Introduction

Significant racial and ethnic disparities in stroke risk factors, occurrence, and outcomes persist worldwide [1,2,3,4,5]. In the USA, approximately 795,000 new and recurring strokes occur annually at an estimated cost of over $30 billion [1]. Blacks are two to three times more likely to have a stroke than Whites and have the highest rate of stroke mortality [1, 6,7,8]. Hispanic Americans also are more likely to suffer a stroke than Whites, and recurrent stroke confers on them a two to threefold increased risk for all-cause mortality [9]. Age-standardized stroke mortality rates in the USA declined 38% from 2000 to 2015. However, since 2012–2013, Black and Hispanic mortality rates have stalled or increased [8]. Among Medicare beneficiaries, Black stroke survivors experience significantly more potentially disruptive care transitions than Whites [10]. Further, compared to Whites, Black stroke survivors have greater physical impairments and more activity limitations at hospital discharge, admission to post-acute care, and following home-based rehabilitation [11,12,13,14]. Data on post-stroke disability among Hispanic survivors are limited but suggest greater disability than among White survivors [15].

The causes of stroke disparities among White, Black, and Hispanic Americans are multiple and complex. Variability in the incidence and treatment of potentially remediable risk factors—systolic blood pressure (SBP), in particular—is frequently cited [1, 7, 16, 17]. Barriers to risk remediation exist at the system, physician and patient levels [18, 19]. These include differential access to medical care and medications [20]; “therapeutic inertia” of physicians, leading to suboptimal treatment regimens [21]; disparities in patients’ knowledge, attitudes, beliefs, and lifestyles [17, 22]; and disparities in the socio-economic determinants of health, including education, income, family circumstances, and stress [23].

This article examines socio-economic and health disparities in a diverse, hypertensive sample of 495 post-stroke Black, non-Hispanic, and Hispanic home health patients at risk of a recurrent stroke due to elevated systolic blood pressure (SBP) [24]. Here, we examine cross-sectional data focused on correlates of SBP—the leading modifiable cause of stroke—and physical function—a key stroke sequela and indicator of post-stroke quality of life. Generally, studies of stroke risk factors have focused on predicting first, not recurrent, strokes and have excluded post-stroke patients like those in this study. Across all health care settings, studies of Black stroke patients are more readily available than studies of Hispanics [2, 13, 14, 24]. In the home care setting, information on both groups is sparse. Our data provide a unique comparison of homebound Black and Hispanic stroke survivors at heightened risk of recurrent stroke absent targeted intervention. We found no published data on this vulnerable group of stroke survivors.

Study Aims

The aims of the analyses were to examine the correlates of baseline SBP and physical function in an ethnically diverse, hypertensive sample of post-stroke Black, and Hispanic home health care patients.

Conceptual Model

Building on the Andersen behavioral model [25], we posited that several classes of variables would be important multivariate correlates of SBP and physical function. These included health system factors (e.g., practice design and goal orientation of prior chronic care); individual “predisposing” factors (e.g., age, sex, race, marital status); individual “enabling factors” (e.g., income, education, health literacy, perceived self-efficacy); and individual “need/illness level” (e.g., comorbid conditions, stroke history, stroke sequelae such as frontal lobe function, depression, body mass index [BMI], and health behaviors such as smoking, alcohol use, diet).

Methods

Study Population

Study methods have been published elsewhere [26]. The study was conducted at the Visiting Nurse Service of New York (VNSNY), the largest nonprofit Medicare/Medicaid-certified home health organization in the USA. Study patients were drawn from New York City’s four largest boroughs. The majority were admitted from the hospital; all were receiving usual home health services, including physician-ordered skilled nursing and/or therapy services. Approved by the organization’s institutional review board (reference #I12-004), the study enrolled 495 adult patients with a history of stroke or transient ischemic attack (TIA) and SBP > 140 mmHg at enrollment. Initial review of electronic health record new admission data identified potential patients with ICD codes related to post-stroke care (the majority had a code of 438.xx [ICD-9] or I69.xx [ICD-10]). Patients were eligible if during a telephone screening call, they self-identified as Black and/or Hispanic through two separate questions asking about race and ethnic identity. Patients self-reporting as non-Hispanic and White, Asian or other (except Black) were excluded. Following the convention of the US Current Medicare Beneficiary Survey, we classified all patients who self-identified as Hispanic (or Latino) as Hispanic, regardless of what race, if any, they indicated [27]. We then created two mutually exclusive categories that consisted of Black non-Hispanic and Hispanic (any or no race indicated). Throughout our analyses, the terms “Black” refer to Black, non-Hispanic patients and “Hispanic” to anyone who responded Hispanic to the ethnicity question. Excluded were patients with SBP > 200 mmHg or DBP > 120 mmHg; those on dialysis with end-stage renal disease or a kidney transplant; and those with severe heart failure (score of 3 or greater on the New York Heart Association functional classification scale [28]), significant cognitive impairment (dementia diagnosis or unable to pass a 6-item telephone cognitive screen [29]) or upper arm circumference ≥ 38 cm. Eligibility was confirmed through an in-home patient interview and BP assessment.

