Potential bypassing of nearest emergency department by EMS transports

Abstract

Objective

Guidelines recommend emergency medical services (EMS) patients to be transported to the nearest appropriate emergency department (ED). Our objective was to estimate the prevalence of EMS transport to an ED other than the nearest ED (“potential bypassing”).

Data sources

Illinois Prehospital Patient Care Report Data of EMS transports (July 2019 to December 2019).

Data collection/extraction methods

We identified all EMS ground transports with an advanced life‐support (ALS) paramedic to an ED for patients aged 21 years and older. Using street address of incident location, we performed geocoding and driving route analyses and obtained estimated driving distance and time to the destination ED and alternative EDs.

Main outcome and measures

Our main outcomes were dichotomous indicators of potential bypassing of the nearest ED based on distance and time. As secondary outcomes we examined potential bypassing indicators based on excess driving distance and time.

Study design

We used Poisson regression models to obtain adjusted relative rates of potential bypassing indicators by acuity level, primary impression, patient demographics and geographic characteristics.

Principal findings

Our study cohort of 361,051 EMS transports consisted of 5.8% critical, 37.2% emergent and 57.0% low acuity cases transported to 222 EDs. The observed rate of potential bypassing was approximately 34% of cases for each acuity level. Treating the cardiovascular primary impression code group as the reference case, we found small to no differences in potential bypassing rates across other primary impression code groups of all acuity levels, with the exception of critical acuity trauma cases for which potential bypassing rate was 64% higher (incidence rate ratio = 1.64, 95% confidence interval, 1.54–1.74). Compared to zip codes with one ED within a 5‐mile vicinity, potential bypassing was higher in areas with no ED or multiple EDs within a 5‐mile vicinity.

Conclusion

Approximately one‐third of EMS transports potentially bypassed the nearest ED. EMS transport destination may be motivated by factors other than proximity.

What is known on this topic

• Emergency medicine guidelines recommend that acute emergency medical services (EMS) patients be transported to the nearest appropriate emergency department (ED).

• Limited evidence from regionalized EMS systems indicates that nearest appropriate EDs are bypassed frequently.

What this study adds

• Using administrative EMS data from Illinois covering the full spectrum of EMS transports, we found that approximately one‐third of transports potentially bypassed nearest ED.

• Potential bypassing rate was similar across all acuity levels and most clinical conditions, higher in areas (zip codes) with either no ED or multiple EDs in the vicinity, and higher among Black and Hispanic patients.

• Ambulance transports to the ED may be motivated by factors other than proximity and time, such as, patient preference and provider determination.

1. INTRODUCTION

Emergency medical services (EMS) response, care and transport remains a vital conduit to acute care. Nearly 80% of EMS patients are transported to an emergency department (ED), ¹ and represent almost 16% of the 145 million ED visits nationally (2016). ² ED patients who arrive by EMS have a four‐fold higher risk of hospitalization (23.4%) compared to other ED patients (5.9%). ² Yet, little is known about factors that determine the ED destination choice of EMS‐transported patients. The American College of Emergency Physicians recommends that patients “should be transported to the nearest appropriate ED in accordance with applicable laws, regulations, and guidelines.” ³ However, the extent to which proximity—as measured by either transport distance or time—determines ED destination is largely unknown. Limited evidence for patients with suspected major trauma, stroke and ST‐segment elevation infarction (STEMI) indicates that EMS bypassing of the nearest appropriate hospital is common. ⁴ , ⁵ , ⁶ , ⁷ , ⁸

A better understanding of the motivations for ED destination is important because EMS transport affects the timeliness of patient intervention and the hospital where care is received. While bypassing may increase time to initiation of care, a preference‐based bypass to the patient's “home” hospital where they received prior care may contribute toward better clinical assessment. Systems of care directives may also influence ED destination. ⁹ For instance, Illinois state law states that “a person shall not be transported to a facility other than the nearest hospital, regional trauma center or trauma center unless the medical benefits to the patient reasonably expected from the provision of appropriate medical treatment at a more distant facility outweigh the increased risks to the patient from transport to the more distant facility.” ¹⁰ Even where the timeliness of treatment is not affected, EMS transport decisions can affect the destination of the hospital selected and level of care provided. In urban areas with multiple EDs in the vicinity, EMS transport can influence the facility where patients obtain care, even if there is limited impact on transport distance or time. Understanding EMS transport patterns can inform the systematic concentration of demographic subgroups in some hospitals. ⁷ , ¹¹ , ¹² , ¹³

Our aim in this study is to quantify the role of proximity as a determinant of the destination ED. Using a rich population‐level administrative database that includes data on incident location, patient acuity level and primary impression, we estimated the prevalence of potential bypassing of the nearest ED for a wide spectrum of EMS transports. These data provide a more definitive assessment of potential bypassing relative to previous work based on insurance claims data. ⁷ , ⁸ Claims data lack information on patient incident location (often limited to zip code), thereby introducing error in measuring travel distance, particularly in urban areas with multiple EDs in close vicinity. ⁷ , ⁹ , ¹⁴ Claims data lack information on initial patient complaint and acuity. ⁷ , ⁹ Using street address of incident location, we performed geocoding and driving route analyses to estimate driving distance and time to alternative ED destinations and indicators of potential bypassing of the nearest ED destination. We characterize population‐level variation in the rates of potential bypassing by patient acuity, primary impression, socio‐demographics, and geographic characteristics.

