Results from ‘Developing Real Incentives and Volition for Exercise’ (DRIVE): A Pilot Randomized Controlled Trial for Promoting Physical Activity in African American Women

Abstract

Objective.

Motivation is a barrier to physical activity (PA) among African American (AA) women, but past studies have implemented a ‘one-size-fits-all’ approach and have not addressed differences in autonomous motivation. This pilot randomized controlled trial assessed the preliminary efficacy of ‘Developing Real Incentives and Volition for Exercise’, a community and theory-based intervention, which evaluated whether a motivationally-matched (vs. a non-matched) intervention increases daily total PA.

Methods.

68 AA women (50.72 ±13.66 years; 86.8% with obesity) were randomized to an 8-week Challenge-focused program (targeted towards high autonomous motivation) or Rewards-focused program (targeted towards low autonomous motivation). Randomization was stratified by baseline autonomous motivation. FitBits were used during the intervention to promote self-monitoring (both programs) and social connectedness (Challenge program only).

Results.

Both programs retained ≥ 80% of participants. Process evaluation revealed high attendance, dose, and fidelity (both programs). However, contrary to expectations, across all motivational levels (low and high autonomous), the Challenge-focused intervention resulted in a greater increase in total daily PA (primary outcome), with an average increase of 17.9 minutes in the Challenge-focused intervention vs. an average decrease of 8.55 minutes in the Rewards-focused intervention. An exploratory follow-up analysis revealed that engagement with the FitBit mobile app predicted greater PA at post-intervention in the Challenge-focused program.

Conclusions.

A team-based approach targeting social connectedness, enjoyment of PA, and positive intragroup competition is a promising approach for promoting PA among AA women. These findings are used to guide a discussion on best practices for engaging AA women in future behavioral interventions.

Introduction

African American (AA) women are the least active U.S. demographic group, with 37.4% meeting aerobic physical activity (PA) guidelines (≥ 150 minutes/week moderate PA or ≥ 75 minutes/week vigorous PA), vs. 55.1% of White women (Virani et al., 2020). Group-based PA interventions hold tremendous promise, as evidenced, for example, by one systematic review, which found that 92% of group-based PA interventions were successful at increasing PA (Harden et al., 2015). By providing opportunities for building a supportive group climate (e.g., social support, feedback, observational learning), group-based programs may be especially relevant for AA women for whom lack of social support and exposure to other active AA women are frequently reported PA barriers (Harley et al., 2009; Joseph et al., 2015). PA interventions for AA women, including group-based approaches, have increased in recent years (Bland & Sharma, 2017; Jenkins et al., 2017). However, attendance and retention remain critical issues in many group-based programs for AA women (Coughlin & Smith, 2016; Lemacks et al., 2013). Such findings highlight the critical need to partner with communities to develop better solutions for reducing barriers to participation and promoting a supportive group environment to engage AA women in greater PA.

One approach to increasing engagement is to enhance personal relevancy through targeting (at the group-level) or tailoring (at the individual-level) (Kreuter & Haughton, 2006; Resnicow et al., 1999). Culturally-targeted interventions have been effective for improving PA, diet, and weight outcomes among AAs, especially when the intervention addresses cultural or social values (Barrera et al., 2013; Kong et al., 2014). For example, past culturally-targeted interventions for AA women have addressed prominent socio-cultural norms and barriers, such as hair concerns, low social support, and caregiver responsibilities (Joseph et al., 2021; Parra-Medina et al., 2010). Although there is extensive research supporting the importance of addressing cultural differences, little research has evaluated whether other theoretical constructs can be used to develop targeted interventions for AA women. Thus, this study evaluated whether motivation for PA is a useful construct for matching AA women to a targeted group-based intervention.

Motivation is often reported as a barrier to PA among AA women (Joseph et al., 2015; Siddiqi et al., 2011). Past studies have evaluated motivation in terms of motivational readiness and tested the efficacy of tailoring interventions based on one’s motivational stage (Prochaska et al., 1997). However, systematic reviews have found limited support for the efficacy of this approach (Bridle et al., 2005). Alternatively, Self-Determination Theory (SDT) (Ryan & Deci, 2000) emphasizes the importance of different sources of motivation, including distinguishing between controlled (i.e., motivation from external factors, such as extrinsic rewards, approval from others) and autonomous sources (i.e., motivation from internal factors, such as enjoyment, values). Although controlled motivation promotes short-term PA, autonomous motivation is a more robust predictor of long-term PA (Teixeira et al., 2012). As a result, many group-based interventions have targeted autonomous motivation as a key mediator for increasing PA (Gillison et al., 2019).

Past PA interventions have rarely addressed individual variability around individuals’ reasons for initiating PA (e.g., low vs. high autonomous motivation). Although autonomous motivation is an important predictor of long-term PA, AA women experience social-motivational barriers, which may interfere with autonomous motivation, such as low social support, lack of a partner, and low exposure to other physically active AA women (Harley et al., 2009; Joseph et al., 2015). SDT suggests that social-motivational needs (e.g., type of social support needed) may vary depending upon one’s level of competency, autonomy, and relatedness (Ryan & Deci, 2000). Thus, among AA women with low activity levels, those initiating PA for autonomous reasons may have different barriers, needs, and interests than those initiating for controlled reasons.

To explore this idea, we conducted a focus group study with AA women to evaluate whether social-motivational PA needs/interests varied between those with relatively low vs. high autonomous motivation (Allison M. Sweeney et al., 2019). A sample 31 AA women with low self-reported PA levels were recruited to participate (age range 24-69, 43.4±11.4 years). Autonomous motivation for PA was measured prior to the focus group discussion, which was used to categorize women as low, medium, or high in autonomous motivation using a tercile-based approach.

The results from the focus group study showed that among participants with relatively high autonomous motivation, there was greater discussion of wanting PA to be fun and competitive (Allison M. Sweeney et al., 2019). Those with high autonomous motivation also expressed a desire for social connectedness with their peers. These results are consistent with research on intragroup competition, which indicates that competing within groups to achieve a common goal promotes enhanced effort, positive affect, and group cohesion (Harden et al., 2014; Tauer & Harackiewicz, 2004; Wittchen et al., 2013). According to SDT, intragroup competition may be especially effective at engaging those with high autonomous motivation because it capitalizes on intrinsic enjoyment and provides an optimal challenge (i.e., balancing novelty and competency). Furthermore, social connectedness may be especially important for those with high autonomous motivation in order to fulfill psychological needs for relatedness (Ryan & Deci, 2000). Thus, those with high autonomous motivation may benefit more from a team-based program targeting intragroup competition, enjoyment of PA, and social connectedness.

