an overview of reviews or Cochrane Overviews.9 The Joanna Briggs Institute, an international research institute in Australia, uses the term umbrella review, defined as ‘an overview of existing systematic reviews’.10 An umbrella review synthesises the evidence from published systematic reviews, selecting reviews based on predetermined criteria without delving much into the quality of the individual primary studies included in the original systematic review.
A systematic review systematically searches for, appraises and synthesises evidence, usually following specific guidelines.11 Hence, an umbrella review should encompass all similar systematic reviews on a specific topic, crystallising the evidence, in an attempt to assist managers and clinicians to improve their departments in an evidence-based manner.
With this background, this umbrella review aims to summarise the evidence from systematic reviews on the interventions that improve patient flow in EDs.
Methods
We compiled evidence from systematic reviews that analysed quantitative primary studies addressing interventions to improve ED patient flow.
Eligibility criteria
Reviews were eligible if they satisfied the following criteria:
Full-text systematic reviews published between 2000 and 2017 in English language.
Searched at least two electronic databases.
Systematic reviews and meta-analyses of quantitative primary studies (systematic reviews including both quantitative and qualitative data were included only if the data werec analysed separately).
ED must be the primary study site.
Must include any intervention, strategy that targeted ED throughput.
Outcome measures (as metrics of patient flow) must have been defined; described in terms of any time interval, for example, length of stay (ED LOS) and any of its submeasures.
Reviews were excluded if any of the following were present:
Focused on disease-specific conditions.
Intentionally focused on country-specific literature.
Primary focus was ED crowding (eg, outcomes were crowding measures, defined as numerical counts such as number of patients in ED).
Non-systematic reviews.
Qualitative evidence syntheses.
Systematic reviews based on theoretical studies, opinions, editorials, commentary.
Search strategy
A comprehensive search strategy, restricted from January 2000 to April 2017, was used to identify articles. Six databases were searched: Medline via Ovid (1946 to present), EMBASE (1974 to July 2016), CINAHL (1982 to present), Cochrane Library, JBI for Systematic Reviews and Implementation reports, Proquest. Three search concepts were used: ‘emergency department’, ‘patient flow’ and ‘crowding’. Systematic review search filters were applied to the search strategy as outlined by Lee et al 12 and Lunny et al 13 (see online supplementary 1 for sample search strategy).
Supplementary file 1
OpenGrey and Google Scholar were searched for grey literature. Citation tracking was conducted in Google Scholar, Web of Science and Epistemonikos. Reference lists of the included articles were reviewed. Conference proceedings identified in the electronic database search were checked for full-text versions and authors contacted if necessary.
Data extraction and quality appraisal
Two authors (LD and SH) independently reviewed the systematic reviews extracting data using a data extraction form developed by the Joanna Briggs Institute10 and ranked the quality using A Measurement Tool to Assess Systematic Reviews (AMSTAR 2) tool (online supplementary 2).14 Differences were settled after discussions to reach a consensus. The quality appraisal of the primary studies identified in the systematic reviews was extracted from each systematic review. The authors of the umbrella review did not perform a new quality appraisal for these primary studies as an umbrella review usually only includes a quality appraisal of the systematic reviews rather than the quality of the primary studies.
Supplementary file 2
Data synthesis
The results were summarised and presented in a tabular form supported by a narrative synthesis. The results were presented based on each intervention and outcome measure. Given the high heterogeneity across the reviews, no additional statistical analyses were conducted.
Analysis of the appropriateness of the statistical analyses was undertaken in a subset of primary studies, to explore the issue of whether potentially correlated data had been addressed. Measures of patient flow, like measures of ED crowding, may be subject to substantial correlation between individuals, which if not taken into account could lead to the wrong conclusion being drawn. This statistical review was performed by SH and LD.
Results
Results of the search process
Six hundred seventeen articles were retrieved from the six databases. Six studies were found through reference lists and citation searching. Four hundred four articles were screened at the title stage. Thirteen full-text articles were included in the final review. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow chart of the study selection15 is depicted in figure 1.
Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow chart of study selection. SR, systematic review.
Description of included systematic reviews
The publication dates of the 13 reviews ranged from 2006 to 2016.16–28 The publication dates of the primary studies ranged from 1995 to 2015. Six of the reviews used the term ‘crowding’ in their titles but had time interval outcome measures which made them suitable for assessing patient flow.17 18 21 26–28 There were 20 randomised control trials (RCT) and 200 non-RCTs. Of these non-RCTs, 125 studies had before-after (BA) designs. The primary studies originated from 20 countries. Participant numbers totalled over 2 million.
