Stroke Technology Diffusion in Rural Settings: Differential Exposure to Certification Levels by Community Income Levels

Abstract

Background and Objectives:

Although stroke technology and care infrastructure have advanced significantly, it remains unclear whether recent expansions of certified stroke centers have benefited rural patients equitably across income levels. This study assessed whether rural communities of varying income experienced similar gains in access to certified stroke centers and whether such expansions were associated with improvements in acute stroke treatment and outcomes.

Methods:

We conducted a retrospective cohort study using 100% Medicare Provider and Analysis Review (MedPAR) data from January 1, 2009, to December 31, 2019. This study included all Medicare fee-for-service beneficiaries diagnosed with acute ischemic stroke who resided in rural US communities. Communities were classified as exposed if a newly certified stroke center – Acute Stroke Ready Hospital (ASRH), Primary Stroke Center (PSC), Thrombectomy-Capable Stroke Center (TSC), or Comprehensive Stroke Center (CSC) – opened within a 30-minute drive. A community-fixed-effects linear probability model was used to evaluate changes in outcomes following stroke center certification. Primary outcomes included: (1) admission to a stroke-certified hospital, (2) receipt of thrombolytic therapy, (3) receipt of thrombectomy, and (4) one-year mortality.

Results:

Among 590,191 rural stroke patients, 4% of low-income and 22% of high-income patients had access to a nearby certified stroke center in 2009. By 2019, 30% of low-income and 50% of high-income communities gained access to at least one newly certified stroke center; high-income communities were three times more likely than low-income communities to be exposed to a newly certified TSC or CSC (5.4% vs. 1.8%). Exposure to ASRHs increased the probability of thrombolysis by 0.63 percentage points (95% CI: 0.05, 1.22), while exposure to TSC/CSCs increased the probability by 1.39 points (CI: 0.28, 2.49), and thrombectomy by 1.12 points (CI: 0.41, 1.83). No differences in 1-year mortality were observed.

Discussion:

During the study period, high-income rural communities experienced more frequent and higher-tier stroke center expansion than low-income rural communities. These access disparities were associated with differential gains in advanced stroke treatments, suggesting that expansions may have inadvertently widened income-based disparities in rural stroke care. These findings underscore the need for equity-focused implementation strategies, ensuring that infrastructure improvements translate into equitable clinical benefits.

Introduction

Over the past few decades, significant developments in stroke technology have revolutionized the care and outcomes for patients with acute ischemic stroke. These innovations, coupled with the establishment of certified stroke centers, have been associated with reduced mortality^1,2 and improved access to timely interventions.³ These interventions include intravenous thrombolytic therapy, which has been the mainstay of stroke treatment for the past three decades, as well as more recent discoveries in mechanical thrombectomy, which became more widely adopted after several landmark studies in 2015.^4–8 However, the diffusion of stroke technology has not been uniform, and disparities in access remain, particularly for lower-income and rural populations. Prior research⁹ has concluded that living in rural areas is linked to reduced access to healthcare services and stroke-specific treatments, such as intravenous thrombolysis and mechanical thrombectomy.

Existing literature has primarily focused on the patterns of growth and accessibility of stroke certification in affluent urban settings, where these facilities are preferentially concentrated. A study from 2014 reported that only 1% of Americans living in rural settings had less than 60- minute access to a certified Primary Stroke Center, compared to 87% of Americans living in major urban centers.¹⁰ Due to stroke certification being voluntary, urban and affluent areas are adopting certification at a consistently higher rate,¹¹ contributing to improved outcomes for patients in these settings. However, rural hospitals, where patients already face unique barriers to healthcare such as limited resources, fewer specialized staff, and greater geographical barriers to care,¹² have received less attention in stroke certification. As a result of these challenges, the capacity for treating acute stroke in rural hospitals is significantly lower than in urban hospitals.¹² There is limited literature on how the patterns of stroke certification have evolved in rural areas over time, both in affluent and lower-income settings. It remains unknown whether rural stroke centers are following the same trajectory as their urban counterparts or if disparities between affluent and low-income rural communities are widening further.

