Racial Disparities in Stroke Recurrence

Abstract

Background and Objective

There are significant racial disparities in stroke in the United States, with Black individuals having a higher risk of incident stroke even when adjusted for traditional stroke risk factors. It is unknown whether Black individuals are also at a higher risk of recurrent stroke.

Methods

Over an 18-month period spanning 2014–2015, we ascertained index stroke cases within the Greater Cincinnati/Northern Kentucky population of 1.3 million. We then followed up all patients for 3 years and determined the risk of recurrence. Multivariable survival analysis was performed to determine the effect of Black race on recurrence.

Results

There were 3,816 patients with index stroke/TIA events in our study period, and 476 patients had a recurrent event within 3 years. The Kaplan-Meier estimate of 3-year recurrence rate was 15.4%. Age-adjusted and sex-adjusted stroke recurrence rate was higher in Black individuals (HR 1.34, 95% CI 1.1–1.6; p = 0.003); however, when adjusted for traditional stroke risk factors including hypertension, diabetes, smoking status, age, and left ventricular hypertrophy, the association between Black race and recurrence was significantly attenuated and became nonsignificant (HR 1.1, 95% CI 0.9–1.36, p = 0.32). At younger ages, Black race was more strongly associated with recurrence, and this effect may not be fully attenuated by traditional stroke risk factors.

Discussion

Recurrent stroke was more common among Black individuals, but the magnitude of the racial difference was substantially attenuated and became nonsignificant when adjusted for traditional stroke risk factors. Interventions targeting these risk factors could reduce disparities in stroke recurrence.

Stroke is a leading cause of death and disability in the United States, costing an estimated $45 billion dollars annually.¹ This burden is unevenly distributed across society, with important disparities across racial and ethnic groups. In particular, Black race is associated with a higher risk of incident stroke, with a nearly 3-fold increased risk in some age groups even when adjusted for traditional stroke risk factors.^2-5 Race is a social identity with specific cultural context and associated social privileges or disadvantages that affect health through myriad mechanisms relevant to stroke, including experiences of interpersonal racism within and outside the health care system and interlocking systems of inequality influencing material and social risk factors.^6,7 These mechanisms contribute to the higher prevalence of traditional stroke risk factors in Black individuals and have independent effects on stroke risk itself.^8,9 Racial disparities in stroke have been stubbornly persistent^10,11 despite significant improvements in overall stroke-related mortality.¹

The social and structural factors that increase the risk of incident stroke in Black individuals are likely to persist after their incident event and may even worsen.¹² There is also evidence that racial disparities exist in poststroke care.¹³ Thus, Black individuals may be at a high risk for stroke recurrence, which is associated with a high case fatality rate and more severe disability,^14-17 giving it an outsized effect on overall stroke-related morbidity and mortality. Nevertheless, data on whether stroke recurrence is more common in Black individuals are conflicting. Some studies have found a modest effect of Black race on recurrent stroke risk in elderly patients.^17,18 Other work found no independent effect of race on stroke recurrence but relied on self-reported stroke prevalence that could have occurred any time before enrollment.¹⁹ This is potentially problematic because recurrence risk varies over time, occurring most commonly early after stroke.^20,21 To better understand the effect of Black race on stroke recurrence, we studied the 3-year risk of recurrence in a population-based study in the Greater Cincinnati/Northern Kentucky region.

Methods

Standard Protocol Approvals, Registrations, and Patient Consents

This study was approved by institutional review boards at all participating hospitals with a waiver from informed consent.

