a b s t r a c t

Objective: To derive and characterize the performance of various metrics of emergency transport time in assessing for sociodemographic disparities in the setting of residential segregation. Secondarily to characterize racial disparities in emergency transport time of suspected stroke patients in Austin, Texas.

Data sources: We used a novel dataset of 2518 unique entries with detailed spatial and temporal information on all suspected stroke transports conducted by a public emergency medical service in Central Texas between 2010 and 2018.

Study design: We conducted One-way ANOVA tests with post-hoc pairwise t-tests to assess how mean hospital transport times varied by Patient race. We also developed a spatially-independent metric of emergency transport urgency, the ratio of expected duration of self-transport to a hospital and the measured transport time by an am- bulance.

Data collection/extraction: We calculated ambulance arrival and destination times using sequential tempo- rospatial coordinates. We excluded any entries in which patient race was not recorded. We also excluded entries in which ambulances’ routes did not pass within 100 m of either the patient’s location or the documented hospital destination.

Principal findings: We found that mean transport time to a hospital was 2.5 min shorter for black patients com- pared to white patients. However, white patients’ transport times to a hospital were found to be, on average,

4.1 min shorter than expected compared to 3.4 min shorter than expected for black patients. One-way ANOVA testing for the spatially-independent index of emergency transport urgency was not statistically significant, indi- cating that average transport time did not vary significantly across racial groups when accounting for variations in transport distance.

Conclusions: Using a novel transport urgency index, we demonstrate that these findings represent race-based variation in spatial distributions rather than Racial bias in emergency medical transport. These results highlight the importance of closely examining spatial distributions when utilizing temporospatial data to investigate geographically-dependent research questions.

(C) 2021

1. Introduction

Abundant research demonstrates clear disparities in acute ischemic stroke treatment and outcomes by race, gender, and age [1-7]. Geographic factors can also influence Access to healthcare and health outcomes [8,9]. For example, the regional incidence of stroke is highest in the eleven contiguous states constituting the Stroke Belt [10-12]. Nu- merous additional studies demonstrated geographic variability in stroke incidence, healthcare access, adherence to Performance metrics, and prevalence of patient risk factors [13-22]. Availability of emergency

* Corresponding author at: Department of Neurology, Dell Medical School, 1601 Trinity Street, Building B, The University of Texas at Austin, Austin, Texas 78712, USA.

E-mail address: [email protected] (D. Paydarfar).

medical services (EMS), a crucial component of stroke care, also varies by location [23].

EMS decisions impact patient outcomes in stroke via differentiation of AIS from stroke mimics, Large vessel occlusion detection, and trans- port efficacy. A systematic review found that pre-hospital delays are a larger contributor to overall delays in care than were In-hospital delays and are associated with race, age, and socioeconomic status [24,25]. While the past few decades have seen substantial reductions in in-hospital delays, there has been comparably little improvement in pre-hospital delays [26]. Moreover, despite the importance of EMS transport, few studies have reported durations of the various stages of EMS transport [27].

Disparate Emergency medical care is particularly significant given the abundant literature showing decreased benefit of interventions as time from stroke onset increases [28-30]. Permanent brain tissue

https://doi.org/10.1016/j.ajem.2021.07.013

0735-6757/(C) 2021

damage from ischemia grows exponentially thus rendering most valu- able the immediate time after stroke onset [31]. Additionally, stroke treatments are time-sensitive, and the 2019 AHA/ASA Guidelines do not recommend the use of intravascular thrombolysis outside of a three- to 4.5-h window or endovascular thrombectomy outside of a 24-h window [32]. Lastly, models using data from large, Multi-center trials have suggested that EMS decisions in the field can influence out- comes in AIS patients [33,34].

Investigating disparities in emergency medical transport can prove challenging because geography is intimately associated with race and SES [35]. Segregation refers to systematic variations in the extent to which two or more groups are physically apart and is most commonly discussed in the context of residential segregation-systematic differ- ences in where persons primarily reside [36,37]. Of note, residential seg- regation most commonly follows racial/ethnic or income lines but can involve any sociodemographic trait. The United States has a long history of profound residential racial segregation, and most communities in America show at least some degree of race- or income-based residential segregation [36,38]. However, the particular spatial patterns of residen- tial segregation vary by location and result from long-standing patterns of discrimination-both individual and institutional-against particular groups [37,39]. For example, one heavily segregated locale may have members of one ethnic group clustered into pockets; in another locale, low-income residents may overwhelmingly live on one side of the city [40]. Regardless of the particular spatial patterns, residential segrega- tion will produce systematic variations in the spatial distribution of a population. This sociodemographic variation in the spatial distribution of the study population can confound attempts to assess for treatment disparity when the outcome of interest is spatially-dependent.

