Effect of Stroke Education Pamphlets vs a 12-Minute Culturally Tailored Stroke Film on Stroke Preparedness Among Black and Hispanic Churchgoers: A Cluster Randomized Clinical Trial

Olajide Williams; Jeanne Teresi; Joseph P Eimicke; Amparo Abel-Bey; Madeleine Hassankhani; Lenfis Valdez; Luisa Gomez Chan; Jian Kong; Mildred Ramirez; Joseph Ravenell; Gbenga Ogedegbe; James M Noble

doi:10.1001/jamaneurol.2019.1741

. 2019 Jul 1;76(10):1211–1218. doi: 10.1001/jamaneurol.2019.1741

Effect of Stroke Education Pamphlets vs a 12-Minute Culturally Tailored Stroke Film on Stroke Preparedness Among Black and Hispanic Churchgoers

A Cluster Randomized Clinical Trial

Olajide Williams ^1,^✉, Jeanne Teresi ^2,³, Joseph P Eimicke ³, Amparo Abel-Bey ¹, Madeleine Hassankhani ¹, Lenfis Valdez ¹, Luisa Gomez Chan ¹, Jian Kong ³, Mildred Ramirez ³, Joseph Ravenell ⁴, Gbenga Ogedegbe ⁴, James M Noble ¹

¹Department of Neurology, Columbia University Medical Center, New York, New York

²Columbia University Stroud Center, New York State Psychiatric Institute, New York

³Research Division, Hebrew Home at Riverdale, Bronx, New York

⁴Department of Population Health, NYU School of Medicine, New York, New York

Accepted for Publication: April 25, 2019.

^✉

Corresponding Author: Olajide Williams, MD, MS, Department of Neurology, Columbia University Medical Center, 710 W 168th St, Sixth Floor, New York, NY 10032 (ow11@cumc.columbia.edu).

Published Online: July 1, 2019. doi:10.1001/jamaneurol.2019.1741

Author Contributions: Dr Williams had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Williams, Ravenell, Ogedegbe.

Acquisition, analysis, or interpretation of data: Williams, Teresi, Eimicke, Abel-Bey, Hassankhani, Valdez, Gomez Chan, Kong, Ramirez, Noble.

Drafting of the manuscript: Williams, Eimicke, Hassankhani, Valdez, Noble.

Critical revision of the manuscript for important intellectual content: Williams, Teresi, Abel-Bey, Gomez Chan, Kong, Ramirez, Ravenell, Ogedegbe, Noble.

Statistical analysis: Williams, Teresi, Eimicke, Kong, Ramirez.

Obtained funding: Williams, Ogedegbe.

Administrative, technical, or material support: Williams, Teresi, Abel-Bey, Hassankhani, Valdez, Gomez Chan, Ramirez, Ravenell, Noble.

Supervision: Williams, Teresi, Abel-Bey, Gomez Chan, Noble.

Conflict of Interest Disclosures: Dr Williams reported receiving grants from the National Institute of Neurological Disorders and Stroke (NINDS) of the National Institutes of Health (NIH). Dr Teresi reported receiving grants from the NIH. Ms Gomez Chan reported receiving grants from the NINDS. Dr Ravenell reported receiving grants from the NIH. Dr Noble reported receiving grant U54NS081765 from the NINDS. No other disclosures were reported.

Funding/Support: This study was funded by grant U54NS081765-01 from the National Institute of Neurological Disorders and Stroke of the National Institutes of Health (Dr Ogedegbe [principal investigator] and Dr Williams [multiple principal investigator]).

Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Data Sharing Statement: See Supplement 3.

Additional Contributions: We thank Janhavi Mallaiah, MBBS, MPH (Columbia University), for assistance with manuscript preparation and submission and Ian Ellis James (Electric Black Productions) for the production of the stroke films. Mr James received compensation for producing the stroke films.

^✉

Corresponding author.

PMCID: PMC6802046 PMID: 31260028

Key Points

Question

What are effective ways to improve stroke preparedness of black and Hispanic community residents at risk for stroke?

Findings

In this cluster randomized clinical trial involving 312 at-risk black and Hispanic churchgoers, there was no significant improvement in stroke preparedness of participants exposed to the intervention (culturally tailored 12-minute stroke films) or usual care (conventional stroke education pamphlets) at 12 months. However, subgroup analysis revealed sustained significant improvement among low-education subsamples, favoring the intervention.

Meaning

Among black and Hispanic churchgoers, no significant improvement in stroke preparedness was observed in response to culturally tailored stroke films or stroke pamphlets at 1-year follow-up.

This cluster randomized clinical trial of stroke preparedness among black and Hispanic churchgoers evaluates the effect of culturally tailored 12-minute stroke films on stroke preparedness vs the usual care practice of distributing stroke education pamphlets.

Abstract

Importance

Black individuals and Hispanic individuals are less likely to recognize stroke and call 911 (stroke preparedness), contributing to racial/ethnic disparities in intravenous tissue plasminogen activator use.

Objective

To evaluate the effect of culturally tailored 12-minute stroke films on stroke preparedness vs the usual care practice of distributing stroke education pamphlets.