This study seeks to examine the diffusion of stroke center certification in rural settings between 2009 and 2019, with a specific focus on income-based disparities. By analyzing trends in access, treatment, and outcomes over this period, we aim to determine whether patterns of stroke center growth have shifted (or self-corrected) to address inequities in stroke care. Has the distribution of stroke care technology diffused in such a way that it is helping the least resourced or is it worsening these disparities? Ultimately, this work aims to inform policies that can close critical gaps in stroke care, particularly for rural populations, ensuring equitable access to life-saving stroke technology and treatment.

Methods

Study population and data.

Our study population includes all Medicare fee-for-service patients who reside in rural communities and were admitted to hospitals for ischemic stroke between January 2009 and December 2019, where rural communities were defined as ZIP codes that were identified as rural areas per the Federal Office of Rural Health Policy (FORHP) under the Health Resources & Services Administration. FORHP first defines rural areas using Census tract criteria, where rural tracts are those with RUCA codes 4-10. It then mapped Census tracts to ZIP code tabulation areas and selected the rural status with the higher population count as the rural approximation for a given ZIP code. A ZIP code is identified as a rural community if it appears in FORHP’s ZIP Code Approximation Excel file.¹³ Mean population count based on Census 2010 among these rural ZIP code communities is 3557 (average count in urban communities is 15,872). Geography was defined at the patient’s residential 5-digit ZIP level, the smallest geographic unit available across all databases in our analysis, that allowed uniform linkage to facility service files and standard ZIP-based rurality/community measures. We use the internal longitude and latitude coordinates provided by the US Census based on the ZIP code tabulation area (ZCTA), which allows us to obtain road-network drive-time calculations from ZIP centroids. Following prior literature,^14–16 a Medicare patient is included in our patient cohort if they have the primary diagnoses in one of the following International Classification of Diseases (ICD) codes: 433.x1, 434.x1, or 436 (version 9), or I63 (version 10). We focus on ischemic stroke patients (as opposed to the broader stroke population that also includes hemorrhagic stroke) since ischemic strokes account for 87% of all strokes,¹⁷ with the major developments in treatment options focused on ischemic stroke rather than hemorrhagic.

At the patient level, we obtained data from the 100% Medicare Provider and Analysis Review (MedPAR), which captures demographics, admission date, mailing ZIP code, diagnoses of comorbidities, and procedure codes and dates; and Medicare beneficiary summary files, which contain vital statistics for identifying death dates. At the hospital level, we identified the stroke certification status for hospitals and the dates certified using a database that we collected from national accrediting bodies and state accreditation agencies.⁸ In addition, we obtained additional hospital characteristics, such as ownership, teaching status, and longitude and latitude of the hospital’s location, from the American Hospital Association and the Healthcare Cost Report Information System. At the community, i.e., ZIP code level, we used 2010 US Census data and American Community Surveys from 2011 to 2019 to identify geographical coordinates, income, and other demographic information for each ZIP code community, and FORHP data to identify rural communities. Lastly, we developed a driving time database via web queries that capture driving time between a community’s center point to nearby stroke centers based on the pair’s longitude and latitude information.¹⁸

Defining exposure to newly certified stroke centers for a given community.

Our study design follows a difference-in-differences framework where we compare changes in outcomes in a community before and after exposure to newly certified stroke centers within a 30-minute driving time (treatment group) relative to changes in outcomes in a community without exposure to newly certified stroke centers during the same period (control group). We define reasonable geographic access as within 30 minutes. Within 30 minutes denotes driving time only (one-way road travel from residential ZIP to the nearest qualifying facility); other prehospital intervals are not included. We use a 30-minute threshold as a widely applied benchmark for reasonable geographic access in health services research.^19–23 This measure captures potential geographic access, not realized prehospital time.

Of note, hospitals can be certified at different levels. At the entry level, hospitals can be certified as Acute Stroke Ready Hospitals (ASRH), whose primary focus is on emergency stroke treatment to stabilize patients and to establish transfer protocols with higher-level stroke centers. Hospitals can also be certified as Primary Stroke Centers (PSC), with more structured inpatient resources dedicated to stroke care compared to ASRHs. Finally, hospitals can also be certified as Thrombectomy-Capable Stroke Centers (TSC) or Comprehensive Stroke Centers (CSC), which have the capability to perform mechanical thrombectomy and have more specialized stroke resources. The methods behind obtaining hospital certification/designation from the four CMS-approved national certifiers are well-described in one of our prior studies.²⁴ Briefly, we obtained yearly certification data from the Joint Commission (TJC), Det Norske Veritas (DNV), Accreditation Commission for Health Care (ACHC), and the Center for Improvement in Healthcare Quality (CIHQ), as well as from state health departments. Certification was coded annually; when multiple designations occurred in the same year, we applied a prespecified priority rule and retained the highest level, per prior work.²⁴

