Background Primary care focused unscheduled care centres (UCC) co-located with major EDs have been proposed as a solution to the rise in ED attendances. They aim to reduce the burden of primary care patients attending the ED, hence reducing crowding, waits and cost.
This review analysed available literature in the context of the impact of general practitioner (GP) delivered, hospital-based (adjacent or within the ED) unscheduled care services on process outcomes, cost-effectiveness and patient satisfaction.
Methods A narrative literature review of studies published between 1980 and 2015 was undertaken. All study types were reviewed and included if they reported a service model using hospital-based primary care clinicians with a control consisting of standard ED clinician-delivered care.
Results The electronic searches yielded 7544 citations, with 20 records used in the review. These were grouped into three main themes: process outcomes, cost-effectiveness and satisfaction. A paradoxical increase in attendances has been described, which is likely to be attributable to provider-induced demand, and the evidence for improved throughput is poor. Marginal savings may be realised per patient, but this is likely to be overshadowed by the overall cost of introducing a new service.
Conclusions There is little evidence to support the implementation of co-located UCC models. A robust evaluation of proposed models is needed to inform future policy.
- emergency care systems, emergency departments
- cost effectiveness
- emergency care systems, primary care
- management, cost efficiency
Statistics from Altmetric.com
- emergency care systems, emergency departments
- cost effectiveness
- emergency care systems, primary care
- management, cost efficiency
ED attendances have steadily increased over the last decade in the UK1 and internationally. One theory explaining this increased demand suggests that this is attributable mainly to patients with problems more suited to primary care2 and that diverting such patients away from the ED may improve access and care across the system. To this end, several models of hospital-based unscheduled care services have been developed that primarily use a workforce consisting of general practitioners (GPs) or other primary care clinicians. These have been implemented at significant cost in many cases, but with little evaluation of effectiveness in the context of local health services. In many instances, the introduction of alternative and untested forms of urgent care has failed to reduce ED attendances.3
This review will seek to analyse the available literature in the context of the impact of GP delivered, hospital-based (adjacent or within the ED) unscheduled care services on process outcomes, cost-effectiveness and patient satisfaction.
In order to explore the evidence supporting a hospital-based urgent care model, we undertook a literature review of existing research. The search strategy was based on the variables upon which the theoretical benefits of an UCC are premised.
A search strategy was designed with the following three-part question based on the study objectives.
For patients presenting to an ED with non-urgent problems (population) and managed by hospital-based primary care professionals (intervention) or emergency physicians (control), are there systematic differences in cost-effectiveness, process outcomes and patient satisfaction (outcomes)?
We searched for specific process outcomes related to the time from arrival to first diagnostic contact with a clinician (waiting time (WT)), time to treatment or intervention (treatment time (TT)) and total time in the ED (length of stay (LOS)).
We searched the following databases for articles between 1980 and February 2015: Business Source Premier, CINAHL, Cochrane Library, DARE, EMBASE, HTA, MEDLINE, NHS Evidence, NHS EED, PsycINFO and SCOPUS. We chose the lower date limit as the concept of ED patients being better suited for primary care only became topical around the mid-1980s. This was supported by a PubMed ‘Results by Year’ keyword scoping search, which demonstrated a low annual publication rate until 1989.
In addition, Google Scholar was searched along with OpenGrey, UK Economic and Social Research Council registry, National Centre for Primary Care Research and Development, King's Fund, Nuffield Trust, NHS Commissioning Board (NHS England), Primary Care Foundation, College of Emergency Medicine, Agency for Healthcare Research and Quality and NHS Institute for Innovation and Improvement.
References from key publications were hand searched and the urgent care leads from local commissioning groups were asked for any other key articles or unpublished data/reports.
Broad search terms were used as this has been shown to increase search sensitivity.4
The heterogeneity and largely theoretical basis of many existing or proposed service configurations resulted in an evolving literature base that reported variable outcomes. An approach loosely based on a realist approach to evidence synthesis was undertaken. This method is described elsewhere,5 and we have used some of its basic underlying theory in preference to a traditional systematic review. In particular, our searches were purposive and hinged on the theoretical basis and proposed outcomes of service models. We included all types of evidence and attempted to search for explanations of why models worked or not. This approach is appropriate for interventions where knowledge of the theoretical basis of models is key to developing contextually effective services.