Alternatively, the results from the focus group study revealed that among participants with relatively low autonomous motivation, there was greater discussion of the need for a partner to reinforce accountability and provide instrumental support. Social support facilitates behavior change through a variety of mechanisms, including enhanced self-efficacy, problem-solving, and positive self-perceptions (Anderson et al., 2006; Wilcox & Vernberg, 1985). Given that those with low autonomous motivation tend to have low levels of competency (Ryan & Deci, 2000), partner-based social support may be especially important for promoting confidence in initiating PA.

Participants with low autonomous motivation in the focus group study also expressed a desire for financial incentives (Allison M. Sweeney et al., 2019). This is consistent with the idea that those with low autonomous motivation are focused on extrinsic (rather than intrinsic) rewards (Ryan & Deci, 2000). Financial incentives have been shown to promote at least short-term changes in PA (Mitchell et al., 2013; Patel et al., 2016). According to General Interest Theory (GIT), when a task requires time to acquire proficiency, rewards help people to link an otherwise unpleasant task to their interests (Eisenberger et al., 1999; Eisenberger & Cameron, 1996). Thus, those with low autonomous motivation may benefit more from a partner-based program that targets competency and interest in PA through partner-based social support and financial incentives.

Drawing from our prior focus group results and research on intragroup competition, social support, and incentives, the present study evaluated whether it is beneficial to address differences in individuals’ motivation for initiating PA. Given challenges in previous PA interventions with attendance and retention (Coughlin & Smith, 2016), this study focused on leveraging AA women’s baseline sources of motivation in order enhance early engagement. Thus, this study explored whether it is useful to tailor intervention strategies towards differences in motivation (low vs. high autonomous) as a starting point during the initiation stage of behavior change.

We developed two motivationally-targeted interventions: a Challenge-focused program targeted towards high autonomous motivation and Rewards-focused program targeted towards low autonomous motivation. Participants were randomized to a program that was matched (high autonomous and Challenge; low autonomous and Rewards) or not-matched (high autonomous and Rewards; low autonomous and Challenge) to their baseline autonomous motivation. In a previous feasibility study, we demonstrated that it was feasible to deliver both programs with high acceptability (Sweeney et al., 2020). Refining this work further, the primary aim of the ‘Developing Real Incentives and Volition for Exercise’ (DRIVE) pilot randomized controlled trial was to evaluate the preliminary efficacy of using a motivational-matching approach for promoting a clinically meaningful increase in daily total PA (≥ 10 minutes of total PA/day) (Moore et al., 2012).

We hypothesized that receiving a motivationally-matched vs. a non-matched program would result in a greater pre-post change in daily total PA (primary outcome). Additionally, we report the feasibility of program implementation through process evaluation, participant feedback, and engagement with program-provided Fitbits (secondary outcomes).

Methods

Study Overview

The study used a group cohort design with assessments at baseline and post-intervention (8 weeks). The study was implemented between 2019-2020 and included 8 treatment groups across 4 cohorts, with group size ranging from 8-10 members.¹ The study was implemented at a community center in Sumter, SC, which the research team has collaborated with on previous studies (Sweeney et al., 2020; Sweeney et al., 2019; Wilson et al., 2015). The community center director was our community partner and provided the team with formal feedback on the research approach and assisted with recruitment (e.g., invitations to speak at community events). The community center offers a free afterschool program from which the research team recruited caregivers of children attending the afterschool program. We offered the program in the evenings (following the afterschool program), which allowed us to deliver the program in a convenient, familiar, and accessible community setting.

Participants were AA women with low PA levels recruited through flyers and community events. Interested participants completed a phone screening during which they received information about eligibility and participation. Eligibility criteria included: 1) ≥ 21 years old; and 2) engaging in < 150 minutes of moderate to vigorous PA per week for the last three months (self-reported with the International Physical Activity Questionnaire (Craig et al., 2003) and an item “How many weeks have you consistently maintained engaging in this level of moderate to vigorous activity?”) Exclusion criteria included: 1) having a cardiovascular or orthopedic condition that would limit PA (self-reported with the Physical Activity Readiness Scale) (National Academy of Sports Medicine, 2019); or 2) uncontrolled blood pressure (systolic >180 mmHg/diastolic >110 mmHg) (measured by research staff prior to enrollment).

Eligible participants completed a group orientation session, which involved reviewing information about the study aims, incentives, and randomization procedure. Participants who enrolled completed a blood pressure assessment, the Behavioral Regulation in Exercise Questionnaire (BREQ-3) (Markland & Tobin, 2004), and provided written informed consent. Consistent with the ORBIT model (Czajkowski et al., 2015), this study evaluated preliminary efficacy through clinical significance. Thus, the sample size was guided based on feasibility, which was informed by our previous pilot feasibility study (Sweeney et al., 2020).² The study was approved by the University of South Carolina’s Institutional Review Board and was registered prior to recruitment on clinicaltrials.gov (NCT # 03873051). This study has not been published previously and this will be the first manuscript with this dataset. We report how we determined our sample size, all data exclusions, manipulations, and the primary measure in the study (total daily PA; pre-registered). The dataset and materials are available on reasonable request.

Randomization

Autonomous motivation for PA was measured at enrollment and used to randomize an even distribution of motivation levels across interventions. The BREQ-3 consists of subscales measuring sources of motivation using a 6-point scale including amotivation, low autonomy (introjected, external motivation), and high autonomy (intrinsic, integrated, identified), and showed high internal consistency (α = .84). An autonomy index score was created by applying a weight to each subscale and summing the weighted scores, with higher scores indicating greater autonomous motivation (Wilson et al., 2007). The scale ranges from −30 to 30. Because median splits have numerous limitations (Aiken & West, 1991) and there were no existing cut points, participants were placed into terciles, such that low = −1.5 to 10.42, medium = 10.75 to 16.1 and high = 16.5-24. Half of the participants within each tercile were randomly assigned to one of two evenings and evenings were then randomized to a condition using a computer-generated randomized algorithm by the PI.