2. METHODS

2.1. Data and study population

We obtained the Illinois Prehospital Patient Care Report Data of EMS transports from July 1, 2019 to December 31, 2019 from the Illinois Department of Public Health. ¹⁵ By state mandate, all EMS providers submit abstracts of all emergent EMS transports. The abstracts contain incident street location, time, disposition, destination facility, patient primary impression and acuity, and patient demographics. ¹⁶ We included only 911 transports to a hospital or freestanding ED and excluded transports to behavioral health and psychiatric hospitals, children's hospitals, federal medical centers, and out‐of‐state destinations. We included ground transports with an advanced life‐support (ALS) paramedic for patients aged 21 years and older, and excluded transport types identified as “interfacility transport” or “standby” (Table S1). As in many states, Illinois has introduced regionalization of transport and care of patients with suspected STEMI to selected hospitals, although it has not been adopted by all regions and EMS systems. Minimum guidelines for the transport and care of patients meeting criteria for stroke and serious trauma are, however, explicitly set forth in Illinois law. ¹⁷ , ¹⁸ , ¹⁹ They are thus in effect in all regions and EMS Systems although they may be expanded upon beyond the state‐imposed minimums at these sub‐state levels. We identified the subgroup of such transports to compare potential bypassing relative to other transports.

EMS agencies in Illinois are publicly and privately owned, although a majority are “housed within municipally‐funded fire departments.” ²⁰ Oversight of their operations is governed by state statute and state administrative codes, and administered by several entities, including the Illinois Department of Public Health's Division of EMS and Highway Safety, EMS medical directors and disciplinary review boards.

2.2. Transport distance and time

Our objective was to estimate the driving distance and time to the actual destination in comparison with alternative ED destinations. With 222 EDs in the study data, we obtained the estimates in two steps. First, using the latitude and longitude of each incident location and all the EDs, we obtained the shortest geographic distance (“as the crow flies”) and selected the nearest five EDs. ²¹ Second, we obtained the estimated driving time and distance to the five EDs; if the actual destination ED was not among the top five EDs, we obtained the estimated distance and time for the destination ED. We used the two‐step approach to economize on costs since the second step incurs a charge per route search. We used CDX ZipStream to authenticate all street address locations and ArcGIS StreetMap Premium (along with ArcGIS Network Analyst) to obtain the estimated driving time and distance. ²² , ²³ The estimated driving routes and times are based on historical traffic data and the driving speeds represent the average over the recent 2 years for typical “peak‐time” (7 a.m. to 7 p.m. on weekdays) and “non‐peak‐time” conditions. ²⁴

2.3. Potential bypassing indicators

Our main outcomes were two dichotomous indicators of potential bypassing (0/1) based on the comparison of the actual destination ED with the nearest ED, one based on driving distance and the other on driving time (Figure S1). In the absence of established benchmarks, we also defined secondary indicators based on the extent of excess transport distance and time. Specifically, based on driving distance, we defined two dichotomous (0/1) indicators of potential bypassing wherein the estimated driving distance to the actual ED destination exceeded the nearest ED destination by 3 or 5 miles. Similarly, we defined two indicators of potential bypassing using excess travel time thresholds of 5 or 10 min.

2.4. Covariates and subgroups

Patient acuity at incident scene was based on two fields indicating the initial and final acuity assessment by the EMS paramedic. Precedence was given to the initial assessment field, wherein patient status was categorized as “Critical,” “Emergent,” or “Lower acuity.” If the initial assessment was not recorded (30% of cases), then the final assessment field was used. The cases with both assessments missing were excluded (see Appendix S1). Provider's primary impression field indicated the patient condition as an ICD‐10‐CM code or as free text. All the conditions (N = 910) were reviewed and grouped into 11 categories by consensus among three co‐authors (JS, WEB and JF): behavioral health, cardiovascular, gastroenterology, general complaint, infectious disease, neurology, pain, respiratory, substance use, major trauma, and others. We used patient demographic information on age, sex and race/ethnicity. We identified several measures of geographic differences. First, using the geocoded location of all EDs, we identified the number of EDs within a 5‐mile vicinity of each zip code. Second, to examine the relative distance between proximate EDs, we defined the difference in the distance (or time) to the nearest and second‐nearest ED. We categorized the geographic location into four rural–urban categories based on the zip code of pickup location (metropolitan, micropolitan, small town and rural). ²⁵ As cases in Chicago constitute a sizable share of overall cases, we identified such cases in all our analyses. To identify the subgroups of patients suspected of STEMI, stroke and serious trauma we used the EMS protocol field.