All article types were retrieved and reviewed. No formal quality assessment was undertaken. While our searches were not confined to UK models, studies were excluded if the setting was not broadly comparable to a UK setting; free at the point of access and similar in terms of outcomes or theoretical basis for the service. No language restrictions were used. (The full search strategy is appended as an online supplementary file.)
The electronic searches yielded 7544 citations, with 68 further citations from conference abstracts and unpublished reports. Also, 42 records were retrieved from hand searching references and 38 records from searches of organisational sites, reports and experts. And, 82 records remained after screening by title and abstract of which 20 were used in the review. Four studies were reported in two or more records. The study flow diagram is appended as an online supplementary diagram (Appendix 2).
The majority of studies (eight) were undertaken in the Netherlands with four from England. The rest were carried out in Australia, Ireland, Spain, Sweden and Switzerland.
Results have been presented according to three key theoretical outcomes identified during the review; process outcomes, patient satisfaction and cost-effectiveness.
The review found significant heterogeneity in the models and methodology used in their evaluation. Four broad service configurations emerged (box 1), which mirrored those identified in another review6 with variable overlap between configurations. The majority of models were not available for the same periods as the ED, and most models operated primarily outside of the usual opening times of GP practices. Triage of suitable patients was undertaken by ED nurses in the majority of studies.
Current UCC models
Within the ED footprint
Patients attend the ED and are triaged into separate streams (urgent or non-urgent/primary care). The primary care stream is staffed by primary care clinicians.
Alongside the ED
The primary care service is distinctly available next to/close to the ED. Patients choose themselves or are redirected from the ED towards the primary care service and vice versa.
Primary care front-end screening/filtering ED patients
Primary care practitioners triage or filter non-emergency (non-ambulance) patients at the front end of the ED, either in person or telephone, thus limiting ED access.
Care provided within the ED jointly with ED staff to all patients who attend. Some overlap with primary care front-end filtering model.
Adapted from Carson et al.6
Studies that reported outcomes of interventions using primary care in or adjacent to an ED are summarised in table 1.
The majority of studies described the addition of a GP to manage minor health conditions, with direct substitution for usual ED staff only explicitly described in one paper. One study evaluated eight EDs with co-located primary care/walk-in centre (WiC) services using traditional EDs as the control.7 Two studies evaluated an integrated GP/ED model, although these functioned only out of hours and the level of integration was variable.8 ,9
Most of the included citations were variations on an uncontrolled before and after study, with one quasi-randomised evaluation,10 two randomised controlled trials (RCTs)11 ,12 and one pragmatic prospective evaluation.13
Impact on attendances
A Swedish study found that the introduction of a GP surgery adjacent to an ED increased the proportion of patients presenting to the ED with non-urgent complaints from 22% to 33%, with a 27% overall increase in the number of visits.14
Two multisite studies reported on the effect of the introduction of new service configurations on attendances. The first, by Kool et al,8 compared traditional Dutch EDs and integrated emergency posts (IEPs)—staffed additionally with a GP, GP assistant or nurse with patients triaged to the appropriate clinician. They found that the proportion of self-referred patients (patients bypassing telephone triage) in the usual ED increased from 53% to 58% (p<0.05), while in the IEP group, there was a reduction from 62% to 46% (p<0.05). When the total attendances for the system (ED and GP) were considered, there was an 11% increase in attendances at the IEPs. Notably, for the same period, there was a 5% reduction in attendances at the control sites.
Other studies examining similar reconfigurations in GP collaboratives (GPCs) with EDs also demonstrated a reduction in self-presenting ED patients.9 ,15 ,16 ,17 However, this may be explained by the gatekeeper role adopted in these models. An increase in overall attendances (ED and GPC) across the system was consistently observed.