Procedure

Both interventions met for 90-minutes, including a baseline session (week one), eight intervention weeks (weeks two-nine), and one post-intervention session (week ten). In week one, facilitators led a group discussion, including an ice breaker, program expectations, and ground rules. Participants completed baseline measures (anthropometrics, survey) and received an accelerometer to wear for seven days to assess baseline PA. After returning the accelerometer, participants received a Fitbit in week two, which was used during the intervention to promote self-monitoring (both programs) and social connectedness (Challenge program only). In weeks two-nine both interventions followed the same structure: 1) feedback on PA goals; 2) health curriculum; 3) PA session; and 4) behavioral skills training (e.g., goal-setting, problem-solving, action-plans) (see Table 1). Participants received an accelerometer to assess post-intervention PA between weeks eight and nine and completed all final measures in week ten.

Table 1.

Overview of key components in the Challenge- and Rewards-focused interventions

	Challenge-focused program	Rewards-focused program
Feedback on PA Goals	Feedback on group-based PA goal	Feedback on individual-based PA goals
	• Review whether collective team goal was met • Receive personalized summary of weekly Fitbit data • Discuss successes, barriers & practice problem-solving • Reinforcement of positive affect and PA	• Review whether individual goals were met • Receive personalized summary of weekly Fitbit data • Discuss successes, barriers & practice problem-solving
Health Curriculum	Health Curriculum Discussion (same in both programs)
	• Weekly learning objectives • Key content • Open-ended questions and group discussion
PA Session	Intragroup competitive physical activities	Group Walking
	• Warm-Up • Intragroup physical activities (e.g., relays, calisthenics challenges)	• Warm-Up • Non-competitive group walking
Goal-Setting	Team-based collective goal-setting	Individual-based goal-setting
	• Group sets a team-based collective PA goal (e.g., 8,000 steps 3 days/week) • Fitbits used to track PA, team-based goal, communicate with group members, and view the group leaderboard • No incentives for meeting weekly goals • Group discussion of anticipated barriers and solutions • Develop a plan for group support (e.g., Fitbit mobile app communication) • Values reflection task	• Select a partner, receive personalized feedback from facilitator, and set a personal weekly PA goal • Fitbits used to track PA and individual-based goal. Not connected to group on the Fitbit mobile app. • Financial incentives for meeting weekly goals • Discussion of anticipated barriers and solutions with facilitator and partner • Develop an action plan with partner for social support

Intervention Overview

The DRIVE program integrated core components from SDT (Ryan & Deci, 2000), Social Cognitive Theory (SCT) (Bandura, 2004), and GIT (Eisenberger et al., 1999), but was also adapted based on our focus group study (see Supplementary Materials Table S1). The Challenge-focused program used a team-based model that targeted social connectedness, positive intragroup competition, enjoyment and valuation of PA, and collective self-efficacy though intragroup competitive physical activities, engagement with the Fitbit mobile app (sending messages, participating in the group leaderboard), collective team goals, and a focus on enjoyment and values. The Rewards-focused program used a partner-based model that targeted partner-based social support, individual self-efficacy for PA, and interest in PA through a walking program, individual-based goals with support from a partner, and performance-based financial incentives.

Group Sessions

Two trained facilitators delivered each program. The lead facilitator was a White female with a PhD in Psychology, and all co-facilitators were White female graduate students in psychology or public health. Facilitators underwent training focusing on behavioral skills related to PA (e.g., goal-setting, problem-solving), motivational interviewing skills, strategies for creating a positive social climate, and cultural competency. The facilitators used weekly guides, which included key content and discussion prompts. Both programs delivered a health curriculum, which focused on PA, but also integrated topics that were identified as important from our focus group study (see Supplementary Materials Table S2). After completing the weekly health curriculum, participants completed 30 minutes of PA (intragroup competitive physical activities or group walk), and behavioral skills training (team or partner-based goals and problem-solving).

In both programs, participants received feedback on their goals, including weekly summaries of their Fitbit PA results and discussing both progress and areas for improvement with the group. In the Challenge-focused program this discussion was centered around participants’ contribution to the collective team goal, whereas the Rewards-focused program discussed their individual goals. In the Challenge-focused program discussion also involved reflecting on positive affective experiences related to their PA (e.g., feelings of accomplishment) and the relationship between PA and personal values (a values-reflection task).

If participants missed a session, a makeup session was offered by phone. Makeup sessions were completed prior to the next group session, lasted 30-45 minutes and included completing key content from the facilitator guide. The goal of the makeup session was to provide feedback (e.g., successes, barriers, problem-solving), deliver key content from the health curriculum, and engage the participant in goal-setting for the upcoming week (individual-based or update on team goal).

Structure of incentives

Participants received equivalent incentives in both interventions ($60), but the structure of the incentives differed. For the Challenge-focused intervention, incentives were provided for completing baseline ($20) and post-intervention measures ($40), whereas for the Rewards-focused intervention incentives were provided for meeting weekly PA goals and measures. The different incentive structure was used to test whether receiving small incentives was an effective approach for engaging those with low autonomous motivation in the Rewards program. This approach was based in part on previous studies finding that small incentives are useful for promoting health behavior change (Leahey et al., 2017).

Challenge-focused intervention (targeted toward high autonomous motivation)

Intragroup competitive physical activities.

The PA session included a warm-up before completing intragroup competitive physical activities, including relay or tag-based games adapted from Zarrett et al., (2021), as well as calistenic, and chair exercise challenges (see Supplementary Materials for an example). These activities required participants to work together in groups, with activities ranging from collaborative (working as one large group) to competitive (competing in teams). By incorporating new activities weekly, while also building upon previous skills (e.g., improving walking speed), the intragroup activities were designed to provide an optimal challenge.

Team-based collective goal-setting.

Participants worked together to develop a collective team-based PA goal. The aim of this approach was to create group accountability and promote positive group competition. The facilitators provided guidance to promote a goal that was specific, measurable, attainable, relevant, and time specific (SMART) (Doran, 1981). Participants also created a personal goal to plan how they would pursue the team goal (e.g., “walk for 20 minutes three days/week”). Prior to each meeting, the facilitators calculated the number of days each member met the team goal using Fitbit data and shared the team average with the group. Facilitators also addressed behavioral skills related to goal-setting by guiding a group discussion around anticipated barriers for the upcoming week, potential solutions, and plans for providing group support (e.g., using the Fitbit mobile app to send messages and track the group’s progress).