2.5. STEMI cohort

Not all EDs may be appropriate destinations based on the provider's assessment of patient needs and the ED/hospital capabilities. As an illustrative case, we examined ED destination patterns for patients with ST‐elevation myocardial infarction (STEMI) as guidelines recommend direct transport to a hospital performing percutaneous coronary intervention (PCI) procedures; we note that not all EMS regions in Illinois had a STEMI plan in 2019. ⁶ , ¹⁹ , ²⁶ For the subgroup of STEMI patients in our data we obtained indicators of potentially bypassing the nearest PCI‐capable hospital. As a determinant of potential bypassing we identified a dichotomous indicator of whether the nearest ED to the patient pickup location was PCI‐capable. As potential destination EDs, we identified the subgroup of Illinois hospitals with PCI‐capability in the Illinois Inpatient Discharge data (2018). ²⁷ We used Agency for Healthcare Research and Quality (AHRQ)'s Inpatient Quality Indicators (IQI) protocol to calculate PCI volume at each hospital using International Classification of Diseases (ICD‐10) procedure codes. ²⁸ Hospitals with at least one PCI procedure performed were categorized as PCI‐capable.

2.6. Reason for ED destination

The EMS abstract includes a paramedic‐reported reason for ED destination field, with values including, “closest facility” and “patient's choice.” We examined potential bypassing rates across cases by the reported reason overall and by race/ethnicity.

2.7. Analysis

We obtained summary rates of potential bypassing by acuity status for EMS transports grouped by patient demographics, primary impression and availability of multiple EDs in the zip code vicinity. We examined the characteristics of the cases within and outside Chicago. We used multivariate regression models to obtain adjusted differences in the frequency of potential bypassing across transport subgroups by patient and location characteristics (see the Estimation Models section in Appendix S1). Our preferred specification was the Poisson model as the estimates—in terms of incidence rate ratio—indicate the relative risk of potential bypassing in subgroup comparisons (Equation (S1a) in the Estimation Models section in Appendix S1). This model was estimated for all main and secondary outcomes and for all acuity subgroups. For all models we obtained robust standard errors clustered at the EMS provider level. ²⁹ As supplementary analysis, we used fixed effects version of the Poisson model to estimate relative risk among transports within the same hospital service area (HSA) and served by the same EMS provider (Equation (S1b) in Estimation Models in Appendix S1). We used the main Poisson model to estimate differences in potential bypassing rates for the subgroup of patients suspected of STEMI, stroke and serious trauma. We examined the STEMI cohort separately and estimated the likelihood of transport to the nearest PCI‐capable center.

To assess the sensitivity of the estimates to regression model specification, we used linear probability and logit regression models of relative differences in the frequency of potential bypassing (see the Estimation Models section in Appendix S1). We assessed the sensitivity of estimates to (a) alternative grouping of acuity based on the availability of initial and final acuity measures, and (b) exclusion of trauma cases. As a complementary assessment of the decomposition of the variance in potential bypassing among transports within and between geographic areas (HSA) we used a variance components model to estimate the share (%) of the overall variance in potential bypassing due to within‐area differences (see the Estimation Models section in Appendix S1). A similar approach was used to obtain the share (%) of overall variance among patients served by the same EMS provider. Data processing and statistical analyses were conducted using Stata version 16.1 during July 2020–October 2021. ³⁰ We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. ³¹

3. RESULTS

Our overall study cohort of 361,051 EMS transports to 222 EDs were grouped into three acuity levels: critical (5.8%), emergent (37.2%), and low acuity (57.0%) (Table 1). Over 12% of patients were 85 and older. Race/ethnicity information was missing for over 12% of the cases. In nearly 28% of the cases, there were no EDs within a 5‐mile vicinity of the zip code of the incident location, and in 13.9% of the cases there were four or more EDs. Primary impression of cardiovascular, neurological, respiratory, and trauma complications accounted for over 48% of all, and 74% of critical, cases (see Table S2a). The difference in distance (and time) between the nearest and the second‐nearest ED is the smallest in areas with multiple EDs in the 5‐mile vicinity and in metropolitan areas (Table S2b). We note that the median distance (and time) is longer in areas with one ED (4 miles and 7.5 min) than in areas with no EDs (2.0 miles and 2.6 min). Similarly, the distance and time were longer in micropolitan areas than those in metropolitan and rural areas. The demographic profile of patients within and outside Chicago differed, particularly in the racial/ethnicity composition (Table S2c). Non‐Hispanic White patients comprised 59.4% of cases outside Chicago and 24.3% of Chicago cases. Race/ethnicity was missing in 16.9% of cases outside Chicago and 0.5% of Chicago cases. In 88.1% of Chicago cases there were two or more EDs within a 5‐mile vicinity of the incident location; in contrast, this figure was 19.2% for cases outside Chicago.

TABLE 1.