The second multisite study was undertaken by Salisbury et al in England. They examined the impact of co-located WiCs and found that the number of patients increased during the study period at hospitals, independently of the presence of a co-located WiC (813/month at intervention sites; 95% CI −30.3 to 1655, p=0.06, and 270/month at control sites; 95% CI −114 to 655, p=0.17). The authors suggested that there was a greater increase in patient numbers at sites with co-located WiCs but there was wide variability between individual sites.7
Process time measures
Five studies reported the effect of the introduction of a primary care service or stream (staffed by GPs) on process times. The outcome measures of relevance to this review were LOS, WT and TT.
Length of stay
An English study, conducted by Salisbury et al,7 reported compliance with the English 4 h target, which measures the proportion of patients with an ED transit time under 4 h.22 In this study comparing eight EDs with co-located WiCs and eight traditional EDs, there was no significant difference in compliance with the 4 h target or patient LOS. The authors commented that the WiC concept was implemented in a limited fashion and with much variability between sites.
Jiménez et al18 evaluated the substitution of a GP for the usual ED clinician for low-acuity patients in a Spanish ED. They found a reduction in mean LOS for these patients of 36% (95% CI −53% to −19%) (119 min to 76 min, p<0.01) with the intervention. A Dutch study, conducted by Kool et al,8 reported a 14 min reduction in LOS from 116 min to 102 min (p<0.05) after introduction of an IEP (GP service along with the ED). Interestingly, the authors also reported a 26 min rise in LOS at the control ED (p<0.05) during the same period. In a later Dutch study, Boeke et al looked at the addition of a GP to the ED for patients who did not require emergency help. They reported a 24 min reduction in mean process time (93 min to 69 min, p<0.001).19 A study comparing a new hospital integrated general practice (HGP) in a Swiss ED with a historical control found that LOS was halved (120 min to 60 min, p<0.001) with the new HGP.21 This reduction was independent of whether the HGP was staffed by ED (general medicine) residents or GPs, who were an additional resource compared with the historical control.
Three studies specifically reported WT as an outcome, although several included the perception of WT as a measure of patient satisfaction. Jiménez et al18 found a 20% reduction in WT (95% CI −40 to −5) from a mean of 84 min to 67 min (p<0.05) in a low-acuity area when staffed by GPs compared with ED residents. Wang et al21 found no difference in WT after the introduction of a GP service for low-acuity patients (25 min in both, p=0.065). This contrasts with an earlier study where the addition of GPs to the ED increased the average WT for admitted patients and those with urgent or emergent complaints from 35 min to 40 min (14%), while reducing the WT for non-urgent cases from 50 min to 37 min.14 This was attributed to the increase in demand resulting from the introduction of GPs. More recently, a statistical model used on an Australian ED data set found that emergency patients waited 1.5 min less in EDs with a co-located GP clinic.20
Jiménez et al18 found no difference between ED clinicians and GPs in the proportion of patients receiving any type of treatment. There was, however, a 25% reduction in the time to treatment (TT) with the introduction of a GP (95% CI −40 to −5). Boeke et al19 found that 48.5% of the patients in the GP group (intervention) received no treatment compared with 40.5% in the usual care (ED) method (p=0.0013). The reduction in mean TT between usual ED care and GP care was 25 min (p<0.001). In the usual care system, 20.1% of all ED patients had been treated within 20 min, and 57% had been treated within 1 h. In the intervention group, 55.8% of all patients received treatment within 20 min, and 80.1% received treatment within 1 h. In the interpretation of the time differences, the GP was an addition and not a substitution for ED clinicians.
Six studies10 ,12 ,13 ,18 ,19 ,21 reported more diagnostic testing (predominantly radiography and blood tests) by ED clinicians than GPs in the ED or adjacent service. One study using GPs within the ED11 found no difference in the use of investigations (RR 1.06; 95% CI 1.00 to 1.13) between ED and GP clinicians. Two studies were RCTs and, while Gibney et al11 reported similar baseline patient characteristics for GP and ED groups, there were some differences in characteristics between groups in the study by Murphy et al12 despite the same randomisation methodology. This was also evident in the non-randomised studies although the effects were unclear.
Considering radiography alone, GPs ordered significantly fewer X-rays than emergency physicians (EPs)10 ,12 ,13 ,19 ,21 in all but one study,11 which found a tendency to more X-ray requesting by GPs, although this was not significant (% relative difference −7, 95% CI −15 to 1).