The first four weeks focused on familiarizing participants with the goal-setting approach. In the last four weeks to further emphasize valuation of PA, goal-setting integrated a values-reflection approach intended to help participants connect their PA goals to their broader life aspirations. Participants identified their most important values (Resnicow et al., 2015) and completed an exercise that involved reflecting upon why one engages in PA (Freitas et al., 2004). The activity is structured so that participants relate PA to increasingly broad reasons that help them to connect PA to their values and life aspirations. This task was completed in week five, and in the following weeks participants reflected on their personal reasons for engaging in PA each week.

Fitbit Mobile App Group Engagement.

Participants were connected to their group members through the Fitbit mobile app to encourage social connectedness outside of the in-person sessions. The app allowed participants to see each other’s progress, send messages, and view their ranking compared to other members via the group leaderboard. As part of the goal-setting procedure, participants developed a plan for providing group support, which typically involved plans to use the app to communicate, monitor the group’s progress throughout the week, and engage in friendly competition using the app’s leaderboard feature. The researchers tracked the usage of this feature to ensure that it was used in the Challenge-focused program only.

Rewards-focused intervention (targeted towards low autonomous motivation)

Group walk.

After a warm-up, participants completed a group walk inside the gymnasium. Walking was selected because previous focus groups revealed that walking was a preferred activity (Sweeney et al., 2019).

Individual-based goal-setting and partner-based support.

Participants selected a partner from the group. Each week, the facilitators worked directly with participant pairs in developing individual PA goals. Using a contingency contracting approach (Sykes-Muskett et al., 2015), participants completed a worksheet that included receiving feedback, specifying a goal, brainstorming strategies for anticipated barriers, and creating an action-plan (when, where, and how to execute the goal). This worksheet was created by the authors and was informed by principles of contingency contracting, SMART goal-setting, and implementation intentions (Gollwitzer, 1999). Participants were also informed what the upcoming incentive would be for meeting their weekly goal. After developing their weekly PA goals, the facilitators guided partners in developing an action-plan for supporting each other (e.g., phone call to check-in).

Incentives.

Participants received incentives for baseline ($10) and post-intervention ($10) measures, with the remaining $40 distributed as incentives for meeting PA goals. Drawing from learning theory (Thaler, 1981), incentives were delivered frequently (every 2 weeks) and varied in size ($8-12), which is consistent with previous behavioral interventions that have used small and variable incentives ($1-$10) for meeting behavioral goals (Leahey et al., 2017).

Process Evaluation

Attendance.

Attendance was calculated using the average proportion of participants in attendance per week, with an a priori goal of ≥ 75% in attendance each session.

Systematic Observation.

Process evaluation was used to evaluate both interventions (reach, dose, fidelity). Undergraduate and graduate students underwent training, including male, female, African American, and White individuals. This involved completing the same intervention training as the facilitators. Additionally, process evaluators completed a certification process, which involved listening to recordings of previous sessions, completing practice evaluations, and achieving high interrater reliability (r ≥ .80). Quality control checks were conducted at the beginning of each cohort to confirm that the systematic observations were completed correctly. Systematic observations were completed for all group sessions.³ Process evaluators collected systematic observational data, which involved completing a checklist to evaluate whether core intervention components were delivered and implemented as planned. Evaluators received a copy of the facilitator guide to evaluate whether each session was implemented as planned.

The a priori goal for dose was for each program component to be delivered ≥ 75% of the time (e.g., “Most (≥75%) of the participants engage in ≥ 30 minutes of PA during PA session”). The a priori goal for fidelity was to average ≥ three at both the facilitator and group-level (four-point scale, 1 = none, 2 = some, 3 = most, 4 = all). The a priori goals for dose and fidelity were based on a previous group-based intervention (Wilson et al., 2022). At the facilitator level, items were used to assess communication skills (“Facilitator(s) use open-ended question to elicit reflections and input from participants”), social support (“Facilitator(s) reinforce positive interactions between participants”), behavioral skills (“Facilitator(s) aid participants in identifying and overcoming barriers towards skill development and goal attainment”), and session content (“Facilitator(s) covered key content as outlined in the facilitator’s guide”). At the group-level, items were used to assess behavioral skills (“Participants provide one another with encouragement and positive reinforcement towards achieving PA goals”), group climate (“Participants share personal stories related to working on PA goals”), group climate during PA (“Participants display signs of excitement or fun [cheering or clapping]”), and positive communication (“Participants have meaningful nonverbal interaction with one another [good eye contact]”).

Mid-Point Participant Surveys.

Participants in both programs completed a self-report paper and pencil survey at mid-point, including items assessing the acceptability of program components (e.g., PA sessions, goal-setting approach, use of Fitbits, interactions with facilitators and group members). These items were created by the authors to assess specific components of each program using a six-point agreement scale and 5-point frequency scale.

Measures

Demographics.

Participants completed a baseline survey, including the following demographic items: employment status, education, annual household income, marital status, and number of children living in their home.

Fitbits.

Participants received a Fitbit (Flex 2), which provided estimates of daily steps during the 8-week intervention. Fitabase (Small Steps Labs LLC) was used to compile participants’ weekly PA data. While Fitbits were used to promote self-monitoring in both programs and social connectedness (Challenge-focused program only), estimates of PA from Fitbits are less reliable than accelerometers (Reid et al., 2017). Thus, estimates of daily steps were collected in both programs to assess different aspects of engagement and not as an objective PA measure. In the Challenge-focused program, participants’ Fitbit accounts were monitored (with permission) by the research team which allowed us to track how frequently and with whom participants exchanged messages on the mobile app. In the Rewards-focused program, Fitbits were used simply as a self-monitoring tool with no opportunity to communicate with group members.

Accelerometry-assessed PA.

