Review results

Results of the literature search

The Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) flow diagram (Figure 1) details the study selection process.

FIGURE 1 The PRISMA flow diagram. Adapted from Chambers et al.1 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. This figure includes minor additions and formatting changes to the original text. (Figure not reproduced here; see the original article.)

After deduplication of the searches from the different databases, there were 8870 references from the main database search and 550 from the telemedicine/telehealth search. Citation searching of Web of Science identified 52 new references and Google Scholar identified 63 new references.

All titles and abstracts were screened by one of the review team with a subset (about 10%) of the titles and abstracts being screened by two reviewers. Any queries were resolved by discussion. A similar process was followed for final decisions on inclusion/exclusion based on full-text documents. Calculation of the kappa coefficient demonstrated good agreement between reviewers [K = 0.729, 95% confidence internal (CI) 0.542 to 0.916]. Studies were excluded at the full-text stage for a variety of reasons (e.g. they covered access to services generally and not emergency care in particular; the intervention was not relevant, e.g. public access defibrillators; or the study discussed changes to services without relating outcomes to travel time or distance).

Characteristics of included studies

We included 44 publications in the review, reporting 43 unique studies. Of these, 12 evaluated the effects of an intervention or change to the health-care system (reconfiguration studies); 30 examined associations between travel distance or time and outcomes in the absence of a specific intervention (association studies); and two evaluated interventions to mitigate the effects of being at distance from an emergency care facility (mitigation studies).

Reconfiguration studies

Table 1 summarises the basic characteristics of the included reconfiguration studies. Six of the included studies were conducted in the USA, with just two2,4 being from the UK. The remaining studies were conducted in other European countries; there were no studies in this group from Canada, Australia or New Zealand.

Six of the included studies focused on ED reconfiguration, providing data on patients with many different types of emergency conditions. Three looked specifically at patients with AMI requiring access to percutaneous coronary interventions (PCI). Two studies examined the effects of service changes involving specialist trauma centres and one looked at the effects of maternity unit closures in France (see Table 1).

The studies used a variety of observational designs, with before–after and cohort designs being most common. Knowles et al.2 and Mustonen et al.11 were the only studies with independent control sites where no reconfiguration had taken place.

Association studies

The largest group of association studies focused on AMI (10 studies), followed by trauma and maternity (seven studies each). Four studies examined general populations and three dealt with stroke [including one study also included in the myocardial infarction (MI) group15]. One study included people with ruptured abdominal aortic aneurysms (AAAs).

Six studies were derived from UK settings. These included two studies of stillbirth risk and travel time in Cumbria;16,17 studies of risk and travel time for life-threatening conditions18 and specifically for MI19 and for ruptured aortic aneurysms;20 and a study of the relationship between distance and social deprivation as influences on ED attendance.21 Other studies were conducted in the USA, Canada, Japan and various European countries. Most studies used a cohort or cross-sectional design with comparisons between distance or time categories and no separate control group. Sample size and study duration varied substantially between studies, with the largest covering whole-country populations, for example Switzerland or the Netherlands. Tables 2–5 summarise the study characteristics.

Mitigation studies

We identified two studies that met our inclusion criteria and provided evidence regarding interventions that might mitigate the effects of being at a distance from an ED (Table 6). The interventions evaluated were first a specialist stroke ambulance and second a co-ordinated state-wide initiative to improve care for rural patients experiencing an AMI. These two studies represented contrasting approaches to mitigation. The specialist stroke ambulance study45 was performed in an urban setting (in Berlin) and the intervention appeared to be of value in reducing time to thrombolysis for patients located within 18 minutes’ travel time from the ambulance base, making its relevance to more rural areas uncertain. The Wyoming intervention of co-ordination between hospitals and other health-care providers covered a largely rural state with long distances between UEC facilities.46 The study reported a reduction in total ischaemic time following introduction of the new system.

A study of public access defibrillators47 was fully considered for eligibility as being potentially mitigating. However, in the final analysis this study was excluded on the grounds that survival of out-of-hospital cardiac arrest was unlikely to be affected by distance to the ED.

We examined the use of telehealth as a potentially mitigating intervention by means of a separate review of systematic reviews (see Chapter 4).

Risk-of-bias assessment

Results of the risk-of-bias assessment are presented in Tables 21–23 (see Appendix 3). Many of the studies were inherently at high risk of bias because of lack of an independent control group. In the reconfiguration group, the most common design was before–after and only four studies compared outcomes between settings with and without changes in distance/time.2,5,9,11 Association studies generally used a cohort or cross-sectional design (see Table 2).