For each community, we first identified all hospitals that are within a 30-minute driving time for each year-quarter, regardless of their stroke center status. We then defined three post-exposure indicators, based on whether the community had exposure to a newly certified (1) ASRH, (2) PSC, or (3) TSC/CSC. Take exposure to an ASRH as an example, a post-exposure indicator turns from 0 to 1 on and after the year-quarter that a hospital from the 30-minute set of a community attained ASRH status in that year-quarter; likewise, for exposure to PSC and TSC/CSC. We combined TSCs and CSCs in one category because there are very few TSCs, especially in rural communities. In a sensitivity analysis, we increased the drive time threshold to 60 minutes so we can capture changes for remote rural communities.

Determining a community’s income level.

Our descriptive analysis first explored where stroke center expansions occurred in rural communities. Specifically, we examined a community’s baseline stock of stroke centers and exposure to newly certified stroke centers by a community’s income levels. We used the 2010 Census, which reports median family income at the ZIP code community levels, to categorize communities into low-, medium-, and high-income. A community was designated as low-income if its median family income was at the lowest quartile of income distribution among all rural communities; high-income if its median family income was at the upper quartile; and medium if within the interquartile. The income level categorization is time-invariant so that communities do not move in and out of income categories.

Outcomes.

We examined actual access, treatment received, and health outcomes. For access, we examined whether a patient was admitted to a certified stroke center. We examined two types of treatment outcomes during the care episode, based on their ICD-9 or ICD-10 procedure codes and dates: whether a patient received thrombolytic therapy and whether a patient received mechanical thrombectomy. The list of procedure codes for each treatment is provided in supplemental materials. For health outcomes, we examined 1-year mortality.

Statistical methods.

Our multivariate analysis followed a difference-in-differences framework, in which we implemented a linear probability model that incorporate both community and time fixed effects, where the key independent variables were the three post-exposure variables defined above (exposure to ASRH, PSC, and TSC/CSC). Using thrombectomy as an example, the coefficient estimate from the post-ASRH indicator represented changes in the probability of receiving thrombectomy after the treatment communities were exposed to newly certified ASRH relative to the reference communities that did not experience changes to their stroke capacities nearby. This design allows us to align a “time zero” of exposure to each community’s specific certification event, thereby addressing the staggered timing of treatment. The community fixed effects were critical because they removed time-invariant unobserved differences across communities (such as underlying differences in patients’ health, cultures, and socioeconomic conditions). In addition, we controlled for individual patients’ demographics (sex, 5-year age groups, race, ethnicity) and comorbidities so that we could compare comparable patients between treatment and control communities.^25,26 Lastly, we included year dummies (i.e., time fixed effects) to capture macro trends that are common to all communities over time.

Role of the Funding Source.

The funders had no role in the design and conduct of the study; reporting, collection, management, analysis, and interpretation of the data.

Standard Protocol Approvals, Registrations, and Patient Consents.

This cohort study was approved by the Institutional Review Board of the National Bureau of Economic Research and followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines.

Data Availability.

The dataset analyzed in the study is not publicly available due to legal and privacy regulations concerning patient-level data.

Results

Our final study population included 590,191 stroke patients who lived in rural communities. Figure 1, Panel A documents the disparities in treatment and outcomes across income levels, illustrating that stroke patients in rural low-income communities consistently had a lower probability of receiving thrombolytic therapy compared to those from high-income areas. Specifically, the gap was 1.9 percentage points in 2009, which widened to 2.9 percentage points in 2018. Similarly, Figure 1, Panel B shows that low-income rural stroke patients had a lower probability of receiving thrombectomy than high-income residents throughout the entire study period, with the gap growing from 0.16 of a percentage point in 2009 to 0.75 of a percentage point in 2019. Low-income rural stroke patients also consistently experienced higher 1-year mortality rates compared to stroke patients in rural high-income communities throughout the study period (Figure 1, Panel C).

Figure 1.

