|Accueil | Aperçu | Revues | Soumettre | Nous Contacter | English|
Rapid access to emergency services is essential for emergency care sensitive conditions such as acute myocardial infarction, stroke, sepsis, and major trauma. We sought to determine US population access to an emergency department (ED).
The National Emergency Department Inventories (NEDI) – USA was used to identify the location, annual visit volume, and teaching status of all EDs in the US. EDs were categorized as 1) any ED, 2) by patient volume, and 3) by teaching status. Driving distances, driving speeds, and prehospital times were estimated using validated models and adjusted for population density. Access was determined by summing the population that could reach an ED within the specified time intervals.
Overall, 71% of the US population has access to an ED within 30 minutes, and 98% has access within 60 minutes. Access to teaching hospitals was more limited, with 16% having access within 30 minutes and 44% within 60 minutes. Rural states had lower access to all types of EDs.
Although the majority of the US population has access to an ED, there are regional disparities in ED access, especially by rurality. Future efforts should measure the relationship between access to emergency services and outcomes for emergency care sensitive conditions. The development of a regionalized emergency care delivery system should be explored.
Time-sensitive interventions – such as coronary revascularization in acute myocardial infarction1, 2, fibrinolytic therapy for acute ischemic stroke3, early goal directed therapy in sepsis4, and trauma center care for severe injury5 – highlight the importance of timely, universal access to care for emergency care sensitive conditions. Receipt of these interventions is contingent upon access to an appropriately equipped and staffed emergency department (ED) or an ED that can triage, stabilize, and rapidly transport patients to more definitive care.
Understanding which EDs have adequate resources to care for patients with emergency care sensitive conditions is difficult. A common perception is that higher volume teaching EDs located at referral hospitals provide more comprehensive care than their smaller community counterparts.6 No centralized data collection system exists that characterizes ED capabilities or the resources available within their parent hospitals. This lack of knowledge regarding individual EDs is an important barrier for researchers and health services planners alike. Researchers cannot determine the relationship between access and resources for emergency care sensitive conditions, and planners cannot develop systems to efficiently deliver patients to the most appropriate level of care.
Although details in terms of ED capabilities and resources are as yet unavailable for the nation, the basic distribution and characteristics of EDs in the United States (US) has been recently compiled.7 To date there have been no population-based estimates of access to these EDs. We therefore sought to generate national estimates of access to various types of EDs within 30, 45, and 60 minutes. These findings have important policy implications for future ED categorization, credentialing, and the regionalization of emergency care.8
The traditional framework describing access to care includes five important domains: 1) availability, 2) accessibility, 3) accommodation, 4) affordability, and 5) acceptability.9 For the subset of conditions requiring prompt intervention to optimize outcome, accessibility to emergency care represents the most important domain. A basic understanding of the location of emergency care facilities relative to the population requiring prompt access to these resources is an essential fist step in developing an intelligently designed emergency care delivery system.
ED information was obtained from the National Emergency Department Inventories (NEDI). NEDI is an Emergency Medicine Network (EMNet)10 project that maintains an inventory of all EDs in the United States. NEDI was developed in response to the absence of a centralized mechanism by which to identify ED factors potentially associated with patient outcomes, including annual number of ED visits and presence of post-graduate residency training programs. The 2003 NEDI-USA database was built on the framework of the 2003 American Hospital Association Annual Survey of Hospitals, the 2003 Verispan Marketing Group’s Hospital Profiling Solution Database, and independent data collected by the EMNet group.7
In order to calculate access to specific types of EDs, all sites were characterized in terms of annual visit volume and teaching status. In our previous work7 we categorized EDs as “higher” volume if they treated more than 1 patient per hour, 24 hours per day, 365 days a year. Here we maintain this description but further classify higher volume EDs based on hourly census and distinguish EDs that treat an average of 1+ patients per hour (≥ 8,760 patients/year), 2+ patients per hour (≥ 17,520 patients/year), and 3+ patients per hour (≥ 26,280 patients/year). Teaching hospitals were identified by membership in the Council of Teaching Hospitals (COTH).11 To be classified as a teaching hospital by the COTH, hospitals must have a documented affiliation with an accredited medical school, and must sponsor, or participate significantly in at least four approved active residency programs, not necessarily including an emergency medicine residency.