The general characteristics of the systematic reviews are presented in table 1. The majority of the reviews were graded as moderate to high quality based on the AMSTAR 2 score. Many of the primary studies were weak, mostly belonging to the BA study design. The systematic reviews conducted by Elder et al,19 Georgiou et al 20 and Jennings et al 23 did not present quality assessments of the primary studies. The review by Bond et al 17 presented a quality assessment but an interpretation of the scores was not provided. The publication agency for that review was not able to provide further information on the quality assessment.
General characteristics of the systematic reviews
A summary of the quality appraisals of the primary studies and the AMSTAR 2 scores is presented in online supplementaries 3 and 4.
Review findings
Description of interventions
The 13 systematic reviews evaluated 26 interventions: full capacity protocols, computerised provider order entry (CPOE), scribes, streaming, fast track and triage. Interventions with similar characteristics were categorised as follows: diagnostic services, assessment/short stay units, nurse-directed interventions, physician-directed interventions, administrative/organisational and miscellaneous. A description of the interventions based on the information presented in the study(s) that assessed it is presented in table 2.
Description of interventions
Statistical evidence from primary RCT studies
The correlation of observations in the ED is a potential issue in the statistical analyses of the reviews and primary studies.29 30 Many standard statistical tests assume that the observations are independent.29 30 An independent observation assumes, for example, that the waiting time of one patient is not correlated with the waiting time of another but this is unlikely to be true in the ED since patients arriving at similar times are also likely to have similar waiting times. Therefore, it is important to consider the dependent nature of the observations when analysing data. Using tests that do not consider dependency or correlation may result in the incorrect estimation of the p value with misleading conclusions.29
Ming et al 24 discussed the correlation issue in their review. Since only one systematic review made reference to the issue, the statistical tests used in a subset of primary studies were examined. Given the substantial number of primary studies that would have to be assessed together with the complexity of the statistical issue, the decision was made to focus only on RCTs. RCTs have stronger study designs that can provide reliable evidence once analysed appropriately. While non-randomised designs are likely to be at an even greater risk for correlation and clustering issues, these designs, particularly the BA studies, are already at high risk of bias even if analysed appropriately. In each systematic review, RCTs that assessed a flow metric were extracted and included. Fifteen RCTs assessed the outcome measures of interest and 14 articles were located (S1-14) (see online supplementary 5 for the statistical review of RCTs).
Supplementary file 5
Summary of findings
A summary of findings for each intervention, based on each outcome measure, is presented in tabular form together with a narrative synthesis. Overlap of primary studies in reviews assessing the same intervention is highlighted in the summary tables.
The summary of findings for full capacity protocols, CPOE, scribes, streaming, fast track, triage, diagnostic services, assessment and short stay units are presented in table 3; nurse-directed and physician-directed interventions are presented in tables 4 and 5; administrative/organisational and miscellaneous interventions are presented in table 6.
Summary of effects of interventions
Summary of findings for nurse-directed interventions
Summary of findings for physician-directed interventions
Summary of findings for administrative/organisational and miscellaneous interventions
Full capacity protocols
This was evaluated in one BA Canadian study from one systematic review. The full capacity protocol significantly improved ED LOS for all admitted patients.28 However, as the review was based on one weak quality study, in abstract form, it is difficult to draw conclusions.
Computerised provider order entry
Two reviews examined the effect of CPOE on patient flow.17 20 The results were derived from studies conducted in the USA and Canada. Bond et al reported a decrease in ED LOS in two non-RCT studies and an increase seen in one BA.17Two BA studies in the review by Georgiou et al reported decreases in LOS (−1.94 hours, 95% CI 0.79 to 3.09 hours; −30 min, 95% CI 28 to 33 min) while two reported increases in LOS (17.4, 95% CI 8.7 to 26.2 min; 36 min, 95% CI 26 to 46 min).20 The review by Georgiou et al concluded that CPOE had inconsistent effects on ED LOS.20
Scribes
The impact of scribes on patient flow was examined in one review that compared services with scribes with those without.22 The settings included six academic and two community EDs across the USA (six), Canada (one) and Australia (one). The primary studies were based on non-RCT designs and those assessing LOS had a high (one) and moderate (four) risk of bias. Meta-analyses performed by the review authors found that scribes had no difference on ED LOS and provider to disposition time.