Study population characteristics by patient acuity

Characteristic	All	Critical	Emergent	Lower acuity
Number of EMS transports (% by acuity)	361,051 (100)	20,784 (5.8)	134,315 (37.2)	205,952 (57.0)
Age group, %
21–44	27.8	23.8	22.9	31.4
45–64	30.0	30.3	29.2	30.5
65–84	29.8	33.6	34.0	26.6
85+	12.4	12.3	13.8	11.5
Female, %	51.2	44.8	51.7	51.6
Race/ethnicity, %
White, non‐Hispanic	49.4	45.9	51.8	48.1
Black, non‐Hispanic	29.7	30.8	26.1	31.9
Hispanic	7.3	7.8	7.1	7.3
Asian, non‐Hispanic	1.2	1.3	1.3	1.0
Others	0.3	0.4	0.3	0.3
Unknown	12.2	13.7	13.3	11.3
Number of EDs in 5‐mile vicinity of incident location, %
0	27.8	28.2	31.8	25.2
1	33.3	33.0	34.7	32.4
2–3	25.1	25.9	22.7	26.5
4 or more	13.8	13.0	10.8	15.9
Difference in distance to the nearest and second‐nearest ED from incident location, mean (SD), miles	3.7 (5.7)	3.2 (5.0)	3.8 (5.6)	3.6 (5.7)
Difference in distance to the nearest and second‐nearest ED from incident location, %
<0.5 mile	19.8	19.1	17.7	21.2
[0.5, 1.0) mile	15.4	15.0	14.9	15.7
[1.0, 2.0) miles	18.4	20.7	18.6	18.1
[2.0, 4.0) miles	22.3	23.5	22.9	21.8
4.0+ miles	24.1	21.6	25.9	23.1
Urban/rural location, %
Metropolitan area	90.0	93.0	89.7	89.9
Micropolitan area	5.0	2.9	4.6	5.5
Small town	3.8	3.0	4.4	3.5
Rural	0.6	0.6	0.7	0.6
Unknown	0.6	0.5	0.6	0.5
Chicago cases, %	28.6	33.3	22.5	32.1

Note: (1) As much of our statistical analyses are performed separately by acuity groups, we have provided the summary statistics for these groups. (2) We have used both the continuous and categorized version of the difference in distance to the nearest and second‐nearest ED measure in our analyses. (3) For summary statistics on the primary impression (i.e., medical complaint) see Table S2a.

3.1. Potential bypassing rates

We found that 35.0% of EMS transports were not to the nearest ED based on estimated driving distance, and this frequency exceeded 34% among all acuity levels (Table 2). A similar pattern was found using estimated driving time as the criterion. Allowing for a margin of 3 miles or 5 min of additional transport reduced the rate of potential bypassing to a range between 10.8% and 14.6% across all acuity levels. And allowing for a margin of 5 miles or 10 min reduced the potential bypassing rate range to 5.1%–6.9%. Patterns in the observed (unadjusted) potential bypassing rates across subgroups were consistent between distance‐ and time‐based indicators (Table S3).

TABLE 2.

Observed rates (%) of potential bypassing

Bypassing threshold	All	Critical	Emergent	Lower acuity
Distance
>0 miles	35.0%	38.0%	35.4%	34.5%
>3 miles	11.1%	11.3%	11.5%	10.8%
>5 miles	6.8%	6.3%	6.7%	6.9%
Time
>0 min	36.2%	39.6%	36.3%	35.8%
>5 min	12.4%	14.6%	12.9%	11.9%
>10 min	5.2%	5.3%	5.1%	5.3%

Note: Here we report the prevalence (%) of potential bypassing based whether the ED destination for the patient was the nearest (quickest) ED destination identification based on comparing the driving distance (time) to alternative EDs as described in the methods section. We obtained prevalence for alternative choice of distance and time thresholds.

Among critical acuity cases, the relative potential bypassing rate was 64% higher for major trauma cases (incident rate ratio [IRR] = 1.64, 95% confidence interval [CI], 1.54–1.74) compared to critical acuity cardiovascular cases (Table 3). The potential bypassing rate was higher in zip code locations with no ED (IRR = 1.54, 95% CI, 1.29–1.84) and locations with two or three EDs (IRR = 1.18, 95% CI, 1.02–1.38) compared to zip codes with one ED in the 5‐mile vicinity. Larger difference in the distance between the nearest and the second‐nearest ED was associated with reduced likelihood of potential bypassing; relative to patients for whom this difference was less than 0.5 mile, IRR was 0.31 for those with at least 4‐mile difference (95% CI, 0.28–0.34). Compared to metropolitan areas, transports from small towns and rural areas were more likely, and those from micropolitan areas were less likely, to experience potential bypassing. For a majority of critical acuity cases—grouped by primary impression, age, and sex—the potential bypassing rates were similar. The rate was higher among (non‐Hispanic) Black (IRR = 1.17, 95% CI, 1.06–1.29) and Hispanic (IRR = 1.11, 95% CI, 1.01–1.22) patients, relative to (non‐Hispanic) White patients. A similar pattern was found for emergent and low acuity cases; however, differences by race/ethnicity were not significant. We found largely similar patterns in potential bypassing rates for cases within and outside Chicago (Tables S4a and S4b). A notable difference was by race/ethnicity; relative to White patients, potential bypassing rates were higher among Black, Hispanic and (non‐Hispanic) Asian patients in Chicago, but the corresponding differences were not significant for patients outside Chicago. The pattern in potential bypassing defined using driving time was similar to that using driving distance (Table S4c).