With regard to each question on the evaluation tool, the following comments can be made about the overall quality of the included literature. Most of the included studies were clear about the temporal relationship of the variables of interest (i.e. which was the ‘cause’ and which the ‘effect’; Q1), although the issue was sometimes confused by the use of linked data sets. Similarity between populations being compared (Q2) varied across the studies, with some being clinically heterogeneous. Similarly, it was often unclear whether comparison groups were being treated similarly other than the intervention or exposure of interest (Q3). This was related to differences over time as well as to studies recruiting clinically diverse populations. Absence of a separate independent control group (Q4) was noted in most of the studies and similarly few studies carried out measurements at multiple time points before and after an intervention or exposure (interrupted time series design; Q5). Completeness of follow-up (Q6) did not show a clear pattern across studies; this question was not applicable for studies with measurement at a single time point. Most studies measured outcomes in a standard (Q7) and reliable (Q8) way, although again some exceptions were identified. Statistical analysis (Q9) was judged to be appropriate with the exception of one study.42 However, as with all observational studies, the possibility of unmeasured confounders affecting the results could not be ruled out.

Results by condition

For each condition, the reconfiguration studies are considered first, followed by the association studies, concluding with an overall summary for each group.

General urgent and emergency care

Of the nine studies of general UEC populations, five focused on adults only.2,7,9,18,42 Four studies included data on patients of all ages,4,8,11,21 but only one reported separate data for adults and children.21

Reconfiguration studies

Six studies evaluated the effects of reconfigurations that increased distance to the ED for general UEC populations. One study9 differed from the others in that it involved subgroups of people with specific conditions (i.e. MI, stroke, sepsis, asthma or chronic obstructive pulmonary disease) rather than general populations of people attending the ED. Details of the methods and findings of the studies are summarised in Table 7.

Five of the studies reported at least some data on patient outcomes (including hospital admissions) associated with ED reconfiguration or closure.2,4,8,9,11 Two of these were controlled studies2,11 and two were from UK settings.2,4 The three studies that reported on mortality2,9,11 found no evidence of increased mortality rates associated with ED reconfiguration. A limitation of this evidence is the fact that one study measured only inpatient mortality9 and another only documented mortality as a secondary outcome.11

The other main patient outcome reported in these studies was hospital admissions, reported in three studies.2,4,8 Hansen et al.8 found a reduction in hospital inpatient admissions in a municipality where the ED changed from 24-hour to daytime only and then closed, requiring residents to travel further to use ED services. However, this reduction was only seen in women and its clinical significance was unclear. In the UK ClosED study, there was no consistent effect on emergency admissions.2 The study found some evidence of a decrease in emergency admissions and admissions considered potentially avoidable, but this was not considered statistically reliable. In their case studies of reconfiguration in the English NHS, Roberts et al.4 presented limited data on emergency admissions. In Burnley, these admissions remained broadly constant despite an increased distance to the nearest ED, while in Kent emergency admissions increased in areas where new facilities were introduced. The limited evidence from these studies makes it difficult to draw firm conclusions about the effect of ED reconfigurations on hospital admissions.

Most studies also reported health system outcomes. The most short term of the included studies7 reported on the effects of closure of an ED in Boston, MA, USA, on a neighbouring ED. This study only measured outcomes related to the ambulance service for 6 weeks before and after the closure. There was a significant increase in ambulance turnaround times (time the ambulance is out of service and unable to respond to another call) and a decrease in the volume of patients attending the remaining ED but no change in the volume of patients transported by EMS vehicles. Further research would be needed to assess whether or not these changes were sustained in the longer term. In the UK, Knowles et al.2 found evidence of an increased burden on ambulance services (increased activity, e.g. increases in emergency ambulance incidents) following closure or downgrading of EDs, reflecting increases in the distances patients had to be transported to reach the nearest ED.

Knowles et al.2 found no consistent effects of ED closure or downgrading on UEC attendance. There was some indication of a decrease in attendances on average across the sites studied but this was not considered statistically reliable. A similar pattern (statistically unreliable evidence of a decrease) was found for attendances by patients who were discharged without treatment or further investigation.2

Two studies in Scandinavian countries investigated patients’ use of alternative services following closure of a local ED. Hansen et al.8 (Denmark) found that patients used services located up to 40 km away. The authors noted that free access to ambulance services and high levels of car ownership meant that distance was not a serious barrier to people accessing urgent or emergency care. Similarly, in Finland the availability of alternative secondary care ED facilities was considered to protect against any risk of increased mortality following closure of a local primary care ED.11 This study reported a negative relationship between distance and ED use by local residents, suggesting that ED closure can be used to encourage patients to access other more appropriate services.

Overall, the reconfigurations evaluated in these studies had relatively small effects on travel times/distance and hence few negative effects on outcomes for patients were reported in the studies that evaluated them. However, even in a relatively small country such as England, there are localities where travel distances are considerably longer than average and where service reconfigurations might be expected to have most impact. Furthermore, general populations of people attending an ED include many people with less time-sensitive conditions who, by definition, are at lower risk of being negatively affected by increased travel times. Effects on health system outcomes varied across studies and settings but the included studies did not suggest that ED closure or reconfiguration causes significant problems for neighbouring services. There was, however, evidence of an impact on local ambulance services, as highlighted by El Sayed et al.7 and Knowles et al.2

Association studies

Only three included studies examined the association between travel times and outcomes in UEC populations not restricted by condition (Table 8). Two of these involved patients with diverse but serious conditions,18,42 whereas the third examined all visits to EDs in the West Midlands region of England.21