Design, Setting, and Participants

Cluster randomized clinical trial between July 26, 2013, and August 16, 2018, with randomization of 13 black and Hispanic churches located in urban neighborhoods to intervention or usual care. In total, 883 congregants were approached, 503 expressed interest, 375 completed eligibility screening, and 312 were randomized. Sixty-three individuals were ineligible (younger than 34 years and/or did not have at least 1 traditional stroke risk factor).

Interventions

Two 12-minute stroke films on stroke preparedness for black and Hispanic audiences.

Main Outcomes and Measures

The primary outcome was the Stroke Action Test (STAT), assessed at baseline, 6 months, and 12 months.

Results

In total, 261 of 312 individuals completed the study (83.7% retention rate). Most participants were female (79.1%). The mean (SD) age of participants was 58.57 (11.66) years; 51.1% (n = 159) were non-Hispanic black, 48.9% (n = 152) were Hispanic, and 31.7% (n = 99) had low levels of education. There were no significant end-point differences for the STAT at follow-up periods. The mean (SD) baseline STAT scores were 59.05% (29.12%) correct for intervention and 58.35% (28.83%) correct for usual care. At 12 months, the mean (SD) STAT scores were 64.38% (26.39%) correct for intervention and 61.58% (28.01%) correct for usual care. Adjusted by education, a post hoc subgroup analysis revealed a mean (SE) intervention effect of 1.03% (0.44%) (P = .02) increase per month in the low-education subgroup (about a 10% increase in 12 months). In the high-education subgroup, the mean (SE) intervention effect was −0.05% (0.30%) (P = .86). Regarding percentage correct, the low-education intervention subgroup improved from 52.4% (7 of 21) to 66.7% (14 of 21) compared with the other subgroups.

Conclusions and Relevance

No difference was observed in stroke preparedness at 12 months in response to culturally tailored 12-minute stroke films or conventional stroke education pamphlets. Additional studies are required to confirm findings from a post hoc subgroup analysis that suggested a significant education effect.

Trial Registration

ClinicalTrials.gov identifier: NCT01909271.

Introduction

The age-adjusted incidence of a first ischemic stroke is higher among black and Hispanic individuals compared with persons of white race/ethnicity,¹ yet they are less likely to use ambulance services, which is associated with delayed hospital arrival and ineligibility for intravenous tissue plasminogen activator use for acute stroke.² Outcomes are largely dependent on patient recognition of stroke symptoms and immediate presentation in the emergency department.³ Stroke preparedness, or the ability to recognize that a stroke is occurring and appropriately call 911, is poor among black individuals and Hispanic individuals compared with whites and may be partly responsible for some of the racial/ethnic disparities observed in acute stroke treatment.^4,5 Most community stroke preparedness interventions have been evaluated with short-term outcomes of behavioral intent, despite the tendency for these effects to decay over time.

Narrative communication has been shown to lower blood pressure in black individuals with hypertension,⁶ increase HIV awareness and testing among Hispanic individuals,⁷ and improve stroke preparedness among black and Hispanic youth.⁸ Therefore, in an effort to improve and sustain stroke preparedness among at-risk black and Hispanic adults, a culturally tailored narrative communication intervention composed of 2 professionally produced 12-minute stroke films was created and successfully piloted.⁹ A church setting was targeted because churches represent valuable delivery channels for behavior change programs in minority communities and experiential storytelling is a cornerstone of sermons. The effectiveness of these stroke films was then tested against the usual care practice of distributing conventional stroke education pamphlets at church-based health fairs. We hypothesized that intervention participants would demonstrate greater stroke preparedness at 12 months compared with usual care. Details of the trial protocol and methods have been described in a previous report¹⁰ and can be found in Supplement 1.

Methods

Design

We designed a 2-arm cluster randomized clinical trial with randomization of churches to intervention or usual care. A simple randomization method using a software program (SAS, version 9.4; SAS Institute Inc) to generate random numbers was used to balance the number of churches within the project (Supplement 1). Institutional review board approval for this study was obtained from Columbia University, and written informed consent was obtained from each participant. This study followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guidelines.

Intervention Condition

We developed the following two 12-minute stroke films on stroke preparedness for Hispanic and black audiences using professional actors and a professional director under the guidance of our transdisciplinary research team¹⁰: (1) “Derrame Cerebral,” which is a telenovella targeting Hispanic individuals (Video 1), and (2) “Gospel of Stroke,” which is a musical film targeting black individuals (Video 2) (eFigure in Supplement 2). As reported in the trial protocol,¹⁰ qualitative data guiding the development of the films included proportional participation from black individuals and Hispanic individuals. The films were culturally tailored using the ecological validity model¹¹ and a Narrative Performance Scale¹² and were guided by the health belief model.¹³ The final intervention was designed to improve stroke recognition and calling 911 (eg, a slight or even transient slurring of the speech should be enough to precipitate action).

Video 1. “Derrame Cerebral”.

Download video file^{(79.2MB, mp4)}

Open in a new tab

A telenovella targeting Hispanic individuals.

Video 2. “Gospel of Stroke”.

Download video file^{(60.6MB, mp4)}

Open in a new tab

A musical film targeting black individuals.