Similar results were found for blood investigations (haematology and biochemistry predominantly) with GPs ordering significantly less than EPs in all10 ,12 ,18 ,21 but two studies.11 ,19 In the two studies reporting microbiology separately, there was a trend to less ordering by GPs in one10 with no difference in the other.13
There was no statistically significant difference in prescribing behaviours between sessional GPs and regular EPs in two studies.10 ,13 The RCTs, however, reported marginally more prescribing by GPs in one11 and significantly more in the other.12
Follow up rates
Seven studies reported disposal outcomes ranging from referral to inpatient teams, admission and outpatient follow-up.9 ,11 ,12 ,13 ,18 ,23 Complete data sets were missing from one study, which reported that the number of patients referred to their GPs for aftercare increased by 17% in the GP group, and the number of patients referred to outpatient clinics decreased by 17%.19 One study examined the effect of an integrated ED/GPC and found an increased follow-up rate of 5.8% (OR 1.015, 95% CI 1.013 to 1.017).9
GPs admitted significantly fewer non-urgent patients to hospital than EPs in three studies.10 ,12 ,13 Two authors found that the proportion of admissions made by either type of clinician was not significantly different,11 ,18 while one study found a 20.2% increase in admission rate (p<0.05) after implementation of an integrated GPC/ED emergency care access point model.9
Three studies demonstrated that GPs made significantly fewer referrals to hospital specialists/consultants,10 ,12 ,13 with one reporting the converse.11 An integrated GP/ED service resulted in a 6.5% increase (from 3.5% to 10%, p<0.001) in specialist consultations.21
One study reported no difference in ED re-attendance rate by patients seen by GPs versus EPs.12 A more recent paper reported a reduction in overall re-attendance rate from 2% to 0.5% (95% CI −6% to −14%) with a fast-track area in the ED for low-acuity patients but no difference in re-attendances by patients seen by a GP (3.1%) compared with an ED resident (3.4%) (95% CI −90 to 73). None of the re-attendances resulted in admission.18
Between 18% and 50% of patients attended their GP surgery for the same problem within 7–30 days of their index ED attendance.12 ,13 ,23 This proportion did not differ significantly if the patient was seen by a GP or ED clinician on the ED attendance.
Six studies reported cost-effectiveness or provided an estimate of costs associated with primary care services in an ED (table 2). The only multisite study, published in 2007 by Salisbury et al in the UK, reported a year-on-year total cost increase of 22% in the intervention group (co-located ED/WiC) and 10% in the control group. This difference was mainly due to the expected increase in clinician cost. There was no significant difference in costs per patient, even when admission costs were included in a sensitivity analysis.7
A study by Van Uden et al in the Netherlands calculated a higher per capita cost of an integrated primary care/ED out-of-hours model than the previously separate services (€11.47 and €10.54, respectively). They also reported a reduction in funding to the ED due to reduced activity despite unchanged overall ED costs.24
Two older UK studies12 ,23 suggested some cost benefit with the introduction of a GP/primary care stream. As these were carried out two decades ago, the cost estimates are likely to be different to the present; however, the authors suggested only marginal cost savings of between £0.58 and £7.60 per patient, excluding admission costs.