Nicholl et al.18 and Murata et al.42 both focused on in-hospital mortality and both found an association between distance to hospital and mortality risk. Nicholl et al.18 looked at people transported by ambulance who were considered to have a potentially life-threatening condition. Distances travelled ranged from 0 to 58 km (median 5 km). Increased risk of death was expressed as an odds ratio (OR) (1.02 per km, 95% CI 1.01 to 1.03 per km).18 Murata et al.42 expressed the association between distance and mortality as a correlation coefficient (R). Values of R for AMI and brain infarction (stroke) were 0.315 (p < 0.01) and 0.233 (p < 0.001), respectively.42 The studies differed in their populations of interest: Nicholl et al.18 studied people with life-threatening conditions whereas Murata et al.42 focused more broadly on ‘acute diseases’ and included people with non-emergency conditions, such as pneumonia. The small number of studies and differences in populations and statistical methods make it difficult to gain an overall estimate of the strength of the association.

The study by Rudge et al.21 examined the association between distance and ED attendance, an important health system outcome, and the influence of socioeconomic circumstances on this association. This study analysed data on over 1.4 million ED visits in the West Midlands during the financial year 2007–8. The authors’ regression model indicated that ED attendance declines with distance by 1.5% (95% CI 1.2% to 1.8%) for adults and 2.2% (95% CI 1.7% to 2.6%) for children. Levels of ED attendance increased with increasing socioeconomic deprivation. ED attendance was also affected by proximity to a minor injury unit (MIU), suggesting that provision of MIUs can modify ED attendance. ED attendance was more sensitive to increased distance in deprived neighbourhoods compared with less deprived areas. This large single study adds to the findings of studies in the reconfiguration group2,8,11 by assessing the role of deprivation in the complex relationship between distance and ED attendance in the general UEC population.

Acute myocardial infarction

Reconfiguration studies

Two studies, reported in three publications, investigated the effects of reconfigurations or closures that increased travel distance on outcomes for patients with AMI,5,12,13 and Table 9 summarises these studies. The two papers by Shen and Hsia12,13 appear to use the same data, and the 2016 paper13 may be considered as an update to the 2012 publication,12 although the later paper does not cite the earlier one.

All of the publications reported some negative effects of increased travel distance on mortality of patients with AMI. Avdic5 compared areas of Sweden with and without ED closures over the period 1990–2010. Overall, there was an increase in mortality of approximately 2 percentage points for each extra 10 km of travel distance. The effect was driven by out-of-hospital deaths and was most pronounced in the first year after an ED closure. This led the author to suggest that ambulance services in particular need to be prepared for the impact of ED closures.

Shen and Hsia in their two publications12,13 analysed data on driving times and outcomes associated with ED closures in the whole of the mainland USA. The first paper12 covered the period 1996–2005. The majority of the study population (89.2%) did not experience any increase in driving time to the nearest ED during the study period. There was a small increase in mortality associated with small increases in driving time (up to 10 minutes), the effect being strongest for those with limited access to hospitals at baseline (e.g. in rural or mountainous areas). Increases in driving time of > 30 minutes were associated with increased long-term (up to 1 year) mortality rates, although increases of this magnitude only affected 0.2% of the study population. Most changes were relatively short term and mortality rates 3 years after ED closure were generally similar to those before the closure.

The later publication from these authors13 analysed similar data for the period 2001–11. This publication presented information on increases in absolute mortality risk. Patients whose driving time increased by 10–30 minutes had an increased risk of 90-day mortality by 1.6 percentage points (95% CI 0.53 to 2.67 percentage points) and 1-year mortality by 2.05 percentage points (95% CI 0.96 to 3.14 percentage points). Corresponding figures for those with > 30 minutes longer driving time were 6.58 percentage points (95% CI 2.49 to 10.68 percentage points) and 6.52 percentage points (95% CI 1.69 to 11.35 percentage points), respectively. The authors noted that 6.58 percentage points represented a 30% relative increase in 90-day mortality in this population.

Association studies

Ten included studies (Table 10) explored the relationship between ED distance and MI outcomes in the absence of reconfiguration. Of these, eight reported the presence of an association and two reported unclear findings. Six studies investigated mortality alone or in combination with other outcomes. Other main outcomes related to travel time or distance were type of treatment strategy, adverse events during transportation and total ischaemic time (time from symptom onset to start of treatment, which was evaluated in two studies). All the studies used a cohort-type design.

Two studies by Balamurugan et al.,23,24 set in Arkansas, USA, analysed data on AMI cases from 2008 to 2012 and from 2012 to 2013, respectively. The earlier study23 focused exclusively on the relationship between driving time to the nearest PCI-capable hospital and MI mortality. Mortality risk in patients from the neighbourhoods with the longest driving distances were 26% higher than those with the shortest distances after adjusting for a range of confounding factors. For comparison with UK studies, the median geodesic distance (shortest path) and driving time were 12.8 miles [interquartile range (IQR) 3.6–30.1 miles] and 28.3 minutes (IQR 9.6–58.7 miles), respectively. This compares with a median distance of 5 km (3.1 miles) in a typical UK study.18 The second study by these authors24 looked at patients who survived to discharge as well as those who died and considered a range of factors affecting mortality risk, including comparing hospitals with and without a continuous PCI capability. The authors found a 9.2% increase in the odds of death for every 10 miles of additional distance to the nearest hospital (with or without continuous PCI capability) (OR 1.092, 95% CI 1.009 to 1.181). Admission to a hospital without continuous PCI capability was the other factor associated with increased odds of death (OR 1.64, 95% CI 1.15 to 2.34).