There was a statistically significant but small increase in the number of patients seen per hour. There were no pooled results comparing the effect of scribes in academic versus community EDs, so it is unclear if the type of ED setting affected the results. The review concluded that evidence was limited for the use of scribes.22
Streaming
Streaming was assessed by one review whose studies were conducted in Australia (two) and the USA (one).25 The primary studies were all moderate-quality BA designs. Pooled results from these studies showed decreased ED LOS and waiting time. One primary Australian study examined the effect of streaming in the different triage categories and found improved ED LOS for lower acuity patients (14 and 18 min less for level 4 and 5 patients, respectively).25 Although streaming had a positive effect on flow metrics, the review concluded that there was weak evidence to support its use.25
Fast track
Three reviews examined the effect of fast track on flow metrics.17 21 25 Studies were conducted in the USA (seven), Canada (seven), the UK (five), Australia (five) and one each from New Zealand, Saudi Arabia, Turkey and Spain.
Pooled results from Oredsson et al found that fast track reduced both ED LOS and waiting times.25 These results for ED LOS were based on seven moderate (two RCT, five BA) and three low (BA) quality studies while those for waiting times were based on six moderate (one RCT, five BA) and three low (BA) quality studies. In the study by Bond et al, 15 primary studies showed improved ED LOS and 8 showed improved waiting times.17 The quality of these studies was not known. The results from Guo et al also showed decreases in ED LOS and waiting times. These were based on low (BA) quality primary studies. The reviews by Oredsson et al and Bond et al concluded that there was moderate evidence to support the use of fast track.17 25
Three RCTs assessed the fast track intervention. Two were cluster RCT designs but there was no evidence to suggest that a cluster analysis was performed (S7, S8). The third RCT was an individual-level RCT that used appropriate statistical analyses but did not consider clustering in the analysis (S14).
Triage
The use of triage systems was assessed by one review with studies conducted in the USA (three) and the UK (two). The quality of these studies is not known. The results were mixed—two BA studies showed a decrease in waiting times while three studies (two controlled clinical trial (CCT), one BA) showed an increase. The review concluded that the results were inconclusive.17
Diagnostic services
Three reviews assessed diagnostic services which included point-of-care testing17 25 and advanced triage.21 Point-of-care testing was evaluated in the USA (three), the UK (one) and Canada (one); all three reviews showed a reduction in ED LOS. The review by Oredsson et al had three moderate (one RCT, two BA) and two low (one RCT, one BA) quality primary studies and concluded that there was limited evidence to support use of point-of-care testing.25 Guo et al assessed advanced triage in one good quality cohort study, which showed a reduction in LOS.21
Two individual-level RCTs assessed point-of-care testing (S12, S13). The statistical tests used were considered appropriate for the design but did not consider clustering/correlation of the data.
Assessment and short stay units
Three reviews examined assessment and short stay units.17–19 Studies were conducted in the USA (one), Canada (three), New Zealand (one) and Saudi Arabia (one). Short stay units showed a reduction in ED LOS for treat-and-release patients from a BA study.17 Bullard et al assessed rapid assessment zones and found shorter ED LOS based on one RCT and BA study both rated as low quality.18 The authors concluded that there was insufficient evidence to support rapid assessment zones.17 18
Nurse-directed interventions
Nurse-directed interventions consisted of various interventions relating to nursing activities. Four reviews contributed to this category.19 23 25 27 The primary studies were conducted in Australia (eight), the UK (six), Canada (five), the USA (three) and one study each in New Zealand, Hong Kong, Singapore, the Netherlands and Sweden.
Two systematic reviews from Jennings et al 23 found that nurse practitioners led to shorter waiting times and LOS. Those findings were based on low-quality studies and the authors concluded that the evidence was limited.
Rowe et al 27 examined the impact of triage nurse ordering. The primary studies compared nurse-initiated X-rays with ED physician-initiated X-rays. The primary studies assessing the ED LOS were all weak three RCT, one CCT, two case-control (CC), three cohort and three BA). One RCT found a statistically significant reduction in ED LOS with triage nurse ordering.27 Oredsson et al looked at nurse-requested X-rays and found a decrease in ED LOS/waiting times based on three RCTs.25 The primary studies by Oredsson et al assessing ED LOS were moderate (one RCT) and low (one RCT) quality while those assessing waiting times were moderate (one RCT) quality. The review concluded that evidence was limited.25
Four of the primary studies assessing nurse-directed interventions were RCTs. One used a cluster RCT design (S6) and three were individual-level RCTs (S9–S11). There was no evidence to suggest that any of the RCTs performed an analysis that considered clustering/correlation.
