"Worth the Walk": Culturally Tailored Stroke Risk Factor Reduction Intervention in Community Senior Centers.

Background Racial/ethnic minority older adults have worse stroke burden than non-Hispanic white and younger counterparts. Our academic-community partner team tested a culturally tailored 1-month (8-session) intervention to increase walking and stroke knowledge among Latino, Korean, Chinese, and black seniors. Methods and Results We conducted a randomized wait-list controlled trial of 233 adults aged 60 years and older, with a history of hypertension, recruited from senior centers. Outcomes were measured at baseline (T0), immediately after the 1-month intervention (T1), and 2 months later (T2). The primary outcome was pedometer-measured change in steps. Secondary outcomes included stroke knowledge (eg, intention to call 911 for stroke symptoms) and other self-reported and clinical measures of health. Mean age of participants was 74 years; 90% completed T2. Intervention participants had better daily walking change scores than control participants at T1 (489 versus -398 steps; mean difference in change=887; 97.5% CI, 137-1636), but not T2 after adjusting for multiple comparisons (233 versus -714; mean difference in change=947; 97.5% CI, -108 to 2002). The intervention increased the percent of stroke symptoms for which participants would call 911 (from 49% to 68%); the control group did not change (mean difference in change T0-T1=22%; 99.9% CI, 9-34%). This effect persisted at T2. The intervention did not affect measures of health (eg, blood pressure). Conclusions This community-partnered intervention did not succeed in increasing and sustaining meaningful improvements in walking levels among minority seniors, but it caused large, sustained improvements in stroke preparedness. Clinical Trial Registration URL : http://www.clinicaltrials.gov . Unique identifier: NCT 02181062.

RCT Entities:   Population Interventions Outcomes

Clinical Perspective

What Is New?

We tested a new low‐cost intervention using in‐house case managers at community senior centers to administer an 8‐session, 1‐month, culturally tailored behavioral intervention to black, Latino, Chinese, and Korean‐American seniors.

The intervention caused short‐term improvements in changes in daily steps, but the improvements were small and unfortunately not sustained 2 months after the intervention; however, this same intervention was successful in causing sustained improved stroke knowledge, specifically increased reported intention to call 911 in response to stroke symptoms.

This is the first study to improve stroke knowledge specifically among older Chinese and Korean Americans.

What Are the Clinical Implications?

Because stroke knowledge is particularly low among racial/ethnic minorities, improved stroke knowledge and preparedness among racial/ethnic minorities might shorten delays between stroke symptoms and receipt of medical care for stroke.

Though there are many factors contributing to disparities in stroke outcomes, whether this intervention's sustained improvement of stroke knowledge and preparedness among ethnic/racial minority seniors can help reduce stroke outcome disparities warrants further investigation.

This low‐cost, easily generalizable community‐partnered intervention successfully increased stroke preparedness, but did not cause sustained improvements in walking behavior; whether other low‐cost interventions can cause sustained improvements in walking behavior is an important area of investigation.

Introduction

The racial/ethnic minority population aged >65 years is expected to more than triple between 2012 and 2050.1 Stroke risk increases with age, and stark racial/ethnic disparities exist in stroke risk burden, incidence, and outcomes. Latino, Asian, and blacks have elevated risk of stroke incidence or stroke mortality compared with non‐Latino whites.2, 3, 4 Behavioral interventions could reduce racial/ethnic disparities in stroke outcomes by decreasing risk factors and increasing knowledge of symptoms that should trigger urgent response.5

One major, modifiable risk factor for stroke is physical inactivity.6, 7, 8 Physical activity independently reduces stroke risk and decreases other cardiovascular risk factors such as hypertension, hyperlipidemia, and obesity.9, 10 Walking is accessible, low cost, and the most popular form of exercise for US adults (including minority older adults).11, 12 Still, less than one third of Americans aged 75 years and older meet federal activity guidelines,4 and racial/ethnic minorities and immigrants tend to be especially sedentary.13, 14 Previous pedometer interventions, including culturally tailored interventions among older racial/ethnic minority adults, have had promising results increasing walking,10, 15 but interventions that require hiring trained staff are challenging to sustain.