A large national cohort study in Switzerland (4.5 million participants and > 19,000 AMI deaths over 8 years) also found an association between driving time to hospital and risk of mortality from AMI.15 The association was strongest for driving time to a university hospital [hazard ratio (HR) for highest vs. lowest quintile 1.19 (95% CI 1.10 to 1.30) for men and 1.10 (95% CI 1.01 to 1.20) for women aged > 65 years)]. This study also examined stroke mortality as discussed below (see Stroke).

A smaller study (3608 patients) in Italy looked at patients with ST elevation myocardial infarction (STEMI) who underwent PCI between 2009 and 2013.26 The authors defined system delay on treatment (PCI) as travel time to hospital plus door-to-balloon time (DTBT) within the hospital. Among those with ≤ 90 minutes of symptom delay, in line with current guidelines, travel time above the median (14 minutes) was associated with significantly higher odds of 30-day mortality (OR 2.46, 95% CI 1.25 to 4.86). There was a significant interaction between travel time and DTBT, such that patients with DTBT below the median only experienced a survival benefit if travel time was also below the median.

ST-elevation MI was also the focus of a small study by Svensson et al.29 in Sweden, who evaluated treatment with a thrombolytic agent (reteplase) in the ambulance before hospital admission. The median interval between the onset of symptoms and the start of thrombolysis was 1 hour 44 minutes in urban areas versus 2 hours 14 minutes in rural areas (p = 0.03). Median time of transport from the scene of the incident to hospital was 17 vs. 33 minutes. There was no difference between the urban and rural groups in 30-day mortality but the rural patients had more symptoms of heart failure. One-year mortality was higher in the rural patients (17 vs. 5%; p = 0.02).

The only UK study in this group was performed by Wei et al.19 in Tayside, Scotland. They analysed data on 10,541 patients with AMI between 1994 and 2003. Survivors were followed for a median of 3.2 years. After adjustment for known risk factors, distance from home to hospital was associated with risk of death before arrival at the hospital and after discharge but not with death in hospital. Compared with < 3 miles, ORs for pre-hospital death were 2.05 (95% CI 1.00 to 4.21) for > 9 miles and 1.46 (95% CI 1.09 to 1.95) for 3–9 miles. The corresponding adjusted HRs for death during follow-up were 1.90 (95% CI 1.19 to 3.02) and 1.27 (95% CI 0.96 to 1.68).

In summary, all of the studies that evaluated mortality found a positive association between increasing travel time/distance and increased mortality following AMI. Variation in populations, settings, statistical methods and length of follow-up make it difficult to estimate an overall effect measure, as discussed further in Strength of evidence.

Turning to other outcomes, Langabeer et al.27 and Postma et al.28 related distance to total ischaemic time [i.e. time from symptom onset to start of treatment (in hospital)]. For patients with STEMI undergoing PCI in the study by Langabeer et al.,27 total ischaemic time and travel distance were higher for patients who were transferred to a different hospital to receive PCI than for those who did not need to be transferred. However, over the whole sample, distance was not significantly associated with total ischaemic time. A similar finding was reported by Postma et al.28 for STEMI patients undergoing PCI in the Netherlands: a longer distance from home to hospital was associated with longer total ischaemic time in patients referred for PCI from a non-PCI ‘spoke’ hospital but not for those transported by ambulance directly to the PCI centre.

The PCI must be performed promptly for optimum results (90 minutes for first-contact patients and 2 hours for transferred patients).22 Andersson et al.22 analysed data on > 27,000 patients who underwent PCI for STEMI in Michigan, USA, between 2010 and 2016. They examined the relationship between transport time to a PCI-capable hospital and choice of treatment (reperfusion) strategy, comparing primary PCI with a drug therapy strategy. Use of the latter decreased during the study period and overall 95% of patients received primary PCI. However, the study demonstrated a clear relationship with distance as primary PCI was used in 97% of patients living < 1 hour from the hospital but only 48% of those with estimated transport times over an hour. Clinical outcomes did not differ significantly between the two strategies.

Transport to hospital involves a risk of adverse events causing a patient’s condition to deteriorate. Bussières et al.25 studied patients with STEMI who were transported to a PCI centre by paramedics in a rural part of Québec, Canada. Increased transport time (comparing 0–14, 15–30 and > 30 minutes) was not associated with an increased risk of clinical adverse events during transport, but patient age was identified as a significant risk factor in this study.