Population information was obtained using data from the US Census Bureau12 local and state estimates, and trends in deliverable addresses from the US Postal Service (Claritus Inc, Ithaca, NY).13 Our main geographic units of analysis were block groups. A block group is a geographic unit containing 600 to 3,000 people that does not cross state or county boundaries. Each block group’s population was assigned a point in space (a centroid) that was nearest to most of its residents. Population estimates and population-weighted centroids for 208,649 block groups were calculated for 2003. The location (longitude-latitude coordinates) of these population-weighted centroids were then compared with the longitude-latitude coordinates representing EDs. Block group population access calculations were aggregated to compute estimates of access for the entire country, the four Census Bureau regions (Northeast, Midwest, South, and West), the 9 Census Bureau divisions, and all 50 States and the District of Columbia. State population totals and state estimates of rurality were based on US Census Bureau calculations. Rural was defined as territory, population and housing units not classified as urban.
Access was calculated by summing the population of the block groups that could reach an ED by ground ambulance within the specified pre-hospital time period. We chose to generate access estimates for 30 minutes, 45 minutes, and 60 minutes as these time intervals provide a reasonable range of transport times that might still permit timely intervention in critical diseases. Each block group was linked to the nearest ED, and redundant access to nearby EDs was not considered in these analyses. The populations or land areas of block groups that could reach an ED within the time period specified were never counted more than once in the summation formula for access. Similarly, we did not account for the fact that many block groups could be assigned to the same ED. All programming code was written and tested using Compaq Visual Fortran Version 6.6 (Compaq Computer Corporation, Houston, TX), and then translated into C++ (Microsoft Corporation, Seattle, WA) and then validated by comparing output from the Fortran code with output from the C++ code.
To calculate ambulance driving times, we used an average urban driving speed of 20.1 mph, an average suburban driving speed of 47.5 mph, and an average rural driving speed of 56.4 mph.14 Drives were classified as urban, suburban, or rural by averaging the population densities (residents per square mile) of the block group surrounding each ED and the block group of origin and then determining whether this average population density fell into the highest, middle, or lowest third among all US block groups. The population densities of intervening block groups were not considered. We then added 1.4, 1.4, and 2.9 minutes to account for the average time from receipt of emergency call to departure in urban, suburban, and rural areas, respectively.14 An additional 13.5, 13.5, and 15.1 minutes in urban, suburban, and rural areas, respectively, were added to account for the average time spent at the scene14 as has been done previously.15 For our access calculations, we permitted the crossing of state lines to arrive at the closest ED.
We could not explicitly determine the location of ground ambulance depots. We therefore estimated the time from ground ambulance depot to the scene by multiplying the time from scene to ED by an empirically determined constant: 1.6, 1.5, and 1.4 for urban, suburban, and rural drives, respectively, to account for differences in travel time between the geographic regions. All driving distances were estimated using previously validated mathematical models of actual road travel.16, 17 These distance estimates have been previously demonstrated to correlate well with actual road travel distances in the US. This analysis was considered exempt from full review by the institutional review board at the University of Pennsylvania.
The 2003 NEDI-USA database identified 4,809 hospitals with general receiving EDs; the sum of all ED visits was 113.3 million visits, which is consistent with the 113.9 million ED visit estimate from the sample of hospitals that participate each year in the US National Hospital Ambulatory Medical Care Survey.18 The median number of annual visits was 18,089. About one-third (n=1,358) of EDs treated fewer than 8,760 patients per year. Of the 3,451 EDs that saw 8,760 or more patients per year, about one-quarter (29%) were in a nonurban setting.
Overall, 71% of the US population has access to an ED within 30 minutes, 94% within 45 minutes, and 98% within 60 minutes. The Northeast had the greatest access within 30 minutes (76%), followed by the West (71%), the Midwest (70%), and the South (68%). Access was much less variable within 60 minutes, ranging from 99.5% in the Northeast to 97% in the West (Table 1, Figure 1).