Physician-directed interventions
Physician-directed interventions assessed the role of physicians in triage. Five reviews contributed to this category.16 19 24–26 The study settings included the USA (19), Australia (5), the UK (3), Canada (3), Hong Kong (2) and 1 each in Northern Ireland, Jamaica, Sweden and Singapore.
Meta-analyses on triage liaison physician compared with nurse-led triage showed statistically significant reductions in ED LOS.26 These findings were based on 3 strong (1 RCT, 2 CCT), 2 moderate (1 ITS, 1 BA) and 14 (1 RCT, 2 CCT, 1 cohort, 10 BA) weak quality primary studies. Two RCTs examining senior doctor triage found statistically significant decreases in ED LOS while one showed a statistically non-significant increase.16 Meta-analyses also showed reductions in waiting times for senior doctor triage.16 The results for ED LOS for senior doctor triage were based on four strong (three RCT, one BA), nine moderate (one CCT, two cohort, six BA) and six weak (one RCT, one cohort, four BA) quality primary studies. The results for waiting times were based on one strong (RCT), five moderate (two cohort three BA) and seven weak (one RCT, one cohort, fiveBA) quality studies. Although senior doctor triage showed improvements in flow metrics, the study concluded that the evidence was not strong enough.16
Team triage was assessed by three reviews which all found decreased ED LOS and waiting times.24–26 Ming et al compared team triage with single nurse triage and found non-significant reductions in ED LOS in four RCTs which were all assessed as low quality.24 Rowe et al performed a subanalysis on four non-RCT studies, comparing team triage and single physician triage and found a statistically significant reduction in ED LOS with team triage.26 These results were based on weak quality primary studies (one cohort, three BA). The primary studies from Oredsson et al assessing ED LOS consisted of three moderate (one RCT, one CCT, 1BA) and one low (RCT) quality. Those assessing waiting times from Oredsson et al consisted of one moderate (BA) and two low (BA) studies. Ming et al 24 and Oredsson et al 25 both concluded that the evidence to support the use of team triage was limited.
Of the primary studies assessing physician-directed interventions, five were RCTs. Four of the RCTs used a cluster randomised design that used appropriate cluster analyses considering clustering and correlation (S1–S4). The fifth RCT was a cluster randomised design but there was no evidence to suggest that a cluster analysis was performed (S5).
Administrative and organisational interventions
Administrative and organisational interventions included a range of strategies such as increasing clinical and non-clinical staff numbers, increasing cubicles/treatment rooms, structural reorganisation, implementation of coordinators.17 21 Studies were conducted in the USA (seven), Australia (three), Spain (two), Canada (two) and one each in Hong Kong, Israel, Sweden and Switzerland. Overall, there were improvements in ED LOS and waiting times. However, these results were based only on BA studies rated as either good or low quality by Guo et al.21 The reviews concluded that there was insufficient evidence to support these interventions.17 21
Miscellaneous
Bond et al assessed electronic tracking boards, dedicated ED radiology staff and bedside registration.17 These studies were all US-based BA designs; all three interventions reduced ED LOS, triage to treatment and triage to room times.
Discussion
This umbrella review summarised evidence from systematic reviews and meta-analyses on interventions that improve ED patient flow. Overall, the evidence supporting the effectiveness of the interventions was weak (as reported by the systematic review authors). Only one intervention had moderate evidence to support its use—fast track. However, one review author noted that, although the evidence was sufficient, there were other factors such as physical limitations in the ED, limited human resources and cost-effectiveness that could affect the implementation of fast track.17
The interventions were not standardised with different terms possibly representing the same intervention. For example, Oredsson et al 25 examined nurse-requested X-rays, an activity performed by nurse practitioners19 23 and seen in triage nurse ordering.27 In some instances, the same primary studies provided evidence for a range of interventions as seen with senior doctor triage, triage liaison physician, physician-assisted triage and team triage.16 19 24–26 Reviews that included paediatric settings did not differentiate between adult and paediatric EDs to determine if this affected the intervention effect. The heterogeneity in the intervention and control groups could affect how interventions were implemented in different settings, a factor which may affect the ability to generalise findings.