TABLE 3.

Differences in potential bypassing rate (>0 mile threshold): incidence rate ratio

Subgroup	Critical acuity		Emergent acuity		Lower acuity
	IRR	95% CI	IRR	95% CI	IRR	95% CI
Reference group potential bypassing rate (%)	40.0%	[36.8%, 43.2%]	46.10%	[44.5%, 47.7%]	55.50%	[53.6%, 57.2%]
Primary impression (reference: cardiovascular)
Behavioral health	0.93	[0.81, 1.07]	0.9	[0.84, 0.97]	0.93	[0.88, 0.99]
Gastroenterology	0.94	[0.77, 1.16]	0.98	[0.93, 1.03]	0.94	[0.89, 0.99]
General complaint	1.02	[0.92, 1.13]	0.97	[0.94, 1.00]	0.95	[0.91, 0.99]
Infectious disease	1.02	[0.88, 1.19]	1.03	[0.97, 1.10]	0.97	[0.90, 1.04]
Neurology	1.03	[0.98, 1.08]	0.97	[0.94, 1.00]	0.96	[0.92, 0.99]
Pain	1.18	[1.03, 1.34]	1.06	[1.01, 1.11]	0.96	[0.90, 1.02]
Respiratory	0.92	[0.81, 1.03]	0.95	[0.91, 1.00]	0.96	[0.92, 1.01]
Substance use	0.9	[0.83, 0.97]	0.84	[0.81, 0.87]	0.83	[0.79, 0.87]
Trauma/injury	1.64	[1.54, 1.74]	1.13	[0.99, 1.29]	0.91	[0.87, 0.95]
All others	0.94	[0.85, 1.03]	1.02	[0.98, 1.06]	0.97	[0.93, 1.01]
Age (reference: 21–44)
45–64	0.98	[0.94, 1.03]	1	[0.96, 1.03]	1.04	[1.02, 1.07]
65–84	0.99	[0.94, 1.04]	1.03	[0.99, 1.06]	1.09	[1.05, 1.13]
85+	0.95	[0.89, 1.01]	0.93	[0.89, 0.98]	1.02	[0.98, 1.08]
Sex (reference: male)
Female	0.96	[0.93, 1.00]	1	[0.98, 1.02]	1.01	[0.99, 1.02]
Race/ethnicity (reference: non‐Hispanic White)
Black, non‐Hispanic	1.17	[1.06, 1.29]	1.07	[0.97, 1.18]	1.04	[0.97, 1.13]
Hispanic	1.11	[1.01, 1.22]	1.02	[0.91, 1.14]	1.04	[0.89, 1.22]
Asian, non‐Hispanic	0.99	[0.88, 1.11]	0.78	[0.72, 0.84]	0.88	[0.80, 0.97]
Others	0.96	[0.77, 1.20]	0.8	[0.65, 0.99]	0.95	[0.85, 1.06]
Unknown	0.86	[0.73, 1.02]	0.83	[0.71, 0.98]	0.78	[0.67, 0.90]
Number of EDs in 5‐mile vicinity of incident location (reference: 1 ED)
0	1.54	[1.29, 1.84]	1.61	[1.40, 1.86]	1.49	[1.29, 1.71]
2–3	1.18	[1.02, 1.38]	1.22	[1.08, 1.38]	1.09	[0.96, 1.25]
4 or more	1.17	[0.96, 1.43]	1.28	[1.10, 1.48]	1.31	[1.12, 1.53]
Difference in distance to the nearest two EDs from incident location (reference: <0.5 mile)
[0.5, 1.0) mile	0.89	[0.83, 0.96]	0.83	[0.76, 0.92]	0.77	[0.68, 0.89]
[1.0, 2.0) miles	0.7	[0.65, 0.75]	0.63	[0.52, 0.77]	0.54	[0.37, 0.79]
[2.0, 4.0) miles	0.54	[0.47, 0.62]	0.5	[0.38, 0.66]	0.45	[0.27, 0.73]
4.0+ miles	0.31	[0.24, 0.39]	0.29	[0.23, 0.38]	0.26	[0.18, 0.36]
Urban/rural location (reference: metropolitan area)
Micropolitan area	0.62	[0.29, 1.34]	0.85	[0.47, 1.53]	0.68	[0.32, 1.46]
Small town	1.27	[1.00, 1.61]	1.57	[1.22, 2.02]	1.31	[0.99, 1.72]
Rural	1.72	[1.45, 2.05]	1.61	[1.31, 1.98]	1.72	[1.35, 2.18]
Unknown	0.44	[0.28, 0.68]	0.43	[0.21, 0.84]	0.6	[0.23, 1.54]
Geography (reference: Chicago city)
Outside Chicago city	1.15	[0.98, 1.36]	0.92	[0.80, 1.05]	0.75	[0.65, 0.86]