Overall, access to EDs with higher volume (1+ visit/hour) was similar with 68% of the population having access within 30 minutes, 90% within 45 minutes, and 95% within 60 minutes. A greater range of access was observed across census regions for these EDs with the Northeast having the greatest access within 30 minutes (76%) followed by the West (68%), Midwest, and South (65%). Access estimates for 60 minutes were similar, ranging from 99.1% in the Northeast to 92% in the Midwest (Table 1, Figure 2). Population access to EDs treating 1+, 2+, and 3+ patients per hour is displayed in table 2.
Access to teaching hospital EDs was much more limited, with only 16% of the population having access within 30 minutes, 32% within 45 minutes, and 44% within 60 minutes. Access to a teaching hospital ED demonstrated the most regional variability. The Northeast had the largest population with access within 30 minutes (31%), followed by the Midwest (17%), South (12%), and West (9%). Access to a teaching hospital within 60 minutes ranged from 36% in the South to 67% in the Northeast. (Table 1, Figure 3).
Statewide access to any ED within 30 minutes ranged from 48% (VT) to 86% (DC), and access to any ED within 60 minutes ranged from 81% (AK) to 100%. Statewide access to higher volume EDs ranged from 33% (SD) to 86% (DC) for 30 minutes, and 45% (SD) to 100% for 60 minutes (Table 1). Access to EDs treating 1+, 2+, and 3+ patients per hour is displayed for each state in table 2.
Access to teaching hospitals varied considerably among states, with a range of access between 0% (AK, ID, MT, NV, WY) to 72% (DC) access for 30 minutes, and 0% (AK, ID, MT, NV, WY) and 100% (DC) access within 60 minutes. In the large majority of states (98%), half of the population did not have access to a teaching ED within 30 minutes, and in three-quarters (73%), less than half of the population had access within 60 minutes (Table 1).
Finally, we examined the relationship between rurality and timely access to emergency services. Even in the four states with more than 50% of the population living in rural areas, population access within 60 minutes remained good (97% to 99% for any ED, and 93% to 98% for a higher volume ED). There was more variability associated with ED access within 30 minutes. A strong linear relationship existed between the percentage of the state population living in rural areas and access to any ED within 30 minutes (R2 = 0.59). Access to higher volume EDs within 30 minutes for the most urban states ranged from 74% to 86%, and from 45% to 54% for the most rural states.
We provide a population level analysis of access to EDs. As such, the estimated travel times may not be directly applicable on the individual level. We believe our inventory of EDs to be comprehensive, however it remains possible that EDs have been omitted. We think it is more likely, however, that we have included clinics that self-describe as an “emergency department” but that are not open round-the-clock or to all comers – key characteristics of an ED. This possibility would lead us to overestimate ED access. It is also possible that EDs with volumes near the volume thresholds that we defined may have enough annual variability to lead to some misclassification bias. In addition, ED closure, hospital closure, and change in COTH affiliation since 2003 may have resulted in overestimations or underestimations of population access.
Our study has potential limitations related to our access estimates. We calculated access based on where people live, not where they might be at the time of onset of symptoms requiring emergency services. Our estimates may underestimate population access as many Americans live in remote areas but may often spend time in more densely populated regions with better ED access. The drive times in our study were derived from a meta-analysis of prehospital times for trauma, and ambulances with medical emergencies may not travel at the same speed as ambulances with trauma patients. Medical patients may also be more likely to arrive by private vehicle than the injured population. There may also be limitations associated with our assumptions about ambulance care in the US. We assume adequate prehospital coverage throughout the US despite the fact that there is tremendous variability in the efficiency of prehospital systems with many rural areas depending on volunteer services. Geographical barriers such as rivers and mountains were not taken into consideration in the analysis. We believe that although this may be a concern in small unit analysis (i.e. a county or city), this would not impact the state and national estimates that we present here. We did not include non-geographic factors in our driving analyses including limited 911 service, inclement weather, traffic congestion, or rates of ED closure and/or diversion. We hope that by using actual ambulance drive times specific to population density, we have partly controlled for these unmeasured confounders. We did not consider the use of helicopters to transport patients to the hospital as has been done in the trauma population as most medical diseases are not regionalized using prehospital triage criteria. Finally, we did not take into consideration redundancy of ED access, and we did not attempt to control for ED crowding and/or diversion. In some highly crowded regions, redundancy is likely essential, whereas in other areas, redundancy may represent unnecessary inefficiencies and/or a maldistribution of resources. Despite these potential shortcomings, our analysis provides the first and only benchmark for population access to EDs.