A 2005 Spanish study showed that the average cost per patient was €116.99 lower with a GP staffed low-acuity area compared with usual ED care (p=0.005).18
A 2012 Dutch study by Bosmans et al reported direct (TT, treatment/investigation and outcomes after discharge) and indirect costs associated with process times. The authors calculated incremental cost-effectiveness ratios associated with process time, patient satisfaction and correct diagnosis. Total costs per patient were €71 lower in a model with lower-acuity patients streamed to a GP (95% CI −121 to −23) and based on GP staffing between 1000 and 1700. Process time costs in the GP model were lower than in the usual ED model (mean difference −€4, 95% CI −4 to −3). The cost-effectiveness analysis showed that the GP streaming model was dominant (more effective, less expensive) in comparison with usual ED care for process time and patient satisfaction. The GP model was considered cost-effective in comparison with usual ED care for ceiling ratios between €0 and €1363 for the number of correct diagnoses.25
Eleven studies were identified that included patient satisfaction as an outcome measure; however, only six of these actually reported any detail of measuring or recording differences between control and intervention groups (table 3).8 ,18 ,19 ,23 ,25 ,26 One study reported patient satisfaction in two papers.19 ,25 Four studies found no difference in satisfaction between usual ED care and addition of a GP/primary care stream.8 ,18 ,23 ,26 One study found that the triage telephone contact prior to being seen resulted in higher patient satisfaction than self-referral. Of note, this study also reported significantly lower staff satisfaction.8
Input or demand
An understanding of the demand profile and number of patients with primary care conditions expected to attend an ED is fundamental to forecasting the resources they require and its associated cost. There is significant variation in the reported estimates of primary care suitable patients attending EDs.27 ,28 ,29 This difference may also be due to the increasing proportion of primary care cases over time.2 Some of the variation is undoubtedly due to the conceptual issue of what could be treated by a GP or ED clinician and what is best treated by a specific specialty. A review of the literature on methods for categorising ED visits as either urgent or non-urgent found 51 methods of doing so (17 conducted prospectively in triage). Comparisons of methods of categorisation in the same population showed variability in levels of agreement, further highlighting a lack of reliability and reproducibility.30
Supply (provider)-induced demand
In health economics, supplier-induced demand describes the amount of demand that exists beyond what would have occurred in a market in which patients are fully informed.31 In the context of unscheduled care in a health system that is free at the point of access, it is perhaps more accurate to describe ‘provider-induced demand’; it can be simply put as—‘if you build it they will come’. In other words, once resources or services are available, overutilisation will occur even if outcomes or quality is uncertain.32 Although this term is more often used to describe the behaviour of clinicians when faced with professional uncertainty or financial targets, it can also describe health-seeking behaviour and demand for unscheduled healthcare; for example, when barriers to access are removed.
In a study of the impact of a WiC on demand on other services, an increase in attendances at a minor injury unit (MIU) was attributed to the fact that the MIU was co-located with the WiC.33 Point estimates in another study of co-located WiC found a non-significant increase in throughput at sites with EDs co-located with a WiC.7 Aggregate data would support this theory with a significant increase in total attendances at all types of unscheduled care facilities since the introduction of WiCs, but with no change in the trend of ED attendances.34
Caution must be exercised, therefore, in estimating the proportion of patients suitable for screening out into a primary care service as this is likely to increase once the service is operational. This effect is likely to be compounded if any new service does not accept patients 24 h/day as the additional patients attending out of hours may worsen the burden on the ED.
Throughput and process
Much of the theoretical basis for removing primary care-type patients from the ED is predicated on their use of resources, which are better directed towards emergencies. This idea appears intuitively robust since as input into the system is reduced, the amount of resources per patient increases. This reduction, therefore, should have a positive impact on WTs and throughput times. Although the evidence suggests that a reduction in these indices is apparent with the addition of primary care service/clinician,18 ,20 this finding is not universal. A large multisite Canadian study by Schull et al35 analysed four million ED visits and demonstrated that low-complexity ED patients are associated with a negligible increase in ED LOS and WT for other ED patients. It follows that reducing the number of low-complexity (primary care) ED patients is unlikely to reduce waiting or throughput times for other patients. A Cochrane review based on three of the UK studies in this review concluded that there was insufficient evidence of any effectiveness on crowding or flow by provision of care to non-urgent patients by GPs rather than ED clinicians.36
The risk of bias being introduced is greater in single-site studies, for example, due to the Hawthorne effect,37 where simply the knowledge of being evaluated increases productivity. However, a simpler reason for the improved performance by the addition of a primary care service was because of the extra clinicians introduced. Arguably, this has the same effect, regardless of the type of clinician, and occurs simply because there are more clinicians managing the same input.
Assuming that any new service is resource neutral (in order to fulfil the overall mandate of long-term cost savings), there is little evidence of improvement in throughput from streaming primary care attendances out of the general ED population.