Given the importance of minimizing time to treatment in AIS, we

aimed to derive metrics of emergency medical transport time that can be calculated with readily available data and are amenable to compari- son over time and space. We then characterized the performance of these metrics in a residentially segregated community. We hypothe- sized that spatially-dependent time metrics would be biased by sociodemographic variations in the spatial distribution of the patient population. As a case study in which to test this hypothesis, we used a large dataset documenting emergency medical transport times of suspected stroke patients in Austin, Texas, a city with a long history of residential racial segregation.

1. Materials and methods
   1. Study setting

This was a retrospective study performed using a novel dataset from a public EMS in Central Texas with detailed temporospatial information on all ambulance runs for which a stroke alert was initiated between Oc- tober 2010 and December 2018. Fig. 1 shows a map of the EMS service area with patient race and their locations at the time of EMS request for each ambulance run and clearly demonstrates residential racial seg- regation of the city’s population. White patients are widely dispersed across the EMS service area, while Black and Hispanic patients are more tightly concentrated around the eastern aspect of Interstate 35 (a major highway running through downtown Austin, TX) (Fig. 1). Of note, the three comprehensive stroke care centers (CSC) within the EMS service area are all located in downtown Austin relatively close to Interstate 35. As mentioned above, we hypothesized that race-based variation in the geographic distribution of the study popula- tion would bias spatially-dependent metrics of emergency transport time. Lastly, though there are multiple stages of the pre-hospital time- line for suspected stroke patients, this analysis specifically examined emergency transport time from a patient’s location to a hospital. The institutional review board at the University of Texas at Austin approved this study.

• 1. Study population

Each record in the dataset contained detailed demographic data along with sequential geographic coordinates and timestamps automat- ically recorded by a navigation device that documented the ambulance’s route from time of dispatch to the patient’s arrival at a hospital. To val- idate documented emergency transport times, we first calculated the Euclidean distance with Vincenty’s ellipsoid formula between all serial geographic coordinates and the patient’s location and hospital destina- tion using Matlab version 9.0 (The MathWorks Inc., Natick, MA, USA). We then identified the latest time at which the ambulance was within a 10 m (m) radius of the patient’s location as time of ambulance depar- ture and the earliest time at which the ambulance was within a 10 m ra- dius of the hospital as time of ambulance arrival; these times were used to calculate duration of transport. We included the 10 m buffer zone to account for systematic error in the positional accuracy of the ambu- lances’ navigation systems [41].

Certain records had no documented temporospatial data for por- tions of the ambulances’ routes. These data gaps often occurred as the ambulance was reaching its hospital destination. Most hospitals in the EMS service area were in urban settings; the surrounding infrastructure may have interfered with ambulances’ navigation systems. To deter- mine if there was racial variation in the distribution of data gaps, we implemented a one-way ANOVA and found that the length of the data gaps did not vary significantly by patient race (F(4, 3854) = 0.597, p = 0.665). This lack of race-based variation further suggests systematic error in the recording of ambulance routes.

We excluded 1332 entries missing temporospatial data for longer than one minute (Fig. 2). We additionally excluded 295 entries in which the ambulances’ routes did not arrive within 100 m of either the patient’s documented location at time of stroke or their hospital des- tination (Fig. 2). Patient race was either self-reported or documented by EMS providers during the encounter. Due to small sample sizes, patients whose race was listed as either “Native American” or “Pacific Islander” were recoded as “Other” for race; 435 patients with no documented race or ethnicity were excluded from the final analysis (Fig. 2).

• 1. Emergency transport time metrics

We defined four readily calculable emergency transport time met- rics that could be used to compare ambulance runs of varying distances. Measured transport time (mTT) to hospital, the simplest metric, was cal- culated as duration in minutes between an ambulance’s departure from the patient’s location and its arrival at a hospital. To compare mTT with a patient’s expected self-transport time (eTT), we estimated road travel time in minutes with ArcGIS Desktop version 10.8 (Environmental Ser- vices Research Institute, Redlands, CA, USA) using coordinates for the patient’s location and their hospital destination. This estimation as- sumes average traffic speeds and does not account for routine traffic de- lays as these may change cyclically over varying time periods. Additionally, the route used by ArcGIS in its estimations can differ from that used by EMS personnel.

EMS personnel are given discretion to select routing and use special traffic permissions. Of note, EMS personnel may shorten their transport time in various ways: increasing travel speed, utilizing traffic permis- sions, or selecting alternate routes. However, the salient metric is hospi- tal transport time irrespective of how transport time is shortened. To evaluate the effect and possible differential application of these traffic permissions, we calculated the difference between expected and mea- sured transport times to a hospital (dTT):

dTT 1/4 eTT -mTT

To more readily compare the duration of EMS transport to ex- pected duration of self-transport, we defined a factor u that

Fig. 1. A map depicting the location of each EMS call contained in the dataset colored by the patient’s race/ethnicity. These data were derived from documented patient locations.

