Socioeconomic Influence on Emergency Medical Services Utilization for Acute Stroke: Think Nationally, Act Locally

Abstract

Background and Purpose:

Rates of emergency medical services (EMS) utilization for acute stroke remain low nationwide, despite the time-sensitive nature of the disease. Prior research suggests several demographic and social factors are associated with EMS use. We sought to evaluate which demographic or socioeconomic factors are associated with EMS utilization in our region, thereby informing future education efforts.

Methods:

We performed a retrospective analysis of patients for whom the stroke code system was activated at 2 hospitals in our region. Univariate and logistic regression analysis was performed to identify factors associated with use of EMS versus private vehicle.

Results:

EMS use was lower in patients who were younger, had higher income, were married, more educated and in those who identified as Hispanic. Those arriving by EMS had significantly faster arrival to code, arrival to imaging, and arrival to thrombolytic treatment times.

Conclusion:

Analysis of regional data can identify specific populations underutilizing EMS services for acute stroke symptoms. Factors effecting EMS utilization varies by region and this information may be useful for targeted education programs promoting EMS use for acute stroke symptoms. EMS use results in more rapid evaluation and treatment of stroke patients.

Introduction

Recent Centers for Disease Control data indicates that despite nearly 4 decades of decline in stroke mortality rates, this decline has begun to level off and even reverse in many “Stroke Belt” states in the Southeast United States.^1,2 This has renewed the drive for large-scale efforts to identify and reduce barriers to rapid treatments in this time-sensitive disease.^3,4 Despite the recent extension of treatment windows^5-7 efficacy of all currently available treatments for acute ischemic stroke (AIS) are time-dependent, with earlier treatment resulting in greater likelihood of favorable outcome.^8,9

Late presentation to the emergency department (ED) remains a major barrier to treatment in acute ischemic stroke (AIS). Activation of Emergency Medical Services (EMS) has been shown to reduce symptom onset-to-arrival times^10,11 (pre-hospital delay), and may be a factor in arriving within the tPA treatment window¹². Additionally, EMS transportation for AIS has been shown to reduce in-hospital delays, including time to emergency physician evaluation, time to computed tomography (CT) imaging, time to neurologist evaluation and time to treatment,^10,13,14 likely due to pre-hospital notification and streamlined triage and ED processes. Rates of thrombolytic use among those who were otherwise eligible on arrival are also higher in patients presenting via EMS compared to private vehicle.^14,15 Moreover, many regions have developed pre-hospital EMS routing protocols in an attempt to deliver large-vessel occlusion (LVO) stroke patients to Endovascular Capable or Comprehensive Stroke Centers for timely advanced therapies. Patients not utilizing EMS services could potentially further incur significant delays if interfacility transfer is needed for such procedures, or when a higher level of care is required.¹⁶

Many investigators in the field have suggested the importance of designing community outreach and education programs to target those less likely to activate EMS services for stroke symptoms.^11-15,17-20 Therefore, associating socioeconomic status (SES) and neighborhood demographic data with EMS usage could be used to target resource-limited educational programs to the populations which would most benefit.

The goal of the current study was to determine if SES and demographic factors were associated with EMS utilization by patients with stroke symptoms in the primary referral regions for Duke University Hospital and Duke Regional Hospital in Durham, NC. Our findings can then inform future regional education efforts targeting the identified groups that are underutilizing EMS services.

Methods

Setting and Population

Durham, North Carolina and the surrounding Durham County has a population of approximately 270,000 and is part of the Raleigh-Durham-Cary Combined Statistical Area with a population of over 1.7 million. Durham and neighboring Granville, Person, Orange and Alamance counties comprise the primary catchment area of Duke University hospital (DUH) and Duke Regional Hospital (DRH), both within the city of Durham. DUH is a 957 bed academic tertiary care center and Comprehensive Stroke Center caring for >1000 stroke patients by discharge diagnosis yearly; DRH is a 369 bed regional academic-affiliate community hospital and Primary Stroke Center caring for >300 stroke patients by discharge diagnosis yearly.