Note: (1) Estimates reported are from a Poisson regression model of the dichotomous potential bypassing indicators based on distance (with >0 mile threshold) (see Equation (S1a) in the Estimated Models section in Appendix S1). All covariates are categorical. Covariate estimates are reported in terms of incidence rate ratio (IRR) using the reference group for each measure category. As all covariates are categorical, the IRR indicates the relative risk of potential bypassing in comparison with the reference group. For instance, IRR = 1.11 estimate for Hispanic patients (critical acuity) indicates a 11% higher risk relative to that among non‐Hispanic White patients. (2) Separate models were estimated for each acuity level group. (3) The first row indicates the potential bypassing rate for the reference cohort: Adult males aged 21–44 years, non‐Hispanic White, with a primary impression of a cardiovascular complaint, with a pick‐up location in the Chicago area with one ED within a 5‐mile vicinity, characterized as metropolitan and with a difference of <0.5 mile in the distance between the nearest and second‐nearest ED.

Potential bypassing using alternative thresholds, both by distance (3‐mile and 5‐mile excess distance) and driving time (5‐ and 10‐min excess time) yielded similarity in rates among a majority of primary impression, age and sex groups (Table S4d–g). The higher rates of potential bypassing for major trauma cases and in zip codes with no ED in the 5‐mile vicinity persisted. For the critical acuity patients suspected of STEMI, stroke and trauma (identified by EMS protocol used), the potential bypassing rate (based on distance) was higher for trauma and STEMI cases and similar to stroke cases relative to other critical acuity cases (Table S4h).

3.2. Provider‐ and area‐level clustering

Variation in potential bypassing occurred predominantly among patients served by the same EMS provider and living in the same HSA (Table 4 and S5). Within‐EMS variance accounted for 76.6% and 85.2% of the total potential bypassing variance using distance‐based and time‐based measures, respectively. Adjustment for care provided by the same EMS provider and within the same area (HSA) yielded little differences in the potential bypassing patterns by clinical and patient characteristics (Table S6).

TABLE 4.

Share of within‐cluster variance in rate of potential bypassing (distance‐based)

Bypassing threshold	Share (%) of within‐cluster variance out of total variance
	All	Critical	Emergent	Lower acuity
EMS provider clustering
>0 miles	76.6%	78.2%	75.7%	76.7%
>3 miles	71.3%	72.1%	70.5%	70.4%
>5 miles	76.1%	80.5%	75.6%	74.7%
Hospital service area clustering
>0 miles	85.2%	83.4%	83.2%	83.2%
>3 miles	74.9%	83.4%	76.2%	72.7%
>5 miles	75.9%	86.7%	76.2%	72.7%

Note: Each number indicates the proportion of the total variance in the potential bypassing indicator accounted for by the within‐EMS provider variance or within‐HSA variance. Each number was obtained from a separate linear regression model (for details see Equations (S4a) and (S4b) in the Estimation Models section in Appendix S1).

3.3. Sensitivity analysis

Alternative estimation model specifications—linear and logit—provided similar patterns of potential bypassing (distance‐based) (Tables S7a, S7b, and S8). In particular, a higher frequency of potential bypassing was found among critical acuity trauma cases, Black patients, areas with no EDs in a 5‐mile vicinity and rural areas. The magnitudes of differences are also similar. For instance, based on the linear probability model for critical acuity cases (Table S7a), the adjusted frequency of potential bypassing for the subgroup with trauma was 25.4 percentage points higher than that for the reference group with the primary impression of cardiovascular problems (48.3 percentage points), indicating a risk ratio of 1.53. The corresponding odds ratio estimate (3.33) from the logit regression is equivalent to a risk ratio of 1.60 (see the Estimation models section in Appendix S1). In determining patient acuity, we used initial acuity measure for majority of the cases (70.1%), but for the remaining cases—for which initial acuity was not documented in the EMS data abstracts—we used the final acuity measure. We estimated the main estimates separately for the two groups (Tables S9a and S9b). As nearly all of the cases without initial acuity measure were within Chicago, the two groups (and the resulting estimates) largely reflect the within and outside Chicago cases (Tables S4a and S4b). Analysis of sensitivity of estimates to the exclusion of trauma cases indicated similar patterns in potential bypassing across demographic and locational characteristics (Table S10).

3.4. STEMI cases

We identified 945 suspected STEMI cases and 86 hospitals with PCI capability (Table S11a). Rates of potential bypassing the nearest PCI‐capable hospital were 53.2% (distance‐based) and 49.3% (time‐based). We found no significant differences in potential bypassing rates across subgroups by patient characteristics (Table 5). Non‐PCI‐capability of the nearest ED was associated with higher risk of potential bypassing based on distance (IRR = 2.5, 95% CI, 2.2–2.9) or time (IRR = 3.4, 95% CI, 2.9–3.3) (Tables 5 and S11b).

TABLE 5.