Fig. 2. Study exclusion criteria.

represents the ratio of EMS transport time to self-transport time. If S represents the speed at which the vehicles travel and D the road dis- tance they travel:

Table 2

Summary statistics of all transport time metrics by patient race/ethnicity. These statistics represent the final dataset.

SEMS 1/4 uSself

D D

	Mean	SD		Mean	SD		Mean	SD		Mean	SD
Asian	13.8	7.54		17.5	7.71		3.68	3.66		1.40	0.356
Black	10.9	5.48		14.3	6.27		3.36	2.49		1.39	0.488
Hispanic	12.1	5.36		15.8	6.06		3.64	3.28		1.36	0.310
Other	12	5.63		15.6	5.64		3.62	2.42		1.40	0.330
White	13.4	6.75		17.5	7.95		4.13	3.30		1.36	0.278

Race mTT (min) eTT (min) dTT (min) u

mTT 1/4 u eTT

1 1

mTT 1/4 u eTT

u 1/4 eTT

mTT

We can thus characterize u as a novel, unitless index of emergency transport time urgency. Expected self-transport time (eTT) is generally longer than emergency transport time (mTT), so we expect that u will be greater than one with larger numbers indicating more urgent trans- port. For example, if self-transport to the hospital takes 30 min but is completed by EMS in 20 min, u will be 1.5.

Lastly, as emergency transport time is obviously dependent on a pa- tient’s physical distance from the hospital, we evaluated each metric’s relationship with the patient’s road distance from the hospital. To do this, we calculated hospital transport distance in kilometers (km) with ArcGIS using coordinates for the patient’s location and the hospital to which they were transported. We then assessed the relationship be- tween hospital transport distance and the emergency transport time metrics with Pearson’s correlation coefficient.

• 1. Statistical analysis

By comparing mean emergency transport time metrics across racial groups, we can determine if EMS personnel differentially apply special traffic permissions. For example, consider two groups A and B with var- ious distributions of a time metric. If there is a statistically significant difference between the average value of the time metric for groups A and B, we can conclude that the urgency of EMS transport is significantly different across groups. If there is no systematic difference in the ur- gency of EMS transport for each group, then the mean value of the time metric should be equivalent across groups. Thus, the test of equiv- alent group mean values of a time metric is an adequate test of the null hypothesis of no Racial disparity in EMS transport.

To determine if the metrics of hospital transport time varied by pa- tient race/ethnicity, we implemented one-way ANOVA tests with the various metrics of emergency transport time as outcomes and patient race as the covariate. Statistically significant omnibus tests were followed with post-hoc pairwise t-tests with Bonferroni-Holm adjust- ments [42,43]. All statistical analyses and figures were completed in R version 3.2 (R Foundation for Statistical Computing, Vienna, Austria).

Table 1

Sample characteristics of the dataset.

Initial Sample Final Sample

Demographics Count Percent Count Percent

Race	Asian	69	1	40	2
	African-American	667	16	401	16
	Caucasian	2477	61	1505	60
	Hispanic	739	18	482	19
	Other	139	4	90	3
Gender	Female	2387	52	1323	53
	Male	2189	48	1192	47
Age	18 years and younger	25	1	6	1
	19 to 64 years	2058	45	1092	43
	65 years and older	2482	54	1417	56

1. Results
   1. Demographic characteristics of the study population

The final dataset included 2518 unique ambulance runs. Table 1 shows the demographic characteristics of the initial and final samples. The majority of patients in the dataset were White (60%). Black patients and Hispanic patients constituted 16% and 19% of the dataset respec- tively. More than half of the patients in the final dataset were female (53%). The majority of patients were over the age of 65 years (56%), and 43% of patients were between 19 and 64 years of age. Only six pa- tients in the final dataset were younger than 18 years of age.