Data Sources

Patients were identified from prospectively collected acute stroke code logs from both Duke University Hospital and Duke Regional Hospital between January 2014 and May 2017, which includes all patients for whom the stroke code alert system was activated during that period. These patients were then cross referenced with existing medical record data within the Duke Decision Support Repository using the DEDUCE query system, a locally-developed tool for data identification and extraction of Duke Health System electronic medical records, details of which have been previously published.²¹

The stroke code logs were used to identify individual patient’s mode of arrival (EMS vs private vehicle), date and time of presentation, last known normal (LKN) time, ED arrival time, door-to-imaging time, tPA use and associated door-to-needle treatment times, and EVT treatment use and associated door-to-groin puncture times. DEDUCE system queries were used to obtain patient demographics (date of birth, sex, race/ethnicity, marital status), home address, longitude/latitude of home address, census block/block group. After obtaining home addresses for patients using DEDUCE, socioeconomic data for individual patients were derived using population-based estimates from US Census and 2017 American Community Survey (ACS) data from the geocoded home residences of a given patient. Possible SES-index values range from 0-100, with higher values for the composite score representing higher SES levels. This included population-based estimates for block group median household income, primary English language use, education (measured as dichotomized percentage of population with Bachelor’s degree or higher) and employment status. The Agency for Healthcare Research and Quality (AHRQ) validated SES-index score was also used as a composite estimate of SES for individual patients.²² This index is based on individual census block and includes 7 census variables: median household income, median value of owner-occupied dwellings, percentage of people >25 years of age with less than 12th-grade education, percentage of people >25 years of age who have completed >4 years of college, percentage of people unemployed, percentage living below poverty level, and percentage of households that average over 1 person per room. Possible AHRQ SES-index values range from 0 to 100, with higher values for the composite score representing higher relative SES levels. For our study, calculations of the AHRQ SES-index are based on 2017 updated American Community Survey data.

Driving distances to presenting hospital were obtained using GoogleMaps Distance Matrix API from individual patients’ home addresses; this tool provides travel distance and time for a set of origins and destinations, based on recommended routes and taking into account historical traffic conditions. Home addresses were used as assumed location of stroke occurrence. The study protocol was approved by the Duke University Health System Institutional Review Board (IRB).

Analysis

The initial dataset included 1871 Emergency room stroke-code patient encounters. Patients were excluded if their age was <18, stroke code was in-house (not in the emergency department), or mode of arrival could not be determined. Private vehicle (PV) use was defined as any non-EMS private transportation, including taxi or any other form of transportation from the scene. Patients with home address outside of a 50-mile radius from Durham, NC or those with only PO Box addresses were excluded as the assumption of home address as location of stroke event was then unlikely to be valid. This also reduced likelihood that patients presenting from interhospital transfer to the ED were included in the dataset. Multiple encounters for the same individual patient occurred in 84 instances; for these cases only an individual’s first chronological presentation was kept in the analysis dataset. After exclusions and censure of observations with missing data, the final analysis population was 1360 encounters.

For statistical tests, an alpha of 0.05 was considered statistically significant. Univariate differences between demographic and categorical SES variables were evaluated using Chi-square tests. Comparisons of means between groups (PV vs. EMS arrival) were made using Wilcoxon Rank Sum test for distance, education, employment, language and time delay variables. Logistic regression analysis was conducted to identify factors associated with mode of arrival to the ER. Factors in this analysis included, sex, race, marital status, age, median income estimates, employment, distance to hospital and English as primary language. Descriptive maps were created using R packages for mapping and ACS 2017 block group data, including educational attainment, median household income, AHRQ SES-index and EMS use percentage, which are displayed in Figure 1, superimposed with patient location color coded by mode of arrival. All statistical analyses were completed using R ver. 3.6.0 (www.R-project.org).

Figure 1.

Maps of SES Measures by Census Block.

Results

Among the 1360 patients (977 DUH, 383 DRH) presenting to the ED for whom the stroke code system was activated, 963 (70.8%) arrived via EMS. A larger proportion of patients at DUH used EMS than at DRH (72.8% DUH vs 65.8% DRH, p = 0.010). Patient characteristics overall and by arrival method are presented in Table 1. Within our population, 54.7% were female, while 51.8% were white, 42.9% African American, 1.6% Asian and 3.8% Other. The average age was 65+16 years (mean+SD) and patients were evenly distributed among the age categories examined. The median patient-level AHRQ SES index was 50.65 (range 39.17 to 59.81) with quartiles of 47.26 (25%), 50.65 (50%), and 53.98 (75%).

Table 1.

Patient Characteristics By Arrival Method.

