Rural Stroke Patients Have Higher Mortality: An Improvement Opportunity for Rural Emergency Medical Services Systems

Abstract

Purpose:

Early recognition and prompt pre-hospital care is a cornerstone of acute stroke treatment. Residents of rural areas have worse access to stroke services than urban residents. The purpose of this study is to (1) describe US trends in rural-urban stroke mortality and (2) identify possible factors associated with rural-urban stroke case-fatality disparities.

Methods:

This study was a nationwide retrospective cohort study of stroke admissions. Primary exposure was rurality of patient’s residence. Primary outcome was hospital encounter death. Secondary outcome was discharge to a care facility or home health care. Univariable and multivariable logistic regression estimated the odds of mortality by subject rurality among stroke subjects.

Findings:

Rural stroke subjects had higher mortality than non-rural counterparts (18.6% rural vs. 16.9% non-rural). After adjustment for patient and hospital factors, patient rurality was associated with increased odds of mortality (aOR = 1.11; 95%CI 1.06 – 1.15; P < 0.001). For the secondary outcome of discharge to home, rural stroke visits were less likely to be discharged to a care facility than non-rural stroke visits (aOR 0.94; 95%CI 0.91 – 0.97; P < 0.001). Results were similar after adjusting for thrombolytics administration and transfer status.

Conclusions:

Rural stroke patients have higher mortality than urban counterparts likely due to their increased burden of chronic disease, lower health literacy and reduced access to prompt pre-hospital care. There may be an opportunity for emergency medical services (EMS) systems to assist in increase stroke awareness for both patients and clinicians and establish response patterns to expedite emergency care.

Introduction:

Stroke is a major source of morbidity and mortality in the United States (US) with 795,000 people affected annually making it the 5^th leading cause of death and a major cause of disability.¹

Rural patients have an increased stroke incidence and mortality ², and since 19.3% of Americans live in rural areas, understanding these rural-urban disparities is an important part of improving US stroke outcomes.³ The causes of rural-urban disparities in stroke burden are thought to be multi-factorial, including increased stroke risk factors in rural patients and differences in acute management at rural hospitals.^4–7 In addition, people living in rural areas are poorer, less educated, and are medically underserved with fewer non-elderly adults having health insurance.² Educational and socioeconomic differences raise the possibility that rural populations are less knowledgeable in recognizing the signs and symptoms of stroke.^8–10 This disparity in health literacy may lead to delayed activation of emergency medical services (EMS), which may make patients less eligible for time-sensitive treatment.^{11, 12}

Even when rural patients initiate emergency care, systemic care delays contribute to rural-urban stroke disparities.^{2, 7, 13} The American Heart Association/American Stroke Association recommends rapid activation of the 911 system and priority dispatch for suspected stroke patients; however this prioritization requires specialized dispatch systems that may not be used in rural areas.¹⁴ Pre-hospital notification has decreased time-to-imaging and time-to-thrombolysis, ¹⁵ but advanced training and standardized destination protocols are rarely used in rural EMS systems.^{16, 17} Rural residents are often far from specialized stroke care, as only 1% of people living in rural areas reside within 60 minutes of a Primary Stroke Center.^{2, 18} Urban hospitals are twice as likely as rural hospitals to administer intravenous thrombolysis. Since 2001, intravenous thrombolytic usage for ischemic stroke in urban hospitals has quadrupled (1.2% to 4.9%) while it has grown at a much slower pace in rural hospitals (0.9% to 1.6%).⁷

The purpose of this study is to test the hypothesis that resident demographic factors and hospital care (i.e. interhospital transfer and thrombolytic administration) do not fully explain the rural-urban disparities in stroke mortality. If rural patient factors and hospital care do not explain disparities, other potentially modifiable factors such as EMS systems can be targeted to alter rural stroke mortality. The objectives of this study are to (1) identify the rural-urban disparity in stroke mortality that is not explained by existing rural demographic risk factors and hospital-level care considerations and (2) compare rural-urban functional outcomes, using discharge-to-home as a proxy for functional independence.