Participants wore omnidirectional Actical accelerometers on an elastic belt over the right hip for 7 days (baseline and post-intervention). Data were recorded in 1-minute epochs, with 60-minutes of consecutive zeros defined as non-wear (Wilson et al., 2015). Previously established cut points (Trumpeter et al., 2012) were used to convert raw data into time spent in different PA intensities (LPA: 100 – 1074; MVPA: ≥ 1075), which were summed to index total daily PA. To evaluate preliminary efficacy, the present study focused on clinical significance, which was defined as increasing daily total PA by ≥ 10 minutes from baseline to post-intervention. This is based on previous research which has found that a PA level equivalent to brisk walking for 10 minutes/day is associated with 1.8 additional years of life expectancy (Moore et al., 2012).

Anthropometrics.

Height and weight were measured using a medical-grade Seca scale and wall-mounted height board using a protocol from a previous trial (Wilson et al., 2015).

Analysis Plan

First, descriptive analyses were used to assess reach (enrollment rate, attendance, retention), dose and fidelity (via systematic observation), and participant feedback. Second, to evaluate preliminary efficacy, pre-post changes in daily total PA were calculated by comparing predicted treatment means across interventions and levels of autonomous motivation (matched and non-matched conditions). Given that this was a pilot study, null hypothesis testing is not appropriate (Czajkowski et al., 2015), and thus we calculated estimates, SEs, and 95% CIs to determine whether a clinically meaningful change occurred between conditions. We expected that those in a motivationally-matched (vs. non-matched) program would show a greater mean increase in total daily PA.⁴ Third, to evaluate engagement during the intervention period we assessed changes in average daily steps across the 8-week intervention (both interventions). Finally, an exploratory follow-up analysis was conducted to evaluate whether group communication on the Fitbit mobile app related to post-intervention total PA (Challenge program only). Analyses were completed using R (R Core Team, 2021).

To calculate predicted treatment means, mixed models were conducted, which included random intercepts and random slopes for treatment group and time. First, a model was conducted with the full sample with fixed effects of time, intervention condition, baseline autonomous motivation, age, season, day of week (to account for weekday vs. weekend differences), and a time*intervention interaction term. To evaluate differences by motivation, separate models were conducted by sub-setting the data by autonomous motivation level. All covariates were grand-mean centered and only time and intervention condition were dummy-coded. With all other variables being centered, by varying the reference group for time and intervention condition, the model intercept was used to compute predicted means, SEs, and 95% CIs for each timepoint and condition. We report the effect size for the time*treatment effect in each model to aid in interpretation of the treatment means. To account for missing accelerometry wear, the mixed models used variance weighting by the inverse of daily wear time proportions for each day during which the accelerometer was worn for ≥ 30 minutes. This approach results in those with a higher proportion of missing wear time being down-weighted (vs. those with less missingness) and improves precision in estimating PA (Yue Xu et al., 2018). See the Supplementary Materials for additional information on data cleaning and effect size calculations.

Results

Recruitment, Participant Characteristics and Retention

A total of 111 participants completed the phone screener, with 102 agreeing to attend a group orientation session. Of these participants, 29 were no-shows and two did not enroll, resulting in an enrollment rate of 71%. Three participants dropped before the first week (due to availability), resulting in a final sample of 68 (see Supplementary Figure S1). As shown in Table 2, the sample consisted of AA women, with an average age of 50.72 ± 13.66 and an average BMI of 37.91 ± 8.47. Retention, defined as completing post-intervention measures, was somewhat higher in the Challenge (97.3%) than the Rewards-focused (83.9%) program, but not significantly different (χ²(1, N = 68) = 2.29, p = 0.13).

Table 2.

Participant characteristics at baseline

	Rewards (N = 31)	Challenge (N = 37)	Total (N = 68)
Age	48.55 (13.58)	52.54 (13.64)	50.72 (13.66)
Baseline BMI	37.53 (9.39)	38.23 (7.73)	37.91 (8.47)
Married	42%	49%	45.60%
Number of children	1.04 (1.26)	0.97 (1.28)	1.00 (1.26)
Education
Grades 9-11	1 (3.2%)	0 (0%)	1 (1.47%)
High School	4 (12.9%)	6 (16.2%)	10 (14.7%)
College 1-3 years	7 (22.6%)	12 (32.4%)	19 (27.9%)
College 4 years	11 (35.5%)	8 (21.6%)	19 (27.9%)
Grad training or professional degree	7 (22.6%)	10 (27.0%)	17 (25%)
Unreported	1 (3.2%)	1 (2.70%)	2 (2.9%)
Income
< 10,000	2 (6.5%)	1 (2.7%)	3 (4.4%)
10-24,999	4 (12.9%)	6 (16.2%)	10 (14.7%)
25-39,999	6 (19.4%)	9 (24.3%)	15 (22.1%)
40-69,999	9 (29.0%)	8 (21.6%)	17 (25.0%)
> = 70,000	6 (19.4%)	11 (29.7%)	17 (25.0%)
Unreported	4 (12.9%)	2 (5.4%)	6 (8.9%)

Feasibility of Program Implementation (Attendance, Dose, Fidelity)

In both interventions, the average percentage of participants in attendance each week exceeded our goal of 75% (Rewards: 79.0%; Challenge: 88.8%). With the inclusion of makeups, attendance rates increased to 87.8% in the Rewards program and 96.8% in the Challenge program. As shown in Table 3, across both programs adequate dose was achieved (ranging from 93-99%). At the facilitator level, fidelity was achieved for communication with participants, social support, behavioral skills and session content across both interventions. At the group level, fidelity was reached for group climate, group climate during PA, and communication skills. Group climate during PA assessed whether participants worked together and provided encouragement (e.g., high-fives, cheering), and as expected these scores were higher for the Challenge than the Rewards-focused program. At the group-level, fidelity for behavioral skills did not meet the a priori criteria of ≥ 3 across all cohorts. A closer look at these items revealed that although participants provided one another with encouragement (M_Rewards = 3.24, M_Challenge = 3.32), scores were lower for helping one another to overcome barriers (M_Rewards = 2.84, M_Challenge = 2.41).

Table 3.