	Total, N = 1360	EMS	PV	P-value
Age, n (%)				<0.001
<50	249 (18.3)	140 (56.2)	109 (43.8)
50-59	248 (18.2)	164 (66.1)	84 (33.9)
60-69	304 (22.4)	212 (69.7)	92 (30.3)
70-79	279 (20.5)	213 (76.3)	66 (23.7)
>80	280 (20.6)	234 (83.6)	46 (16.4)
Sex, n (%)				0.349
Female	744 (54.7)	519 (69.8)	225 (30.2)
Male	616 (45.3)	444 (72.1)	172 (27.9)
Race, n (%)				0.001
Caucasian	704 (51.8)	487 (69.2)	217 (30.8)
African	583 (42.9)	436 (74.8)	147 (25.2)
American	22 (1.6)	15 (68.2)	7 (31.8)
Asian	51 (3.7)	25 (49.0)	26 (51.0)
Other
Ethnicity, n (%)				<0.001
Non-	1315 (96.7)	944 (71.8)	371 (28.2)
Hispanic	27 (2.0)	9 (33.3)	18 (66.7)
Hispanic	18 (1.3)	10 (55.6)	8 (44.4)
Unknown
Marital Status, n (%)				0.001a
Single	705 (51.8)	528 (74.9)	177 (25.1)
Partnered	644 (47.4)	426 (66.2)	218 (33.8)
Other	22 (0.8)	9 (81.8)	2 (18.2)
Distance to Hospital, median miles (IQR)	–	8.9 (4.7 - 20.4)	8.8 (4.8 – 19.2)	0.799b
English Language, % (IQR)	–	88 (81 – 93)	88 (81-94)	0.720b
Bachelor’s degree & Above, % (IQR)	–	29 (15 – 53)	34 (18 – 56)	0.005b
Employed, % (IQR)	–	62 (52 – 70)	62 (53 – 71)	0.555b
Median Income, n (%)				<0.001
Q1	339	256 (75.5)	83 (24.5)
Q2	339	258 (76.1)	81 (23.9)
Q3	334	229 (68.6)	105 (31.4)
Q4	348	220 (63.2)	128 (36.8)
AHRQ SES-Index, n (%)				0.005
Q1	340	248 (72.9)	92 (27.1)
Q2	331	245 (74.0)	86 (26.0)
Q3	344	252 (73.3)	92 (26.7)
Q4	345	218 (63.2)	127 (36.8)

Abbreviations: EMS, Emergency Medical Services; PV, private vehicle; IQR, Interquartile Range.

^aP-value by Chi-Square test comparing Single vs Married/Partnered only.

^bby Wilcoxon Rank Sum Test.

Based on the univariate analysis (Table 1.), there was no difference in the rate of EMS usage between the sexes (p = 0.349). The rate of EMS usage was significantly different between age groups (p < 0.001), with the older groups using EMS more frequently. EMS use between racial groups was also significantly different (p = 0.001) with African Americans most likely to arrive by EMS (74.8%) followed by Caucasians (69.2%), then Asians (68.1%) and then those identifying as ‘other’ (49%). Hispanics were far less likely to use EMS services than non-Hispanics (33.3% vs 71.8% respectively; p < 0.001). Patients who were single were more likely to use EMS compared to those who were married or partnered (74.9% vs 66.1% respectively, p = 0.001). Within the study population, 644 patients identified as married/partnered, 275 of whom were female and 369 of whom were male. Within this group, 60.4% of married female patients arrived by EMS versus 70.5% of married male patients (p = 0.007).

In the univariate analysis, EMS use rates also varied significantly by median income, demonstrating that patients living in areas with lower median income were more likely to use EMS than those living in high income areas (p < 0.001). Patients living in areas with lower average educational attainment (measured as percentage of population with Bachelor’s degree or higher) were more likely to use EMS than those from areas with higher education (p = 0.005). Distance from patients’ homes to the hospital was not correlated with EMS use. EMS use rates did not vary with either employment rates or English language use in our population. The AHRQ SES-index was associated with EMS use similarly to income, with those with higher SES-index utilizing EMS less (p = 0.005).

In our logistic regression analysis (Table 2), factors significantly associated with EMS service use included single marital status (OR 1.39, p = 0.015), Non-Hispanic ethnicity (OR 3.26, p = 0.030), age (OR 1.42 for every 10 year increase in age, p=<0.001), median income (OR 0.887 for every $10,000 increase in average median household income, p < 0.001), and employment (OR 1.13 for every 10% increase in average household employment, p = 0.037).

Table 2.