Trend in treatment and 1-year mortality for ischemic stroke patients between 2009 and 2019 in rural communities, by income levels

The rest of our analysis focused on examining whether and how the expansion of stroke centers has contributed to the persistent and widening gaps across income levels over time that were observed in Figure 1. Table 1 shows the characteristics of the patient population. Within this rural population, 525736 (89%) were White, 47643 Black (8%), 4236 Hispanic (1%), and the remaining 1% other races. Approximately 55% were female with the remaining 45% male. Overall, only 43% of patients in rural communities were exposed to newly certified stroke centers within a 30-minute drive time during the study period. The most common stroke center opening nearby was PSC (24% had access), followed by ASRH (15%), then TSC or CSC (3%). Online supplemental Table e1 further breaks down the patient characteristics by the level of certification to which that community was exposed, and detailed facility characteristics are described in prior work.²⁷

Table 1.

Descriptive Statistics of Rural Patient Characteristics

	N	%
N	590,191
Lives in a community where a hospital within a 30-minute drive received stroke certification during study period (2009-2019)
Overall	252,086	43%
By highest certification level
ASRH	88,157	15%
PSC	144,515	24%
TSC or CSC	19,414	3%
Income Levels
Low Income	161,758	27%
Medium Income	298,015	50%
High Income	130,418	22%
Patient demographics
White	525,736	89%
Black	47,643	8%
Hispanic	4,236	1%
Other non-white races	5,650	1%
Female	322,909	55%
Age distribution at time of admission
65-69 years	97,835	17%
70-74 years	106,858	18%
75-79 years	112,291	19%
80-84 years	111,689	19%
85+ years	161,518	27%
Patient clinical conditions
Recurring stroke	43,333	7%
Transfer	104,341	18%
Peripheral vascular disease	55,132	9%
Pulmonary circulation disorders	18,521	3%
Diabetes	194,927	33%
Kidney failure	102,753	17%
Liver	4,964	1%
Cancer	22,189	4%
Dementia	55,082	9%
Valvular disease	52,139	9%
Hypertension	495,864	84%
Chronic pulmonary disease	102,921	17%
Rheumatoid arthritis/collagen vascular	16,353	3%
Coagulation deficiency	19,582	3%
Obesity	48,920	8%
Substance use	12,139	2%
Depression	58,181	10%
Psychosis	38,173	6%
Hypothyroidism	102,566	17%
Paralysis and other neurological disorder	340,639	58%
Ulcer	1,945	0%
Weight loss	25,260	4%
Fluid and electrolyte disorders	130,933	22%
Anemia (blood loss and deficiency)	66,918	11%
Access, Treatment, and Health Outcomes
Admitted to stroke hospital	348,490	59%
Received thrombolytic therapy during hospitalization	55,892	9%
Received thrombectomy	13,752	2%
1-year mortality	176,194	30%

Abbreviations: ASRH, Acute Stroke Ready Hospital; PSC, Primary Stroke Center; TSC, Thrombectomy Capable Stroke Center; CSC, Comprehensive Stroke Center.

To investigate the pattern of stroke center expansion over time, Figure 2 shows that in 2009, only 4% of patients in low-income rural communities had access to a stroke center within a 30-minute drive time, whereas 22% in high-income communities already had access to a stroke center within a 30-minute drive time. The baseline access disparity widened over the study period as more newly certified stroke centers tended to occur near high-income communities. As illustrated in Figure 3, Panel A, patients living in high-income communities experienced a greater expansion of newly certified stroke centers compared to patients in low-income communities, even with an initially higher baseline availability of stroke-certified hospitals within a 30-minute driving time, as shown in Figure 2. The rest of the panels in Figure 3 show systematic differences in levels of certification across income levels. Panels C and D both show that high-income communities have the highest rate of exposure to newly certified PSCs or TSC/CSCs, followed by medium-income, then low-income communities. Panel D shows that by 2019, there was a 3-fold difference in access to newly certified stroke centers at the highest level: 5.4% of high-income communities had a TSC/CSC that opened nearby, whereas only 1.8% of low-income communities had access to this level of stroke center. Comparing the low-income trends across Figure 3, Panels B through D reveals that low-income communities were more likely to be exposed to entry-level certifications than higher-level ones: by 2019, 12% were exposed to ASRHs, but less than 2% to TSC/CSCs.