Gaps in stroke knowledge may also contribute to stroke disparities; racial/ethnic minority groups tend to know less about stroke than non‐Latino white adults.16, 17 For example, stroke awareness/preparedness (ie, ability to identify and respond to stroke symptoms) is lower across racial/ethnic minorities,18, 19 including Asians,20 and older adults.17 Among patients hospitalized for stroke, Asians, Latinos, and black women have lower rates of 911 utilization than their white counterparts.21 Increasing stroke knowledge in these communities could improve treatment response time and promote self‐efficacy to change personal behavior to decrease stroke risk.

This study tested the effectiveness of a potentially sustainable, culturally tailored, 1‐month intervention to increase walking and stroke knowledge among Latino, Korean, Chinese, and white seniors. The intervention provided stroke education and drew on motivational psychology theories to increase self‐efficacy and the perceived benefits of walking. Strong community partnership enabled cultural tailoring of the intervention curriculum,22 which can increase the impact of health interventions.12, 23 To facilitate long‐term sustainability, the intervention focused on training in‐house case managers at community senior centers to administer the program.

The primary aim was to test whether the intervention increased walking in this high‐risk population. As secondary aims, we examined whether the intervention improved stroke knowledge, self‐efficacy, positive beliefs about exercise, and clinical health indicators such as blood pressure. We also explored effects on other health‐relevant outcomes, including quality of life.

Methods

Study Design and Participants

“Worth the Walk” was a single‐blind, randomized, controlled trial (RCT). The full protocol was published previously,24 and Data S1 summarizes minor protocol changes. This community‐partnered participatory research project aimed to be sustainable, including 4 Los Angeles community‐based, senior service organizations primarily serving Latino, Korean, Chinese, and black older adults. At least 2 case managers at each organization completed full‐day trainings and demonstrated proficiency to facilitate intervention sessions. The intervention became part of regular senior center programming with the intention that it could continue beyond the funded study period.

Researchers and site staff collaborated to recruit 2 sequential cohorts for each racial/ethnic group (Table S1). Site staff made presentations at the senior centers and invited interested seniors to complete screening interviews with trained bilingual research staff. Inclusion criteria included self‐reported history of hypertension, age 60 years or older, and ability to walk (assistive devices allowed) and to sit in a class setting. Subjects had to self‐identify as 1 of the 4 racial/ethnic demographic groups, communicate in that ethnic‐specific language (English, Spanish, Mandarin Chinese, or Korean), and be available to attend all study sessions.

Approximately 1 week after screening, eligible seniors participated in on‐site, 1‐on‐1 data collection interviews. Trained bilingual research staff collected interview data at baseline (T0), after the 1‐month intervention (T1), and 2 months after the intervention concluded (T2; 3 months after baseline). Participants enrolled from October 2014 through May 2016; the last follow‐up data collection sessions ended September 2016. The University of California, Los Angeles institutional review board approved the trial design, and participants provided written informed consent before data collection. Study data are available from the corresponding author upon request.

Randomization and Blinding

After T0 data collection, participants were randomized to 1 of 2 study arms: immediate intervention or wait‐list control. Participants were randomized using the Research Electronic Data Capture (REDCap) web application permuted block randomization, with randomized block sizes stratified by sex and race/ethnicity. Data collection staff were blind to assignment.

Intervention Group

Trained site case managers facilitated 8, 1‐hour intervention sessions, held twice‐weekly over 1 month, promoting walking and stroke knowledge to reduce risk burden. The curriculum content combined aspects of social cognitive theory and attribution theory to motivate change in walking behavior.25, 26, 27 Sessions 6 and 7 were culturally tailored to each racial/ethnic group to enhance relevance and impact, using insight gained from collaboration with racial/ethnic‐specific community action boards and 12 previously conducted focus groups.22 Additional curriculum information is available in Data S1. Participant retention was encouraged through attendance monitoring and telephone reminders.

Wait‐List Control Group

During the data collection period, groups received the same frequency of contact from research staff (eg, both groups received reminder calls to wear pedometers) and the same incentives (pedometer and $75 total honoraria). The wait‐list control groups received the intervention after final (T2) data collection.