Logistic Regression Model: Social Predictors of EMS Utilization.^a

Variable	Estimate	Odds ratio (95% CI)	P-value
(intercept)	-2.604	-	0.012b
Sex: male	0.230	1.26 (0.98 – 1.62)	0.074
Race: Black/AA	0.200	1.22 (0.43 – 3.18)	0.691
Race: Caucasian	-0.144	0.87 (0.31 – 2.20)	0.769
Race: Other	-0.155	0.856 (0.25 – 2.88)	0.803
Marital Status: Single	0.326	1.39 (1.07 – 1.80)	0.015b
Marital Status: Other	0.687	1.99 (0.48 – 13.57)	0.397
Ethnicity: Not Hispanic	1.180	3.26 (1.15 – 9.82)	0.030b
Ethnicity: Unknown	0.531	1.67 (0.44 – 6.59)	0.455
Agec	0.035	1.421 (1.31 – 1.54)	<0.001b
Median Incomed	-1.20e-5	0.892 (0.84 – 0.94)	<0.001b
Employede	0.012	1.133 (1.01 – 1.26)	0.037b
Distance	0.009	1.01 (0.10 – 1.02)	0.137
English Language	-0.004	1.00 (0.98 – 1.01)	0.604

Abbreviation: AA, African American.

^aN_EMS = 963 (70.8%).

^bp < 0.05.

^c for every 10 years increase in patient’s age.

^d for every $10,000 increase in median household income.

^e for every 10% increase in percentage of employment estimate for household.

We analyzed both pre-hospital and in-hospital time delays by method of arrival to the ED (Table 3). The overall average duration from last known normal (LKN) to ED arrival was 1.4 hours (SD 4.34), with no difference between those arriving via EMS or private vehicle (p = 0.364). Importantly, in-hospital delays were significantly reduced for patients arriving by EMS. Arrival to code (p < 0.001), arrival to CT scan (p < 0.001), and arrival to tPA administration (p < 0.001) time intervals were all significantly faster for those patients arriving by EMS compared to those arriving by private vehicle. When examining mode of arrival by time of day, we found higher rates of EMS use during ‘off-hours’ (6 PM to 6 AM), with particularly high rates of EMS utilization after midnight and before 8 AM (Figure 2).

Table 3.

Analysis of Time Delays By Arrival Method.^a,b

Time interval	Total	EMS	PV	P-value
LKN to Arrival, h	1.40 (0.82 – 2.60)	1.32 (0.83 – 2.43)	1.55 (0.75 – 3.13)	0.364
Arrival to Code, h	0.17 (0.05 – 0.38)	0.10 (0.00 – 0.27)	0.35 (0.23 – 0.55)	<0.001
Arrival to CT, h	0.43 (0.30 – 0.65)	0.37 (0.27 – 0.57)	0.61 (0.47 – 0.80)	<0.001
Arrival to tPA, h	0.96 (0.69 – 1.37)	0.90 (0.66 – 1.24)	1.18 (0.87 – 1.56)	<0.001

Abbreviations: LKN, Last known normal; CT, computed tomography; tPA, tissue plasminogen activator; EMS, Emergency Medical Services; PV, private vehicle.

^aValues are median (Interquartile range) unless otherwise noted.

^bp-value calculated using Wilcoxon rank sum test.

Figure 2.

EMS Use by Time of Arrival.

Maps showing the geographic distribution of specific demographic and SES factors were developed in hopes of further informing potential outreach targets by identifying geographic areas with lower EMS use or factors contributing to lack of EMS use. Specifically, we examined distribution of median income, educational attainment and AHRQ SES-index within the census blocks represented in our patient population, for comparison against a map of EMS usage rates (Figure 2). Unfortunately, obvious correlations were difficult to observe with the naked eye in these visualizations, due to limited granularity in the available data, despite clear associations within the statistical analyses.

Our final study population excluded 98 encounters (stroke codes) which were recurrent events for individual patients already in the study population. This was done to reduce potential bias as repeat presentations could no longer be considered independent events - patients within our health system receive stroke education that includes instruction to call 911 for stroke-like symptoms. However, these recurrent presentations were reviewed. Among these, 63 patients had a first presentation via EMS; 51 returned a second time by EMS while 12 came back by PV. There were 18 patients who had a first presentation by PV, and of these 11 had a second presentation by EMS and 7 by PV. Three patients had >2 presentations to the ED for stroke-like symptoms during the study period (a total of 10 encounters) with decidedly mixed EMS use, and all with at least 1 PV presentation after the initial event. While this suggests a trend toward EMS use with recurrent events, second event EMS use was 76.5% - only slightly higher than our overall population EMS use rate.