• 1. Primary outcomes of the ANOVA tests

Table 2 shows summary statistics for all transport time metrics grouped by patient race. One-way ANOVA tests showed that mTT varied significantly by race (F(4, 2513) = 14.47, p < 0.0001). Group means varied significantly for Asian and Black (p = 0.036), Hispanic and Black (p = 0.031), Black and White (p < 0.0001), and Hispanic and White patients (p = 0.001). Thus, average mTT was shortest for Black patients, then Hispanic patients, White patients, and longest for Asian patients (Table 2). eTT also varied significantly by patient race (F(4, 2513) = 18.53, p < 0.0001). Group means varied significantly for His- panic and Black (p = 0.018), Black and White (p < 0.0001), and His- panic and White patients (p < 0.0001). Though average eTT were slightly longer than average mTT, the same race-based pattern was ob- served in which the average eTT was shortest for Black patients, then Hispanic patients, White patients, and longest for Asian patients (Table 2). Thus, Asian and White patients had, on average, longer trans- port times to a hospital relative to Black and Hispanic patients.

dTT also varied significantly by patient race (F(4, 2513) = 5.914, p < 0.0001). Group means varied significantly for Hispanic and White (p = 0.03068) and Black and White patients (p = 0.00015). However, the race-based pattern seen for mTT and eTT was reversed for dTT; Black patients had the smallest average dTT followed by Hispanic pa- tients, Asian patients, and then White patients with the largest differ- ence (Table 2). Thus the difference between emergency transport time and estimated self-transport time was largest for White patients and smallest for Black patients. Finally, in contrast to the previous metrics, one-way ANOVA testing for u (F(4, 2513) = 0.883, p = 0.473) was not statistically significant, indicating that the average value of u did not vary significantly across racial groups.

• 1. Spatial dependency of emergency transport time metrics

Fig. 3 depicts the relationships between hospital transport distance and the four metrics of hospital transport time along with Pearson’s cor- relation coefficient and associated p-value. mTT (r = 0.89, p < 0.001) and eTT (r = 0.96, p < 0.001) have a strongly positive correlation with hospital transport distance (Fig. 3A and B). dTT has a moderately positive correlation (r = 0.43, p < 0.001) with hospital transport dis- tance (Fig. 3C), while u has a weakly negative correlation (r = -0.14,

Fig. 3. Scatterplots depicting relationship between hospital transport distance (km) and the various hospital transport time metrics with Pearson’s correlation coefficient and associated p-value. A) Measured hospital transport time (min). B) Expected hospital transport time (min). C) Difference in expected and measured hospital transport time (min). D) u, a spatially-independent metric of transport urgency.

p < 0.001) with hospital transport distance (Fig. 3D). Notably, u shows greater variation over shorter hospital transport distances (Fig. 3D). Lastly, Fig. 4 shows the spatial density-as estimated by a bivariate nor- mal kernel-of the patient population across the EMS service area subset by patient race.

1. Discussion

We investigated the performance of four emergency transport time metrics in assessing for racial disparities in emergency transport of suspected stroke patients in the setting of stark residential racial

segregation (Figs. 1 and 4). We hypothesized that time metrics most strongly correlated with distance would be most sensitive to group- level variations in the spatial distribution of the patient population. Ad- ditionally, previous studies demonstrated that road distance from a pa- tient’s location to a hospital can significantly impact patient outcomes, so it is essential to control for hospital transport distance when assessing for transport disparity [44,45]. However, hospital transport distance will vary systematically by sociodemographic characteristics in a residentially segregated community; patients of particular sociodemographic groups may live, on average, further or closer to a hospital relative to patients of other sociodemographic groups. Thus

Fig. 4. Density plots depicting the spatial distribution of each racial group with darker colors indicating greater population density. Blue points indicate major hospitals. These data were derived from documented patient locations.

transport time metrics that strongly correlate with transport distance have minimal utility in a residentially segregated community, as the spatial concentration of population subgroups relative to hospital loca- tion will bias these metrics.

Both mTT and eTT strongly correlate with hospital transport distance (Fig. 3A and B); this is to be expected, as the strongest determinant of transport time is transport distance. Thus, in the presence of group- level variation in hospital transport distance-as is seen in Austin,

Texas-both mTT and eTT are not readily comparable metrics of emer- gency transport time. dTT also substantially correlates with transport distance (Fig. 3C) because the possible range of dTT varies with hospital transport distance. As hospital transport distance increases, the poten- tial time saved by emergency transport relative to self-transport also in- creases. Over short transport distances, dTT is necessarily constrained, as EMS personnel have less distance over which to exercise special traffic permissions. Notably, the impact of dTT varies by transport duration; a

Fig. 5. Maps depicting the standard distance (transparent circle with black outline) and mean center (opaque point) of each racial/ethnic group. Red points indicate major hospitals. These data were derived from documented patient locations.

dTT of 5 min is more impactful for a 10-min trip than for a 30-min trip. Thus dTT does not permit ready comparison of trips of varying hospital transport distances.

Because it can be defined as a unitless index of the ratio of eTT to mTT, we described u as representative of emergency transport urgency, the factor by which emergency transport time differs from expected dura- tion of self-transport. In contrast to mTT, eTT, and dTT, u is minimally cor- related with hospital transport distance (Fig. 3D). It is thus less sensitive

to the group-level variation in spatial distributions that would be seen in residentially segregated populations and readily permits comparison of trips of varying distances.