Methods:

Study Design & Setting

The analysis was a retrospective cohort study of hospital admissions in the US between 2012 to 2016 using the Nationwide Inpatient Sample (NIS). The NIS is a nationally-representative sample of all-payer US hospital admissions collated by the Health Care Utilization Project (HCUP) at the Agency for Healthcare Research and Quality (AHRQ), and it includes patient demographic, diagnosis, and procedure information by visit for non-federal, short-term acute care hospitals in contributing states.¹⁹ This study was not considered human subjects research by the local institutional review board due to its de-identified structure and public availability. This research project is reported according to STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) guidelines.²⁰

Participants

The NIS contains administrative claims on approximately 8 million acute care hospitalizations per year. Visits in the NIS are sampled from State Inpatient Databases (44 – 47 states each year), comprising over 95% of the US population in the sampling frame with a 20% stratified sample of US hospital discharges.¹⁹ The NIS data from 2012 to 2016 were combined to include 35.8 million unweighted visits. Hospital visits of adult stroke patients were included in the study. Pediatric patients (<18 years old) and patients with a discharge disposition of transfer were excluded. Stroke was defined by a previously validated combination of discharge diagnosis codes (International Classification of Diseases[ICD]-9: 430.xx, 431.xx, 434.xx, or 436.xx; ICD-10: I60, I61, I63, or I64) at any position in the diagnostic codes.²¹ Patients transferred in to a facility were included in the study.

Variables

Exposure.

The primary exposure was the rurality of the subject’s residence. Rurality was determined by the National Center for Health Statistics (NCHS) urban-rural classification scheme. “Rural” was defined as a subject residing in a county considered to be non-metropolitan (micropolitan/not metropolitan or micropolitan counties). “Non-rural” included subjects residing in a county with a metropolitan area (central metropolitan/fringe metropolitan/metropolitan with 250,000-999,999 people/metropolitan with 50,000 to 250,000 people). An additional analysis was performed using six categories of rurality, as defined by the NCHS categories: central counties of metropolitan area with population of one million or greater, fringe of counties metropolitan area with population of one million or greater, counties in metro areas of 250,000 to 999,999 people, counties in metro areas of 50,000 to 250,000 people, micropolitan counties, and not metropolitan or micropolitan counties.

Outcome.

The primary outcome was death during the hospital encounter. Secondary outcomes were a composite of death or discharge to an institution (e.g. skilled nursing facility, long-term acute care hospital) to approximate an outcome of poor neurological function at discharge and discharge to a care facility (non-home) care.

Covariates.

Subject-level and visit-level covariates included age, sex, race, primary payer, transfer status, mortality risk, co-morbidities, type of stroke, and admission status. Age was categorized into four groups: 18-44 years, 45-64 years, 65-84 years, and 85 or more years. Transfer status was a binary indicator of subjects with a hospital admission originating from an inter-hospital transfer. Mortality risk was estimated using proprietary software (3M Health Information Systems, St. Paul, MN) assigning All Patient Refined-Diagnosis Related Groups and classified into five groups: Extreme Likelihood of Dying, Major Likelihood of Dying, Moderate Likelihood of Dying, Minor Likelihood of Dying, and no class specified.¹⁹ This is a proprietary mortality risk estimation calculated from age, diagnoses, and procedures and has been used to estimate severity in previous studies of stroke patients in administrative claims data.^22–25 Co-morbidities were defined according to the 29 reduced Elixhauser co-morbidities (NIS does not include cardiac arrythmias).²⁶ The type of stroke was dichotomized as ischemic or non-ischemic; ischemic stroke was defined using diagnostic codes (ICD-9: 434 or ICD-10: I63).²¹ Tissue plasminogen activator (tPA) administration was defined by procedure codes during the admission (ICD-9-CM: 99.10 or ICD-10-PCS: 3E03017, 3E03317. 3E04017, 3E04317, 3E05017, 3E05317, 3E06017, AND 3E06317).^{18, 27, 28} Admission status was defined as elective or non-elective. Facility-level covariates included hospital region, rurality and teaching status, inpatient volume, and control/ownership of the hospital. Rural hospitals were defined by the Core-Based Statistical Area of the hospital county and included hospitals in micropolitan and non-core counties. Inpatient volume was categorized into deciles based on the total number of discharges reported by the hospital. Hospital annual volume tertiles were defined as low (less than 2,453 total discharges per year), middle (2,453 – 4,677 discharges), and high (4,678 or more discharges). Control/ownership of the hospital was categorized as: government- nonfederal, private- not-for-profit, or private – investor-owned. To preclude identification of hospitals in small strata, two collapsed categories (government/private and private) were also included (for a total of five categories of hospital ownership). A dummy categorical variable for year was included to account for temporal trends. Patients with missing data on in-hospital death, rurality of residence, or both were excluded.