Usual Care Condition

Distributing conventional stroke education pamphlets in English and Spanish is a common approach to public stroke education. However, the long-term effects of these pamphlets on stroke preparedness are unknown, and our study design may help close this knowledge gap. We used 3 sets of stroke education pamphlets in English and Spanish describing the 5 cardinal stroke symptoms¹⁴ and the importance of calling 911. One set is from the National Institute of Neurological Disorders and Stroke, one is from the American Stroke Association, and one is from the National Stroke Association.

Eligibility Criteria

Black and Hispanic churchgoers were eligible for the study. Participants had to be at least 34 years old with no history of stroke and at high risk for stroke (defined as a history of ≥1 stroke risk factor, such as hypertension, type 2 diabetes, tobacco use, abdominal obesity, heart disease, and high cholesterol).

Implementation Procedures

Recruitment and follow-up occurred between July 26, 2013, and August 16, 2018. Intervention individuals were gathered in the church sanctuary for a single screening of the stroke film. After the formal screening, a DVD of the film was given to participants, who were then asked to watch the film again and share it with family and friends. Usual care individuals were given the 3 stroke pamphlets in their preferred language at a specially organized church health fair (to mirror community practice), and they were asked to review these “potentially lifesaving” pamphlets carefully and then share them with family and friends.

Sample, Recruitment, and Retention

Figure 1 shows the CONSORT flow diagram for study recruitment and retention. Twenty-five churches were approached, 19 expressed interest, and 14 were recruited and randomized. Although the trial protocol specified that 7 churches per arm were to be recruited, 1 usual care church dropped out after randomization due to a protocol violation (the church provided an additional stroke lecture to congregants). Given that there was overrecruitment of individuals at the remaining sites, no replacement church was selected. The final analyzed sample consisted of 13 black and Hispanic churches located in urban neighborhoods (7 randomized as intervention and 6 as usual care); there were 161 individuals enrolled in the intervention arm and 151 individuals enrolled in the usual care arm. In total, 883 congregants were approached, 503 expressed interest, 375 completed eligibility screening, and 312 were randomized. Sixty-three individuals were ineligible (younger than 34 years and/or did not have at least 1 traditional stroke risk factor). Churches were given $1000 for agreeing to participate and to provide on-site space for the study, and each participant was given $120 (50% at the beginning, 25% at 6 months, and 25% at the end of the primary outcome data collection period). Details of power and sample size calculations can be found in the earlier report¹⁰ and in the trial protocol (Supplement 1).

Masking

Investigators (O.W., J.T., J.P.E., J.K., and J.M.N.) involved in outcome analysis were masked to group assignments. However, allocation concealment was not practical among research coordinators (A.A.B., M.H., L.V., and L.G.C.) who were involved in church recruitment, survey administration, pamphlet distribution, and coordination of the film screenings.

Measures

Stroke Preparedness

The primary outcome was self-reported behavioral intent to call 911 based on the validated Stroke Action Test (STAT), assessed at baseline, 6 months, and 12 months.¹⁵ The STAT score reliably predicts the actions that lay people would take if a stroke occurs. The STAT measures behavioral intent to call 911 for suspected stroke symptoms using hypothetical stroke scenarios, in which the stroke diagnosis is not provided. Each scenario features 1 of 21 stroke symptoms and 7 distracter symptoms and the appropriate response to them. The 21 stroke symptoms are derived from the 5 cardinal “suddens” stroke symptoms.¹⁴ There are 21 because the 5 cardinal symptoms are presented in a variety of scenarios. The STAT was administered to participants in person by research coordinators at church sites at baseline, who read each question aloud and recorded the response of participants. This initial face-to-face approach facilitated the development of rapport with participants to help mitigate attrition. The same approach was used over the telephone at 6 months and 12 months. Comparisons of in-person and telephone interviews have been shown to be accurate within acceptable margins of error for gathering quantitative research data, including differences in mode of response.^16,17 The ordinal α reliability estimate for this sample was .956; McDonald ω total was 0.959, and the explained common variance was 53.70%.

Stroke Experience

Stroke experience was evaluated using a composite of 4 items. These included measuring whether or not a participant has ever had or been told by a physician that he or she had a stroke, ministroke, or transient ischemic attack).

Education

Low education was defined as 0 to 11 years of education. High education was defined as 12 or more years of education. We chose to cut the education distribution at the median, recognizing that education is poorly measured in black individuals and Hispanic individuals.

Time

The length of time from baseline was computed in months. The mean (SD) time from baseline to the 12-month follow-up was 11.93 (1.33) months, with a median of 12 months, a minimum of 10 months, and a maximum of 17 months.

Analytic Aim

The primary aim was to evaluate the effects of the short stroke film compared with the usual care approach of distributing stroke education pamphlets. The primary outcome was stroke preparedness using the STAT over 12 months, assessed at baseline, 6 months, and 12 months. A secondary aim was to examine the influence of sociodemographic characteristics, including race/ethnicity and education and their interactions with intervention status, on STAT performance.

