Patients Living in Impoverished Areas Have More Severe Ischemic Strokes

Abstract

Background and Purpose

Initial stroke severity is one of the strongest predictors of eventual stroke outcome. However, predictors of initial stroke severity have not been well-described within a population. We hypothesized that poorer patients would have a higher initial stroke severity upon presentation to medical attention.

Methods

We identified all cases of hospital-ascertained ischemic stroke (IS) occurring in 2005 within a biracial population of 1.3 million. “Community” socioecomic status (SES) was determined for each patient based on the % below poverty in the census tract in which the patient resided linear regression was used to model the effect of SES on stroke severity. Models were adjusted for race, gender, age, pre-stroke disability, and history of medical comorbidities

Results

There were 1895 ischemic stroke events detected in 2005 included in this analysis; these cases were 22% were black, 52% were female, and the mean age was 71 years (range 19–104). The median NIHSSS was 3 (range 0–40). The poorest community SES was associated with a significantly increased initial NIHSSS by 1.5 points (95% CI 0.5–2.6 p<0.001) compared with the richest category in the univariate analysis, which increased to 2.2 points after adjustment for demographics and comorbidities.

Conclusions

We found that increasing community poverty was associated with worse stroke severity at presentation, independent of other known factors associated with stroke outcomes. SES may impact stroke severity via medication compliance, access to care, cultural factors, or may be a proxy measure for undiagnosed disease states.

Introduction

Many patient-related factors are known to influence functional outcome in ischemic stroke patients^1–3. Chief among these is the initial stroke severity^4–6. Recent studies of stroke outcome have demonstrated that the initial National Institutes of Health Stroke Scale Score (NIHSSS) upon presentation for medical attention is the most important predictor of outcome, and modeling outcome using the initial NIHSSS alone is more predictive than models that include patient demographics and comorbidities⁷. However, predictors of initial stroke severity itself have not been well-described within a population. Some non-population based studies have suggested that heart failure, dementia, renal insufficiency, and atrial fibrillation (and treatment of atrial fibrillation) may impact stroke severity^{6, 8–10}.

Socioeconomic status (SES) has been shown to affect access to care, medication compliance, disease incidence, and chronic risk factor management in stroke and many other disease processes^11–21. Given these effects of poverty on pre-morbid conditions and that incidence of stroke is higher among lower SES groups, we postulated that poverty would also negatively influence the symptom severity upon presentation to medical attention. “Community SES”, or the socioeconomic status of the neighborhood in which one resides, is a well-validated proxy variable for estimating an individual’s SES^22–24, and useful when individual income and/or educational level data is not available. Therefore, we hypothesized that stroke patients living in impoverished areas would have more severe initial stroke severity upon presentation to emergency medical attention, even after controlling for other described predictors of stroke severity and outcome.

Methods

The Greater Cincinnati/Northern Kentucky (GCNK) region includes two southern Ohio counties and three contiguous Northern Kentucky counties that border the Ohio River. Only residents of the five study counties are considered for case ascertainment. There were 17 hospitals in the GCNK region in 2005. Previous studies have documented that residents of the five counties who have a stroke exclusively seek care at these hospitals rather than at hospitals in the outlying region.²⁵ This study was approved by the Institutional Review Board at all participating hospitals.

The GCNK Stroke Study involved ascertainment of all stroke events that occurred in the population in calendar year 2005. Details of previous study periods’ case ascertainment have been previously published.²⁶ In 2005, screening was identical to the techniques used in previous study periods. All area residents who were either inpatients or discharged from the emergency department with primary or secondary stroke-related International Classification of Disease, 9^th Revision (ICD-9) discharge diagnoses 430–436 at the 17 acute-care hospitals in the study region were screened for inclusion. All events were cross-checked to prevent double counting.

Once potential cases were identified, a study research nurse abstracted information regarding stroke symptoms, physical exam findings, past medical/surgical history, medication use prior to stroke, social history/habits, pre-hospital evaluation, vital signs and emergency room evaluation, neurological evaluation, diagnostic test results (including lab testing, EKG and cardiac testing, and neuroimaging of any type), treatments, and outcome. To clarify, medical history was recorded as noted on admission (i.e. a history of hypertension was only counted if documented in the medical record as being present prior to the stroke event). Stroke severity was estimated using a validated method of retrospective NIH Stroke Scale score (rNIHSS) obtained from review of the physician exam as documented in the emergency department evaluation.^{27, 28} Classification of race/ethnicity was as reported in the medical administrative record. The research nurse made a determination as to whether a stroke or TIA may have occurred and consulted with study physicians for any questionable cases. If the nurse abstractor was unsure whether or not a stroke occurred, the event was abstracted so a study physician could determine whether the event met stroke criteria.