Statistical Methods

Descriptive statistics were used for subject-, visit-, and facility-level covariates. Chi-square and Wilcoxon-Mann-Whitney test were used to compare covariates by rural residence. Univariable logistic regression predicted the odds of mortality by rurality of residency among stroke patients. Multivariable modeling was based on theory (ie previous literature and clinical guidelines) and statistical significance. Variables were screened for model inclusion (p<0.20). Multicollinearity was assessed using Variance Inflation Factors. Survey weighting was used to account for the complex NIS sampling design, including accounting for unobserved visits with weighting to produce national estimates.²⁹

Effect Modification.

Four variables were specified a priori as hypothesized modifiers of the relationship between rurality and mortality in stroke: primary payer, stroke type, hospital location/teaching status, and transfer status. Interaction terms between each potential effect modifier and the primary exposure (rurality) were tested in separate models, adjusted for covariates. If a statistically significant interaction was found, a stratified analysis was performed to estimate the odds ratio for rural and non-rural visits. One additional pre-planned subgroup analysis was performed to evaluate the role of hospital volume on the relationship between rurality and mortality among stroke admissions. Based on visualization of effects in deciles, alternate volume categorizations, including tertiles, quartiles, and quintiles were performed to select the volume categorization that best described data patterns. The subgroup analysis was performed after grouping subjects by rurality (based on patient residence) and hospital volume tertiles.

Sensitivity Analysis.

Sensitivity analyses were performed (1) defining the study population as only subjects with a non-elective admission status and (2) re-defining ischemic stroke using a less-specific definition (ICD-9: 434 or 463 or ICD-10: I63 or I64).²¹

All statistical analyses were completed using SAS (version 9.4; SAS Institute, Inc., Cary, NC).

Results:

Description of Study Participants

There were 1,865,310 weighted inpatient visits for stroke included in the study sample (Figure 1). Approximately 17.9% (95% CI 17.5 – 18.3%) of stroke admissions were subjects from a rural county. Visits from rural subjects were more likely to be white race, Medicare primary payer, and transferred from another facility than visits from non-rural subjects (Table 1). Visits at lower volume hospitals, rural hospitals, Midwest and South hospitals, and government hospitals were more common among visits of rural stroke subjects (Figure 2).

Figure 1.

Flowchart of Study Subjects.

Table 1.

Subject Demographics and Clinical Characteristics by Subject Rurality.