Outcome Assessments

Survey instruments were forward‐ and back‐translated into Spanish, Korean, and Chinese. All black participants and 3 Latino participants completed the interview in English; all other Latino participants completed the interview in Spanish. Korean‐ and Chinese‐American participants were interviewed in Korean or Mandarin Chinese, respectively. Data were collected by trained interviewers by REDCap on iPads.

Primary outcome: mean daily steps

After the screening, all participants were instructed to wear a (provided) Fitbit Zip pedometer28 daily until T2 to record “normal everyday walking levels”; both intervention and control group participants continued to use the pedometer until the T2 follow‐up appointment. At each interview, data were downloaded from the previous 7 days. Research staff telephoned participants reminding them to wear their pedometers. Research staff computed mean daily steps when at least 3 days of data were recorded in the week preceding each interview; only days with over 50 steps were included.

Secondary self‐reported health outcomes

The study adapted the Stroke Action Test,29 which measures intended response to descriptions of stroke and other disease symptoms (ie, intent to call 911 immediately versus less urgent responses). Specifically, stroke preparedness was defined as the percent of 17 descriptions of stroke symptoms for which participants reported they would call 911 immediately (eg, sudden facial weakness, sudden trouble seeing in 1 eye, or sudden arm weakness). Participants were also asked to list 3 risk factors associated with stroke. An adapted chronic disease self‐efficacy scale assessed confidence in one's ability to exercise and do different tasks and activities managing stroke risk.30 Outcome expectations for exercise were measured through agreement with statements such as “exercise makes me feel better physically.”31

Secondary clinical health outcomes

Seated systolic and diastolic blood pressure were measured with a standard protocol using automated devices (Omron HEM‐907XL; Omron Healthcare, Inc Hoofddorp, Netherlands). Three measurements were taken with a 5‐minute rest between each; we analyzed the average.32 Researchers measured height twice (cm) at baseline and weight twice (kg) at each time point. We used these averages to create body mass index scores at each time point. At T0 and T2, fingerpricks provided capillary blood samples for CardioChek Lipid Panel test strips to measure nonfasting cholesterol and dried blood spots for glycated hemoglobin and C‐reactive protein assays conducted by the University of Washington Department of Laboratory Medicine (Seattle, WA). The CardioChek provided measures of total cholesterol and high‐density lipoprotein cholesterol, which were used to calculate non‐high‐density lipoprotein cholesterol.33 For each dried blood spots assay, linear regression equations were used to convert the directly measured analyte into a blood‐equivalent (for % glycated hemoglobin) value or a plasma‐equivalent (for C‐reactive protein) concentration.

Exploratory outcomes

At each interview, participants completed the Medical Outcomes Study 12‐item Short Form34 to assess health‐related quality of life, the 9‐item Patient Health Questionnaire to assess depressive symptomology,35 and a survey of current limitations in activities of daily living36 to assess disability. At T0 and T2, participants reported healthcare utilization (number of physician visits and nights in a hospital) in the previous 3 months.

Statistical Analysis

We justified the planned sample size of 240 participants in the published study protocol.24 We conducted intention‐to‐treat analyses evaluating differences in change scores from T0 to T1 and T2.37 To control for multiple comparisons, Bonferroni adjustments were applied to significance thresholds (eg, P<0.025 for the 2 primary steps/day analyses and P<0.0018 for the remaining 27 secondary and exploratory outcome analyses).

When change‐score differences were observed between groups, we also examined the pre/post change score for each group. To preserve intention to treat, we used the multiple imputation by chained equations (MICE) procedure in Stata/IC (version 15.1; StataCorp LP, College Station, TX) to fill in missing values for both continuous and binary outcomes (50 imputation sets). To test robustness of these change‐score analysis results, we also conducted sensitivity analyses (with comparable Bonferroni‐adjusted significance thresholds) using: (1) ANCOVA models predicting the postintervention outcomes adjusting for the baseline level of the outcome and (2) repeated‐measures mixed‐effects modeling for outcomes measured across each of the time points.