Results of the intervention implementation (dose and fidelity)

	Cohort 1		Cohort 2		Cohort 3		Cohort 4
	Challenge	Rewards	Challenge	Rewards	Challenge	Rewards	Challenge	Rewards
Dose
Session content (15 items, %)	97%	94%	98%	93%	99%	97%	93%	96%
Fidelity
Facilitator Level, M (SD)
Communication	3.99 (0.04)	3.59 (0.13)	4.0 (0.00)	3.94 (0.14)	3.95 (0.11)	3.9 (0.28)	3.91 (0.14)	4.00 (0.00)
Social Support	3.62 (0.53)	3.31 (0.37)	3.79 (0.27)	3.88 (0.35)	3.75 (0.35)	3.56 (0.64)	3.15 (0.34)	3.5 (0.58)
Behavioral Skills	3.71 (0.19)	3.83 (0.21)	3.73 (0.22)	3.71 (0.12)	3.71 (0.26)	3.8 (0.18)	3.56 (0.27)	3.76 (0.18)
Session Content	3.88 (0.35)	3.88 (0.35)	4 (0.00)	3.88 (0.35)	3.88 (0.35)	3.75 (0.46)	4 (0.0)	4 (0.0)
Group Level, M (SD)
Behavioral Skills	3.00 (0.60)	3.64 (0.48)	2.57 (0.53)	3.12 (0.74)	3.5 (0.60)	2.75 (0.93)	2.90 (0.22)	2.33 (0.58)
Group Climate	3.19 (0.65)	3.86 (0.24)	3.43 (0.35)	3.94 (0.18)	3.5 (0.53)	3.81 (0.26)	3.10 (0.22)	3.33 (0.29)
Group Climate during PA	3.17 (0.61)	1.95 (0.65)	3.75 (0.66)	1.22 (0.34)	3.81 (0.35)	1.77 (0.94)	NA	NA
Communication skills	3.41 (0.42)	3.58 (0.25)	3.53 (0.35)	3.75 (0.31)	3.82 (0.31)	3.67 (0.54)	3.29 (0.45)	3.4 (0.17)

Note: A priori criteria for dose ≥ 75% and a priori criteria for fidelity was ≥ 3

Group climate during PA was not assessed in Cohort 4 give that the majority of sessions were delivered virtually

Participant Feedback

Across both programs, there was high acceptability of the Fitbits (Challenge: M=5.66±0.92, Rewards: M=5.96±0.19) and high levels of agreement that the DRIVE program was beneficial and informative (see Supplementary Materials, Table S3). In the Challenge-focused program, participants indicated that the PA games were fun (M=5.69±0.75), but there was greater variability in the extent to which the activities were perceived as challenging (M=4.80±1.56). Participants expressed high acceptability for the team-based goal-setting approach (with all items averaging ≥ 5 on a 6-point scale). In the Rewards-focused program, participants indicated high levels of acceptability for the walking program and goal-setting approach (with all items averaging ≥ 5 on a 6-point scale). Participants indicated that their partners were supportive (M=5.59±0.83), with 77.8% indicating that they communicated with their partner at least 2-3 days/week outside of in-person sessions. Finally, there was some variability around whether the financial incentives were viewed as helpful (M=4.67±1.25).

Preliminary Efficacy

As shown in Table 4, the predicted treatment means for the overall model revealed that at post-intervention, participants in the Challenge- vs. Rewards-focused intervention engaged in 173.13 (SE = 10.36) [95% CI = 152.79, 193.47] vs. 138.98 (SE = 11.47) [95% CI = 116.47, 161.49] minutes of daily total PA, respectively. These predicted means reflect an average increase of 17.9 minutes/day of total PA in the Challenge-focused intervention and an average decrease of 8.55 minutes/day of total PA in the Rewards-focused intervention. The time*treatment effect in the overall model yielded an effect size of Cohen’s d = 0.710.

Table 4.

Predicted treatment means and standard errors for total physical activity across treatment programs, motivational levels, and time points.

		Challenge-focus		Rewards-focus
		M (SE)	95% CI	M (SE)	95% CI
Across all motivational levels	Baseline	155.23 (12.66)	[130.39, 180.08]	147.53 (13.17)	[121.68, 173.38]
	8 weeks	173.13 (10.36)	[152.79, 193.47]	138.98 (11.47)	[116.47, 161.49]
High Autonomous Motivation	Baseline	168.98 (24.50)	[120.77, 217.18]	157.45 (26.93)	[104.47, 210.44]
	8 weeks	192.39 (23.10)	[146.93, 237.84]	148.45 (26.28)	[96.75, 200.15]
Low Autonomous Motivation	Baseline	145.28 (39.33)	[67.89, 222.66]	155.69 (40.36)	[76.28, 235.09]
	8 weeks	150.85 (37.63)	[76.81, 224.89]	122.51 (36.44)	[50.82, 194.21]
*Medium Autonomous Motivation	Baseline	165.99 (21.98)	[122.78, 209.21]	125.50 (25.53)	[75.28, 175.71]
	8 weeks	181.35 (21.41)	[139.24, 223.46]	131.54 (26.51)	[79.40, 183.68]

Note

^*denotes exploratory analyses

Next, separate models were conducted for each of the motivational levels. Among those with high autonomous motivation, the matched (Challenge-focused) program was associated with an average increase of 23.41 minutes compared to an average decrease of 9 minutes in the non-matched (Rewards-focused) program. The time*treatment effect yielded an effect size of Cohen’s d = 0.85. Alternatively, among those with low autonomous motivation, the matched (Rewards-focused) program was associated with an average decrease of 33.18 minutes compared to an average increase of 5.57 minutes in the non-matched (Challenge-focused) program. The time*treatment effect yielded an effect size of Cohen’s d of 1.06. Among those with medium autonomous motivation, the Challenge-focused program was associated with an average increase of 15.36 minutes compared to an average increase of 6.04 minutes in the Rewards-focused program. The time*treatment effect yielded an effect size of Cohen’s d of 0.22).

Fitbit Results

Changes in Average Daily Steps During the Intervention

From the beginning to the end of the 8-week intervention period, participants in the Challenge-focused program demonstrated a greater increase in their average daily steps than the Rewards-focused program (M_Change of 1899.44 vs. 177.43) (see Table 5). Among those with high autonomous motivation, the matched (Challenge-focused) program maintained a higher average daily step count across the 8-week program than the non-matched (Rewards-focused) program (Grand M of 10100.24 vs. 8341.14), but the average change across the intervention period was similar across both programs (Challenge: M_Change=354.38 vs. Rewards: M_Change= 368.265). Alternatively, among those with low autonomous motivation, the non-matched (Challenge-focused) program demonstrated a greater increase in daily steps than the matched (Rewards-focused) program (Challenge: M_Change=3803.63 vs. Rewards: M_Change= −152.53). Finally, those with medium motivation also demonstrated a greater increase in average daily steps in the Challenge-focused program (Challenge: M_Change=1777.84 vs. Rewards: M_Change=750.844).