Another potential factor limiting generalisability was the overlap of interventions. The multifaceted interventions were based on the implementation of combined strategies. Since no direct comparisons were made between the single intervention and the combination of strategies it is unknown which one was responsible for the observed effects. This was also a factor in fast track, which in some studies was either nurse or doctor led and in others was combined with streaming or rapid assessment zones.17 25 Again it is unclear which factor (nurse-led or doctor-led fast track, streaming or assessment zones) contributed to the effect.
A 2011 overview examined interventions to mitigate ED crowding.31 Although the overview did not meet criteria for inclusion in the umbrella review, it did measure flow metrics and identified additional interventions that are worth mentioning. These included bedside ultrasound, computerisation, clinical decision and observation units, bed coordination and multifaceted interventions (eg, UK 4-hour target). These interventions also showed benefits to improving flow metrics but like the interventions identified in the umbrella review, there was still insufficient evidence to support the implementation of any of the interventions.31
Although this umbrella review identified interventions that could improve patient flow, an understanding of how and why these interventions produced (or did not produce) their desired effect, is still unclear. This is important because the studies were conducted in countries with different models of emergency care. The majority of studies were in countries with developed emergency care systems and a dedicated emergency medicine specialty (the USA, the UK, Australia, Canada). Thus, generalising the findings to other models of ED care may still be difficult; an exploration of the mechanism underlying the intervention or the patient flow process may be beneficial.
Lastly, the uncertainty surrounding the appropriate use of statistical tests in the cluster RCTs affects the conclusions drawn on the effectiveness of the intervention. The RCTs using individual patient designs appeared to use appropriate tests; however, the potential importance of clustering/correlation in individual patient RCTs is an issue that should be considered in future trials of patient flow.32 This is particularly important for the fast track intervention, which was the only intervention with evidence supporting its implementation but for whom clustering/correlation was not considered in the RCTs that examined the intervention.
Limitations
There are several limitations to this review. Measures of patient flow were not standardised across the included systematic reviews. The most common outcome measures were ED LOS and waiting times. Two primary studies from one review presented different definitions of ED LOS (arrival to physical departure vs triage to physical departure). This was not unexpected since there is no universal definition for patient flow and crowding terms and measures.
Although the majority of the systematic reviews were graded as either high or moderate quality, within the systematic reviews there was a predominance of weak primary studies and study designs. Many of the systematic review findings were based on primary studies with non-RCT designs; almost two-thirds were BA studies, which are known to produce bias.33 The Cochrane Effective Practice and Organisation of Care (EPOC) guidance recommends against the inclusion of uncontrolled BA study designs in systematic reviews.33
Some systematic review findings were based on a small number of primary studies and several reviews included abstracts rather than peer-reviewed full-text articles. Some systematic reviews examining the same intervention had overlap of the primary studies contributing to the outcome measure. Thus, it was not always new evidence being presented for each intervention.
The authors of the systematic reviews also noted the high heterogeneity seen with study settings, designs, populations, interventions and outcome measures, which prevented the pooling of results and performance of meta-analyses.
Conclusion
The evidence to support implementation of the majority of the interventions was considered weak. Future studies should distinguish between non-flow (crowding) and flow and the respective measures. Stronger study designs are also required, as well as an exploration of the patient flow process, how these interventions work and why some interventions work in some settings and not others. Furthermore, the issue of correlation of observations when conducting statistical analyses should be considered in all future studies. ED patient flow is a complex phenomenon and a greater understanding of the patient flow process could assist in the development of effective interventions.
Supplementary file 3
Supplementary file 4
Supplementary file 6
Acknowledgments
The authors would like to thank Dr Dawn Teare for her advice with the statistical review of the primary RCT articles. The authors would also like to thank Dr Brian Rowe who answered questions about his published studies.
References
Footnotes
Contributors LDeF, SG, RO’H, PT contributed to the concept and design of the review. LDeF conducted the umbrella review. LDeF, SG, RO’H, PT contributed to the analysis and interpretation of the data. LDeF and SH conducted the quality appraisal and the statistical analysis of the primary RCT studies. LD drafted the article; all authors contributed and revised the content of the drafts and approved the final version. Dr Dawn Teare provided advice on the statistical review of the primary RCTs. Dr Brian Rowe provided information on his published articles.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient consent Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
Linked Articles
- Primary survey