Differences in potential bypassing rate (>0 mile threshold): STEMI cases

Subgroup	Adjusted rate of differences
	IRR	95% CI
Reference group potential bypassing rate (%)	32.50%	[26.0%, 40.8%]
Age (reference: 21–44)
45–64	0.93	[0.78, 1.10]
65–84	0.91	[0.77, 1.08]
85+	0.78	[0.63, 0.97]
Sex (reference: male)
Female	0.97	[0.87, 1.08]
Race/ethnicity (reference: non‐Hispanic White)
Black, non‐Hispanic	1.11	[0.76, 1.64]
Hispanic	0.56	[0.22, 1.46]
Asian, non‐Hispanic	1.1	[0.59, 2.07]
Unknown	1.01	[0.87, 1.16]
Number of EDs in 5‐mile vicinity of incident location (reference: 1 ED)
0	1.20	[1.06, 1.35]
2–3	1.25	[0.92, 1.71]
4 or more	1.10	[0.19, 6.42]
Urban/rural location, % (reference: metropolitan)
Micropolitan area	1.06	[0.23, 4.89]
Small town	1.48	[1.14, 1.91]
PCI‐capability of nearest ED (reference: PCI capable)
Not PCI‐capable	2.5	[2.19, 2.85]

Note: (1) Estimates reported are from a Poisson regression model of the dichotomous potential bypassing indicators based on distance (with >0 mile threshold) (see Equation (S1a) in the Estimated Models section in Appendix S1). All covariates are categorical. Covariate estimates are reported in terms of incidence rate ratio (IRR) using the reference group for each measure category. As all covariates are categorical, the IRR indicates the relative risk of potential bypassing in comparison with the reference group. For instance, IRR = 0.78 estimate for patients aged 85 years and above indicates a 22% lower risk relative to that among patients aged 21–44. (2) We estimated the model combining all STEMI cases since all cases were identified using a STEMI protocol. (3) Some covariate groups were not included due to zero or few STEMI cases.

3.5. Reasons for destination ED

Among all EMS transports, “closest facility” was the most frequently reported reason for destination ED (68.4%), followed by “patient's choice” (21.0%), “family choice” (2.3%), “protocol” (1.8%), “regional specialty center” (1.4%) and “diversion” (0.6%) (Table S12a). The frequency with “closest facility” was higher among Black, Hispanic and Asian patients (all >79%) relative to White patients (59.8%). Adjusted for acuity, primary impression, demographic and locational factors, this frequency remained higher among non‐White groups (Table S12b). Among cases with “closest facility” as the reported reason, the potential bypassing rate was about 25% for every acuity level (Table S12c). And among cases with “patient's choice” as the reported reason, the potential bypassing rate exceeded 55% for every acuity level.

4. DISCUSSION

Using administrative data for all patients in Illinois transported by EMS to an ED, our study highlights three findings. First, approximately one‐third of the cases involved potential bypassing of the nearest ED destination. Among patients with suspected STEMI, the rate of potential bypassing of the nearest PCI‐capable hospital exceeded 49%. Second, with some exceptions, potential bypassing rates were similar across subgroups of cases by acuity level, primary impression, age, and sex. This similarity in relative differences among subgroups remained consistent under alternative criteria of potential bypassing. Incident locations in zip codes with no ED within a 5‐mile vicinity had over twice the rate of potential bypassing relative to zip codes with one ED within a 5‐mile vicinity. However, potential bypassing rates were generally higher for Black and Hispanic patients for most outcome indicators. Third, there was sizable discordance between our indicator of potential bypassing with EMS‐reported reason for choosing the destination ED; 25% of transports for which the reported reason for destination ED was “closest facility,” appeared to have not been transported to the closest facility.

A remarkable feature of our findings was the consistency of similarity in potential bypassing rates across virtually all patient impression conditions at all three acuity levels. In the absence of a benchmark threshold for categorizing bypassing, we defined six indicators of potential bypassing, using three thresholds by distance (0, 3, and 5 miles of excess transport) and time (0, 5, and 10 min of excess transport). Across all the six indicators, we found a consistent pattern of potential bypassing across the three acuity levels and 11 primary impression conditions. The main exception was major trauma cases of critical acuity.

Multiple geographic characteristics were associated with the likelihood of potential bypassing. Of particular importance is the difference in distance (and time) between the nearest and second‐nearest ED from the pick‐up location. This difference was inversely associated with the likelihood of potential bypassing. An unexpected pattern is that, while the likelihood of potential bypassing is higher in areas with multiple EDs within 5 miles of the incident triggering the EMS response, potential bypassing was higher in areas with no EDs within this 5‐mile radius compared to those with one ED within this radius. A plausible explanation is that the difference in distance between the nearest and second‐nearest ED was lower in areas with no EDs than in areas with one ED. A similar pattern may underlie the lowest likelihood of potential bypassing in micropolitan areas, lower than that in metropolitan and rural areas; that is, the difference in distance between the nearest and second‐nearest ED was the highest in micropolitan areas.