	Total		Rural		Non-Rural
Characteristics	N 1,747,660	% (CI) 100 -	n 303,310	% (CI) 17.4 (16.9 – 17.8)	n 1,444,350	% (CI) 82.6 (82.2 – 83.1)
Age
18 – 44 years	132,905	7.6 (7.5 – 7.7)	19,110	6.3 (6.0 – 6.6)	113,795	7.9 (7.7 – 8.0)
45 – 64 years	618,740	35.4 (35.2 – 35.6)	102,755	33.9 (33.4 – 34.3)	515,985	35.7 (35.5 – 36.0)
65 – 84 years	765,640	43.8 (43.6 – 44.0)	142,480	47.0 (46.5 – 47.4)	623,160	43.1 (42.9 – 43.4)
85+ years	230,375	13.2 (13.0 – 13.3)	38,965	12.8 (12.5 – 13.2)	191,410	13.2 (13.1 – 13.4)
Female	857,085	49.0 (48.9 – 49.2)	147,875	48.8 (48.4 – 49.2)	709,210	49.1 (48.9 – 49.3)
Payer
Medicare	983,940	56.3 (56.0 – 56.6)	187,285	61.8 (61.2 – 62.3)	796,655	55.2 (54.9 – 55.5)
Medicaid	181,415	10.4 (10.2 – 10.6)	25,210	8.3 (8.0 – 8.6)	156,205	10.8 (10.6 – 11.0)
Private Insurance	403,700	23.1 (22.8 – 23.4)	60,355	19.9 (19.5 – 20.3)	343,345	23.8 (23.5 – 24.1)
Self-Pay	116,950	6.7 (6.5 – 6.9)	19,755	6.5 (6.2 – 6.8)	97,195	6.7 (6.5 – 6.9)
No Charge	10,210	0.6 (0.5 – 0.6)	1,105	0.4 (0.3 – 0.5)	9,105	0.6 (0.6 – 0.7)
Other	50,645	2.9 (2.8 – 3.0)	9,550	3.1 (2.9 – 3.4)	41,395	2.9 (2.8 – 3.0)
Race/Ethnicity
White, Non-Hispanic	1,178,435	67.4 (66.9 – 68.0)	250,910	82.7 (81.9 – 83.6)	927,525	64.2 (63.6 – 64.8)
Black, Non-Hispanic	294,475	16.9 (16.5 – 17.2)	31,150	10.3 (9.6 – 10.9)	263,325	18.2 (17.8 – 18.6)
Any Race, Hispanic	157,285	9.0 (8.7 – 9.3)	10,290	3.4 (3.1 – 3.7)	146,995	10.2 (9.8 – 10.5)
Transfer Admission Source
Transfer from Acute Care Hospital	203,635	11.7 (11.2 – 12.1)	71,475	23.6 (22.5 – 24.6)	132,160	9.2 (8.7 – 9.6)
APR-DRG Risk of Mortality
Extreme	394,605	22.6 (22.4 – 22.8)	67,990	22.4 (22.0 – 22.8)	326,615	22.6 (22.4 – 22.8)
Major	341,840	19.6 (19.4 – 19.7)	59,850	19.7 (19.4 – 20.1)	281,990	19.5 (19.4 – 19.7)
Moderate	537,120	30.7 (30.5 – 30.9)	93,645	30.9 (30.5 – 31.3)	443,475	30.7 (30.5 – 30.9)
Minor	474,025	27.1 (26.9 – 27.3)	81,085	27.0 (26.6 – 27.4)	392,220	27.2 (26.9 – 27.4)
Co-morbidities
Congestive Heart Failure	216,115	12.4 (12.2 – 12.5)	39,315	13.0 (12.7 – 13.3)	176,800	12.2 (12.1 – 12.4)
Paralysis	130,890	7.5 (7.4 – 7.6)	19,615	6.5 (6.2 – 6.7)	111,275	7.7 (7.6 – 7.8)
Chronic Pulmonary Disease	273,340	15.6 (15.5 – 15.8)	55,390	18.3 (17.9 – 18.6)	217,950	15.1 (14.9 – 15.2)
Diabetes Mellitus without Chronic Complications	439,270	25.1 (24.9 – 25.3)	80,605	26.6 (26.2 – 27.0)	358,665	24.8 (24.6 – 25.1)
Diabetes Mellitus with Chronic Complications	151,685	8.7 (8.5 – 8.8)	22,995	7.6 (7.3 – 7.9)	128,690	8.9 (8.7 – 9.1)
Coagulopathy	102,700	5.9 (5.8 – 6.0)	15,705	5.2 (5.0 – 5.4)	86,995	6.0 (5.9 – 6.1)
Deficiency anemias	223,725	12.8 (12.6 – 13.0)	34,630	11.5 (11.1 – 11.7)	189,095	13.1 (12.9 – 13.3)
Alcohol abuse	87,765	5.0 (4.9 – 5.1)	14,395	4.7 (4.6 – 4.9)	73,370	5.1 (5.0 – 5.2)
Drug abuse	65,650 (895)	3.8 (3.7 – 3.8)	8,650	2.9 (2.7 – 3.0)	57,000	3.9 (3.9 – 4.0)
Ischemic Stroke Type	1,439,715	82.4 (82.1 – 82.7)	253,510	83.6 (83.1 – 84.1)	1,186,205	82.1 (81.8 – 82.4)
Elective Admission Status	107,330	6.1 (5.9 – 6.3)	31,300	10.3 (9.9 – 10.8)	76,030	5.3 (5.1 – 5.5)

APR-DRG = All Patients Refined Diagnosis Related Groups.