Statistical Analysis

Analysis of longitudinal data was performed on an intent-to-treat basis and adjusted for clustering of individuals within churches using SAS PROC MIXED.¹⁸ The statistical tests were based on trajectories (slopes) estimated from observed data points over 3 waves of data. Significance tests were 2 tailed, and the α level was set at .05 for each primary outcome. The end points entered into the models were treated as continuous variables and did not require prior transformation based on graphical inspection of the distribution of the outcomes and of the residuals from the models. Adjusted means (SEs) of the baseline and follow-up end points were estimated. Modeling of group heterogeneity in cluster and residual variances was not necessary. The number of clusters was 13 churches. The mean cluster size per church was 24, with a minimum of 18, a maximum of 31, and a median of 25.

Sensitivity analyses were conducted using both examination of the mean change since baseline¹⁹ and an analysis of covariance (ANCOVA) mixed-model approach, adjusting for clustering within churches. In addition, because the sample had a higher mean education than was observed in previous studies of similar interventions in our target populations, education was treated as a moderator. While few non-Hispanic black individuals (25 [15.7%]) had less than a high school education, about half of the Hispanic individuals (74 [48.7%]) had less than a high school education. Because race/ethnicity and education were confounded, the focus of the analyses was education. However, the role of race/ethnicity was also evaluated, despite the modest sample size for black individuals with low education. To examine the moderating role of education, interaction terms for intervention status by education subgroup were constructed. The difference in slopes between the randomization groups within the low-education and high-education subgroups was examined using contrast estimates. Similar estimates were also computed for the other subgroup analyses. Additional analyses were conducted examining the interaction of race/ethnicity, education, and intervention status. The best covariance structure was selected based on examination of the Akaike information criterion²⁰ and the Schwarz-Bayesian information criterion.²¹ An unstructured covariance structure was observed to be superior and was thus used in all analyses.

Results

Overall Intervention Effects on Stroke Preparedness

Of the 312 individuals randomized, 261 completed the study (83.7% retention rate). The treatment groups were equivalent on all demographic and key covariates at baseline (Table 1). The treatment groups were also equivalent on measures of physical function and activity and quality of life (eTable 1 in Supplement 2). Some small, significant differences were observed on measures of self-reported food intake and health literacy in the direction of somewhat greater literacy in the intervention group than in the usual care group (eTable 2 in Supplement 2). The mean education was about 12 years in both treatment groups; 32.0% (n = 99) had less than a high school education. Most participants were female (79.1% [n = 246]). The mean (SD) age of participants was 58.57 (11.66) years. About half of the participants were non-Hispanic black (51.1% [n = 159]), and the remainder were Hispanic (48.9% [n = 152]) (eTables 3, 4, and 5 in Supplement 2).

Table 1. Baseline Demographics by Treatment Group^a.

Variable	Total (N = 311)	Usual Care (n = 151)	Intervention (n = 160)	P Value
Sex, No. (%)
Male	65 (20.9)	38 (25.2)	27 (16.9)	.07
Female	246 (79.1)	113 (74.8)	133 (83.1)	.07
Age, mean (SD), y	58.57 (11.66)	58.94 (11.27)	58.22 (12.05)	.59
Age, y, No. (%)
<40	18 (5.8)	7 (4.6)	11 (6.9)	.64
40-49	57 (18.3)	27 (17.9)	30 (18.8)
50-59	94 (30.2)	49 (32.5)	45 (28.1)
60-69	80 (25.7)	37 (24.5)	43 (26.9)
70-79	52 (16.7)	28 (18.5)	24 (15.0)
80-89	10 (3.2)	3 (2.0)	7 (4.4)
Race/ethnicity, No. (%)
Non-Hispanic black	159 (51.1)	73 (48.3)	86 (53.8)	.34
Hispanic	152 (48.9)	78 (51.7)	74 (46.3)	.34
Marital status, No. (%)
Single/never married	79 (25.4)	34 (22.5)	45 (28.1)	.26
Married	113 (36.3)	60 (39.7)	53 (33.1)	.23
Domestic partnership	15 (4.8)	8 (5.3)	7 (4.4)	.70
Divorced	39 (12.5)	20 (13.2)	19 (11.9)	.72
Separated	23 (7.4)	11 (7.3)	12 (7.5)	.94
Widowed	42 (13.5)	18 (11.9)	24 (15.0)	.43
Education, y, No. (%)				.64
Low, 0-11	99 (32.0)	50 (33.3)	49 (30.8)
High, ≥12	210 (68.0)	100 (66.7)	110 (69.2)
Education, mean (SD), y	11.99 (4.06)	11.84 (4.06)	12.13 (4.07)	.54
Employment status, No. (%)
Working now	164 (52.7)	80 (53.0)	84 (52.5)	.84
Looking for work, unemployed	32 (10.3)	17 (11.3)	15 (9.4)
Disabled, permanently or temporarily	29 (9.3)	15 (9.9)	14 (8.8)
Retired	76 (24.4)	36 (23.8)	40 (25.0)
Keeping house	8 (2.6)	2 (1.3)	6 (3.8)
Other	2 (0.6)	1 (0.7)	1 (0.6)
Annual family income for the past 12 mo, $, No. (%)^b
<15 000	90 (30.9)	49 (35.0)	41 (27.2)	.31
15 000-34 999	87 (29.9)	39 (27.9)	48 (31.8)
35 000-49 999	79 (27.1)	39 (27.9)	40 (26.5)
≥50 000	35 (12.0)	13 (9.3)	22 (14.6)
Unknown	11 (0.0)	8 (0.0)	3 (0.0)
Refusal	9 (0.0)	3 (0.0)	6 (0.0)

Open in a new tab

^{^a}

t Tests were used to compare continuous variables, and likelihood χ² tests were used to compare categorical variables.