Prior literature on ED proximity of EMS transported patients is limited. To our knowledge, the only previous study used national Medicare claims data (2006–2012) to examine the ED destination of EMS‐transported patients residing in the same zip code. Defining the most common ED destination for (White) patients residing in each zip code as the focal destination ED, the study found that over 38% of all patients were not transported to the focal destination ED in their residence zip code. ⁷ A similar frequency was found among the subgroup of patients with high‐risk conditions, including STEMI, stroke and major trauma. A limitation of that study is that, since the claims data did not include the patient incident (street) location, the centroid of the patient residence zip code was used to assess destination ED transport, which may limit the precision of estimated transport distance. Also, our study captures patient acuity and primary impression in the prehospital setting, which may differ from the discharge diagnosis reported in the claims data and is likely to more accurately reflect the decision‐making process of the EMS personnel. Nevertheless, it is notable that our study's estimate of a potential bypassing rate of 35% (distance‐based) is consistent with the 38% estimate from the previous study.

An important implication of our findings is that the ED destination of EMS‐transported patients may be motivated by factors other than distance and time. These may include patient and family preference, provider determination based on systems of care directives and traffic conditions, and system‐related factors, such as, ambulance diversion. Note that our study estimates of the potential bypassing rate may well be an underestimate of the share of patient preference (and other factors), since some patients may have a preference for the nearest ED. Evidence of alternative motivations has also been examined in the context of regionalization of specialized care, such as serious major trauma, STEMI and stroke, given the persistence of “under‐triage,” wherein patients are not transported to the preferred regional centers. ⁴ , ⁵ , ⁶ , ³² In a study of STEMI patients in North Carolina (2008–2010), 36% of the cases were not transported to a PCI‐capable hospital as recommended by the EMS protocol. ⁶ Another study of 61 EMS agencies in the western United States (2006–2008) found that 28% of serious major trauma patients were transported to a non‐trauma center. ⁴ That study found that the most common reason indicated by the EMS provider for selecting the destination ED was patient or family choice (50.6%), followed by closest facility (20.7%) and specialty resource center (15.2%). ⁴ Patient or family preference may be influenced by prior experience and a desire to be transported to the “home” hospital. The role of patient or family preference has also been identified in the context of care‐seeking among rural patients. ⁸ Provider determination of suitable destination ED, based on systems of care directives and local traffic conditions have also been recognized as a source of differential ED destinations. ⁹

Recognition of motivations other than proximity of ED destination, in turn has implications for a number of well‐known facets of health services utilization in the United States. First, there is compelling evidence of concentration of Black and Hispanic patients in a small proportion of hospitals, even for high‐risk conditions (e.g., STEMI) for which a large proportion of patients arrived by EMS. ¹² Patient or provider preference may account for systematic differences in destination EDs, even involving bypassing of proximate EDs. ³³ Second, the evaluation of relationships between EMS transport destination and patient outcomes becomes more complex in the presence of patient preferences. For instance, bypassing the nearest ED in favor of the patient's “home” hospital, where the patient has a prior history of outpatient and inpatient care, may represent improved continuity of care, which in turn may be associated with improved inpatient and post‐discharge outcomes. ³⁴ , ³⁵ One strand of literature has treated diverse ED destinations for patients from the same zip code as random assignment to evaluate hospital quality ⁹ ; however, even if the assignment of the ambulance provider is random, destination may be systematically influenced by patient preferences. Third, in the presence of patient preferences, the impact of ambulance diversion and regionalization may be heterogeneous across patients residing in the same area. ³⁶ , ³⁷ A recent study found that regionalization of STEMI transport in California was associated with a smaller improvement and Black and Hispanic patients, relative to White patients, in access to PCI hospitals and receipt of PCI treatment; such differences may arise from differences in potential bypassing by race/ethnicity. ³⁷

4.1. Limitations

We recognize several limitations of the study. We lacked clinical details of patient symptoms and severity indicators (e.g., vital signs) that may influence destination decisions. We could not link the EMS record to the ED and hospitalization records to evaluate the EMS provider severity assessment. Not all EDs have specialist services appropriate for all patients. Our destination ED analysis for STEMI patients considered the setting wherein all Illinois EMS regions adopted STEMI systems of care; as of 2019, some regions had not yet done so. Our estimates of driving distance and time are based on historical data on driving routes and may not accurately reflect actual traffic conditions. The sizable proportion of missingness of the race/ethnicity indicator (approximately 12%) is a limitation in making subgroup comparisons.

4.2. Conclusions

In conclusion, our findings indicate that that a sizable proportion of EMS transports are not to the nearest (or quickest) destination ED. Under alternative thresholds by driving distance and time, we found similarity in the frequency of potential bypassing among critical and low acuity cases, and across a wide range of suspected medical conditions. These findings suggest that EMS transport destination may be motivated by factors other than proximity, including patient preference, provider determination and system‐level factors (e.g., ambulance diversion). Given the sizable prevalence of potential bypassing, it will be important to assess the underlying mechanisms and the relative role of the contributing factors. Whether and how potential bypassing affects patient outcomes also needs to be studied.

CONFLICT OF INTEREST

The authors declare no conflicts of interest.

Supporting information

Appendix S1. Supporting Information.

Hanchate AD, Qi D, Stopyra JP, Paasche‐Orlow MK, Baker WE, Feldman J. Potential bypassing of nearest emergency department by EMS transports. Health Serv Res. 2022;57(2):300-310. doi: 10.1111/1475-6773.13903