Figure 2.

Hospital Demographics by Subject Rurality.

Univariable Analysis

Hospital mortality was 17.2% (95%CI 17.0 – 17.4%) among the stroke subjects. Rural stroke subjects had a higher mortality than non-rural subjects (18.6% rural vs. 16.87% non-rural), uOR = 1.14; 95%CI 1.11 – 1.17; P < 0.001) (Table 2).

Table 2.

Univariate and Multivariable Logistic Regression Results.

	Model 1: Unadjusted			Model 2: Adjusted*			Model 3: Adjusted* with Transfer Status and tPA
	uOR	95%CI	p-value	aOR	95%CI	p-value	aOR	95%CI	P-value
MORTALITY
Rural vs. Non-Rural	1.14	1.11 – 1.17	<0.001	1.11	1.06 – 1.15	<0.001	1.06	1.02 – 1.11	0.006
MORTALITY OR NON-HOME DISCHARGE
Rural vs. Non-Rural	0.98	0.96 – 1.00	0.0378	0.94	0.91 – 0.97	<0.001	0.93	0.90 – 0.96	<0.001
NON-HOME DISCHARGE
Rural vs. Non-Rural	0.88	0.86 – 0.90	<0.001	0.90	0.87 – 0.92	<0.001	0.91	0.88 – 0.94	<0.001

Model 1 represents the unadjusted association between rurality and mortality.

Model 2 represents the association between rurality and mortality adjusted for patient and hospital factors.

Model 3 represents the association between rurality and mortality adjusted for two factors commonly thought to contribute to excess rural stroke mortality, tPA and inter-hospital transfer, as well as patient and hospital factors

^*Adjusted for: age, sex, race, primary payer, mortality risk, co-morbidities, type of stroke, admission status, hospital region, rurality and teaching status, inpatient volume, control/ownership of the hospital, and year.

Rurality and Mortality

After adjustment (age, sex, race, primary payer, mortality risk, co-morbidities, type of stroke, tPA administration, admission status, hospital region, rurality and teaching status, inpatient volume, control/ownership of the hospital, and year) in a logistic regression model, subject rurality was associated with an increased odds of mortality (aOR = 1.11; 95%CI 1.06 – 1.15; P < 0.001) (Table 2). The association between categories of rurality (central metropolitan, fringe metropolitan, metropolitan (250K – 1M), metropolitan (50K-250K), micropolitan, and non-metropolitan/micropolitan) and mortality is shown in Figure 3.

Figure 3.

Dose-response Relationship between Rurality and Mortality.

Rurality and Secondary Outcomes (Mortality/Disability & Discharge to Home)

For the secondary outcome, rural stroke visits were less likely to experience a composite outcome of death and disability than non-rural stroke visits (aOR 0.94; 95%CI 0.91 – 0.97; P < 0.001) (Table 2). Rural subjects were also less likely to be discharged to a facility after hospitalization (aOR 0.90; 95%CI 0.87 – 0.92; P < 0.001).

Effects of Hospital Interventions: tPA, Hospital Volume, and Inter-hospital Transfer

The source of 13.9% (95% CI 13.2 – 14.5) of stroke admissions was inter-hospital transfer. Adjustment for transfer status and tPA attenuated the increased odds of mortality for rural subject visits (aOR=1.06; 95%CI 1.02 – 1.11; P = 0.006), meaning that transferred rural patients had less of an increased mortality than those who weren’t transferred (Table 2). Inter-hospital transfer did not modify the relationship between rurality and mortality when included as an interaction term in the multivariable model (P = .063 for rurality by transfer interaction term). The effect of patient rurality on mortality among stroke patients differed by volume of the hospital, defined as tertile of annual inpatient volume (P = .001). While non-rural stroke subjects had similar mortality based on hospital volume tertile, rural stroke subjects had higher odds of mortality in lower volume hospitals (Figure 4A). Further, in both rural and non-rural hospitals, patient rurality was associated with increased odds of mortality compared to non-rural patients in non-rural hospitals (rural hospitals: aOR = 1.46, 95%CI 1.36 – 1.56, p < 0.001; and non-rural hospitals: aOR = 1.43, 95%CI 1.18 – 1.74, P < 0.001) (Figure 4B). Interestingly, non-rural patients in rural hospitals also experienced increased mortality compared to their non-rural peers in non-rural hospitals (aOR= 1.06; 95%CI 1.02 to 1.11; P = 0.006).