Data Collection

Primary baseline data were collected during in-home interviews conducted by trained, supervised research interviewers. Data on type of most recent stroke were obtained from patients’ hospital discharge records.

Measures

SBP measurements were the average of 3 in-home readings taken by research interviewers using a validated, automated oscillometric BP device (Microlife Watch BP, Golden, CO, USA). NINDS Common Data Elements (CDEs) were used for (1) demographics (age, sex, race, education, marital status); (2) health behaviors (tobacco, alcohol, and drug use); (3) symptoms of depression (Patient Reported Outcomes Measurement Information System® (PROMIS)-depression—non-somatic depressive symptoms) [30,31,32,33,34]; and (4) physical function (PROMIS-physical function—patient-reported difficulty in performing activities of daily living (ADLs) and instrumental activities of daily living (IADLs) such as walking, climbing stairs, doing indoor/outdoor chores) [35]; and the standardized Barthel Index (patient-reported need for assistance with ten ADLs) [36, 37]. Additional measures were (1) the Charlson Comorbidity Index (patient illness severity and mortality risk) [38]; (2) the CHEW health literacy scale (patient-reported problems, confidence, and need for assistance with health information and medical records [39]; (3) the Lorig self-efficacy in chronic disease management scale (patient-reported confidence in doing activities/tasks related to their chronic condition) [40]; (4) the beliefs about medications (BMQ) scale (patients’ beliefs about the necessity of medications or their potential harm) [41, 42]; (5) the Frontal Assessment Battery (FAB), (patients’ frontal lobe functional performance of tasks in six domains—conceptualization, mental flexibility, motor programming, sensitivity to interference, inhibitory control, and environmental autonomy) [43]; and (6) two scales of the Patient Assessment of Chronic Illness Care (PACIC) instrument—the practice design/organization subscale (patients’ perceptions about the extent to which their care was “organized” and provided necessary information) and the goal setting/intervention tailoring subscale (patient-reported support for setting goals and managing their health) [44]. All scales are published, validated measures with good reported reliability, which was also observed for this sample.

Analytic Approach

The analysis, using baseline data, consists of bivariate descriptive statistics and multivariate models that describe associations between selected independent and dependent variables. In accord with the Andersen model, independent variables, in addition to race/ethnicity, include other predisposing, enabling, and need/illness variables such as age and sex, health literacy and efficacy, stroke history and sequelae, and other potential predictors of selected outcomes. Dependent variables are SBP and two measures of physical function.

Psychometric analyses were conducted to examine the performance of the measures for the study sample. Internal consistency was estimated using a version of Cronbach’s alpha for estimation of ordinal, polytomous items. Ordinal alpha was calculated based on polychoric correlations. Additionally, we estimated McDonald’s omega total, a reliability estimate derived from factor analyses that reflects the proportion of total common variance explained. The explained common variance is a measure of dimensionality.

Bivariate analyses were conducted to examine baseline characteristics of the Black and Hispanic study populations and differences between them. Phi correlation coefficients were used for binary variables and Pearson correlation coefficients for ordinal and continuous variables. Multivariate analyses were conducted to evaluate the association between the selected dependent variables (SBP and 2 physical function measures) and independent variables. The primary analysis was hierarchical; however, we also conducted stepwise analysis to examine the variables that were the most explanatory. The EM algorithm was used to impute missing data in the independent variables. Collinearity diagnostics were examined and sensitivity analyses were performed removing potentially collinear variables. Tests of significance included t tests within the multivariate analyses to test the significance of the association between the independent and dependent variables. Analyses were performed using IBM SPSS Statistics version 24 (IBM Corp, 2016) and the R psych package [45].

Results

Internal Consistency

Most reliability estimates were in the moderate to high range (0.658–0.976) (Table 1). The Lorig self-efficacy, PROMIS depression, and physical function measures evidenced the highest reliability (ordinal alpha 0.976, 0.953, and 0.941, respectively). The ordinal alpha estimate for the standardized Barthel Index was 0.808. Other ordinal alpha reliability estimates for selected study measures were Chew health literacy (0.772), beliefs about medications (0.797), PACIC practice design scale (0.751), and PACIC goal setting scale (0.875). The FAB evidenced the lowest reliability (0.658), perhaps attributable to its measurement of multiple underlying constructs, which can result in lower estimates.