We provide national, regional, and statewide access estimates to of EDs by volume and by teaching status. We demonstrate that the large majority of the US population has access to an ED within 60 minutes, and demonstrate variability by region and state for higher volume EDs and teaching EDs. Living in a rural area is a key driver of these results as we observed a strong linear relationship between the population of a state living rurally, and overall resident access to emergency care.
The central hypothesis of emergency care is that rapid diagnosis and early intervention in acute illness or acutely decompensated chronic illness improves patient outcomes.19 The time sensitive nature of many conditions including MI, stroke, sepsis, and major trauma have been well described.1–5 Professional organizations have developed guidelines intended to standardize the care of these conditions, including the appropriate use of specialized levels of care, such as cardiac catheterization and interventional radiology.20–23 Despite this, fundamental details about the emergency care system, including population access to EDs and the resources available upon ED arrival, are largely unknown. A notable exception to this knowledge gap is the trauma system.24, 25
The US trauma system has been recognized by the Institute of Medicine as a model for the rapid delivery of emergency care.8 Trauma centers are embedded within EDs, but are independently accredited to meet the core explicit requirements necessary for immediate diagnostics and treatment of severe injury. Trauma systems have used principles of operations research26 to ensure that severely injured patients are rapidly delivered to an appropriate facility either by allowing emergency medical service providers to bypass the closest hospital, or via stabilization outside of the trauma center followed by rapid interhospital transfer. These systems improve outcomes5, and allow for intelligently planned expansion to ensure optimal patient access to trauma care. The development of systems to ensure adequate access to appropriately resourced facilities for other, non-trauma emergency care sensitive conditions has been limited.
In addition to the regional variability in ED access, we demonstrate a relationship between ED access and rurality. These findings are not surprising as we expect dense urban areas to have more hospitals and shorter transport times. However, there are unique challenges associated with attracting emergency physicians27 and practicing emergency medicine in rural areas28 that are related to the importance of these findings. The disparity in emergency care access that we describe may be associated with poor outcomes for time critical diseases. Our data support efforts to advance testable pilot solutions to increase population access to emergency services for rural Americans. These solutions include subsidizing rural hospitals29, providing physician incentives to provide services at these hospitals30–32, identifying specialty centers33–36, and improving prehospital and inter-hospital referral networks8, 23, 37, 38; it is likely that multiple approaches will be needed to improve inequities in access. An improved understanding of the role that physical location of an ED, the resources of the ED, and the mechanism by which patients can be efficiently delivered to the appropriate setting is essential in the ongoing planning and development of the US healthcare system.
In conclusion, we provide estimates of access to emergency care in the US. The vast majority of Americans can arrive in a timely manner to an ED, but there are geographical inequities in timely access to emergency services and gaps in our knowledge of where appropriate interventions for emergency care sensitive conditions are available. Future efforts should be directed towards better understanding the capabilities of EDs to provide comprehensive emergency services. An ED categorization scheme that addresses these issues, coupled with our access data, would provide a framework for the development of regionalized care delivery systems for emergency care sensitive conditions.
The study was supported by the Robert Wood Johnson Foundation and the National Library of Medicine (R21LM008700)
We are grateful for the statistical and programming support of Justin Williams, Jianguo Li, Karl Dailey, Vicky Tam, and Marlen Kokaz.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.