^{^b}

“Unknown” and “Refusal” categories were excluded from significance tests.

At baseline, the mean STAT scores of the treatment groups were equivalent. Moreover, using unadjusted t tests, no statistically significant end-point differences for the STAT were seen at follow-up periods. Significance tests were reported because randomization was at the church level rather than at the individual level. The intervention and usual care groups started out equivalently: the mean (SD) baseline STAT scores were 12.40 (6.11) for the number of correct responses and 59.05% (29.12%) correct in the intervention group and 12.25 (6.05) and 58.35% (28.83%) in the usual care group. These increased to a mean (SD) of 13.52 (5.54) and 64.38% (26.39%) in the intervention group and 12.93 (5.88) and 61.58% (28.01%) in the usual care group at 12 months. The differences were nonsignificant and are summarized in Table 2. More participants who completed the 12-month assessment were female, were non-Hispanic black, and had a higher mean education compared with those who did not complete the 12-month assessment (eTable 6 in Supplement 2). Results from the primary and sensitivity analyses for the primary aim related to the intervention effects are summarized in eTables 7, 8, 9, 10, and 11 in Supplement 2.

Table 2. Observed STAT Number of Questions Answered Correctly and Percentage of Questions Answered Correctly by Treatment Group Over Time^a.

Variable	Mean (SD)			P Value
Variable	Total (N = 311)	Usual Care (n = 151)	Intervention (n = 160)	P Value
No. of questions answered correctly
Baseline	12.33 (6.08)	12.25 (6.05)	12.40 (6.11)	.83
6 mo	13.13 (5.78)	12.80 (5.74)	13.43 (5.82)	.37
12 mo	13.22 (5.71)	12.93 (5.88)	13.52 (5.54)	.41
Percentage of questions answered correctly
Baseline	58.71 (28.93)	58.35 (28.83)	59.05 (29.12)	.83
6 mo	62.59 (27.57)	61.12 (27.42)	63.94 (27.73)	.40
12 mo	63.97 (27.20)	61.58 (28.01)	64.38 (26.39)	.41

Open in a new tab

Abbreviation: STAT, Stroke Action Test.

^{^a}

t Tests were used to compare the STAT between treatment groups at each time point.

Evaluation of Education as a Moderator

Realizing the differing education distributions between the previous study samples and this one, we decided to perform a post hoc subgroup analysis (by education) and hypothesized that this variable might be a key driver of the intervention effects. The low-education subgroup (0-11 years) started out with fewer correct responses and a lower percentage correct (mean [SD], 11.31 [7.24] and 53.84% [34.49%] for the intervention group and 10.28 [6.09] and 48.95% [29.01%] for the usual care group) compared with the high-education subgroup (≥12 years) (mean [SD], 12.90 [5.53] and 61.43% [26.35%] for the intervention group and 13.24 [5.82] and 63.05% [27.70%] for the usual care group) (Table 3).

Table 3. Observed Outcomes by Treatment Group and Education Subgroup for the STAT Number of Questions Answered Correctly and Percentage of Questions Answered Correctly^a.

Variable	Education	Time	Usual Care		Intervention		P Value
Variable	Education	Time	No.	Mean (SD)	No.	Mean (SD)	P Value
No. of questions answered correctly	Low	Baseline	50	10.28 (6.09)	49	11.31 (7.24)	.45
		6 mo	43	10.58 (5.88)	41	14.27 (6.32)	.007
		12 mo	42	10.93 (6.22)	36	14.00 (5.82)	.03
	High	Baseline	100	13.24 (5.82)	110	12.90 (5.53)	.67
		6 mo	89	13.81 (5.39)	104	13.10 (5.61)	.37
		12 mo	87	13.93 (5.51)	93	13.33 (5.45)	.47
Percentage of questions answered correctly	Low	Baseline	50	48.95 (29.01)	49	53.84 (34.49)	.45
		6 mo	43	50.45 (28.04)	41	67.94 (30.08)	.007
		12 mo	42	52.04 (29.62)	36	66.67 (27.72)	.03
	High	Baseline	100	63.05 (27.70)	110	61.43 (26.35)	.67
		6 mo	89	65.94 (25.76)	104	62.36 (26.73)	.35
		12 mo	87	66.34 (26.24)	93	63.49 (25.96)	.47

Open in a new tab

Abbreviation: STAT, Stroke Action Test.

^{^a}

t Tests were used to compare the STAT between treatment groups at each time point within education distributions. Low education was defined as 0 to 11 years of education. High education was defined as 12 or more years of education.

Adjusted by education (eTables 12, 13, 14, 15, 16, and 17 in Supplement 2), the primary analysis showed that there was a mean (SE) intervention effect of 0.22 (0.09) (P < .02) increase per month on the STAT in the low-education subgroup (eTable 14 in Supplement 2). In the high-education subgroup, the mean (SE) intervention effect was −0.01 (0.06) (P = .84). These findings can be contrasted with the high-education subgroup, in which minimal change was noted across the intervention and usual care groups. Examining the mean change, Figure 2 shows the model-based means for the low-education and high-education subgroups. These results were confirmed in sensitivity analyses treating education as a continuous variable (eTable 18 in Supplement 2).