In summary, most of the included studies suggest that increased distance or travel time is associated with increased risk of mortality for patients with MI (some studies specifically recruited those with STEMI). Transfer between hospitals may increase risk by lengthening delays in starting treatment (increased total ischaemic time) and time to reach hospital may influence the choice of treatment strategy. However, the one study that looked at adverse events during transport found no evidence of a link to transport time.25

Mitigation studies

Improved co-ordination of services may improve quality of care for patients in rural areas and mitigate the effect of being at a distance from an ED or specialist centre. A before–after study by Langabeer et al.46 evaluated an initiative in the rural US state of Wyoming and reported increased use of primary PCI and reduced total ischaemic time for patients with STEMI (see also the study by Svensson et al.29 discussed above under Association studies, involving pre-hospital electrocardiogram (ECG) and administration of a thrombolytic agent in the ambulance).

The role of telehealth as a mitigation intervention is discussed in Chapter 4.

Trauma

Reconfiguration studies

Two included studies dealt with reconfigurations affecting trauma care,10,14 and both looked at the closure of trauma centres in California, USA, and their impact on remaining centres (Table 11). The earlier study, by Yaghoubian et al.,14 covered Los Angeles County, whereas Hsia et al.10 analysed data for non-federal trauma centres across the whole state. The time periods covered by the studies also differed (1997 to 2006 for Yaghoubian et al.,14 and 1999 to 2009 for Hsia et al.10). Nevertheless, there is likely to be a degree of overlap in the data, which should be taken into account in interpreting the results.

Both studies identified short-term effects on patient outcomes following trauma centre closures. Yaghoubian et al.14 reported increases in monthly patient volume and injury severity in the year after closure. However, after adjusting for injury severity, the mortality rate was lower in the post-closure period than before and the complication rate was unchanged.14 The authors attributed the relatively favourable outcomes to increased experience resulting from the increased number of patients attending the centre. Importantly, this study also reported on changes to the staffing and organisation of the trauma centre that are also likely to have contributed to successful adaptation to increased demand. Changes included addition of a dedicated nurse practitioner, four surgery residents and an additional trauma attending physician. The resident trauma staff were reorganised into three smaller teams with a view to distributing patient care more evenly. Advances in critical care during the study period were also likely to have played a role in improving outcomes.14

The state-wide study by Hsia et al.10 was distinctive in that the authors were able to identify (based on postal codes) patients whose travel times were affected or unaffected by trauma centre closures. The key finding of the study was that odds of inpatient mortality were significantly higher for patients with increased driving time compared with those experiencing no change (OR 1.21, 95% CI 1.04 to 1.40). The increased odds of death was even more pronounced in the 2 years after a closure (OR 1.29, 95% CI 1.11 to 1.51). By contrast, patients whose driving distance to a trauma centre decreased during the study period had lower odds of death compared with the ‘no change’ group (OR 0.83, 95% CI 0.75 to 0.92). These differences in outcomes were observed against the background of an overall decrease in trauma mortality in California during the study period.10

The results of these two studies suggest that reconfiguration in the form of trauma centre closures may negatively affect outcomes for patients with potentially life-threatening trauma. This effect may be mitigated by improved quality of care and more efficient organisation of services in remaining centres but these centres are likely to experience increased demand and possibly an increase in the proportion of more severely injured patients. This evidence comes from the USA, where services are largely delivered by private providers in an insurance-based system. The English NHS has an organised system of major trauma networks and lower trauma centre capacity than the USA, which makes reconfiguration involving trauma centres unlikely.

Association studies

Seven studies examined the association between distance/time and outcomes for trauma patients in the absence of reconfiguration (Table 12). Most studies used a cohort design and all were conducted in North America (six in the USA and one in Canada). The majority of included studies found a clear association between distance to a trauma centre and outcomes (primarily mortality) for trauma patients.30–34 These studies were performed in rural or mixed settings (often covering a large area such as the whole of a US state) and used a variety of distance/time measures (see Table 12).

Only two studies did not find an association between distance/time and outcomes. In the study by Lerner et al.,35 the time measure was total out-of-hospital time rather than travel time per se. A possible explanation suggested by the authors is that more severely injured patients were prioritised for rapid transport to hospital, leading to those with less severe injuries and a higher probability of survival spending more time in reaching the hospital. In an urban setting, McCoy et al.36 found no association between transport time and mortality for patients with either blunt or penetrating trauma. The authors suggested this may be because patients transported to trauma centres have higher survival rates than those taken to other facilities despite longer transport times.

Maternity

Reconfiguration studies

Only one study from the reconfiguration group examined reconfiguration affecting maternity services.6 In the largely rural region of Burgundy, France, the number of maternity units decreased from 20 to 15 between 2000 and 2009. Mean estimated journey time to the nearest maternity unit increased from 21 to 24 minutes and the maximum time increased from 61 to 72 minutes. The study found that a travel time of ≥ 30 minutes was associated with an increased risk of fetal heart rate anomalies, meconium-stained amniotic fluid, unplanned out-of-hospital births and pregnancy-related hospitalisation. Odds of perinatal death were increased for those travelling furthest (≥ 46 minutes) but the number of deaths was small and the CIs of the OR were wide (OR 1.86, 95% CI 0.69 to 5.01).