Results

Of the 356 people screened, 23% were ineligible, 12% declined or were otherwise unable to participate, and 233 completed T0 and were randomized (120 intervention, 113 control; Figure 1). Participant demographics are described in Table 1; 95% of Latino and all Chinese‐ and Korean‐American participants were immigrants. Ninety percent of randomized participants completed T2. Participants who discontinued the study did not differ from those who completed T2 on sociodemographic or clinical characteristics.

Figure 1

CONSORT flow diagram. AA indicates African American; CA, Chinese American; KA, Korean American; LT, Latino; T0, baseline; T1, immediately postintervention; T2, 2 months postintervention.

Table 1

Demographic and Baseline Health Characteristics

	Total (N=233)	Intervention (n=120)	Control (n=113)
Demographics
Age, y	73.9 (0.4)	74.1 (0.6)	73.6 (0.6)
Female, N (%)	161 (69.1)	82 (68.3)	79 (69.9)
Black, N (%)	55 (23.6)	28 (23.3)	27 (23.9)
Latino, N (%)	63 (27.0)	31 (25.8)	32 (28.3)
Chinese American, N (%)	55 (23.6)	30 (25.0)	25 (22.1)
Korean American, N (%)	60 (25.8)	31 (25.8)	29 (25.7)
Did not complete high school, N (%)	97 (41.6)	43 (35.8)	54 (47.8)
Baseline health status
Mean steps/day	4934 (209)	4548 (292)	5343 (301)
Stroke preparedness	0.51 (0.02)	0.49 (0.03)	0.54 (0.03)
Inactivity as stroke risk factor, N (%)	49 (21.0)	26 (21.7)	23 (20.4)
Disease and exercise self‐efficacy	7.6 (0.1)	7.6 (0.2)	7.6 (0.2)
Outcome expectations for exercise	1.8 (0.3)	1.8 (0.5)	1.8 (0.5)
Systolic BP, mm Hg	124.9 (1.2)	122.9 (1.5)	127.0 (2.0)
Diastolic BP, mm Hg	66.3 (0.7)	65.7 (1.0)	67.0 (1.1)
BMI, kg/m2	28.4 (0.4)	28.2 (0.6)	28.6 (0.6)
Proportion no ADL limitations	0.74 (0.03)	0.67 (0.04)	0.82 (0.04)
Katz comorbidity index score	2.0 (0.1)	2.1 (0.2)	1.8 (0.2)
Non HDL cholesterol, mg/dL	121.1 (3.0)	118.1 (4.4)	124.3 (4.0)
% HbA1c (whole‐blood equivalent)	6.0 (0.1)	6.0 (0.1)	5.9 (0.1)
Log CRP (plasma equivalent)	0.06 (0.04)	0.02 (0.06)	0.11 (0.06)
Physical‐health–related QOL	42.1 (0.7)	41.1 (0.9)	43.3 (0.9)
Mental‐health–related QOL	50.3 (0.7)	49.8 (0.9)	50.8 (0.9)
Depressive symptomology	4.9 (0.4)	5.3 (0.6)	4.4 (0.5)
Visits to physician in past 3 mo	2.3 (0.1)	2.6 (0.2)	2.0 (0.1)
Total nights in hospital in past 3 mo	0.3 (0.1)	0.5 (0.2)	0.2 (0.1)

Mean (SE), unless otherwise specified. ADL indicates activities of daily living; BMI, body mass index; BP, blood pressure; CRP, C‐reactive protein; HbA1c, glycated hemoglobin; HDL, high‐density lipoprotein; QOL, quality of life.

At T0, participants randomized to the intervention tended to have fewer daily steps, more disability, and more physician visits than those in the control arm. Groups were comparable on all remaining outcome measures (Table 1). Pedometer adherence did not differ between the intervention and control groups at any time point: Overall, 62% had valid data for 7 days at baseline, and 10% had less than 3 days. Fifty‐eight percent of participants randomized to the intervention attended 7 or 8 sessions; 77% attended at least 50% of the scheduled classes, and 14% did not attend any sessions.

Intention‐to‐treat analyses evaluating differences in change scores showed that the intervention group had better walking change scores than the control group at T1 (Table 2). At T2, the difference was not statistically significant (Table 3). ANCOVA sensitivity analyses (Table S2) and the repeated‐measures mixed‐effects sensitivity analyses (Table S3) produced identical conclusions. Pre/post mean walking change scores for the intervention group alone were not statistically significant (Tables 2 and 3): mean increase of 489 steps at T1, 95% CI (−13 to 990); 233 steps at T2, 95% CI (−508 to 973; Figure 2).