Fig. 4 clearly demonstrates group-level variation in the spatial distribution of our study population. Asian and White patients are most widely dispersed across the EMS service area, while Hispanic and Black patients are most geographically concentrated and appear to have their highest population density adjacent to the CSCs in

downtown Austin (Fig. 4). The significant differences in mTT and eTT by patient race suggest that White and Asian patients had signifi- cantly longer emergency medical transport times relative to His- panic and Black patients (Table 2). However, analysis of time saved by utilization of EMS special traffic permissions-as represented by dTT-exhibited an inverse pattern wherein Black and Hispanic pa- tients seemed to have significantly less time saved by emergency transport relative to White and Asian patients (Table 2). Thus, it would seem that these results discord insofar as Black patients re- ceived less advantage from these traffic permissions relative to White patients, while White patients had longer average transport times to the hospital relative to Black patients.

We posit that these discordant results actually reflect substantial race-based variation in patient location across the EMS service area which preferentially biases spatially-dependent metrics of emergency transport time. The significant differences in mTT and eTT by race reflect strong race-based variation in the spatial distribution of our patient population. Moreover, because the spatial dispersion of Asian and White patients is much greater (Fig. 4), EMS personnel have more road time over which to utilize special traffic permissions; this explains the greater dTT for Asian and White patients relative to Black and His- panic patients. The significant differences between mean dTT by patient race thus reflect the spatial dependence of dTT (Fig. 3) and the substan- tial group-level variation in our patient population.

The spatially-independent metric u did not vary significantly by patient race suggesting no racial disparity specifically in emergency transport urgency of suspected stroke patients in Austin, Texas. Though we demonstrate how use of a spatially-independent transport time metric contextualizes results initially suggestive of profound race- based disparity, we cannot exclude other disparities in emergency med- ical transport. Additionally, racial segregation has been shown to pro- duce unique socio-ecological environments that drive exposure to environmental risk factors and reflect varying levels of access to healthcare [46-49]. In short, geography strongly influences health behaviors, disease risk, and access to care, but its influence is deeply tan- gled with race and socioeconomic status; these complicated relation- ships further emphasize the importance of careful documentation of spatial distributions [50]. Furthermore, residential segregation is pres- ent in almost all communities in America and causes significant group-level variation in patients’ spatial distributions by the relevant sociodemographic characteristics [51].

Analyses that concern spatial relationships should be sensitive to group-level variation in the spatial distribution of the study population, as group-based differences in patient location can mask evidence of dis- parate care or create the false appearance of equity. By adequately char- acterizing how spatial distributions vary across population subgroups of interest, researchers can provide the necessary context for appropriate interpretation of results. As an example of best practices for spatial data characterization, Fig. 5 demonstrates a comprehensive visual de- piction of group-level variation in the spatial distribution of our patient population by documenting Geometric mean and standard distance, the primary center and dispersion parameters for most spatial distributions, for each population subgroup along with relevant geographic land- marks. We advise researchers using geographic data to conduct detailed exploratory analyses, carefully describe spatial variations in their fea- tures of interest, and use spatially-independent metrics. Given the long history of race- and income-based segregation in the United States, investigations of healthcare disparities that involve spatial rela- tionships should begin with careful attention to the methodological challenges produced by the geographic concentration of underserved populations.

To our knowledge, this is the first study to derive and implement a spatially-independent metric of emergency transport time. By compar- ing measured emergency transport time with the expected duration of self-transport, u permits comparison of varying trips even in the ab- sence of detailed data on infrastructure-not only do traffic patterns,

roads, and other infrastructure considerations change over time, they are difficult to accurately operationalize. Given the heterogeneity in emergency medical services across the country, our focus on one com- munity highlights the importance of local analyses, as national-level data may conceal significant Regional variation in emergency medical care [52]. Furthermore, our study uses individual incidence and destina- tion coordinates which permits more accurate estimation of travel times [44,53,54].

However, the use of u assumes that relative change between mTT and eTT reflects only the urgency of EMS transport. Thus u does not ac- count for disparity in EMS evaluation of a patient nor in the selection of a hospital destination. Moreover, the heteroskedasticity in u when pre- dicted by hospital transport distance likely reflects the outsize impact of small changes in transport time for short trips (Fig. 3D); this suggests that u is not an ideal metric for communities in which patients live im- mediately adjacent to hospitals. Important data limitations include sys- tematic error in the recording of timestamps and patient demographic information.