A limitation of this study was that it did not distinguish between areas with and without maternity closures. As such, it was primarily measuring the association between travel time and outcomes rather than the effect of increased travel times following closure of the nearest unit.

Association studies

In addition to the Combier et al.6 study, seven studies examined associations between travel time to maternity units and maternal or perinatal outcomes (Table 13). Three studies found evidence of an association, including large national cohort studies from Norway37 and the Netherlands.41 The third study to find an association was conducted in rural British Columbia, Canada,39 and the long travel times involved make the findings less applicable to the UK setting.

By contrast, two publications from Cumbria found no evidence of an association between distance to hospital and risk of stillbirth.16,17 It appears that there is some overlap of data between these two publications, with the later report16 covering a longer time period and a larger population (the whole of Cumbria compared with West Cumbria in the earlier publication). Although from a UK setting, these studies are of limited value because the included date range is unlikely to reflect current practice and service delivery. A further study from the USA that found no evidence of an association looked specifically at very low birth weight infants (500–1499 g).38 The authors noted that in this study a high proportion of neonatal deaths occurred within 24 hours of birth and were likely to be related to factors other than access to care.

Finally, a study by Pilkington et al.40 examined data for the whole of France for the period 2001–8. In contrast to the Combier et al.6 study from Burgundy, these authors found that women living closer to a maternity unit had a higher risk of neonatal mortality. They attributed this to many maternity units being located in socioeconomically deprived areas where the risk of poor outcomes is likely to be higher.

In summary, the findings of the review for maternity and neonatal populations were inconclusive. The relationship between distance and outcomes varied between studies, reflecting differences in populations and settings as well as possible changes over time.

Stroke

Reconfiguration studies

None of the reconfiguration studies looked solely at patients with stroke. Hsia et al.9 presented results by condition for a variety of UEC conditions. Patients with stroke who experienced an increase in distance to the nearest ED did not have higher in-hospital mortality compared with those who experienced no increase.

Association studies

Three association studies reported data for patients with stroke (Table 14).15,43,44 Acharya et al.43 and Leyden et al.44 both focused on timely receipt of thrombolytic therapy as a major outcome and both concluded that patients living further from stroke units or hospitals with appropriate facilities were less likely to receive thrombolytic therapy. The Swiss National Cohort study reported by Berlin et al.15 analysed data on stroke deaths. In comparison with the situation with MI (reported in the MI section of this chapter, see Acute myocardial infarction), the relationship between mortality risk and driving time to a university hospital was less pronounced for stroke; the results varied for different age and sex groups and also by the type of statistical model used.

The review thus provides inconsistent evidence on the relationship between distance and mortality for patients with stroke. There is limited evidence from two studies suggesting that more distant patients are less likely to receive prompt thrombolytic therapy.

Mitigation study

The only included mitigation study evaluated the effectiveness of a specialist stroke ambulance in an urban setting (Berlin) and concluded that benefits from reduced time to thrombolysis justify a specialist ambulance up to a travel time of 18 minutes.45

The role of telehealth as a mitigation intervention is discussed in Chapter 4.

Other conditions

One included study looked at the relationship between travel time and survival for patients with a ruptured AAA, which is a condition with a high pre-hospital mortality rate.20 This study in West Sussex, England, found no relationship between travel time and mortality risk. After adjustment for confounders, the OR for survival associated with a 10-minute increase in estimated travel time to the nearest hospital was 0.97 (95% CI 0.70 to 1.34), the wide CI suggesting uncertainty about the true effect. The study authors noted that earlier studies (mainly dating from outside the search period for this review) had found a relationship; they suggested that such studies were biased by the omission of deaths that occurred before reaching hospital.20 Given that Souza et al.20 published their study in 2005, the possible impact of changes in management of ruptured AAA should be taken into account in interpreting the study results.

Summary of reconfiguration studies

Most of the included reconfiguration studies reported on changes in mortality rates following reconfiguration (Table 15). For studies of general UEC populations, there was no evidence that reconfiguration resulting in increased travel time/distance affected mortality rates. This was classed as stronger evidence being derived from studies with control groups. By contrast, there was evidence of increased risk from studies restricted to patients with AMI. Evidence for other conditions was inconsistent or very limited and none of the included studies looked at stroke patients specifically (although people with stroke were an identifiable subgroup in the study by Hsia et al.9).

Evidence on health system outcomes was inconsistent (Table 16), reflecting the diverse outcomes and settings included. El-Sayed et al.7 identified some short-term negative effects on EMS systems following closure of an ED in Boston, MA, USA. More significantly, the closED study in the UK (England) identified some evidence of increased pressure on ambulance services in areas where an ED had been closed or downgraded, including an increase in emergency ambulance incidents.2 Across England as a whole, distances travelled to access emergency care increased only slightly between 2001/2 and 2011/12.4 Case studies analysed by Roberts et al.4 demonstrate how reconfigurations affects travel distances and hospital admissions in specific local contexts.