Population

The Greater Cincinnati/Northern Kentucky Stroke Study (GCNKSS) is a population-based stroke study occurring in 5 contiguous counties in Southern Ohio and Northern Kentucky that abut the Ohio river. The study population size is approximately 1.3 million people, and its composition is similar to the United States overall for Black race, age, education, and income. For this study, race/ethnicity was obtained from the electronic medical record, and our terminology adheres to the 1997 Office of Management and Budget (OMB) standards. This approach has been previously shown to have good agreement with self-report.^22,23

Case Ascertainment

Methods for case ascertainment in the GCNKSS have been previously published in detail.²⁴ In brief, during the 2015 study year, all potential stroke cases in the study region among patients 20 years and older were identified using ICD codes (ICD9 430–436 and ICD10 160–169/G45-46). For the purposes of this study, we increased statistical power by broadening the inclusion criteria to include index strokes in Black individuals that occurred during the last 6 months of 2014. Cases were obtained from all local hospitals, local public health care clinics, and a random sample of local physicians' offices. For each potential case, a trained research nurse abstracted relevant clinical data with a standardized case report form, and these cases were then adjudicated by a trained study physician. Cases were categorized as ischemic strokes, transient ischemic attack (TIA), intracerebral hemorrhages (ICHs), subarachnoid hemorrhages (SAHs), or unknown (patients with a clinical history consistent with stroke/TIA but with no or inadequate imaging)^10,24,25; ischemic strokes were further subtyped into cardioembolic, small vessel, large vessel, other cause (including dissection, cancer, etc), and undetermined cause.²⁶ All TIAs were included in this study. In general, TIAs were defined clinically as focal symptoms that fully resolved within 24 hours of onset; however, patients with MRI findings consistent with acute infarction that met the clinical definition of TIA were categorized as ischemic stroke.

Determination of Stroke Recurrence

To assess for recurrence within 3 years of the index event, potential stroke admissions to all local hospitals were identified in 2015, 2016, 2017, and 2018 using the same ICD codes. Basic demographic data were obtained for all identified presentations, and these demographic data were cross-referenced with the list of patients with incident stroke in 2015 to identify potential recurrent stroke presentations. Clinical data from these presentations were then abstracted to a standardized case report form. Cases were then adjudicated by a study physician in the same manner as described earlier, except TIAs that were not considered a recurrent stroke. Recurrent strokes could be of any subtype (e.g., an ICH that developed in a patient who originally had an ischemic stroke would be considered a recurrent event). In general, any changes in examination or imaging that occurred within 14 days of an incident ischemic stroke or ICH/SAH were not considered stroke recurrence, unless the new event occurred in a different vascular territory, led to a new deficit clearly different from the initial event, or was because of a different stroke mechanism.²⁷ For patients who experienced TIAs, any recurrent event occurring at any time was included (including ischemic stroke, ICH, and SAH). Patient deaths were also assessed over the same period using the National Death Index.

Statistical Analysis

First, univariate analyses were conducted to compare demographic and clinical characteristics between Black and White individuals who had an index stroke during the study period using the Wilcoxon rank sum test for age and the χ² test for categorical and dichotomous variables. The Kaplan-Meier estimation was used to summarize the cumulative incidence of recurrent stroke at 1, 2, and 3 years postincident stroke by race. Equal precision 95% confidence bands were computed by specifying a log-log transformation. Cox proportional hazards models were then used to estimate the hazard ratio (HR) for the risk of recurrence in 2 models that progressed from race, age, and sex factors alone to race, age, sex, and stroke risk factors of hypertension, diabetes, atrial fibrillation, coronary artery disease (CAD), left ventricular hypertrophy (LVH), a history of previous stroke, current smoker, and incident stroke types (ischemic stroke, ICH, SAH, and TIA). Age was included in the model as a continuous factor, and an interaction term for age by race was evaluated. In addition, interaction terms for each stroke risk factor by race were explored along with models stratified by race. Patients who died before recurrence were censored at the time of death. A proportional subdistribution hazards model was used to evaluate the risk of recurrent stroke with competing risk of death from other causes before subsequent stroke. The proportional hazards assumption was checked by evaluating a time by Black race interaction term in the model and by plotting Schoenfeld residuals by time.

Data availability

Data not provided in the article because of space limitations may be shared (anonymized) at the request of any qualified investigator for purposes of replicating procedures and results.