Stroke-trained study physicians reviewed every abstract to verify whether a stroke or TIA had occurred, after taking into account all available information, including imaging reports and, when necessary, review of actual images. Events with transient symptoms with positive DWI imaging are considered ischemic strokes.²⁹ Both study nurses and study physicians undergo extensive training prior to reviewing events, and the study maintains detailed physician and research nurse study manuals that describe screening, abstraction, and reviewing procedures, ensuring a consistency of methodology among study personnel.

Cases of acute ischemic strokes, both first-ever and recurrent, were included in the present analysis. Intracerebral hemorrhage (ICH) and subarachnoid hemorrhage (SAH) events were not included. The onset of stroke symptoms must have occurred within the study time period. A stroke-trained study physician reviewed every abstract to verify whether a stroke or TIA had occurred. The case definition of ischemic requires either a focal neurologic deficit in a defined vascular territory lasting >24 hours (definition adapted from the Classification for Cerebrovascular Diseases III³⁰ and from epidemiological studies of stroke in Rochester, MN³¹) and/or a stroke in the study physicians’ opinion (which took into account all available information, including imaging reports and, when necessary, review of actual images). Events with transient symptoms with positive DWI imaging are considered ischemic strokes. All ischemic stroke cases were further subtyped by the study physician based on all available clinical and radiographic information. Subtype categories were the following: cardioembolic, large-vessel, small-vessel, other, and undetermined cause. The onset of stroke symptoms must have occurred within the study time period. Details of subtype definitions have been previously published³². Charts were screened for an additional 60 days beyond the end of the study period to capture patients who suffered a stroke during the study period but had not yet been discharged.

In addition, for this analysis cases must have resided at home (not including nursing, retirement, or group homes, or jail), as the census tract of residence is less likely to reflect individual SES.

Census tract information was obtained from the U.S. Census Bureau website (www.census.gov accessed in 2010) to allow time for revisions of census figures. For included cases, SES was estimated by aggregate census tract SES measures. Previous studies have verified census tract SES variables as a valid measure of community SES.^{23, 24, 33} There are 346 census tracts in the GCNKSS region, ranging from rural to urban and from deprived to wealthy. Documented home addresses of the patients were geocoded to their census tract through Tele Atlas’s EZ-Locate client (www.geocode.com). Once a patient’s community SES was estimated using the method described above, the SES measure was linked to the remaining clinical information for that stroke patient, so that a patient-level analysis could be performed.

Percentage of the population living below the U.S. definition of poverty within each U.S. census tract was used as the SES variable in this study, as this has been used the most frequently and has been validated.^{23, 24} For this analysis, four categories of “percentage below poverty” were designated a priori: <5% (108 census tracts), 5–10% (104 census tracts), 10–25% (68 census tracts), and >25% (66 census tracts

Data were managed and analyzed using SAS versions 8.02 and 9.2 (SAS Institute, Cary, NC), and SPSS v 18.0 (SPSS Inc., Chicago, IL). Generalized linear models were used to explore the effect of SES on stroke severity quantified using the rNIHSS. Because the NIHSSS is not a truly linear scale, a secondary analysis was conducted in which patients were stratified according to stroke severity and logistic regression was used to model the effect of SES on the odds of severe stroke (defined as NIHSS>10, pre-specified analysis).

Results

There were 2248 ischemic stroke events detected in 2005; 353 cases were excluded: 7 due to age<18, 29 due to missing NIHSSS, 14 cases for incorrect ICD-9 code, and 303 were excluded due to residence within institutions.

Among the 1895 remaining cases, 22.0% were black, 52.2% were female, and the median age was 71 years (range 19–104). The median estimated NIHSSS was 3 (range 0–40). The demographic and vascular risk factor characteristics of the included and excluded ischemic stroke patients, are presented in Table 1, as well as the distribution of community socioeconomic status among these patients.

Table 1.