Table 2

Change in Study Outcomes From Baseline (T0) to Immediately After Intervention Completed (T1)

Outcome	Change From Baseline Unadjusted Mean [95% CI]		Intervention Regression Coefficient Predicting Change
	Intervention	Control	b [95% CI]	P Value
Primary
Steps/day	489 [−13 to 990]	−398 [−834 to 38]	887 [233–1540]	0.008
Secondary, self‐reported
Stroke preparedness	0.19 [0.13–0.25]	−0.02 [−0.07 to 0.02]	0.22 [0.15–0.29]	<0.001
Inactivity as stroke risk factora	0.2 [0.1–0.3]	0.1 [0.03–0.20]	0.4 [−0.2 to 0.9]	0.161
Self‐efficacy	0.30 [0.02–0.58]	−0.1 [−0.4 to 0.2]	0.37 [−0.02 to 0.77]	0.063
Exercise outcome expectations	−0.1 [−0.200 to <0.001]	0.02 [−0.06 to 0.11]	−0.12 [−0.25 to 0.01]	0.072
Secondary, clinical
Systolic BP	−1.2 [−4.5 to 2.0]	−2.8 [−6.2 to 0.6]	1.5 [−3.2 to 6.2]	0.52
Diastolic BP	−0.7 [−2.6 to 1.2]	−2.1 [−3.9 to −0.4]	1.4 [−1.2 to 4.1]	0.27
BMI	−0.02 [−0.16 to 0.13]	0.05 [−0.07 to 0.16]	−0.06 [−0.25 to 0.12]	0.50
Exploratory
Physical‐health–related QOL	−1.0 [−2.7 to 0.6]	−2.1 [−3.7 to −0.5]	1.1 [−1.2 to 3.3]	0.35
Mental‐health–related QOL	−0.1 [−1.9 to 1.7]	0.2 [−1.4 to 1.7]	−0.2 [−2.6 to 2.2]	0.86
Depressive symptomology	0.3 [−0.6 to 1.2]	0.3 [−0.6 to 1.1]	0.02 [−1.2 to 1.2]	0.98
ADL categorya	0.05 [−0.16 to 0.25]	0.12 [−0.05 to 0.29]	−0.01 [−0.58 to 0.57]	0.98

Intervention regression coefficient reflects mean difference between intervention and control conditions in level of change from T0 to T1. Models use regression with imputed values. ADL indicates activities of daily living; BMI, body mass index; BP, blood pressure; QOL, quality of life.

Ordinal logistic regression.

Table 3

Change in Study Outcomes From Baseline (T0) to 2 Months After Intervention Completed (T2)