Figure 4.

Rural versus Non-Rural Mortality in Stroke Patients by Hospital Annual Volume.

Effects of Stroke Type and Primary Payer

The effect of subject rurality on mortality was not different by stroke type (non-ischemic versus ischemic) (P = .552). Primary payer source modified the relationship between rurality and mortality in stroke (P = .012). Within payer categories, only the other category (including Worker’s Compensation, Indian Health Services, CHAMPUS/VA, Elmendorf, Other Government, and Other Miscellaneous) was an effect modifier, indicating rural patients from these payer categories may be especially at risk of increased mortality.

Sensitivity Analyses

Sensitivity analyses were performed using (1) an alternative definition of ischemic stroke²¹ and (2) only non-elective stroke admissions. Results did not change in the sensitivity analyses, as rurality remained similarly associated with increased mortality and with a decrease in the composite outcome of death or non-home discharge (Table S1).

Discussion:

Stroke patients who reside in rural areas have higher mortality than stroke patients who live in urban areas. This result is consistent with previously published work, however, our study adjusted for previously identified factors thought to be causal and there is still a significant disparity. We conclude that there are unidentified factors remaining related to rural stroke care that contribute to the disparity.^{2, 30–32}

Previously, increased rural stroke mortality has been partially attributed to increased patient-level risk factors in rural areas and differences in hospital-level care. Howard et al. suggest the “more detrimental risk factor profile” and higher risk factor burden in rural communities contribute to the rural-urban stroke mortality disparity more than socioeconomic status.^{2, 31} Comorbidities, as well as some proxies for socioeconomic status (payer and race/ethnicity), were included in our adjusted calculation and the rural-urban disparity in mortality remained. Risk factor and socioeconomic differences between rural and urban populations may contribute to the stroke mortality disparity, as previous literature suggests, but our data indicate there are more factors yet to be identified.

Understanding why rural patients experience higher stroke mortality than urban patients is critical for reducing this disparity. By including models that do and do not incorporate thrombolytics and transfer, we have estimated both the total effect of rurality on mortality and the direct effect of rurality on mortality after adjusting for other patient and hospital factors. In the hospital, decreased utilization and delays in thrombolytic administration have been associated with increased stroke mortality in rural patients.⁷ Hospitals that receive transfers often have increased resources and are more likely to be comprehensive stroke centers (CSC), which have been proven to improve stroke patient outcomes.³³ In our study, the mortality disparity remained after adjusting for thrombolytic administration and transfer status. This suggests that while rural patients that are transferred have a better chance at survival compared to patients admitted locally, rural stroke patients still have higher risk of death. This points to additional factors contributing to mortality in rural stroke patients in the rural pre-hospital or hospital environment. Identification of these additional factors and their relative effects on the rural-mortality relationship is important.

In our secondary outcome analysis, we found that rural stroke patients were more likely to be discharged home, regardless of transfer status. This is similar to results of previous studies that suggest rural stroke patients are less often transferred to inpatient rehabilitation.^{2, 13} This finding could be due to multiple factors. Patient and family beliefs, priorities, and financial concerns may influence use of long-term care. Rural families may feel more equipped to take care of stroke survivors at home. Alternatively, the findings of increased mortality and reduced care center discharge may indicate rural-urban differences in end-of-life care decisions.

We propose that pre-hospital systems are contributors to rural stroke mortality.

First, the time from symptom onset to medical response arrival clearly affects stroke outcomes. Early recognition of stroke symptoms and prompt EMS activation are potentially delayed in rural systems as rural populations have been identified as having lower health literacy regarding stroke recognition and management.³⁴ Public health campaigns have been used to increase stroke awareness and recognition, but the campaigns are resource-intensive and may be difficult to scale across geographically vast rural settings.³⁵

Second, rural EMS systems fundamentally struggle to maintain a rural provider’s clinical competency, which may translate into challenges during the management of acute stroke. Rural EMS providers often volunteer to serve their community, sometimes with a primary profession not related to healthcare, and may face time and financial barriers in maintaining their medical knowledge. They may not reliably identify and/or treat stroke patients as they may not be as updated on current stroke care guidelines and see fewer strokes in general.³⁶ There is then potential risk of lack of understanding of blood pressure goals, decreased pre-notification of receiving hospitals and decreased awareness of when to initiate air medical transport.