Results

Characteristics of the Study Population

A flow diagram of study enrollment and recurrence is shown in Figure 1. During the study period, there were a total of 3,816 patients with index stroke/TIA events in the GCNK region and 476 patients with recurrent events. The region is essentially biracial, and <1% of all index stroke cases happened in race ethnic groups other than Black or White non-Hispanic (N = 31). Among patients with index stroke or TIAs, 2,515 (66%) were ischemic stroke (including 163 patients whose symptoms resolved but had imaging evidence of an infarct), 830 (22%) were TIAs, 372 (10%) were ICHs, and 96 (2%) were SAHs. Three patients with index strokes could not be definitively classified because of inadequate data (i.e., the stroke subtype was unknown). Race-stratified characteristics of patients who had index strokes during our study period are summarized in Table 1. In our population, Black individuals who experienced an index stroke were younger, more likely to experience diabetes, hypertension, LVH, previous strokes, and be active smokers; White individuals were more likely to have atrial fibrillation and CAD (p < 0.01). Among subtypes, White individuals were more likely to experience an index TIA and less likely to experience an index ischemic stroke or ICH/SAH (p < 0.01).

Figure 1. Flow Diagram of Study.

Abbreviations: IS = Ischemic Stroke, ICH = intracerebral Hemorrhage, SAH = Subarachnoid Hemorrhage, TIA = Transient ischemic attack.

Table 1.

Demographic and Clinical Characteristics of Study Population Stratified by Race

Risk of Recurrent Stroke

Overall and race-stratified estimates of cumulative recurrent stroke risk are summarized in Table 2, and the Kaplan-Meier curves of recurrence are shown in Figure 2. At 3 years, the cumulative risk of stroke recurrence was 17.4% in Black individuals (95% CI 13.7%–21.9%) compared with 14.7% in non-Black individuals (95% CI 12.5%–17.2%). The overall risk of stroke recurrence at 3 years was 15.4% (95% CI 13.5%–17.6%).

Table 2.

Kaplan-Meier Estimates of Cumulative Risk of Stroke Recurrence, Overall and Stratified by Race

Figure 2. Kaplan-Meier Curve of Stroke Recurrence Over Time, Stratified by Race.

Two multivariable models of recurrent stroke risk were then developed (Table 3). In Model 1, Black race was adjusted only for age and sex; Black race was associated with a higher risk for stroke recurrence in this model (HR 1.34, 95% CI 1.10–1.64). In Model 2, Black race was adjusted for traditional stroke risk factors including hypertension, diabetes, smoking status, age, LVH, and stroke subtype. In this model, Black race was not significantly associated with recurrent stroke (HR 1.11, 95% CI 0.90–1.36). In a sensitivity analysis, we excluded all patients with a history of stroke before the index event in our study and repeated the abovementioned multivariable models; the results were unchanged. A time by Black race interaction term was not statistically significant, and the residual plot showed no evidence of a pattern by time, supporting the assumption of proportional odds.

Table 3.

Multivariable Models of Recurrent Stroke Risk

An interaction term was tested for age and Black race in both models (Table 3). While these terms were not statistically significant, there was a clear trend toward larger effect size estimates associated with Black race at younger ages, particularly in the fully adjusted model. In an analysis stratified by race, age was significantly associated with recurrence only in White individuals (eTable 1, links.lww.com/WNL/C304). We also looked for the evidence of interaction between other traditional stroke risk factors and Black race (eTable 1, links.lww.com/WNL/C304). Two potential interactions were observed. First, CAD was associated with a marginally lower rate of recurrence in Black individuals and a marginally higher rate of recurrence in White individuals, although neither term reached statistical significance. Second, female sex was associated with recurrence in Black but not in White individuals.