Table 1 Internal consistency estimates
Full size table

Characteristics of the Black and Hispanic Study Populations: Bivariate Analyses

Table 2 presents the results of bivariate analyses described below.

Table 2 Bivariate results: patient perceptions and characteristics
Full size table

Health System Organization: Chronic Care

Black participants rated practice design (how well their prior HTN care was organized) significantly lower than Hispanics (3.2 vs. 3.5, respectively, p = 0.006) and lower than a diabetic population sampled by the developer [44]. There were no significant racial/ethnic differences in perceived support for goal setting (Mean 2.7, SD 1.2), which was somewhat lower than in the developer’s sample [44] (PACIC scales range from 1 to 5; higher scores indicate better-perceived care).

Individual Predisposing Characteristics

Of the 495 participants, 345 (69.7%) were categorized Black, non-Hispanic, and 150 (30.3%), Hispanic. Three fifths (61.4%) of Black participants were born in the USA, compared to just one fifth (22.0%) of Hispanic participants (p < 0.001). Participants’ mean age was 67, and about one quarter were 75 years or older. Compared to Hispanics, Black participants were significantly more likely to be female (62.3% vs. 44.7%, p = 0.001).

Individual Enabling Characteristics

Compared to Hispanic participants, Black participants attained more education (11.9 years vs. 9.7 years, p < 0.001), higher income (43% of Blacks vs. 29% of Hispanics reported an annual income > $15,000, p = 0.003) and greater health literacy (p < 0.001). Marital status was not significantly different, although Hispanic participants were somewhat more likely to be married or in a domestic partnership. There was no significant difference in self-efficacy (mean 7.0, SD 2.2); scores were similar to or better than in published national samples [40].

Individual Need/Illness Level

There were no significant differences between Black and Hispanic participants in number, recency, or type of stroke. On average, participants had experienced 1.7 strokes prior to study entry, and a median of 47 days (mean 1.5 years) had elapsed since their most recent stroke. Sixty-five percent had experienced their most recent stroke within the past 90 days, and most strokes were ischemic. Charlson comorbidity scores were similar for both populations (in the 3–4 range associated with an estimated 50% risk of 1-year mortality [38]), as were depression scores (mean 52.1, SD 10.5) and BMI (BMI = mean 30.1, low end of obese). Heart disease affected about a fifth of participants. Hispanic participants were significantly more likely than Black participants to report a prior heart attack (25.5% vs. 15.0%, p = 0.006) or heart failure (29.0% vs. 19.6%, p = 0.027) but significantly less likely to report valvular heart disease (18.4% vs. 26.9%, p = 0.040). Sixty-two percent of study participants had diabetes.

Black participants were more likely than Hispanics to report they had an alcoholic drink at least once/month (23.7% of Black vs. 14.7% of Hispanic participants, p = 0.006). Current or prior smoking affected 15.0% of participants, who smoked less than one cigarette a day on average. There were no significant differences between Blacks and Hispanics in beliefs about medications or stroke symptom knowledge. Beliefs about medications scores (mean 20.3, SD 5.0) were on the high end of negativity (score of 35 indicates maximum negativity) [41, 42], and stroke symptom knowledge (mean 5.6, SD 1.3) was no greater than in a recently reported pre-stroke population of New York City Blacks and Hispanics (mean 5.6, SD 1.48) [46].

Dependent Variables: SBP and Physical Function

SBP

In both bivariate (Table 2) and multivariate (Table 3) analyses, Black participants had significantly higher mean SBP than Hispanics. Unadjusted SBP among all participants was 154.6 mmHg; SBP among Black participants was 155.5 mmHg and among Hispanics 152.4 mmHg (p = 0.020). (DBP, not considered an independent stroke risk factor, was also significantly higher among Black participants (86.7) than among Hispanics (83.6) (p = .012)). In multivariate models, being of Black race/ethnicity was the only statistically significant factor associated with participants’ baseline SBP after controlling for other potential confounders and predictors selected for theoretical reasons. Being Black was associated with a 3.55 mmHg higher SBP relative to being Hispanic (hierarchic regression model) (p = 0.020) (Table 3). The results of stepwise regression were consistent.