EMS providers have indicated they want more education in stroke care and would like feedback from hospitals.³⁷ EMS systems are encouraged to partner with receiving stroke centers for stroke protocols and continuing education that can be presented at staff meetings allowing rural EMS providers to maintain their training. In addition, simulation scenarios can be incorporated into the providers’ training for practical application on concepts taught in the classroom. Feedback loops should be developed so rural EMS systems can better understand their patient outcomes. These initiatives can lead to EMS providers’ more rapid recognition of stroke symptoms which we know is associated with faster door-to-physician, door-to-balloon times, and greater likelihood of receiving thrombolytics – all known to improve patient functional outcomes.^{38, 39}

Third, Jennings et al. found that cardiac arrest patients in an urban area had significantly higher survival rates than patients in rural areas with attribution predominantly assigned to the difference in EMS arrival times.⁴⁰ Stroke is also a time sensitive disease with similar risk factors, and it can be extrapolated that longer EMS arrival times can increase the mortality of these patients as well. A recent systematic review investigating patient outcome and its intersection with rural versus urban EMS noted that rural EMS patients experienced longer wait times for EMS to arrive, experienced longer on-scene times, had longer transport times and overall spent a longer time in pre-hospital setting. Shultis et al. reported a higher stroke death rate in rural areas than in urban areas.^{41, 42} The reviewed articles indicated that patients living in rural areas had lower survival rates from equivalent conditions when compared to those living in urban areas. Many of the issues associated with this differential in survival rates have been associated with prolonged EMS arrival to the scene. Rural EMS systems will continually need to focus on strategies that address these time barriers which may include a minimal paid on-call staff that can initiate response while volunteer providers are mobilizing, commitment to comparable on-scene times as urban services (with the noted challenge that rural EMS providers may not have the same repetition level and professional experience), and development of well-honed mutual aid agreements with air medical services.

Our study has shown that rural pre-hospital care can be a critical piece to improving stroke outcomes. However, further research must be done to better characterize what rural pre-hospital factors are most significant and develop rural-oriented solutions.

There are important limitations to our study. First, this study lacks a stroke-specific measure of severity of illness, such as National Institutes of Health Stroke Scale (NIHSS). As this was unavailable, we used the APR-DRG, an all-disease severity index, but residual confounding may remain. Second, only information related to the discharging hospital was available, without knowing data about other hospitals in which patients received care prior to transfer. Third, there is no information on mode of arrival to the hospital (ie EMS or private vehicle) or the characteristics of those EMS agencies. As EMS is unmeasured in this dataset, differences in EMS use and capabilities associated with rurality are likely included in the reported rurality effects or in the hospital volume effects (as EMS relates to hospital choice). Fourth, we have no data on why some patients did not receive tPA (eg too late, ineligible). Fifth, this dataset used reports at the visit-level rather than the patient-level, yet we report results using patient-level verbiage. As we used a validated definition of incident stroke,⁴³ we assume there are few repeat visits captured in the study cohort. Sixth, a higher number of “elective admissions” were observed in rural patients compared to urban patients and we are unable to obtain clinical details to understand the differences in admission status. Finally, mortality is not the most sensitive outcome for stroke care. Use of patient-centered outcomes sensitive to stroke care, such as neurological disability and return to daily activities are ideal outcomes. Using discharge to a facility, we attempted to approximate the outcome of neurologic disability in this study, but future studies could focus on more robust definitions of this outcome.

Conclusion:

After accounting for known hospital-level drivers of rural stroke mortality, there remained an association between rurality and mortality. Further studies could determine the extent rural pre-hospital care contributes to the mortality disparity and what modifiable risk factors can be addressed at the EMS agency and systems level.

Prior Presentations:

• Oral abstract presentation at the 2019 Great Plains Society for Academic Emergency Medicine meeting (Springfield, IL)

• Poster abstract presentation at the 2020 National Association of EMS Physicians meeting (San Diego, CA)