Figure 2.

Baseline access to stroke centers within 30-minute drive time in rural communities in 2009, by income levels

Figure 3.

Share of patients whose community is exposed to newly certified stroke centers between 2009 and 2019, by certification and income levels

We used community fixed effects models to investigate the extent to which exposure to newly certified stroke centers changed the actual access, treatment, and health outcomes in the treatment community, relative to a community with no such exposure. Figure 4 shows the coefficients from the multivariate model. Panel A shows that having a newly certified stroke center nearby did increase actual admission to stroke centers, ranging from a 15.83 percentage point increase (95% CI: 9.87, 21.79) if the newly certified stroke center was a TSC or CSC to 32.83 percentage points (CI: 30.44, 35.23) if the newly certified stroke center was a PSC. Panel B shows that having a newly certified stroke center near a community increased the probability of receiving thrombolytic therapy, although by a different magnitude depending on the certification level. In communities where the exposure was to an ASRH, the probability increased by 0.63 percentage points (CI: 0.05, 1.22). Given the mean rate of 9%, this represents a 7% relative increase. In contrast, if the exposure was to a TSC or CSC, the probability increased by 1.39 percentage points (CI: 0.28, 2.49) or a 15% relative increase. Panel C shows the probability of thrombectomy only increased significantly if the exposure was to the highest level of certification: exposure to TSC or CSC increased the probability of thrombectomy by 1.12 percentage points (CI: 0.41, 1.83). Given the mean rate of 2%, this represents a 56% relative increase. Panel D shows that while all point estimates were negative when examining 1-year mortality, we did not observe statistically significant changes in 1-year mortality post-exposure to any type of certification level.

Figure 4.

Changes in the probability of outcomes on and after a rural community is exposed to newly certified stroke centers within 30-minute drive time, by the exposure’s certification levels

In a sensitivity analysis where we examined stroke center opening within 60 minutes, our conclusion remains similar (supplemental Figure e1). Overall, the effects were attenuated when using the 60-minute threshold compared to the 30-minute threshold, which is expected since a facility an hour away is unlikely to meaningfully change patients’ care-seeking behavior. One notable difference is that the probability of being admitted to stroke centers only increased by a small amount when there is a newly certified TSC or CSC within 60 minutes. This likely reflects that many of these communities already had access to lower-level certified centers, so the incremental effect on admissions to any stroke center is smaller. However, exposure to TSC/CSC is still associated with statistically significant improvement in treatment and mortality outcomes.

Discussion

Our study of 590,191 ischemic stroke patients found that there has been a persistent and widening gap over time in treatment and long-term mortality across income levels in rural areas; and that this growing disparity was likely due to a combination of two factors: the disparity in geographic access to stroke centers by income, as well as the increased access and treatment probability for communities that were exposed to newly certified stroke centers. In particular, we found that these differential benefits were likely driven by the type of stroke center openings, where low-income communities were more likely to be exposed to newly certified ASRH with limited stroke capabilities than higher-level certified stroke centers, whereas high-income communities were more likely to have TSC/CSC openings nearby. Our multivariate model further demonstrated that exposure to higher-level stroke centers increased the probability of being admitted to stroke centers substantially, receiving thrombolytic therapy (when exposed to PSC or higher level) and thrombectomy (when exposed to a TSC or higher). Taken together, our results suggest that certification growth has not been equitable, particularly for advanced stroke interventions that require more specialized resources, which further exacerbated the disparity in stroke care between low- and high-income communities in rural areas.

Our findings align with prior research showing that thrombolytic therapy has been under-utilized in rural areas (regardless of income level),²⁸ potentially due to structural barriers that have prevented the equitable distribution of these resources. Our study contributes to current literature by showing that within the already under-served rural areas, the natural expansion of stroke care has not occurred evenly across communities with varying levels of income, and that advanced stroke care infrastructure, such as the availability of neuro-interventionalists, standardized stroke protocols,¹² and rapid patient triage²⁹ are still disproportionately concentrated in wealthier rural regions. Given that treatment for ischemic stroke is a time-sensitive procedure with the potential to drastically improve functional outcomes (as opposed to mortality, for which we did not expect nor did we find significant differences), this income-based disparity may contribute to worse long-term recovery in low-income rural stroke patients.