Discussion

National guidelines have recommended public education regarding recognition of acute stroke symptoms and activation of EMS via 9-1-1²³ and increasing public awareness of stroke symptoms and the crucial need of activating EMS is an objective of the Healthy people 2020 project.²⁴ Despite this, EMS utilization remains as low as 38-64% nationwide and has changed little since the introduction of time-sensitive therapies for AIS over 2 decades ago.^{12-14,17,25,26} In fact, a recent survey of stroke in-patients showed 75% of respondents were not aware of the impact of EMS use on stroke treatment and outcomes.²⁷

Regional approaches to targeted education, however, have shown some success with multiple small studies showing public educational efforts reduced pre-hospital delays,²⁸ and improved treatment rates.²⁹ This underutilization of EMS for stroke has been shown to be associated more with those of minority race or ethnicity, and younger age,^11,13,14,18 however some conflicting results among prior investigations exist, suggesting there may be regional-specific or time epoch variations in these trends. Further, relatively little has been published examining the impact of socioeconomic status, or neighborhood characteristics on EMS use.

The goal of our study was to evaluate if location/neighborhood, demographics, and/or socioeconomic factors play a role in EMS utilization for individuals presenting with acute stroke symptoms in our region, in order to identify educational opportunities within communities. It has previously been established that patients with lower SES are impacted more severely by stroke, both in incidence of stroke occurrence, as well as stroke severity, post-stroke disability, and mortality,^19,30,31 which may suggest poorer healthcare access or underutilization of available resources such as EMS. Others have also shown those with insurance, baseline good health and established primary care physician were more likely to call 911.³² Therefore, we initially hypothesized that lower SES individuals would be also less likely to activate EMS, perhaps due to concerns regarding cost of the service, lack of insurance coverage, or less education about the urgency of stroke. However, this was not the case. In fact, we found those with higher income and higher AHRQ SES-index scores were the least likely to utilize EMS in our population, especially in the highest quartile of each.

Overall, our population had a higher rate of EMS use (70.8%) than previously published national-level data.¹⁴ Though encouraging, this may be a result of our selection methodology of identifying all patients for whom the stroke code system was activated, rather than those with a final diagnosis of stroke, as many prior studies have done. Our method is less likely to capture those presenting with stroke very late, but more likely to capture acute stroke mimics. In fact, we found 663 of 1347 (49.2%) patients had clear final diagnosis of stroke (ischemic, hemorrhagic) or TIA; or a mimic rate of 50.8%. Published data suggests this mimic rate is more commonly around 30%^33,34 when determined prospectively by expert reviewers, though determination retrospectively by discharge coding likely also misses a significant number of true strokes with more general ICD10 codes related to ‘weakness’, ‘numbness’, ‘aphasia’ or similar.

In our population, age was associated with EMS use, demonstrating increasing likelihood with increasing age. This is consistent with several prior studies with similar findings.^{11,13,14,26,35,36} Likewise, we found no association between sex and EMS use, consistent with some prior works,^13,25,26 though others have shown women more likely to call EMS.¹⁴ Available literature on racial associations with EMS use for stroke is quite variable. Lacy et al.¹¹ found non-white patients were significantly less likely to use EMS services, and Ekundayo et al.¹⁴ reported similar findings. Several other contemporary studies^13,25,26 found no association between race and EMS use while only an older study by Barsan et al.³⁵ found African Americans were more likely to call 9-1 -1, as was the case in our population. A large study utilizing national-level data by Mochari-Greenberger et al.¹⁸ found white women were most likely to use EMS (62%) followed by black women, and black men (58%) while Hispanic men were least likely to use EMS (52%). Overall, this suggests there is likely regional-specific factors within communities that affect EMS use and supports the position that regional data evaluation and targeted education is necessary to improve these rates for any given locale.

Interestingly, we found those patients who identified as ‘single’ were significantly more likely to arrive via EMS than those married or partnered. This may be result of simple lack of alternative means of transportation when the patient is unable to drive themselves, and living alone has been previously shown to be associated with EMS use.¹³ One study did show only 35% of patients using EMS cited lack of alternative transportation as their reason for use.³⁷ Interestingly, subgroup analysis of married/partnered individuals found that married men were significantly more likely to arrive to the ED via EMS than married women. This may suggest that men’s partners were more likely to call 911 on their behalf.