Demographics, Vascular Risk Factors, and Community Socioeconomic Status for Ischemic Stroke Patients Included and Excluded from this Analysis, GCNK Population

	Excluded Ischemic Stroke Patients (N=353)	Included Ischemic Stroke Patients (n=1895)
Age (standard deviation)	77 (16.5)	69 (14.3)
Black (%)	73 (20.7)	417 (22.0)
Female (%)	249 (70.5)	990 (52.2)
History of HTN (%)	287 (81.3)	1486 (78.4)
History of DM (%)	132 (37.4)	686 (36.2)
History of coronary artery disease (%)	124 (35.1)	659 (34.8)
History of Untreated AFib (%)	58 (16.4)	157 (8.3)
History of Treated Afib (%)	42 (11.9)	125 (6.6)
History of HF (%)	101 (28.6)	321 (16.9)
History of dementia (%)	148 (41.9)	123 (6.5)
History of Prior stroke (%)	143 (40.5)	486 (25.6)
Current smoker (%)	113 (32.0)	1036 (54.7)
Pre-stroke disability mRS <2 (%)	278 (18.9)	358 (78.8)
Estimated NIHSSS Mean (SD)	9.7 (8.5)	5.7 (6.7)
Community SES: percent below poverty
0–(%)		618 (32.6)
6–10% (%)		607 (32.0)
11–25% (%)		453 (23.9)
>25% (impoverished) (%)		217 (11.5)

Figure 1 presents the geographic distribution of census tracts and community poverty status within the 5-county region of the overall GCNK region. While most of the more impoverished areas in the region are near the “downtown” region of urban Cincinnati city proper, there are impoverished rural regions represented within this area as well.

Figure 1.

Geographic Distribution of U.S. Census Tracts and % of Residents Living Below the Poverty Level, Greater Cincinnati/Northern Kentucky Region

Those living in a census tract with >25% poverty had a mean rNIHSS of 6.7 (95CI 5.8–7.6), which was on average 1.5 points higher than those living in census tracts with < 5% poverty (mean NIHSS 5.2, 95CI 4.7–5.7). This difference was highly significant (p=0.006), and was further exaggerated to 2.2 points when adjusting for comorbidities and demographic factors hypothesized to impact stroke severity a priori. (Table 2). Of the pre-specified factors, a higher initial rNIHSS was associated with age, pre-stroke disability, a history of heart failure, untreated atrial fibrillation (defined as no anticoagulation or antiplatelet medication at the time of the event) and, marginally, a history of dementia. A history of hypertension was associated with a slightly lower rNIHSS. All other risk factors and demographics were not statistically significantly associated with severity. An analysis of ischemic stroke subtypes and the impact on stroke severity did not find a significant overall difference in median rNIHSSS between the subtypes of ischemic stroke (small vessel, large vessel, cardioembolic, other known cause and undetermined etiology),(Kruskall-Wallis test, p=0.07. We also did not find a difference in the distribution of ischemic stroke subtypes when stratified by SES (chi-square, p=0.152).

Table 2.

Multivariable Model of Change in Initial NIHSSS Associated with SES, Demographics, and Medical Co-Morbidities

	Change in NIHSSS Points*	95% CI of the Change	P-value
% below poverty within census tract: poorest vs. richest category	2.23	(1.06 to 3.39)	0.000
Age (by decade)	0.30	(0.06 to 0.53)	0.013
Race (Non-black vs. black)	0.64	(−0.19 to 1.48)	0.132
Gender F vs. M	−0.43	(−1.04 to 0.17)	0.162
Pre-stroke disability mRS (per point)	0.44	(0.15 to 0.73)	0.003
History of prior stroke	0.57	(−0.14 to 1.28)	0.113
History of CAD	0.08	(−0.60 to 0.75)	0.823
History of treated Afib	0.59	(−1.82 to 0.24)	0.348
History of untreated Afib	1.88	(0.79 to 2.97)	0.001
History of HTN	−1.36	(−2.12 to −0.61)	0.000
History of CHF	2.07	(1.21 to 2.93)	0.000
History of DM	−0.06	(−0.69 to 0.58)	0.865
History of Dementia	1.11	(−0.16 to 2.38)	0.086
Current smoking (vs. non-smoker)	−0.23	(−0.84 o 0.38)	0.465

^*= a negative value is associated with a lower NIHSS score

We also performed a secondary analysis to explore the association between SES and severe stroke, using a dichotomized NIHSSS. Of the mild-moderate strokes, 22.4% of cases were black, while 20.0% of those with severe stroke were black. There was no statistical difference between these proportions (p=0.403). The odds of rNIHSSS>10 were twice as high for the patients within the poorest SES category compared with the richest (OR 2.0, 95%CI 1.2–3.2, p=0.006) after controlling for demographics and vascular risk factors.