To investigate the possibility that different mortality rates could contribute to apparent disparities in recurrence, we performed a separate competing risk analysis between death and recurrence (eTable 2, links.lww.com/WNL/C304). In this analysis, estimates of the effect of Black race on recurrence did not appreciably change, and Black individuals were at a higher risk for both recurrence and death. Finally, because index stroke subtypes had different rates of recurrence and varied in prevalence between White and Black individuals, we repeated a minimally adjusted model in each index stroke subtype (ischemic stroke, TIA, and ICH/SAH). Black race was associated with a higher HR of recurrence in each subtype, but this only reached statistical significance when all subtypes were combined (eTable 3, links.lww.com/WNL/C304). In a separate analysis, we also added in stroke subtype into the minimally adjusted model, which did not meaningfully attenuate the association between Black race and stroke recurrence (eTable 4, links.lww.com/WNL/C304).

Recurrence Risk Among Ischemic Stroke Subtypes

Because ischemic strokes had the highest risk of recurrence, a separate analysis restricted to patients with index ischemic strokes was then performed (Table 4). We specifically looked at whether the risk of recurrent stroke differed by index ischemic stroke subtypes. The most common index ischemic stroke subtype was undetermined etiology (37%); the remaining subtypes were cardioembolic (27%), small vessel (18%), large vessel (12%), and other determined etiology (6%). Patients with other and undetermined etiology subtypes were significantly more likely to experience recurrent stroke (HR 1.35 when compared with small vessel disease, p = 0.045).

Table 4.

Multivariable Model of Recurrent Stroke Risk Restricted to Index Ischemic Stroke Events (n = 2,515)

Discussion

In this large population-based study, we found significant disparities in the rate of stroke recurrence, with a higher recurrence rate in Black individuals that is consistent over 3 years of follow-up. These data highlight the importance of secondary stroke prevention as part of efforts to reduce disparities in stroke-related mortality. Furthermore, we were able to estimate that the overall risk of stroke recurrence was 9.3%, 12.6%, and 15.4% at 1, 2, and 3 years, respectively. These estimates include all index stroke subtypes in a well-defined population and were derived with validated methods of case ascertainment and adjudication,²⁴ arguing for their internal and external validities. Previous estimates of stroke recurrence have varied considerably in the setting of temporal trends in secondary prevention and methodological differences,²⁰ but our estimate is similar to the rate observed in other contemporaneous cohorts.^28,29 This estimate of recurrence should thus provide a reliable baseline to evaluate the population-level effects of emerging treatments for secondary stroke prevention in future studies.^30,31

Apart from CAD and atrial fibrillation, traditional stroke risk factors were overall more prevalent among Black individuals in our study. Traditional stroke risk factors also generally had the same effect on recurrence regardless of race, although we did find that Black race had marginal interactions with CAD and sex, which should be further explored in other population-based studies. When adjusted for these traditional stroke risk factors, the effect of Black race on stroke recurrence was significantly attenuated. Our results would thus suggest that the higher recurrence rate in Black individuals is predominantly driven by a higher burden of traditional stroke risk factors. This is strikingly different from incident stroke³² and supports findings from a secondary analysis of the Prevention Regimen for Effectively Avoiding Secondary Strokes (PRoFESS) trial along with work in the REGARDS cohort.^19,33 Both these studies were large and included patients of all ages, although the PRoFESS trial excluded patients with recurrent cardioembolic stroke. Conflicting with our results are 2 population-level studies based on ICD codes in administrative databases and a subgroup analysis of the Platelet Oriented Inhibition in New TIA and Minor Ischemic Stroke (POINT) trial.^17,18,34 The subgroup analysis of the POINT trial looked only at short-term recurrence rates and excluded a significant number of stroke patients (including those requiring anticoagulation or carotid endarterectomy) and thus cannot be directly compared with our data. The population-level studies that found an independent effect of Black race on stroke recurrence were conducted in elderly patients, which could suggest that Black race is more strongly associated with recurrence in elderly individuals. This is unlikely. Our results suggest that the opposite is true, with Black race having a larger effect size in young patients, similar to the race-age interaction observed in incident stroke.¹⁹ This effect may not be fully attenuated by traditional stroke risk factors at younger ages, which will need to be explored in future studies. While the race-age interaction term did not reach the traditional threshold for statistical significance of p < 0.05, the ideal threshold for evaluating the significance of interaction terms has been debated.³⁵