Outcome	Change From Baseline Unadjusted Mean [95% CI]		Intervention Regression Coefficient Predicting Outcome
	Intervention	Control	b [95% CI]	P Value
Primary
Steps/day	233 [−508 to 973]	−714 [−1264 to −164]	947 [27–1867]	0.044
Secondary, self‐report
Stroke preparedness	0.18 [0.12–0.24]	−0.03 [−0.07 to 0.02]	0.20 [0.13–0.28]	<0.001
Inactivity as stroke risk factora	0.07 [−0.03 to 0.17]	0.05 [−0.04 to 0.15]	0.07 [−0.52 to 0.66]	0.83
Self‐efficacy	0.23 [−0.08 to 0.55]	−0.36 [−0.67 to −0.04]	0.59 [0.15–1.03]	0.009
Exercise outcome expectations	−0.11 [−0.20 to −0.02]	−0.01 [−0.11 to 0.08]	−0.10 [−0.23 to 0.03]	0.137
Secondary, clinical
Systolic BP	−1.7 [−5.1 to 1.8]	−3.8 [−7.0 to −0.6]	2.1 [−2.6 to 6.9]	0.38
Diastolic BP	−1.2 [−3.2 to 0.8]	−2.7 [−4.5 to −0.9]	1.43 [−1.3 to 4.1]	0.29
BMI	−0.14 [−0.33 to 0.05]	−0.01 [−0.20 to 0.18]	−0.13 [−0.40 to 0.14]	0.35
Non‐HDL cholesterol	1.7 [−7.2 to 10.7]	−10.0 [−19.5 to −0.4]	11.7 [−1.6 to 25.0]	0.083
HbA1c	−0.10 [−0.28 to 0.07]	0.02 [−0.11 to 0.15]	−0.12 [−0.34 to 0.09]	0.26
logCRP	−0.03 [−0.12 to 0.07]	−0.01 [−0.13 to 0.11]	−0.02 [−0.17 to 0.13]	0.81
Exploratory
Physical‐health–related QOL	0.6 [−1.0 to 2.1]	0.3 [−1.4 to 2.0]	0.3 [−2.0 to 2.5]	0.82
Mental‐health–related QOL	−0.8 [−2.5 to 1.0]	−0.3 [−2.2 to 1.6]	−0.4 [−3.0 to 2.1]	0.73
Depressive symptomology	−0.4 [−1.2 to 0.4]	0.02 [−0.9 to 0.9]	−0.38 [−1.57 to 0.81]	0.53
ADL categorya	0.1 [−0.1 to 0.4]	0.2 [0.1–0.4]	−0.1 [−0.7 to 0.4]	0.66
Physician visits	−0.3 [−1.0 to 0.3]	−0.1 [−0.4 to 0.2]	−0.2 [−0.9 to 0.5]	0.54
Nights in hospital	0.01 [−0.6 to 0.6]	−0.1 [−0.4 to 0.2]	0.1 [−0.6 to 0.8]	0.84

Intervention regression coefficient reflects mean difference between intervention and control conditions in level of change from T0 to T2. Models use regression with imputed values. Bonferroni adjustments for multiple comparisons mean the primary steps/day significance threshold is P<0.025 and the significance threshold for the remaining secondary and exploratory outcomes is P<0.0018. ADL indicates activities of daily living; BMI, body mass index; BP, blood pressure; CRP, C‐reactive protein; HbA1c, glycated hemoglobin; HDL, high‐density lipoprotein; QOL, quality of life.

Ordinal logistic regression.

Figure 2

Outcomes with significant intervention effects at each time point, separated by intervention and control group. Bars represent 95% CIs. A, Mean steps per day. B, Stroke preparedness.

Change‐score analyses showed that the intervention improved stroke preparedness. The intervention group indicated that they would call 911 for 49% of presented stroke symptoms at T0. The intervention group's stroke preparedness increased to 68% at T1 and was 66% at T2, whereas stroke preparedness did not change in the control group (Tables 2 and 3). Sensitivity analyses produced identical conclusions (Tables S2 and S3).

The intervention did not affect other secondary or exploratory outcomes (likelihood of listing inactivity as a stroke risk factor, self‐efficacy, outcome expectations for exercise, blood pressure, body mass index, health‐related quality of life, depressive symptomology, and disability). The intervention also did not change biomarkers (non‐high‐density lipoprotein cholesterol, glycated hemoglobin, and C‐reactive protein) or healthcare utilization at T2 (Tables 2 and 3).

Discussion

In this older, racial/ethnic minority sample, intervention participants had better short‐term walking change scores than control‐group participants. The effect was small and not sustained. The intervention caused large, sustained improvements in stroke preparedness. It had no effect on the remaining secondary and exploratory outcomes.

Immediately postintervention, the intervention group walked relatively more compared with the control group. Sensitivity analyses provided strong convergent support for this group difference, though the effect size was below the minimal clinical important difference for steps observed in specific disease states (the pulmonary rehabilitation minimal clinical important difference lies between 600 and 1100 steps/day).38 The observed intervention effect was partly attributed to a downward trend in the control group. Given that the control group walked more than the intervention group at baseline, we cannot rule out the possibility that the observed intervention effect on steps may partially reflect regression to the mean. Participants were able to see their pedometer step counts, which may have acted as an intervention on its own for all participants and, based on novelty, may have promoted more walking than usual at baseline for participants in both arms of the study. Participants receiving the intervention may have maintained their initial level of enthusiasm whereas control participants may have decreased their walking in the absence of the intervention.