Table 5.

Average daily steps during 8-week intervention period measured with Fitbits.

	Across all motivation				Low Autonomous Motivation				Medium Autonomous Motivation				Hiah Autonomous Motivation
	Challenge (N = 36)		Rewards (N = 29)		Challenge (N = 10)		Rewards (N = 12)		Challenge (N = 14)		Rewards (N = 7)		Challenge (N = 12)		Rewards (N = 10)
	M	SD	M	SD	M	SD	M	SD	M	SD	M	SD	M	SD	M	SD
Week 2	8841.27	3739.45	7550.50	3341.09	7194.42	5003.33	7246.24	3667.55	9073.65	3318.94	6942.91	2405.22	10042.64	2513.69	8340.92	3650.68
Week 3	9737.26	3038.40	7993.85	2584.53	8816.39	3366.65	7963.23	3170.48	9633.39	3047.56	8328.86	2316.04	10706.62	2686.42	7796.08	2188.10
Week 4	9947.29	3390.19	7672.07	2587.77	9639.64	3264.78	7651.97	3414.50	9944.60	3647.41	7912.67	1847.26	10206.60	3483.08	7511.74	1982.54
Week 5	9753.60	3167.22	8755.01	2552.06	9523.80	4191.07	8463.38	3307.76	9985.01	3052.82	8826.57	1928.24	9675.12	2531.66	9088.20	1928.24
Week 6	9494.15	2745.80	8181.28	2807.92	9212.51	3016.78	8098.55	3286.53	9672.82	2657.89	8441.57	2705.35	9520.38	2842.63	8089.15	2487.96
Week 7	10820.21	4431.49	7897.41	2984.09	12201.99	4885.24	7029.43	3020.19	10501.25	3676.46	8292.26	3447.97	10040.86	4945.38	8747.60	2552.57
Week 8	11148.88	4016.26	8072.56	2591.32	12465.68	4715.98	7338.20	2575.93	11010.74	3382.73	8904.36	2677.01	10212.71	4128.80	8446.26	2568.17
Week 9	10740.71	5143.68	7727.93	2836.89	10998.05	5393.03	7093.71	2390.41	10851.49	4884.08	7693.75	3142.19	10397.02	5655.99	8709.18	3264.83
Grand Mean	10060.42		7981.33		10006.56		7610.59		10084.12		8167.87		10100.24		8341.14
Mean Change (W9-W2)	1899.44		177.43		3803.63		−152.53		1777.84		750.84		354.38		368.27

Note. Fitbit data was not collected during week 1 of the program, as this week was used for collecting baseline accelerometry data

Fitbit Mobile App Communication (Challenge-focused program only)

In the Challenge-focused program, the majority of participants used the messaging feature on the Fitbit mobile app, with 78.4% of participants sending at least one message during the 8-week program. They were actively engaged in messaging on the app (defined as sending ≥ 1 message/week) for an average of 2.54 weeks (SD = 2.08). Participants exchanged messages at a similar rate, with an average of 9.78 (SD = 19.29) messages sent and 9.76 (SD = 15.61) messages received over 8-weeks. Participants were in contact with an average of 2.46 groups members (SD = 2.24), with group sizes ranging from 7 to 10 members. In a follow-up analysis, we examined whether group communication on the app (i.e., the total number of group members contacted during the program) was associated with post-intervention total PA. A linear mixed effects model (with baseline total PA as a covariate and random intercepts and random slopes for treatment group and subject) revealed that communicating with more group members was associated with greater total PA at post-intervention, Estimate = 10.99, SE = 3.44, p = .004 (see Supplementary Table S4).

Discussion

The present study tested whether receiving a motivationally-matched (vs. non-matched) intervention resulted in a clinically-meaningful increase in total daily PA among African American women. For those with high autonomous motivation, the matched (Challenge-focused) program was associated with a greater increase in total daily PA than the non-matched (Rewards) program. However, contrary to expectations, those with low autonomous motivation also demonstrated a greater increase in total PA in the (non-matched) Challenge-focused program than the matched (Rewards-focused) program. In the Challenge-focused program, those with medium to high autonomous motivation met the a priori criteria of increasing their total daily PA by ≥ 10 minutes, whereas those with low autonomous motivation demonstrated a smaller increase of 5.57 minutes (vs. a decrease of 33.18 minutes in the Rewards-focused program). This may suggest that those with low autonomous motivation require a longer intervention period or greater dose of the Challenge-focused intervention. Additionally, the Challenge-focused program was associated with a greater increase in daily steps (for those with low to medium autonomous motivation) or greater maintenance (for those with high autonomous motivation). In summary, although the motivational-matching approach was not supported, this study provides preliminary support that the Challenge-focused program is a promising approach for engaging AA women in greater PA.

An important question to consider is why the motivational-matching approach did not yield the hypothesized results. Although we reasoned that extrinsic rewards and building competency through partner-based support would be an effective strategy for engaging those with low autonomous motivation, overall participants in the Rewards-focused program were less engaged, as evidenced by somewhat lower attendance and higher attrition. Both interventions exceeded a priori goals for attendance, dose, and most fidelity components, which helps to rule out an alternative explanation that the results were due to poor program implementation. Additionally, in the Rewards-focused program, participants reported feeling supported by their partner and frequent levels of contact, suggesting that the partner-based component was implemented as planned. The incentives in the Rewards-focused program were relatively small which may have impacted the results, but past behavioral interventions have successfully used small incentives to promote behavioral goals (Leahey et al., 2017). Thus, while the process evaluation data indicate that the Rewards-focused program was delivered as planned, the results suggest that those with low autonomous motivation did not benefit from the motivational-matching approach.