Table 3 Hierarchical regression analysis for variables predicting systolic blood pressure
Full size table

Physical Function

In both bivariate (Table 2) and multivariate (Table 4) analyses using the PROMIS and standardized Barthel physical function measures, there were no significant differences between Black and Hispanic participants.The mean unadjusted PROMIS physical function score was 34.0 (SD 7.4) (mean standardized T-score is 50 (SD 10)) (higher scores indicate better function). The mean standardized Barthel score was 70.7 (SD 20.4) on a scale of 0 to 100 (higher scores indicate better function). In hierarchical regression models, seven independent variables consistently appeared as significant predictors of physical function measured by both PROMIS and standardized Barthel scores. The common demographic predictor was sex, with male participants more highly functioning than females (p = 0.015 PROMIS; p = 0.026 standardized Barthel). Age was significant only in the standardized Barthel model, where lower age was associated with better function (p = 0.009). The common health-related variables associated with higher baseline physical function were fewer comorbidities (p = 0.048 PROMIS; p = 0.024 standardized Barthel), lower BMI (p = 0.024 PROMIS; p = 0.043 standardized Barthel), and fewer depressive symptoms (p = 0.035 PROMIS; p = .006 standardized Barthel). Higher FAB scores were significantly associated with function only in the standardized Barthel model (p = 0.005). Additionally, higher health literacy (p < 0.001 PROMIS; p < 0.001 standardized Barthel) and self-efficacy (p = 0.002 PROMIS; p = 0.006 standardized Barthel) and current alcohol drinking (vs. abstinence) (p = 0.001 PROMIS; p = 0.009 standardized Barthel) were significantly associated with better function in both models. For both PROMIS and standardized Barthel, stepwise regression results were consistent.

Table 4 Hierarchical regression analysis for variables predicting physical function and physical limitations
Full size table

Discussion

This study is the first to provide detailed data on Black and Hispanic post-stroke patients receiving home health care. We found that Black participants in our study were significantly more likely than Hispanic participants to be female and to have attained higher education, higher income, and greater health literacy [47]. In this sample, where three fifths of Black participants were born in the USA, compared to only one fifth of Hispanics, we believe the differential in education, income, and health literacy is most likely attributable to the predominantly immigrant status of Hispanic study participants. Why Black participants were predominantly female compared to Hispanic participants is not clear to us, especially in light of limited comparative data. The current literature on age-specific sex differences in stroke risk is primarily limited to white populations and is somewhat inconsistent. Moreover, it suggests that women are protected from stroke at younger ages and that this protection is reduced but not reversed with advancing age [48]. Thus, the age composition of our population does not suggest gender differences in study sample. We do know that a larger proportion of the non-Hispanic, Black population in NYC is female (54.7%) than male (45.2%) and that the female to male difference is smaller in the Hispanic population (51.4% female vs. 48.5% male) [49], but this differential is not large enough to explain the 17.6% gender gap in our population. Clearly, this finding requires further investigation.

Whatever the causal factors, after controlling for differences in the possibly protective factors of sex, income, health literacy, and education—and after controlling for a wide range of other potentially predictive or confounding factors—we found that study participants were characterized by a significant racial/ethnic difference in SBP. Relative to being Hispanic, being non-Hispanic Black was associated with a 3.55 mmHg elevation in SBP, the leading modifiable cause of stroke. The finding of elevated SBP in Black participants is striking in that the main criterion for study entry was SBP > 140 mmHg and < 200 mmHg. Thus, even in this narrowly defined home health care population of stroke survivors with uncontrolled HTN—all of whom are at risk for recurrent stroke—a significant racial/ethnic disparity was detected. The finding is new inasmuch as available studies examining uncontrolled SBP as a stroke risk factor have focused on predicting first strokes and thus have excluded post-stroke patients [50,51,52], while some also have excluded patients with diabetes [52], which affected 62% of our study participants. The finding is clinically as well as statistically significant in that widely cited studies have equated a 10-mmHg elevation in SBP to a 24% increase in stroke risk among Black individuals [51], and every 1-mmHg elevation in SBP to a 1% increase in stroke mortality in the general population [50]. These estimates, however, are based on predicting risks of a first stroke. To develop valid estimates that quantify the role of HTN as a risk factor for stroke recurrence will require future research on post-stroke patients such as those in our sample.

Sixty-five percent of our participants had experienced a stroke in the past 90 days and 40% had experienced more than one stroke at the time of our interview. Over 96% were admitted with medication regimens that included one or more HTN drugs. Yet, their mean SBP of 154.6 was 14.5 mmHg above the level defined as “control.” Uncontrolled SBP in the presence of HTN medications suggests that some combination of inadequate/inappropriate HTN medication regimens, poor physician monitoring and follow-up, and/or barriers to patient adherence is contributing to potentially remediable risk of stroke recurrence in the study population. These factors have been shown to be especially pernicious in vulnerable minority groups with disparate outcomes [19, 21].