Summary of association studies

The association studies found evidence that increased travel time or distance is associated with increased mortality risk for the AMI and trauma populations, whereas for maternity the evidence was inconsistent (Table 17). Where evidence was classified as ‘weaker’, this reflects the absence of true control groups but the large size of some of the studies and plausibility of the association are additional supporting factors. There was also weaker evidence of an association from two studies of patients with a range of conditions typically requiring emergency care.18,42

Other outcomes (Table 18) were reported in smaller numbers of studies. Although evidence for these outcomes was inconsistent for patients with AMI, weaker but consistent evidence was found for adverse maternity outcomes and access to thrombolysis for stroke patients being influenced by distance from specialist services.

Quantitative estimates of the relationship between travel distance or time and mortality risk are summarised in Table 19. Studies varied widely in their methodology. In particular, authors calculated either an effect measure per unit of distance (e.g. 10 miles or 10 km) or between different distance categories (e.g. highest vs. lowest quintile). This, together with the variety of different outcomes measured (e.g. in-hospital mortality, 30-day mortality or mortality at various follow-up points) makes it difficult to pool outcomes across studies.

As an example, two studies from the UK18,19 used different distance measures (km vs. distance categories in miles) and measured mortality at different time points, as well as studying different populations (general life-threatening conditions vs. AMI). Qualitatively, the increase in risk with distance appears greater for the AMI population, but interpretation is complicated by the fact that differences between distance categories were statistically significant for pre-hospital and follow-up mortality but not for deaths in hospital.19 There was more uniformity in UK-based maternity studies, but these were conducted by the same research team in broadly the same setting16,17

Summary of mitigation studies

The two included mitigation studies provide very limited evidence for the effectiveness of improved service co-ordination for STEMI in a US rural setting46 and of a specialist stroke ambulance in an urban setting.45

Results by setting

UK versus international

Possible limitations to the relevance of international evidence to the UK setting need to be taken into account in interpreting the study findings and this topic will be addressed in Chapter 5.

Only two of the 12 reconfiguration studies were conducted in the UK (see Tables 1 and 2). Despite this, UK studies provided the larger part of the evidence related to mortality risk for this patient group following service reconfiguration. Other reconfiguration studies came from western European countries (i.e. Sweden, France, Denmark and Finland) and the USA (six studies). In terms of specific conditions, all the reconfiguration studies of patients with AMI and trauma came from the USA, which differs considerably from the UK in size and health system organisation.

Turning to association studies, the UK was relatively well represented in studies of general UEC populations (two out of three) with included studies covering mortality18 and effects of distance on ED attendance.21 By contrast, only 1 out of 10 studies on AMI came from the UK (Scotland).19 All of the included studies on trauma came from the USA (six studies) or Canada (one study).

Maternity was one area where there was a suggestion of a difference between UK and non-UK studies, with two linked studies from Cumbria16,17 finding no evidence of a link between distance to hospital and neonatal death, in contrast to a number of international studies. There was one UK study of ruptured aortic aneurysm, providing the very limited evidence on this condition included in the review.20

Some European health systems place particular emphasis on pre-hospital diagnosis and triage, for example by doctors working alongside paramedics. We did not include any studies of this type of model in relation to distance, but differences in service organisation should be taken into account when assessing study applicability to the UK context.

There were no included UK studies for stroke or for mitigation interventions other than telehealth.

Urban versus rural

The effects of service reconfiguration and the relationship between distance/time and outcomes may differ between rural and urban areas. Some studies included in the review focused on the needs of rural areas whereas others compared outcomes between rural and urban areas.

None of the reconfiguration studies of general UEC populations was conducted in specifically rural settings but several covered whole countries or US states, including both urban and rural areas. Roberts et al.4 analysed data for England as a whole and highlighted areas where distances travelled to access emergency care were considerably higher than the national average. Unfortunately, the detailed data on distances travelled in this report were accompanied by limited outcome data from a few case studies, making it difficult to draw firm conclusions about rural versus urban differences. The closED study also included sites in diverse areas of England but discussion of issues specific to rural areas was limited.2 Reconfiguration studies for AMI and trauma also mainly covered large areas and provided limited information specific to rural settings.

Combier et al.6 reported on the impact of maternity unit closures in a largely rural part of France. Closure of a number of units between 2000 and 2009 was associated with increases in mean (21 to 24 minutes) and maximum (61 to 72 minutes) estimated journey times. Increased travel time was associated with increased risk of various adverse outcomes (see Maternity). By contrast, two studies of births in a rural English county (Cumbria) found no evidence of an increase in risk of perinatal death associated with increased distance to hospital.16,17 This UK evidence should be interpreted cautiously because of the length of time since the data were collected but advances in perinatal care in recent years give no reason to suspect that the situation is likely to have changed. The only other specifically rural study included (conducted in British Columbia, Canada) is likely to be of limited relevance to UK settings because of the large distances involved.39 The overall evidence for distance to maternity units in rural areas may thus be classified as ‘inconsistent’.