There are several possible explanations for the modest and unsustained success of this intervention on the primary outcome (steps). Although previous randomized, controlled trials using pedometers have increased daily walking, these randomized, controlled trials were typically longer in duration than this 1‐month active intervention and/or excluded participants who used assistive devices.10, 15, 25 Our short and inclusive study protocol promoted generalizability, but may have dampened effect size; it is possible that the intervention might have succeeded in increasing steps if it had been of longer duration. Though we worked closely with our community action boards to make all program materials understandable to people with low health literacy, it is possible that we failed in this regard and the intervention might have been successful in a population with higher levels of formal education. In addition, pedometer nonadherence was a big problem; pedometers allow assessment of behavior unbiased by self‐report, but because participants did not wear them consistently, this added noise to the data and decreased our ability to detect an intervention effect.

The most important finding in terms of addressing stroke disparities outcomes is that the intervention increased intention to call 911 in response to stroke symptoms immediately after the intervention and 2 months later. Stroke awareness is of very strong medical interest, and it is a strength of this study that the successful educational component of the intervention was low cost and community based. Though there are many factors contributing to stroke disparities, increasing behavioral intent to call 911 has been identified as an important potential means to reduce stroke disparities.39 Although intent to call 911 will not always result in behavior changes, population‐level increases in symptom recognition and intent to call 911 is still a desirable outcome; knowledge is a necessary, but not sufficient, condition for improving outcomes.20 Our finding matches past interventions’ success sustaining increased stroke preparedness in black and Latino general populations40, 41 and extends this success to older adults and Korean and Chinese Americans.

The intervention aimed to improve knowledge of stroke risk factors and emphasized the importance of exercise in preventing stroke, so it was surprising that the intervention did not increase participants’ likelihood of reporting physical inactivity as a stroke risk factor in open‐ended questioning. The open‐ended versus closed choice question structure may explain this.17 Another possibility is that pre‐existing beliefs could inhibit thinking about lack of exercise as a stroke risk factor; some Chinese, Latino, and Korean Americans in focus groups believed that overly strenuous exercise could cause stroke.22 Moreover, participants may not report lack of exercise as a risk factor, but still know that moderate exercise can help prevent stroke.

Cultural tailoring using a community‐partnered approach promoted long‐term sustainability; 2 sites continued the program beyond the end of the study period. Nonetheless, the concept of cultural tailoring is complex, with no single “black culture” or “Korean‐American culture” etc; it would be inappropriately presumptuous to generalize findings from this nonrepresentative sample to all members of each minority group across the nation. Additional research is needed to test whether the culturally tailored interventions generalize to US‐born Latino, Korean, and Chinese Americans or beyond Los Angeles, and whether cultural tailoring specifically improved efficacy. We cannot determine whether observed differences between study‐arm participants were attributed to the actual content of the intervention versus the increased “attention” given to the intervention‐arm participants; future research should examine this by building in a stronger attention‐control condition.

Summary

This low‐cost intervention, well integrated into ongoing community programming, corresponded with small but better walking change scores immediately postintervention and successfully caused large improvements in stroke preparedness in this older racial/ethnic minority sample. Further study should assess whether this approach can successfully decrease racial/ethnic disparities in stroke and associated detrimental outcomes.

Sources of Funding

This work was supported by the National Institute of Neurological Disorders and Stroke of the National Institutes of Health (Los Angeles Stroke Prevention/Intervention Research Program in Health Disparities [SPIRP] 1‐U54NS081764), the National Institute on Aging of the National Institutes of Health (1K24AG047899‐02, P30‐AG021684, and P30AG028748), and the NIH National Center for Advancing Translational Science (NCATS) UCLA Clinical and Translational Science Institute grant number UL1TR001881.

Disclosures

None.

Data S1. Supplemental methods.

Table S1. Dates and Sites Enrolled

Table S2. Sensitivity Analysis 1: ANCOVA Results From Multiple Imputation

Table S3. Sensitivity Analysis 2: Repeated‐Measures Mixed‐Effects Results From Multiple Imputation

Click here for additional data file.