Current models of health inequity consider not only individual char- acteristics such as race and SES but also the larger context of a patient’s location [55]. Further research should address the myriad ways in which patient location influences health outcomes in time-sensitive condi- tions. Disparities in the placement of hospitals or EMS stations and urban-planning decisions such as the placement of roads may also influ- ence access to healthcare. Lastly, a large proportion of Americans do not call emergency medical services [56,57]. Racial minorities have been found to be less likely to call EMS [58]. Further research should investi- gate barriers to EMS usage and variation in outcomes between patients who utilize EMS and those who do not.

1. Conclusion

Given the importance of minimizing time to treatment in AIS, we aimed to derive simple, easily calculable metrics of emergency medical transport time and then characterize the performance of these metrics in a residentially segregated community. For a case study in which to test this hypothesis, we used a large dataset documenting emergency medical transport times of suspected stroke patients in Austin, Texas, a city with a long history of residential racial segregation. We found that spatially-dependent metrics were more sensitive to the group- level variation in the spatial distribution of the patient population that is seen in residentially segregated communities. We additionally de- scribed the performance of a spatially-independent metric and pro- posed best practices for working with spatial data in the setting of residential segregation.

Funding

This research did not receive any specific grant from funding agen- cies in the public, commercial, or not-for-profit sectors.

Prior presentations

A poster describing the results of this study was presented at the 2021 International Stroke Conference.

Ethics approval

This was an entirely retrospective study conducted with data ob- tained from an external organization. The Institutional Review Board of the University of Texas at Austin approved this study.

Declaration of Competing Interest

None.

References

1. Faysel MA, Singer J, Cummings C, Stefanov DG, Levine SR. Disparities in the use of in- travenous t-PA among ischemic stroke patients: population-based recent temporal trends. J Stroke Cerebrovasc Dis. 2019;28(5):1243-51.
2. Boehme AK, Siegler JE, Mullen MT, Albright KC, Lyerly MJ, Monlezun DJ, et al. Racial and gender differences in Stroke severity, outcomes, and treatment in patients with acute ischemic stroke. J Stroke Cerebrovasc Dis. 2014 Apr;23(4):e255-61.
3. Agarwal S, Menon V, Jaber WA. Outcomes after acute ischemic stroke in the United States: does residential ZIP code matter? J Am Heart Assoc. 2015 Mar;10:4(3).
4. Persky RW, Turtzo LC, McCullough LD. Stroke in women: disparities and outcomes. Curr Cardiol Rep. 2010;12(1):6-13.
5. Elfassy T, Grasset L, Glymour MM, Swift S, Zhang L, Howard G, et al. Sociodemographic disparities in long-term mortality among stroke survivors in the United States: the REGARDS study. Stroke. 2019;50(4):805-12.
6. Song T, Pan Y, Chen R, Li H, Zhao X, Liu L, et al. Is there a correlation between socio- economic disparity and functional outcome after acute ischemic stroke? Wang X, editor. PLoS One. 2017 Jul 26;12(7):e0181196.
7. Yan H, Liu B, Meng G, Shang B, Jie Q, Wei Y, et al. The influence of individual socio- economic status on the clinical outcomes in ischemic stroke patients with different neighborhood status in Shanghai, China. Int J Med Sci. 2017;14(1):86-96.
8. Guagliardo MF. Spatial accessibility of primary care: concepts, methods and chal- lenges. Int J Health Geogr. 2004;Vol. 3.
9. Mohan JF. Explaining geographies of health care: a critique. Health Place. 1998 Jun 1;

4(2):113-24.