Discussion

We found that ischemic stroke patients who lived in poorer areas presented with significantly higher stroke severity at presentation for medical attention, independent of other reported factors associated with stroke severity and outcomes. In fact, patients living in poorer regions were twice as likely to have a severe stroke (rNIHSSS >10). To our knowledge, predictors of initial stroke severity within a diverse population have not been reported previously.

There are many possible explanations for our findings. SES may impact stroke severity via medication compliance, access to care, cultural factors, or may be a proxy measure for undiagnosed disease states. Most studies of functional outcome after stroke adjust for the initial severity of the stroke event. Yet in doing this, it is possible that some important factors impacting severity, and thereby outcome, are being “adjusted away”. Delays in presentation to medical setting are unlikely to explain our findings, as community poverty has not been previously found to have clinically relevant impacts on time of arrival to medical attention³⁴, and more severe strokes tend to arrive earlier.³⁵ Future studies are warranted to elucidate the mechanism by which poverty appears to be impacting severity.

The magnitude of difference in severity between patients residing in the poorest neighborhoods and those residing in the richest was the largest effect of all of the factors in our model, with a two point increase in the NIHSS. Having a history of heart failure similarly increased severity. All other co-morbidities in the model either had no impact, or increased severity slightly, with the exception of a history of hypertension, which paradoxically was associated with lower severity. One could hypothesize that patients with enough access to medical care to carry a diagnosis of hypertension were more likely to be controlling or treating their risk factors than those who were likely hypertensive but weren’t aware of it. Many other known predictors of functional outcome after stroke did not appear to impact the initial stroke severity. It is likely that co-morbidities such as diabetes or smoking are contributing to complications in an already impaired person after stroke, which may be an entirely different mechanism than what impacts the initial severity of a stroke event.

The amount of change in the rNIHSS associated with poorer SES is relatively modest, only 2.2 points. How meaningful this would be clinically could be of debate. It should be noted that the NIHSSS is a simple, ordinal scale that describes stroke severity, and likely all “points” on the scale are likely not equal in clinical significance. For example, does getting one point for “hemibody numbness” really equal having mild-to-moderate aphasia on neurologic examination? Despite these limitations, the NIHSSS is the best, well-validated measure of severity that exists today, and therefore was our primary measure of severity in this analysis.

A limitation or our analysis is that community SES may not necessarily reflect individual SES, and is a limitation of any analysis that uses aggregate measures of SES. We did not use individual SES because this information is not available in the medical record; even insurance status is not well documented in our community. Community SES is a validated proxy for individual SES^{24, 36, 37}. Additionally, community SES may offer unique contributions, (especially regarding access to medical care) that may be less dependent on individual SES (e.g., traffic patterns, crime, etc.). Prior work has also suggested racial differences in the importance of community effects³⁷. We do note that the excess risks we found are valid only within our population. Another potential bias is that, while patients excluded due to geo-coding difficulties or those already institutionalized were not different from included patients by age, race nor gender, we do not know if they had a different SES than included patients. However, patients with addresses unable to be geo-coded are likely to be of diverse SES, potentially including urban homeless patients and patients using P.O. Box or rural addresses. We did not include pre-morbid medications in our analysis, since we have no way of measuring compliance with the medication regimens. Data on the impact of medications on stroke severity is mixed and inconclusive, even if compliance is assumed.^38–40 An important sub-group of patients that were excluded from this analysis are nursing home residents, for similar reasons as described above. These patients are more likely to be disabled, and the excluded patients do have a higher stroke severity than those included in this analysis (Table 1). In addition, any incidence study that relies on medical contact for counting of events risks missing events that were not recognized by the general public as needing medical attention.

In conclusion, we present factors associated with initial stroke severity, and have found that community socioeconomic status is significantly associated with severity of ischemic stroke. Understanding how socioeconomic factors contribute to initial stroke severity is critical to improving outcomes among stroke patients in the U.S.