Figure 2. — Model-based means from the mean change model analyses are based on repeated-measures mixed-model analyses, adjusting for clustering within churches. An unstructured covariance structure was assumed. Comparison of the usual care and intervention slopes over time in the low-education subgroup (0-11 years) yielded a significant increase of 2.28 (P = .03). There was no significant change between slopes in the high-education subgroup (−0.02; P = .98). STAT indicates Stroke Action Test.

The primary outcome measure (STAT) instrument is composed of 21 stroke hypothetical scenarios. An improvement in the mean scores from 11 to 14 as observed among the low education group represents a 3-point increase (or 13.13% increase), which represents an additional 3 hypothetical scenarios.

Additional ANCOVA analyses showed that there was a mean (SE) 13.13% (4.31%) (P = .003) increase in STAT scores in the low-education intervention subgroup compared with usual care and a negligible effect in the high-education intervention subgroup (eTable 19 in Supplement 2). Examination of collinearity diagnostics revealed the expected collinearity between the education variables. Use of a single education variable produced acceptable collinearity diagnostics.

Although no significant differences in baseline demographics between treatment groups within the low-education subgroup were observed, additional analyses exploring the confounding effects of race/ethnicity on education (eTables 20, 21, 22, and 23 in Supplement 2) revealed that more Hispanic participants than black participants (74 vs 25, respectively) were in the low-education subgroup, and the reverse was observed in the high-education subgroup (132 vs 78, respectively). While the few black individuals with low education precluded definitive analyses by subgroup, we found that Hispanic individuals with low education were significantly more likely to benefit from the intervention than the comparison subgroups, further supporting the primacy of education as the explanatory variable in our post hoc analysis.

Race/Ethnicity, Socioeconomic Status, and Stroke Experience

We examined baseline zero-order correlations between the STAT and measures of socioeconomic status and found that being non-Hispanic black (correlation coefficient, 0.266; P < .001), having more education (correlation coefficient, 0.149; P = .009), and having a higher family income (correlation coefficient, 0.150; P < .01) were significantly associated with answering a greater number (or percentage) of questions correctly. Stroke experience (having a higher score indicated greater stroke experience) was not significantly associated with the number (or percentage) of questions answered correctly (correlation coefficient, −0.027; P = .64).

Use and Sharing of Materials

At the 12-month follow-up, there were no significant differences between treatment groups for the number of times they reviewed or reread the materials, which on average was 2 times (mean [SD], 2.47 [2.34] for usual care and 2.34 [2.04] for intervention; P = .62). However, compared with usual care participants, more intervention participants shared stroke information inside the household (54.0% [61 of 113] vs 31.0% [40 of 129], P < .001) and outside the household (58.9% [76 of 129] vs 33.3% [43 of 129], P < .001) (eTable 24 in Supplement 2).

Discussion

When comparing the effect of a culturally tailored 12-minute stroke film with the usual care practice of distributing stroke education pamphlets, we did not find significant improvements in stroke preparedness or appreciable differences by treatment group at 12 months, even though the stroke films resulted in more shared information. In a post hoc subgroup analysis, the stroke preparedness of individuals with low education improved significantly in response to the stroke film but not in response to the pamphlets, an effect that was sustained at 12 months, although it is important to note that these moderating effects observed are hypothesis generating only.

Limitations and Strengths

Our study has several limitations. First, our subgroup of black participants with low education was small, and our subgroup analyses examining interactions of race/ethnicity, education, and intervention status require confirmation with future research. Regarding the clinical significance of our finding among those with low education, we acknowledge that the STAT measures intent to act and not the action itself (calling 911). Therefore, it is difficult to extrapolate these intentions to clinically meaningful outcomes. However, because ambulance use is the major determinant of early hospital arrival for stroke (a requirement for successful acute stroke treatment), we argue that the improvement among low-education participants (recognizing an additional 3 hypothetical stroke symptom scenarios and calling 911 immediately, representing a 13.13% increase in STAT scores) may have clinically meaningful implications when applied to a large population of high-risk black and Hispanic adults. Second, education was self-reported and may not have been accurately measured due to the variability in education standards across countries of origin, particularly for Hispanic residents born outside the United States. Third, our modest overall study sample size (n = 312) is a limitation, although the high retention rate (83.7% [n = 261]) among individuals from these minority groups is a strength. Fourth, we did not include a cost analysis or cost-effectiveness component, which may limit the generalizability of this approach. Fifth, although allocation concealment was impractical for research coordinators involved in administering the STAT, we do not expect biased assessments of this outcome due to the standardization of the interview process and the rigid nature of the multiple-choice survey responses. Fifth, bias in measurement is always a concern; however, the Hispanic version of the STAT used in this study was developed using qualitative methods and criterion standard translation processes²² to increase the likelihood of measurement equivalence. Nonetheless, given the sample size requirements it was not possible to perform formal tests of measurement invariance.