In contrast to HTN, race/ethnicity was not significantly associated with differences in physical function between our Black and Hispanic post-stroke populations. The factors we found to be associated with patients’ self-reported ability to perform, with or without assistance, a range of ADLs such as bathing, dressing, getting up from a chair, walking inside or outdoors, or doing chores, were generally consistent with the literature on physical function [53]. “Healthier” patients with fewer comorbidities and greater perceived self-efficacy [53], lower BMI [54], less depression [55], and higher FAB scores reported significantly better function. The relationship of FAB scores to function is logical given the well-established involvement of the frontal lobes in motor function [56]. To what degree frontal lobe function was affected by a patient’s prior stroke and a factor leading to or coterminous with better or poorer self-reported function at home care admission cannot be determined from these data. Similarly, the direction of the relationship between depression and function cannot be determined from these cross-sectional data. Possibly, patients in our study who were less depressed in the early period following their stroke were more likely to adhere to early treatment regimens, including hospital-based rehabilitation exercises, and thus experienced better function when interviewed at baseline. Conversely, stroke survivors’ ability to carry out daily activities in the early post-stroke period may have mitigated their depression. Health literacy and perceived self-efficacy also may have operated in more than one way—contributing to greater ability to understand and/or greater inclination to perform recommended rehabilitation exercises and/or resulting from a patient’s residual functional capacity even after experiencing a recent stroke. Our finding on current drinking is consistent with evidence that moderate drinking (1–2 drinks/day) is associated with less risk of ADL decline than either heavy drinking or abstinence [57]; the drinkers in our study generally drank less than 1 drink/day. Lastly, male study participants reported significantly better physical function than females. This finding is consistent with a number of studies that have found females to report more limitations and use of assistance than males, perhaps because disabling conditions such as fracture, osteoporosis, and osteoarthritis are more prevalent among women than men [58].

Our study adds to the emerging literature on racial/ethnic disparities in post-stroke physical function [12, 59]. Most studies have focused on disparities between non-Hispanic Black and non-Hispanic White populations, and few have included data on Hispanics. Our study is unusual in this respect. Multiple factors have been posited to account for observed disparities, including differences in pre-stroke functional capacity, stroke type and severity, stroke-related care, use of rehabilitation services, and other factors in the patient’s environment such as social support networks. However, work on causality is still in an early stage, and causal factors remain uncertain [12, 59].

Our findings should be interpreted in light of several limitations. First, our study targeted two populations with well-documented health disparities relative to non-Hispanic Whites. Thus, it included only Black and Hispanic post-stroke patients with uncontrolled SBP; it excluded racially/ethnically similar patients whose SBP was controlled, and it excluded non-Hispanic White, Asian, and other patients altogether. The advantage of these inclusion/exclusion criteria is that we have collected rich data on at-risk Black/Hispanic patients in an understudied home care population. The disadvantage is that our two racial/ethnic groups can be compared to each other but not to the broader post-stroke population. Second, because all data reported here are cross-sectional, the direction of possible “causal” relationships suggested by the multivariate models cannot be determined. The use of cross-sectional data to explore potential causal relationships is commonly used to generate hypotheses and ideas for future investigation [60]. Third, because the study sample is limited to home health patients served by one organization in a large urban area, it is not nationally representative. Additional research on post-stroke patients in other geographic settings will be necessary to determine to what extent our findings on BP and function are generalizable. Fourth, the functional measures in our study—albeit highly reliable—are all self-reported rather than professionally assessed. Arguably, however, individuals’ perceptions of their capacity to perform or their need for assistance with a range of ADLs/IADLs may be more determinant of the quality of their lives than the clinical assessment of a nurse or rehabilitation therapist.

Conclusion

Each year, approximately 100,000 home care patients report a prior stroke [61]. Home care—with its workforce of skilled nurses, rehabilitation therapists, and home health aides—focuses on providing patient support for more effective chronic care self-management and improved function. Deeper understanding of the vulnerable Black and Hispanic patients served by urban home care organizations should reinforce the role of home health care as a potentially high impact point of intervention for addressing persistent stroke-related disparities in the US population. Given BP disparities, risk of disruptive care transitions and the challenges of BP control in vulnerable populations such as those served by VNSNY, high priority should be given to home-based, culturally tailored interventions designed to link patients to continuous, responsive hypertension care, and help them address barriers to HTN treatment adherence.