SES has previously been shown to explain much of stroke incidence differences between racial groups,³⁸ supporting that these factors should be considered when evaluating care access issues in a given region. However, we identified only 1 small prior study that examined patient income levels as a factor in EMS utilization rates, demonstrating lower income individuals were more likely to use EMS³⁶. Findings from our regional data were similar. When divided into quartiles, our population’s lowest income quartile’s (<$37,646/year) EMS utilization rate was 75.5% compared to 63.2% in the highest quartile (>$70,291/year). Our logistic regression model showed that for every $10,000 increase in median household income for a patient’s home census tract, the likelihood of EMS use was 0.89 times lower. For example, compared to those living in an area with mean income of $40,000/year, those making $60,000/year are nearly 2 times less likely to call EMS for stroke symptoms. We were also able to show that patients from areas with higher education rates were similarly less likely to use EMS services when presenting with stroke like symptoms. Though employment individually did not seem to be associated with EMS use, within the regression model this factor became significant (OR 1.13, p = 0.037), showing a 1.13 time increase in EMS use for every 10% increase in employment. Prior studies have shown that patients who have stroke symptom onset at work²⁶ or first noted by bystanders¹³ are more likely to use EMS services, which may account for this difference seen with increasing employment rates. As expected, a high degree of collinearity was noted between income, education and employment; education was highly correlated with median income, and moderately correlated with employment rates, resulting in a best fit regression model that included income and employment but not education. Similarly, AHRQ SES-index was not included in the model due to collinearity.

Longer travel distances, in general, are considered only to represent a small component of pre-hospital delay, but have been shown to be associated with arrival by private vehicle.¹² In our population, however, there was no association between the assumed distance traveled (home address to hospital) and EMS use, nor was this a significant factor in the regression model. Last known normal to arrival was not different between those arriving by EMS and those by PV. However, those arriving by EMS in our population were evaluated and treated faster, as has been shown in prior studies^10,13,14, again lending strong support to efforts aimed at increasing EMS utilization. Importantly, those arriving ‘off hours’ (6 PM – 7 AM) were more likely to use EMS which may be related to availability of alternative transportation.

Limitations

There are several limitations to our study. First, the study is retrospective in design, though our dataset is based on a prospectively collected database, which included all ED stroke code encounters sequentially occurring during the given time frame. This methodology did result in a notable number of patients excluded from the final analysis due to missing data, which may reduce statistical power of our analysis. Additionally, discharge diagnoses are not known in our study population, and therefore we likely included stroke mimics. As the study goal was to identify patients who have stroke-like symptoms in order to encourage EMS utilization, this approach was felt to be appropriate. The use of stroke-code logs to identify patients for inclusion may itself introduce bias toward EMS use, as those arriving well beyond tPA or thrombectomy treatment windows are unlikely to have the stroke code system activated, and therefore would not be included in our data. It is expected that very late presenting patients may be more likely to arrive by PV, though the currently presented data does not evaluate this. Further, data on stroke severity on arrival was not readily available, which may have played a role in EMS use, which prior studies have indicated are related.^14,26,36

There are several other factors of interest which were not included in our analysis and may effect EMS use. First, insurance coverage data was not available. However, targeting potential patients within the community who are under- or uninsured is unlikely to be a feasible approach for community education. Prior studies have shown patients with private insurance or HMO insurance were less likely to use EMS than those with Medicare or Medicaid,^11,14 and no difference in EMS use between privately insured and uninsured patients.¹⁴ Insurance coverage is also estimated to be a poor indicator of SES or ability to pay for EMS services due to the immense variability between plans and regions. We also lacked information on stroke severity or pre-stroke disability. Previous studies suggest both higher National Institutes of Health Stroke Scale (NIHSS) and disability at baseline is associated with higher EMS utilization,²⁶ and patients with milder symptoms were less likely to present in time for thrombolytic treatment.³⁹

Conclusions

Among patients in our region arriving to the ED for whom a stroke code was called, patients who were younger, had higher income, married or partnered, more educated and those with higher AHRQ SES-index scores were less likely to utilize EMS. Patients identifying as Hispanic were unlikely to arrive by EMS compared with non-Hispanics, while African Americans were most likely, compared with all ethnicities considered. Those arriving by EMS were evaluated and treated faster than those arriving by private vehicle. These results can be used to better plan and target educational efforts to improve EMS utilization for acute stroke symptoms. The variability in prior similar studies, and differences from our own data further supports that this type of analysis should be considered for any given region, as local factors likely modify results.