Our post hoc subgroup analysis is consistent with prior studies^23,24,25 suggesting that use of video is more successful than written materials alone in improving patient recall and comprehension, particularly among low-literacy populations. While these findings may have potential implications regarding the design of stroke preparedness interventions, confirmatory testing in future studies is required.

Conclusions

Our study reveals deficiencies in conventional stroke education pamphlets and culturally tailored short-length stroke films for improving stroke preparedness of racial/ethnic minority groups. More work is needed to address this important public health problem.

Supplement 1.

Trial Protocol

Click here for additional data file.^{(559.4KB, pdf)}

Supplement 2.

eFigure. Two 12-Minute Short Stroke Films Targeting Black and Hispanic Churchgoers

eTable 1. Baseline Characteristics by Treatment Group

eTable 2. Baseline Education by Treatment Group

eTable 3. Distribution of Education at Baseline

eTable 4. Baseline Comparison of Education Level Definitions Between Treatment Groups

eTable 5. Baseline Demographics for Low Education (0-11 Years) Group by Treatment Groups

eTable 6. Baseline Demographics by Completer Status (Completer Defined as Having at Least Baseline and 12-Month Follow-up)

eTable 7. Primary Analyses of the Effects of the Intervention on the Stroke Action Test (n = 309)

eTable 8. Comparisons of Slopes for Intervention and Usual Care Groups

eTable 9. Model Based Means and Standard Errors for the Stroke Action Test by Randomization Group (n = 309)

eTable 10. Sensitivity Analyses of the Primary Aim: Average Change Since Baseline

eTable 11. Sensitivity ANCOVA Analysis for the Stroke Action Test (n = 309)

eTable 12. Model Based Means and Standard Errors for the Stroke Action Test by Education Level and Randomization Group (n = 309)

eTable 13. Slope Analysis (Monthly) for the Stroke Action Test (n = 309)

eTable 14. Slope Analysis Subgroup Comparison (Monthly) for the Stroke Action Test (n = 309)

eTable 15. Sensitivity Analyses of the Moderating Effects of Education Using the Average Change Model: Subgroup Comparison Follow-up vs Baseline for the Stroke Action Test (n = 309)

eTable 16. Model Estimate for Percent of Questions Answered Correctly (SAS Proc Mixed, Unstructured Variance, Adjusted for Clustering)

eTable 17. Overall (Average Change Since Baseline) Comparison Between Intervention and Usual Care by Education (High vs Low)

eTable 18. Sensitivity Analyses Examining the Number of Questions Answered Correctly Treating Education as Continuous Variable

eTable 19. Sensitivity Analyses Using the ANCOVA Model (Baseline as Covariate) Subgroup Comparison for the Stroke Action Test: Follow-ups (n = 309)

eTable 20. Observed Means (Standard Deviations) for the STAT by Education Level and Race/Ethnicity and Randomization Group for the Stroke Action Test, Number of Questions Answered Correctly

eTable 21. Model Based Means and Standard Errors for the Stroke Action Test by Education Level and Race/Ethnicity and Intervention Status (n = 309)

eTable 22. Education and Race/Ethnicity Subgroup Comparison Follow-up vs Baseline for the Stroke Action Test, Number of Questions Answered Correctly (n = 309)

eTable 23. Education and Race/Ethnicity Subgroup Comparison, Follow-up vs Baseline for the Stroke Action Test, Number of Questions Answered Correctly (n = 309)

eTable 24. Observed Means and Percentages for the Share Variables

Click here for additional data file.^{(158.2KB, pdf)}

Supplement 3.