Although there is some evidence of increased resource utilisation when primary care patients are seen in an ED, the magnitude of this overuse is questionable. The Primary Care Foundation reported that most UCCs stated a cost per case of £28–40; however, they could not obtain reliable data to determine a true cost comparison with the ED.6
Although there is evidence of cost-effectiveness of using primary care clinicians in two European studies,18 ,25 their reference costs were significantly different from UK costs, and from each other. In the most recent multicentre English study, no difference in costs per patient was found,7 while the savings reported by two earlier studies were for processes only and used variable and ill-defined reference costs.12 ,23 The ED operates with high fixed costs and relatively low marginal costs. The average cost would, therefore, decrease as the number of patients increases.38 This variability in the magnitude of any cost saving is consistent with findings in the North American and Australian literature, with the suggestion that the true costs of non-urgent care in the ED are relatively low.39 ,40
Capital and recurrent costs add significantly to the stated cost per case in any new service. The potential savings from a diversion of non-urgent visits to primary care are, therefore, likely to be much less than is widely believed.
Some consideration of the destabilising economic impact of removing work from the ED is important. In settings in which ED budgets are based on activity, lower ED volume may result in reduced income but with little or no change in operational cost as these are relatively fixed.24
Patient choice and satisfaction
There is little evidence of increased patient satisfaction from ED-based primary care services, with one study finding that a high percentage of patients actually expressed a preference for care in an established ED compared with that in a new co-located WiC.26 Dissatisfaction was mainly related to the environment, pace and case-mix of the ED, which understandably is not conducive to less urgent needs. Of note, the only study looking at a fully integrated ED/primary care system found no change in patient satisfaction, but significantly reduced staff satisfaction associated with the new service.8 Although this could be explained by the natural response to change,41 it emphasises the importance of ensuring that the new model is sound and does not risk alienating the workforce. Disengaged staff can potentially result in lower productivity and higher patient dissatisfaction.
By blurring the line between emergency and primary care by co-locating services, there is a risk of losing the continuity of care that primary care provides and encouraging ad hoc health-seeking behaviour. This is likely to lead to confusion, longer pathways and lower degrees of satisfaction with the services being used.42
The review was not undertaken as a systematic review as it was apparent that the variability and context of interventions and outcomes became lost in the rigidity of the method. This may have led to some bias in the selection and use of retrieved articles. However, the broad, purposive search strategy allowed a more inclusive range of citations to be used and ensured that the developing evidence was framed within context and reported outcomes. We undertook no formal assessment of quality due to the marked variability in study design, setting and outcomes. This allowed a fuller understanding of the individual components of the models presented.
Unfortunately, this review has no proposed solution to the problem of increasing unscheduled care demand and the system-wide effects this has. There is a suggestion that the solution of a co-located UCC has not taken into account the complexity of the underlying issues43 and perhaps the focus should be on other contributory factors. This review could not consider all of these issues, but supports a fuller exploration of some of the assumptions and theoretical benefits of UCCs in the broader context of the issues facing the health community, such as the aging population, access and integrating health and social/community care for example.
There are significant and unexpected consequences of simply transferring interventions that work in one setting without an understanding of context and the process of change.44 Much of the impetus for implementing co-located UCCs stems from an accepted theoretical basis supported by individual examples of success in other settings. The evidence base, however, suggests that the expected benefits of the introduction of such a service are not a given, with variable outcomes reported.
There is evidence of the unexpected consequence of a paradoxical increase in demand driven by co-locating services that are meant to reduce such demand. Similarly, the theoretical cost savings are not as expected when subject to closer scrutiny, particularly as most reported savings are based on marginal costs without consideration of capital or indirect expenditure. Evidence for a model that is cost neutral across the health community is lacking. Any proposed model, therefore, requires robust evaluation before implementation.
Patients are generally good at deciding where to access care, and inappropriate choices are generally a function of complex socioeconomic factors and shortcomings in the unscheduled care system.3 People attend the ED by choice—removing this option may be unacceptable and more readily addressed by simplifying, adapting and resourcing unscheduled care to manage all potential service users across the system.
Contributors SR conceived the idea and wrote the first draft manuscript. SR and AR conducted the literature searches. SM,SA, CO and AR contributed substantially to subsequent content and drafts.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.