Methodological differences may account for the conflicting results. Unlike the previous population-level studies, we included all index stroke subtypes including TIAs in our analysis; however, even when we restricted our analysis to ischemic strokes, we still did not find an independent effect of Black race on recurrence. Our approach of using physician adjudication for all index and recurrent strokes could also explain some of the discrepancies because other studies have used ICD codes alone. ICD codes have been validated extensively for incident stroke, but less is known about their reliability in measuring stroke recurrence in the United States. There is evidence that ICD codes modestly overestimate recurrence.³⁶ Finally, our minimally adjusted estimate for the effect of Black race on recurrent stroke was 1.34; this estimate is within the 95% confidence intervals of the estimates given in previous studies and^17,19 within the confidence interval of the effect of Black race in our fully adjusted model (0.90–1.36). It is thus possible that sampling variability could account for the discrepancies across studies, which could be addressed with future pooled analyses. Regardless, this would still imply a relatively small effect size of Black race on recurrent stroke compared with its effect on incident stroke, which requires further study.

The lack of an association between Black race and stroke recurrence beyond traditional risk factors should not be taken as evidence that the social determinants of health and structural racism that underlie disparities in stroke incidence are not relevant in stroke recurrence. Indeed, the higher prevalence of stroke risk factors in Black individuals likely reflects the effect of health inequities. For instance, diet and educational attainment partly mediate the higher prevalence of hypertension in Black individuals,³⁷ and social factors are also tightly associated with the development of diabetes.³⁸ Social determinants and structural inequities could also limit the efficacy of secondary stroke prevention in Black individuals, for instance, regarding blood pressure control.³⁹ Thus, while our results provide hope that targeting modifiable stroke risk factors could reduce disparities in recurrence, this is unlikely to be achieved without understanding and addressing the structural inequities that contribute to these risk factors in the first place. There is some evidence that risk factor modification can be achieved, but success depends on organizational approaches to poststroke care that may address some of these disparities.⁴⁰

Risk factors for recurrent stroke in our study were largely similar to those described for initial stroke.^41,42 Two notable exceptions to this were CAD and atrial fibrillation, which have both been associated with the risk of incident stroke but were not associated with the risk of recurrent stroke in our study. Differential effects of these factors on initial vs recurrent stroke have been previously observed¹⁹ and could reflect effective secondary prophylaxis for associated stroke mechanisms; indeed, contemporary studies have found that atrial fibrillation is actually associated with a lower risk of recurrent stroke in patients who are discharged on anticoagulation.²⁸ Atrial fibrillation and CAD were overall more prevalent among White individuals in our study, and future studies should explore whether these risk factors contribute to disparities in stroke recurrence. Meanwhile, the continued importance of other potentially modifiable risk factors for recurrence (e.g., hypertension, diabetes, and smoking) reinforce the potential effect that aggressive medical management can have on long-term outcomes in stroke survivors.

In our population, recurrence was most common among patients with index ischemic strokes. Using TIAs as the referent group, index ischemic stroke was associated with recurrence even after adjustment for traditional risk factors, which has been previously observed.^43,44 Recurrence was most common among patients with undetermined or other determined ischemic stroke etiologies, albeit with a wide confidence interval. This could reflect more evidence-based secondary prevention treatments available for the better-defined stroke subtypes and reinforces the need for a thorough workup of cryptogenic stroke patients, although our results should be considered preliminary. Only 1 contemporaneous study looked at population-level recurrence with subtyping of ischemic strokes, and there was no evidence of an effect of stroke subtype on recurrence risk.²⁹ Directly comparing the 2 studies is challenging because the previous work included index strokes that occurred as early as 2000. There have been major changes to secondary prevention strategies since that time,⁴⁵ which could affect the recurrence rate of specific subtypes. Future work exploring temporal trends in recurrence by ischemic stroke subtype will be needed to confirm our findings.