Data Sharing Statement

Click here for additional data file.^{(20.3KB, pdf)}

References

1.White H, Boden-Albala B, Wang C, et al. Ischemic stroke subtype incidence among whites, blacks, and Hispanics: the Northern Manhattan Study. Circulation. 2005;111(10):1327-1331. doi: 10.1161/01.CIR.0000157736.19739.D0 [DOI] [PubMed] [Google Scholar]
2.Lacy CR, Suh DC, Bueno M, Kostis JB. Delay in presentation and evaluation for acute stroke: Stroke Time Registry for Outcomes Knowledge and Epidemiology (S.T.R.O.K.E.). Stroke. 2001;32(1):63-69. doi: 10.1161/01.STR.32.1.63 [DOI] [PubMed] [Google Scholar]
3.Hacke W, Kaste M, Fieschi C, et al. ; The European Cooperative Acute Stroke Study (ECASS) . Intravenous thrombolysis with recombinant tissue plasminogen activator for acute hemispheric stroke. JAMA. 1995;274(13):1017-1025. doi: 10.1001/jama.1995.03530130023023 [DOI] [PubMed] [Google Scholar]
4.Willey JZ, Williams O, Boden-Albala B. Stroke literacy in Central Harlem: a high-risk stroke population. Neurology. 2009;73(23):1950-1956. doi: 10.1212/WNL.0b013e3181c51a7d [DOI] [PMC free article] [PubMed] [Google Scholar]
5.Hsia AW, Edwards DF, Morgenstern LB, et al. Racial disparities in tissue plasminogen activator treatment rate for stroke: a population-based study. Stroke. 2011;42(8):2217-2221. doi: 10.1161/STROKEAHA.111.613828 [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Houston TK, Allison JJ, Sussman M, et al. Culturally appropriate storytelling to improve blood pressure: a randomized trial. Ann Intern Med. 2011;154(2):77-84. doi: 10.7326/0003-4819-154-2-201101180-00004 [DOI] [PubMed] [Google Scholar]
7.Olshefsky AM, Zive MM, Scolari R, Zuñiga M. Promoting HIV risk awareness and testing in Latinos living on the U.S.-Mexico border: the Tú No Me Conoces social marketing campaign. AIDS Educ Prev. 2007;19(5):422-435. [DOI] [PubMed] [Google Scholar]
8.Williams O, Leighton-Herrmann Quinn E, Teresi J, et al. Improving community stroke preparedness in the HHS (Hip-Hop Stroke) randomized clinical trial. Stroke. 2018;49(4):972-979. doi: 10.1161/STROKEAHA.117.019861 [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Williams O, Leighton-Herrmann E, DeSorbo A, et al. Effect of two 12-minute culturally targeted films on intent to call 911 for stroke. Neurology. 2016;86(21):1992-1995. doi: 10.1212/WNL.0000000000002703 [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Ravenell J, Leighton-Herrmann E, Abel-Bey A, et al. Tailored Approaches to Stroke Health Education (TASHE): study protocol for a randomized controlled trial. Trials. 2015;16:176. doi: 10.1186/s13063-015-0703-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Bernal G, Bonilla J, Bellido C. Ecological validity and cultural sensitivity for outcome research: issues for the cultural adaptation and development of psychosocial treatments with Hispanics. J Abnorm Child Psychol. 1995;23(1):67-82. doi: 10.1007/BF01447045 [DOI] [PubMed] [Google Scholar]
12.Lee JK, Hecht ML, Miller-Day M, Elek E. Evaluating mediated perception of narrative health messages: the Perception of Narrative Performance Scale. Commun Methods Meas. 2011;5(2):126-145. doi: 10.1080/19312458.2011.568374 [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Carpenter CJ. A meta-analysis of the effectiveness of health belief model variables in predicting behavior. Health Commun. 2010;25(8):661-669. doi: 10.1080/10410236.2010.521906 [DOI] [PubMed] [Google Scholar]
14.American Stroke Association. Learn more stroke warning signs and symptoms. https://www.strokeassociation.org/en/about-stroke/stroke-symptoms/learn-more-stroke-warning-signs-and-symptoms. Accessed October 2, 2018.
15.Billings-Gagliardi S, Mazor KM. Development and validation of the Stroke Action Test. Stroke. 2005;36(5):1035-1039. doi: 10.1161/01.STR.0000162716.82295.ac [DOI] [PubMed] [Google Scholar]
16.Bjorner JB, Rose M, Gandek B, Stone AA, Junghaenel DU, Ware JE Jr. Difference in method of administration did not significantly impact item response: an IRT-based analysis from the Patient-Reported Outcomes Measurement Information System (PROMIS) initiative. Qual Life Res. 2014;23(1):217-227. doi: 10.1007/s11136-013-0451-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
17.Bjorner JB, Rose M, Gandek B, Stone AA, Junghaenel DU, Ware JE Jr. Method of administration of PROMIS scales did not significantly impact score level, reliability, or validity. J Clin Epidemiol. 2014;67(1):108-113. doi: 10.1016/j.jclinepi.2013.07.016 [DOI] [PMC free article] [PubMed] [Google Scholar]
18.SAS [computer program]. Version 9.4. Cary, NC: SAS Institute Inc; 2012. [Google Scholar]
19.Diggle PJ, Liang KY, Zeger SL. Analysis of Longitudinal Data. New York, NY: Oxford University Press; 1994. [Google Scholar]
20.Sullivan LM. Repeated measures. Circulation. 2008;117(9):1238-1243. doi: 10.1161/CIRCULATIONAHA.107.654350 [DOI] [PubMed] [Google Scholar]
21.Schwarz G. Estimating the dimension of a model. Ann Stat. 1978;6(2):461-464. doi: 10.1214/aos/1176344136 [DOI] [Google Scholar]
22.Ramirez M, Teresi JA, Ogedegbe G, Williams O. Using qualitative methods to assess the conceptual equivalence of the Spanish and English Stroke Action Test (STAT). Qual Health Res. 2016;26(13):1812-1822. doi: 10.1177/1049732316636362 [DOI] [PubMed] [Google Scholar]
23.Houts PS, Witmer JT, Egeth HE, Loscalzo MJ, Zabora JR. Using pictographs to enhance recall of spoken medical instructions II. Patient Educ Couns. 2001;43(3):231-242. doi: 10.1016/S0738-3991(00)00171-3 [DOI] [PubMed] [Google Scholar]
24.George S, Moran E, Duran N, Jenders RA. Using animation as an information tool to advance health research literacy among minority participants. AMIA Annu Symp Proc. 2013;2013:475-484. [PMC free article] [PubMed] [Google Scholar]
25.Rahman R. Comparison of telephone and in-person interviews for data collection in qualitative human research. Interdisciplinary Undergraduate Res J . 2015;1(1):10-13. https://journals.uic.edu/ojs/index.php/IURJ/article/view/5428/4442. Accessed February 25, 2019. [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials