Study protocol of “Worth the Walk”: a randomized controlled trial of a stroke risk reduction walking intervention among racial/ethnic minority older adults with hypertension in community senior centers

Recruitment takes place at the same 4 senior centers (one for each ethnic group) where in-house WTW case managers have been trained. Members of the research team make brief presentations describing the project during events attended by large numbers of seniors, such as the low-cost hot midday meals provided at all of the senior centers participating in this study. Recruitment flyers our team developed in partnership with our mini-CABs listing the study objectives, eligibility criteria, and staff contact information are also posted and distributed at senior center sites. Potentially eligible seniors are invited to approach/contact study staff individually to be formally screened on pre-determined days at the senior center.

During screening, study staff complete in face-to-face format a screener form listing the eligibility criteria with each prospective participant. Prospective participants are given a FITBIT® Zip pedometer and instructed on its proper use. A letter and reply form (referenced above) is then faxed to the physician indicated by each potentially eligible participant (along with the patient’s signature indicating consent for our staff to contact the physician) describing the intervention and asking him or her to contact the study team within one week stating whether his or her patient has any medical contraindication to participating, including unstable angina, uncompensated heart failure, uncontrolled cardiac arrhythmia, severe aortic stenosis, hypertrophic cardiomyopathy, cardiomyopathy from recent myocarditis, severe pulmonary hypertension, abdominal aortic aneurysm, recent systemic or pulmonary embolus, thrombophlebitis, and severe balance problems. A faxable reply form is provided so that the physician can simply check “yes” or “no.”

Approximately one week after screening has been completed, eligible participants are invited to attend a baseline data collection session at the senior center, during which trained project staff complete the informed consent process with each potential participant individually in a private setting. After participants have provided written documentation of informed consent, staff proceed to collect baseline data (see Table 1).

After baseline data have been collected and participants have dispersed, we utilize a computerized randomization procedure stratified by site and gender to assign participants to either the intervention or the control (wait-list) arm. The randomization procedure is programmed into Research Electronic Data Capture (REDCap), a secure web application for building and managing online surveys and databases [32]. A randomization allocation table was created using random permuted blocks with randomized block sizes 2, 4, 6 and is uploaded into REDCap. Smaller block sizes are used to distribute participants within either group as evenly as possible.

One designated research assistant is un-blinded for the duration of the study and works with senior center staff to schedule the WTW intervention sessions. Senior center staff call participants to inform them of their scheduled intervention session. While the case manager group leaders administer WTW, the same un-blinded research assistant attends sessions to take attendance and monitor fidelity to the curriculum using a standardized checklist of all the major teaching points of each intervention session (see Additional file 1: Appendix A). Other (blinded) project staff members maintain regular telephone contact with participants in both arms of the study to remind them to wear their pedometers but without explicitly prompting physical activity or walking. Project staff who are in the field collecting data are blinded to the participant assignment group for the entire study duration.

All participants in both the intervention and control groups receive the same frequency of contact from study staff (phone call reminders) and the same incentives (pedometers, $25 visa gift card at each data collection time point). We do not refer to the control group as a “wait-list” when speaking with participants and agency staff, but rather refer to that group by the month in which they will start the WTW program, for example, we will compare the “January” group with the “April” group.

Participants assigned to the intervention group immediately attend the group-based interactive behavioral stroke risk reduction walking intervention led by an in-house case manager after baseline (T0). In addition to reminder phone calls prior to each intervention session, efforts to increase retention and sustain participation in the intervention include scheduling transportation for participants when applicable, as well as notifying participants that attendance will be taken at each session.

Participants assigned to the wait-list control group are invited to participate in WTW after follow-up data collection is completed 3 months from baseline (they do not repeat outcome measures at this time). As previously mentioned, control group participants are also given pedometers by blinded RAs, receive the same reminders to wear them, have follow-up data collected at the same time points, and have the same frequency of contact with study personnel as the intervention group throughout the study duration.

Data sources include FITBIT® Zip pedometers [33], surveys, physical exams, and fingerpricks. The surveys, physical exams, and fingerpricks are administered with each individual at the senior centers by trained bilingual study staff. Surveys are interviewer-administered using REDCap [32], accessed using iPads with internet access. Survey instruments were translated into the native languages of our targeted ethnic groups (Spanish, Mandarin and Korean) a priori through a professional translation company; the accuracy of translations was verified by bilingual study staff as well as by seniors with whom the surveys were piloted prior to trial implementation. Pedometer data are downloaded using the iPads and stored/accessed through the secure FITBIT® website at each data collection time point. All study outcomes, collection time points, and data sources are listed in Table 1.

The primary outcome measure is change in walking level from baseline (T0) to 1 and 3 months (T1 and T2) as measured by the FITBIT® Zip pedometers. Physical activity level is supplementarily assessed using the International Physical Activity Questionnaire [34] included in the survey at all three time points. Stroke risk factor knowledge is assessed via survey using the Stroke Action Survey and questions developed by Pancioli et al. [35, 36]. We measure self-efficacy for reducing stroke risk and increasing physical activity. Self-efficacy is best captured by measuring both self-efficacy expectations and outcome expectations [21]. As such, we measure self-efficacy expectations with the General Self-Efficacy Scale [37] and a modified version of the Chronic Disease Self-Efficacy Scale [38] that focuses on self-efficacy for stroke risk reduction; outcome expectations for being physically active is assessed using the Outcome Expectations for Exercise Scale [39].

To explore the relationship between the intervention and biological markers of health, we collect height and weight in light clothing without shoes at all three data collection time points in strict accordance with standardized procedures we have used previously. A random 5 % of participants will have their weight and height measured by two different research assistants so that inter-rater reliability can be calculated. To assess our main secondary outcome measure—systolic blood pressure—trained staff implement a standard, seated blood pressure protocol based on current Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure (JNC 7) guidelines, but allowing a 5-min rest between each measure [40]. Point-of-care CardioChek meters are used to measure cholesterol (using the Lipid Panel strip), which require only a small capillary blood sample (from a fingerprick). We use dried blood spots to obtain assays for HgA1c and CRP, also collected by fingerpricks. So that we can explore whether the intervention influenced healthcare seeking to control stroke risk factors, we ask participants to report on visits to healthcare providers in the previous three months at baseline and 3 months [41].

Other measures being collected are listed in Table 1. These include exploratory outcomes that might be influenced by the intervention such as health-related quality of life [42] and stress [43]. In addition, we will measure constructs such as acculturation level [44] and neighborhood walkability [45] that, while not linked explicitly to specific aims, are considered to be critical covariates likely to be associated with outcomes and can also be used to evaluate the success of the randomization.

We will use standard analytic methods for randomized controlled trials [52]. After comparing baseline characteristics of intervention and control group participants, and accounting for missing data, outcomes will be compared between the two groups. Though we have made unidirectional hypotheses, to allow for the chance that participants enrolled to the control group could have better outcomes than those in the intervention arm, we will use 2-tailed tests of significance for all analyses.

We have a priori selected the outcome of change in steps/day to be the primary endpoint of interest. Analyses of all other outcomes will be adjusted for multiple comparisons [54].

All analyses will be conducted using intention to treat, in which any subject randomized to the intervention arm remains in that arm regardless of whether or not he or she received the intervention, and likewise for the control arm. We will measure level of participation for those randomized to the intervention arm, and will conduct a sensitivity analysis that assesses the stability of the study’s conclusions when an intention to treat analysis versus an analysis that takes into account level of participation in the intervention.

Using t-tests for continuous variables and chi-square tests for categorical variables, we will compare the two groups on baseline characteristics to measure the success of the randomization: 1) sociodemographic characteristics; 2) mean steps/day over 1 week in the window post screening and distribution of pedometers, to the T0 baseline data collection one week later; 3) systolic blood pressure, body mass index, non-HDL cholesterol, HgA1c; 4) stroke knowledge, self-efficacy. Since steps/day is a continuous variable, at each time-point (30 days/T1 and 90 days/T2) we will perform a simple descriptive cross-sectional analysis of mean change in scores between baseline and follow-up in each treatment arm. In our primary analyses, we will test the statistical significance of the unadjusted difference in change-scores between the groups using t-tests for symmetrically distributed data and analogous nonparametric tests such as the Wilcoxon sign rank tests for data that are skewed. Imbalance between the treatment arms of characteristics described above would inflate the standard error, making the unadjusted analyses appropriately conservative; therefore, the unadjusted analyses will serve as the primary results of the trial. If, however, we did identify chance differences between the treatment arms in the distribution of the baseline level of the outcome, then we will also conduct secondary analyses in which we will use multivariate modeling, such as analysis of covariance, to adjust for these factors. The adjusted means of the change-scores from multivariate models will be compared and tested between the two arms, and will also be compared with the unadjusted means to assess the influence of these factors on endpoints. As sensitivity analyses, we will also examine the difference between groups in the absolute steps/day (instead of the change-score), adjusting for baseline. These sensitivity analyses will allow us to compare outcomes between groups without assuming linearity of effect across different baseline levels of physical activity. Using multivariate modeling, we will also test interactions between treatment arm and selected effect modifiers (e.g., age, level of acculturation, etc.) to identify sub-groups of participants who improve more than other groups. Since this trial is not designed with power to examine these subgroups, these analyses will be exploratory and hypothesis-generating in nature. Finally, we will assess changes in steps/day between baseline and the end of the intervention period, and differences between the control and intervention groups in steps/day trajectories over time. These analyses will consist of graphical analysis and parametric statistical testing using repeated measures mixed effects models [55, 56]. To determine the extent to which increases in stroke knowledge and self-efficacy mediate improvements in steps/day, the variation (R-squared) in increase in steps/day is explained by variation in increase in the stroke knowledge and self-efficacy constructs will be calculated [57, 58]. Analysis of the effect of the intervention on all the secondary and exploratory outcomes will follow the same analysis.