Prior work has shown that stroke certification disproportionately benefits urban and affluent areas.^30,31 What are the potential reasons for this? Rural hospitals, in general, tend to be less equipped to become stroke-certified because it requires a minimum threshold of capability that some hospitals may not have. For example, one study reported that 25% of responding hospitals, primarily rural, did not have standardized ED stroke protocols.¹² Urban and affluent hospitals likely have greater financial resources to pursue higher-level stroke certification, whereas lower-income areas, regardless of urban or rural status, may face financial constraints that limit their ability to obtain accreditation.³⁰ Several previous studies^9,28,29 have also reported that rural populations face significant delays in receiving stroke treatment due to geographic barriers to specialized care, limited access to specialists, and longer EMS response times. Our study further demonstrated that the urban-rural disparity was further exacerbated by the income gradient in rural areas.

A key strength of our study is its comprehensive, population-level analysis of stroke certification trends over a decade, allowing us to capture long-term patterns in admission, treatment, and outcomes across rural income groups. Additionally, our study focused specifically on rural communities, an often-overlooked population in stroke research. However, this study has several important limitations. Our analysis used administrative data, which did not allow for a granular understanding of the severity of stroke, as can be seen in the National Institutes of Health Stroke Scale. Neither did our administrative data allow us to obtain information on the utilization of telephone consultation between rural hospitals and certified stroke centers. We also did not have functional outcomes, such as patients’ instrumental activities of daily living, which would have been better measures of health outcomes than the 1-year mortality included in our analysis. While our data precede the widespread adoption of AI-assisted imaging platforms that automatically analyze CT or MRI perfusion scans to identify salvageable brain tissue and guide thrombectomy decisions, these technologies have the potential to reduce diagnostic delays and enhance triage in hospitals with limited stroke expertise. Because our study period predates the widespread implementation of such tools, their impact on transfer patterns and treatment access in rural areas could not be evaluated but represents an important area for future research. We could not adjust for rural community population density, which may correlate with both income and hospital certification likelihood. This study includes only Medicare fee-for-service beneficiaries and excludes those in Medicare Advantage, who may differ demographically and clinically. Additionally, because our analysis uses Medicare fee-for-service beneficiaries, it primarily reflects adults above 65 years old and is not designed to represent rural populations overall. As rural Hispanic communities are, on average, younger, with higher uninsured rates³² and substantial participation in agricultural labor, they are likely underrepresented in our data. Race, ethnicity, and sex in the Medicare data are based on Social Security Administration records; reporting errors in this official record may lead to measurement error in demographic stratifications. Furthermore, there are variations in the definitional requirements to be a stroke center across different geographic regions and states, as not all go through the same certification processes,²⁴ and this underlying measurement noise could have affected the results of our analysis. Similarly, we have measurement errors in our driving time estimation because we use the same geographic coordinates provided by the Census, representing the geographical centroid, for patients from the same ZIP code community.

These results have several important policy implications. Low-income rural communities continue to lack equitable access to higher-level stroke centers, which suggests that voluntary certification programs alone may be insufficient to ensure equitable distribution of stroke care. A previous study highlighted the importance of utilizing telestroke care in rural areas to bridge geographical barriers to accessing stroke treatment and reported that telestroke was associated with higher rates of utilization for thrombolytic therapy and thrombectomy in rural and super rural areas in the United States.³³ Improved infrastructure for telestroke and regionalized stroke care networks³⁴ may be viable policy solutions to improving stroke care access, treatment, and outcomes in rural areas.

Our study found that while rural stroke center certification expanded between 2009 and 2019, high-income communities disproportionately benefited and experienced greater access to newly certified PCSs, TSCs, CSCs, and higher rates of thrombolytic therapy and thrombectomy. In contrast, low-income communities saw more limited certification growth, primarily in ASRHs, which has led to widening disparities in advanced stroke care. These further exacerbated the existing gaps in care between low and high-income communities, even in rural areas.

Supplementary Material

Supplemental Material

Table e1. Descriptive Statistics of Patient Characteristics, by the type of stroke center exposure

Table e2. ICD-9 and ICD-10 codes for identifying treatment outcomes

Figure e1. Changes in the probability of outcome on and after a rural community is exposed to newly certified stroke centers within a 60-minute drive time