1. Borhani NO. Changes and geographic distribution of mortality from cerebrovascular disease. Am J Public Heal Nations Heal. 1965 May;55(5):673-81.
2. Lanska DJ, Kuller LH. The geography of stroke mortality in the United States and the concept of a stroke belt. Stroke. 1995 Jul;26(7):1145-9.
3. Lanska DJ. Geographic distribution of stroke mortality in the United States: 1939- 1941 to 1979-1981. Neurology. 1993 Sep 1;43(9):1839-51.
4. Mullen MT, Judd S, Howard VJ, Kasner SE, Branas CC, Albright KC, et al. Disparities in evaluation at certified primary stroke centers: reasons for geographic and racial dif- ferences in stroke. Stroke. 2013;44(7):1930-5.
5. Mullen MT, Wiebe DJ, Bowman A, Wolff CS, Albright KC, Roy J, et al. Disparities in ac- cessibility of certified primary stroke centers. Stroke. 2014;45(11):3381-8.
6. Freyssenge J, Renard F, Schott AM, Derex L, Nighoghossian N, Tazarourte K, et al. Measurement of the potential geographic accessibility from call to definitive care for patient with acute stroke. Int J Health Geogr. 2018 Dec 12;17(1):1-14.
7. Scott PA, Temovsky CJ, Lawrence K, Gudaitis E, Lowell MJ. Analysis of Canadian pop- ulation with potential geographic access to intravenous thrombolysis for acute is- chemic stroke. Stroke. 1998 Nov;29(11):2304-10.
8. Seabury S, Bognar K, Xu Y, Huber C, Commerford SR, Tayama D. Regional disparities in the quality of stroke care. Am J Emerg Med. 2017;35(9):1234-9.
9. Gonzales S, Mullen MT, Skolarus L, Thibault DP, Udoeyo U, Willis AW. Progressive rural-urban disparity in acute stroke care. Neurology. 2017;88(5):441-8.
10. Sacco RL, Gardener H, Wang K, Dong C, Ciliberti-Vargas MA, Gutierrez CM, et al. Racial-Ethnic disparities in acute stroke care in the Florida-Puerto Rico collaboration to reduce stroke disparities study. J Am Heart Assoc. 2017;6(2):1-10.
11. Koifman J, Hall R, Li S, Stamplecoski M, Fang J, Saltman AP, et al. The association be- tween rural residence and stroke care and outcomes. J Neurol Sci. 2016 Apr;363: 16-20.
12. Kleindorfer D, Xu Y, Moomaw CJ, Khatri P, Adeoye O, Hornung R. US geographic dis- tribution of rt-PA utilization by hospital for acute ischemic stroke. Stroke. 2009;40 (11):3580-4.
13. Roberson S, Dawit R, Moore J, Odoi A. An exploratory investigation of geographic dis- parities of stroke prevalence in Florida using circular and flexible spatial scan statis- tics. PLoS One. 2019;14(8):1-16.
14. Powers WJ, Rabinstein AA, Ackerson T, Adeoye OM, Bambakidis NC, Becker K, et al. Guidelines for the early management of patients with acute ischemic stroke: a guideline for healthcare professionals from the American Heart Association/ American Stroke Association. Vol. 49. Stroke. 2018;2018:46-110.
15. Evenson KR, Foraker RE, Morris DL, Rosamond WD. A comprehensive review of prehospital and in-hospital delay times in acute stroke care. Int J Stroke. 2009 Jun; 4(3):187-99.
16. Kleindorfer DO, Lindsell CJ, Broderick JP, Flaherty ML, Woo D, Ewing I, et al. Commu- nity socioeconomic status and prehospital times in acute stroke and transient ische- mic attack. Stroke. 2006 Jun;37(6):1508-13.
17. Pulvers JN, JDG Watson. If time is brain where is the improvement in prehospital time after stroke? Front Neurol. 2017;Vol. 8:617.
18. Golden AP, Odoi A. Emergency medical services transport delays for suspected stroke and myocardial infarction patients. BMC Emerg Med. 2015 Dec 3;15(1):34.
19. Saver JL, Fonarow GC, Smith EE, Reeves MJ, Grau-Sepulveda MV, Pan W, et al. Time to treatment with intravenous tissue plasminogen activator and outcome from acute ischemic stroke. JAMA. 2013 Jun 19;309(23):2480.
20. Fassbender K, Balucani C, Walter S, Levine SR, Haass A, Grotta J. Streamlining of prehospital stroke management: the golden hour. Lancet Neurol. 2013;Vol. 12: 585-96.
21. Saver JL, Goyal M, Van Der Lugt A, Menon BK, Majoie CBLM, Dippel DW, et al. Time to treatment with endovascular thrombectomy and outcomes from ischemic stroke: a meta-analysis. JAMA - J Am Med Assoc. 2016;316(12):1279-88.
22. Saver JL. Time is brain - quantified. Stroke. 2006;37(1):263-6.
23. Powers WJ, Rabinstein AA, Ackerson T, Adeoye OM, Bambakidis NC, Becker K, et al. Guidelines for the early management of patients with acute ischemic stroke: 2019 update to the 2018 guidelines for the early management of acute ischemic stroke a guideline for healthcare professionals from the American Heart Association/ American Stroke A. Stroke. 2019;Vol. 50 E344-418.