Much of the evidence from association studies of patients with AMI relates to rural areas, including US states that are mainly rural such as Arkansas23,24 and Wyoming,27 and a rural region of Québec, Canada.25 This evidence was again inconsistent in terms of presence versus absence of an association.

The only UK study in the MI association studies group was conducted in Tayside, Scotland, by Wei et al.19 This study found an association between distance and mortality (including both pre-hospital and follow-up mortality). In addition, the authors classified patients’ home addresses into eight rurality categories based on postcode data. The results suggested that outcomes may be influenced by rurality as well as distance per se. The study authors noted that the rurality code may be a surrogate measure of different patient behaviour in different geographical settings (i.e. people in rural areas may behave differently in terms of reacting to symptoms and seeking help).

For trauma patients, a number of studies covering rural or mixed areas of the USA (often whole states) found associations between mortality risk and distance to hospital (generally a trauma centre rather than a normal ED).30–33 The exception was a study by Lerner et al.35 in New York state, which found that total out-of-hospital time was not associated with mortality risk. There were no UK studies in this group and the evidence should be interpreted with this in mind.

Finally, three studies of stroke patients conducted in mixed settings shed some light on urban–rural differences. Two studies focused on timely receipt of thrombolysis. In St Louis city and St Louis county, Missouri (USA), patients living further from the hospital were less likely to receive thrombolysis, but the difference was not explained by differences in time taken to reach the hospital.43 The others suggested that other factors such as type of transport to the hospital or patient awareness of available services may have been important. In a study in South Australia, access to thrombolysis was described as poor in both urban and rural areas.44 In the report of the Swiss National Cohort study, the authors speculated on possible differences between rural and urban populations, including lower awareness of stroke symptoms and more reliance on helicopter transport in rural areas.15

In summary, factors other than increased distance to hospital that may affect the relationship with outcomes have been tentatively identified in a few studies. Only a few studies have focused on UK rural populations or compared rural and urban populations.

Health system context and implementation of service changes

All of the studies included in the reconfiguration group mentioned the context in which changes to services are implemented and the need for planning and co-operation to minimise the impact on patients. However, only six of them provided any substantial information. Broader implications for health services are considered in Chapter 5 (see Implications for service delivery).

The closED study authors sought information about relevant service changes through a documentary analysis of material from NHS trusts, ambulance trusts and commissioning organisations (clinical commissioning groups and primary care trusts).2 No information on how the health service planned for the impact of ED closures or downgrading was obtained from either NHS annual reports or ambulance service sources. Data analysed in the study suggested that ambulance services made changes in anticipation of ED closures as increases in average journey times were lower than expected. The authors identified understanding the processes used by health-care providers and commissioners to minimise adverse effects of implementing service reorganisation as a priority for future research. The study also called for improved consistency of recording and coding of routine data.

In the context of the UK NHS, the case studies provided by Roberts et al.4 provided examples of UEC reconfigurations carried out in England and their effect on travel times. However, there was no background information on the rationale for and implementation of the changes.

Some additional evidence came from non-UK studies. The most detailed information was reported by Yaghoubian et al.14 in their before–after study of a trauma centre closure. Changes made in anticipation of increased patient numbers included additional staffing and changes to team organisation. Increased experience and background improvements in care were identified as factors that supported an improvement in adjusted mortality rates in the post-closure period. A ‘mature’ regional trauma system ensured that remaining trauma centres were able to handle additional patients.

In another study from the USA, El-Sayed et al.7 commented on the importance of early notification of a planned ED closure and ‘continuous and dynamic co-operation’ between hospital and EMS administrations in minimising the impact on EMS services.

In addition to early planning, there was evidence of health systems adapting during the first few years after a reconfiguration. Avdic5 noted that the effect of ED closures on AMI mortality was only statistically significant for the first year, attributing this to attempts to compensate for increased travel distance by, for example, increased investment in emergency services or greater use of preventative care. Shen and Hsia12 identified a 4-year transition period after which outcomes were similar to the pre-reconfiguration period.

Moderating and mediating factors

Key moderating and mediating factors extracted from included studies are summarised in Figure a in relation to patient and health system outcomes. Not all relevant studies are included in the figure, but fuller details of influencing and explanatory factors extracted from included studies are presented in Appendix 4, Table 24.

A point not stressed elsewhere in this report is the influence of measurement factors on the reported travel time/distance to hospital. The included studies differed in their methodological approach according to the availability of data. For example, data on the exact location of incidents were often not available and the patient’s residential postcode or similar was used instead. Studies also differed in whether they measured direct distance to a UEC facility or attempted to assess distance or time using the actual road network.

Patient-related and context-related influences on pre-hospital time identified in Figure a were often found in similar forms in multiple studies. In particular, optimal deployment of EMS services and the choice between use of private or ambulance transport can substantially influence the time taken to reach hospital. Service-related factors that influence outcomes for patients range from pre-hospital initiation of treatment to the range and location of specialist facilities available. These can potentially mitigate the effects of travelling further for emergency care.

Included studies have also evaluated impacts of service changes on broader health system activity and performance, including primary care11 and alternative UEC facilities8 as well as indicators of ambulance service quality.2