The Challenge-focused program may have been more effective across motivational levels due to the focus on collective goal-setting and social connectedness. Previous PA interventions for AA women have included goal-setting and targeted social support (Jenkins et al., 2017). However, collective (rather than individual) goals and the use of technology to further enhance group communication outside of in-person sessions have rarely been implemented in past interventions for AA women. Collective outcomes (vs. individual performance) may be a more culturally-congruent approach for AA women, given the importance of collectivism among some AAs (Carson, 2009; Nobles, 1991). Furthermore, collective goals and positive group interactions are theorized to be key strategies for enhancing group cohesion (Estabrooks et al., 2012). This may be important because group cohesion is a significant predictor of intervention attendance (Smith-Ray et al., 2012), which, in turn, is critical for achieving high dose. In summary, regardless of baseline motivation, AA women may be more likely to benefit from an intervention approach that targets collective goals and social connectedness due to the potential cultural relevancy of these strategies.

Best Practices for Engaging AA Women in Future Interventions

There are several lessons learned from this study that can be used to inform best practices for engaging AA women in future behavioral interventions. First, we found that using community-engaged strategies, including integrating community feedback via focus groups, collaborating with a community partner, and implementing the study in a community-based setting, was essential for achieving high enrollment, attendance, and retention. Previous research has highlighted lack of trust, transportation, and time as prominent barriers to research participation among AA communities (Otado et al., 2015). The present study was implemented in a community center with a free after school program. Many of our participants were caregivers of youth who participate in this afterschool program, and as result they were already at the community center on a regular basis. This made participation in the study relatively convenient, especially given that our study team provided childcare during the group sessions to reduce barriers related to caregiving responsibilities. We propose that to reduce barriers to participation one best practice for future interventions is to deliver programs in partnership with trusted community leaders and within community settings that are familiar and convenient for AA women.

Second, future studies may benefit from targeting positive intragroup competition. Lack of a partner, social support, and exposure to other physically active AA women are frequently reported barriers to PA (Harley et al., 2009; Joseph et al., 2015). The present study offered a unique opportunity to compare different approaches to addressing these social-motivational barriers. While the Rewards-focused program targeted partner-based social support, this approach was not as impactful as the Challenge-focused program. One novel component of the Challenge-focused program was the focus on positive intragroup competition. Previous studies have found that compared to men, women demonstrate increased effort and enjoyment when competing as a group, rather than individually (Wittchen et al., 2013). Despite converging evidence highlighting the value of positive intragroup competition, past PA interventions with AA women have primarily focused on individual-based outcomes and non-competitive approaches to PA, such as walking (Bland & Sharma, 2017). Thus, another best practice for engaging AA women in future behavioral interventions it to integrate opportunities for positive intragroup competition (e.g., through collective goal-setting, intragroup activities).

Third, we found that communication on the Fitbit mobile app was associated with greater PA at post-intervention in the Challenge-focused intervention. While previous studies have focused on the use of Fitbits for self-monitoring and PA adherence (Hartman et al., 2018), there is growing evidence that combining fitness trackers with a mobile/online group platform increases PA (Zhang & Jemmott Iii, 2019). Some studies have found low engagement in group communication when integrating a mobile app with a fitness tracker (Tong et al., 2019). However, such studies have tended to use individual (vs. team-based) goals and have not included in-person contact with group members as in our Challenge-focused program. Thus, future interventions may benefit from pairing in-person sessions with a mobile app that allows for shared goal-setting and ongoing group communication to further strengthen social connectedness among AA women.

Strengths and Limitations

There are several strengths of this study, including high attendance and retention rates, objectively-measured PA, and a focus on an underrepresented population. However, it is important to acknowledge that the study used a small sample size, which limits generalizability. Additionally, there were a handful of participants who knew each other prior to enrolling in the program, which may have impacted the study results to some degree. Furthermore, although our baseline measure of autonomous motivation yielded scores similar to past studies with African American women (Landry & Solmon, 2004), the present study did not include individuals with very low levels of autonomous motivation. Individuals with very low autonomous motivation are challenging to recruit into PA interventions as they often do not view exercise as important. Future research is needed to identify the best practices for reaching and engaging those with very low motivation.

Another strength of the study was the use of a community-engaged approach and a multi-theoretical framework. Eliciting community feedback and directly involving community partners in the study design is critical for enhancing the feasibility and acceptability of interventions (Coughlin & Smith, 2016). This study likely would not have been possible without the support from our community partner. However, it should be acknowledged that the integration of community feedback with existing theoretical models added a layer of complexity that in some cases may have undermined efforts to develop two motivationally distinct programs. For example, although different strategies were used (team-based approach targeting social connectedness vs. partner-based approach targeting competency), both programs included social components that may have enhanced relatedness among group members. Thus, this study highlights some of the challenges that arise from developing interventions that are both theoretically-grounded and community-engaged. We recommend that researchers continue to prioritize community feedback in intervention design, be systematic in mapping how community input and theory are used to inform program elements and anticipate how to resolve potential challenges with integrating these approaches (e.g., using a community steering committee and an expert panel).

Important future directions for this program of research include: comparing the Challenge-focused program to an appropriate contact control, evaluating the impact of the intervention on long-term maintenance of PA, using additional strategies to strengthen community engagement (e.g., involve community health workers/AA women as facilitators, community advisory board), evaluating other outcomes (e.g., blood pressure, weight loss, stress), and integrating implementation science models to promote sustainability. For example, given the novel finding that communication with group members on the Fitbit app predicted greater PA, an important next step will be to evaluate the best practices for sustaining engagement with the mobile app, including communication and collective goal-setting among group members.

Conclusions

Although the motivational-matching approach was not supported, this study provides promising preliminary support for a novel PA intervention for AA women. These results will be used to optimize the Challenge-focused program before proceeding to a full-scale efficacy trial. Racial and ethnic minorities remain majorly underrepresented in behavioral interventions (Haughton et al., 2018), and thus there is a critical need to continue to identify the best practices for reducing barriers to participation and promoting health behavior change among minority communities. The lessons learned from this pilot study, including demonstrating the value of delivering interventions within community-settings, integrating opportunities for positive intragroup competition, and pairing in-person sessions with a mobile app to strengthen social connectedness, offer a useful framework for future researchers aiming to engage AA women in behavioral interventions.

Public Health Significance.

Across differences in motivation, a team-based model is a promising approach for promoting PA among AA women. Delivering programs within community-based setting, using positive group competition, and a mobile app to promote social connectedness are suggested best practices for engaging AA women in future interventions.

Availability of Data and Materials.

The dataset used during the current study is available from the corresponding author on reasonable request.