24. Holodinsky JK, Williamson TS, Demchuk AM, Zhao H, Zhu L, Francis MJ, et al. Model- ing stroke Patient transport for all patients with suspected large-vessel occlusion. JAMA Neurol. 2018;75(12):1477-86.
25. Holodinsky JK, Patel AB, Thornton J, Kamal N, Jewett LR, Kelly PJ, et al. Drip and ship versus direct to endovascular thrombectomy: the impact of treatment times on transport decision-making. Eur Stroke J. 2018;3(2):126-35.
26. Halverson JA, Barnett E, Casper M. Geographic disparities in heart disease and stroke mortality among black and white populations in the Appalachian region. Ethn Dis. 2002;12(4) S3-82-91.
27. Krysan M, Crowder K. Cycle of segregation: Social processes and residential stratifi- cation. Russell Sage Foundation: New York, New York, USA; 2017.
28. Massey DS, Denton NA. The dimensions of residential segregation. Soc Forces. 1988 Dec;67(2):281.
29. Logan JR, Zhang W, Turner R, Shertzer A. Creating the black ghetto: black residential patterns before and during the great migration. Ann Am Acad Pol Soc Sci. 2015;660 (1):18-35.
30. Yang TC, Zhao Y, Song Q. Residential segregation and racial disparities in self-rated health: how do dimensions of residential segregation matter? Soc Sci Res. 2017; 61:29-42.
31. Gibbons J, Yang TC, Brault E, Barton M. Evaluating residential segregation’s relation to the clustering of poor health across american cities. Int J Environ Res Public Health. 2020;17(11).
32. Merry K, Bettinger P. Smartphone GPS accuracy study in an urban environment. PLoS One. 2019;14(7):1-19.
33. Holm S. A simple sequentially rejective multiple test procedure. Scand J Stat. 1979;6 (2):65-70.
34. Keselman HJ, Miller CW, Holland B. Many tests of significance: new methods for controlling type I errors. Psychol Methods. 2011;16(4):420-31.
35. Apparicio P, Gelb J, Dube A-S, Kingham S, Gauvin L, Robitaille E. The approaches to measuring the potential spatial access to urban health services revisited: distance types and aggregation-error issues. Int J Health Geogr. 2017 Dec 23;16(1):32.
36. Mizen A, Fry R, Grinnell DE, Rodgers S. Quantifying the error associated with alterna- tive GIS-based techniques to measure access to health care services. AIMS Public Heal. 2015;2(4):746-61.
37. Williams DR. Racial residential segregation: a fundamental cause of racial disparities in health. Public Health Rep. 2001 Sep 1;116(5):404-16.
38. Inagami S, Borrell LN, Wong MD, Fang J, Shapiro MF, Asch SM. Residential segrega- tion and Latino, Black and White mortality in New York City. J Urban Heal Bull New York Acad Med. 83(3).
39. Bravo MA, Anthopolos R, Kimbro RT, Miranda ML. Residential racial isolation and spatial patterning of type 2 diabetes mellitus in Durham, North Carolina. Am J Epidemiol. 2018;187(7):1467-76.
40. Bravo MA, Batch BC, Miranda ML. Residential racial isolation and spatial patterning of hypertension in Durham, North Carolina. Prev Chronic Dis. 2019 Mar 28;16: 180445.
41. Williams CT, Metzger DS. Race and distance effects on regular syringe exchange pro- gram use and injection risks: a geobehavioral analysis. Am J Public Health. 2010;100 (6):1068-74.
42. Intrator J, Tannen J, Massey DS. Segregation by race and income in the United States 1970-2010. Soc Sci Res. 2016;60:45-60.
43. Andresen EM, Diehr PH, Luke DA. Public health surveillance of low-frequency popu- lations. Annu Rev Public Health. 2004 Apr;25(1):25-52.
44. Wang F, McLafferty S, Escamilla V, Luo L. Late-stage breast Cancer diagnosis and health care access in Illinois*. Prof Geogr. 2008 Jan;60(1):54-69.
45. Ader J, Wu J, Fonarow GC, Smith EE, Shah S, Xian Y, et al. Hospital distance, socioeco-

nomic status, and timely treatment of ischemic stroke. Neurology. 2019;93(8): e747-57.

1. Tung EL, Cagney KA, Peek ME, Chin MH. Spatial context and health inequity: reconfiguring race, place, and poverty. J Urban Heal. 2017;94(6):757-63.
2. Mohammad YM. Mode of arrival to the emergency department of stroke patients in the United States. J Vasc Interv Neurol. 2008 Jul;1(3):83-6.
3. Xirasagar S, Tsai M, Heidari K, Hardin JW, Wu Y, Wronski R, et al. Why acute ische- mic stroke patients in the United States use or do not use emergency medical ser- vices transport? Findings of an inpatient survey. BMC Health Serv Res. 2019 Dec 3; 19(1):929.
4. Mochari-Greenberger H, Xian Y, Hellkamp AS, Schulte PJ, Bhatt DL, Fonarow GC, et al. Racial/ethnic and sex differences in emergency medical services transport among hospitalized US stroke patients: analysis of the national get with the guidelines-Stroke registry. J Am Heart Assoc. 2015;4(8):e002099.