A major strength of this work is its use of the GCNKSS, which is one of the largest population-based studies of incident stroke in Black and White individuals in the United States. This large size makes it uniquely capable of addressing the effect of racial disparities on stroke recurrence. There are also important limitations to this work. First, the GCNKSS has a well-validated sampling method for identifying index outpatient strokes in the population, but we were unable to sample for recurrent strokes that were managed exclusively in the outpatient setting. Because almost all patients in our population are hospitalized for stroke workups,⁴⁶ this is unlikely to significantly affect our results. Second, we required recurrent strokes within 14 days of an index ischemic stroke to be in a new location, from a different mechanism, or have clearly new/different symptoms. Because recurrent strokes are most common early after stroke, this could lead to a more conservative estimate of recurrence rate; however, it also reduces the risk of misclassifying examination fluctuations in the acute period as recurrent events. Third, there is a risk of misclassification bias if recurrent events were missed because of patients migrating out of the 5-county study region. Because this migration is not likely to vary by race, it would most likely bias our results toward the null and not affect the central findings. Fourth, index strokes from this study occurred before widespread availability of direct oral anticoagulants and other interventions for secondary stroke prophylaxis such as prolonged cardiac rhythm monitoring to detect occult atrial fibrillation.⁴⁷ Continued surveillance of stroke recurrence at the population level will be critical to understanding how this burden is distributed across society. Another potential limitation arises from our approach of ascertaining strokes in Black individuals for an additional 6 months to increase our sample size; while unlikely over such a short period, this could confound our results if temporal trends in stroke recurrence are different in Black and White individuals. Finally, we cannot definitively determine whether stroke risk factors changed in individual patients between index and recurrent events. This would not affect our primary finding of a higher burden of stroke recurrence in Black individuals but could change the effect sizes of the variables measured in our fully adjusted models.

Glossary

CAD

coronary artery disease

GCNKSS

Greater Cincinnati/Northern Kentucky Stroke Study

ICHs

intracerebral hemorrhages

LVH

left ventricular hypertrophy

OMB

Office of Management and Budget

TIA

transient ischemic attack

Appendix. Authors

Study Funding

National Institute of Neurologic Disorders and Stroke, R01NS030678-26.

Disclosures

D. Robinson: Reports relevant funding from the National Institute of Neurologic Disorders and Stroke (R01NS030678-26). R. Stanton: Reports relevant funding from the National Institute of Neurologic Disorders and Stroke (R01NS030678-26). H. Sucharew: Reports relevant funding from the National Institute of Neurologic Disorders and Stroke (R01NS030678-26). Kathleen Alwell: Reports relevant funding from the National Institute of Neurologic Disorders and Stroke (R01NS030678-26). Mary Haverbusch: Reports relevant funding from the National Institute of Neurologic Disorders and Stroke (R01NS030678-26). F. D. L. Rios La Rosa: Reports no disclosures relevant to the manuscript. S. Ferioli: Reports relevant funding from the National Institute of Neurologic Disorders and Stroke (R01NS030678-26). E. Coleman: Reports no disclosures relevant to the manuscript. A. Jasne: Reports no disclosures relevant to the manuscript. J. Mackey: Reports relevant funding from the National Institute of Neurologic Disorders and Stroke (R01NS030678-26). M. Star: Reports no disclosures relevant to the manuscript. S. Demel: Reports no disclosures relevant to the manuscript. E. Mistry: Reports relevant funding from the National Institute of Neurologic Disorders and Stroke (R01NS030678-26). S. Slavin: Reports relevant funding from the National Institute of Neurologic Disorders and Stroke (R01NS030678-26). K. Walsh: Reports relevant funding from the National Institute of Neurologic Disorders and Stroke (R01NS030678-26). D. Woo: Reports no disclosures relevant to the manuscript. B. Kissela: Reports relevant funding from the National Institute of Neurologic Disorders and Stroke (R01NS030678-26). D.O. Kleindorfer: Reports relevant funding from the National Institute of Neurologic Disorders and Stroke (R01NS030678-26). Go to Neurology.org/N for full disclosures.