The remotely delivered lifestyle activity intervention was guided by social cognitive theory and self-determination theory and used motivational interviewing (MI; Figure 2). Social cognitive theory emphasizes self-efficacy, behavioral capability (knowledge and skills to perform a behavior), self-monitoring, and outcome expectancies to promote behavior change [65,66]. Self-determination theory posits that intrinsic versus extrinsic motivation leads to a greater commitment to behavior change and maintenance of change [67]. Intrinsic motivation is more likely when 3 core psychological needs are met: autonomy, competence, and relatedness [67,68]. MI is a person-centered method of communication to help resolve ambivalence related to behavior change, enhance intrinsic motivation to change, and brainstorm ways to overcome barriers [69].

The goal was to encourage participants to be more active throughout the day—a whole-of-day approach—without regard for minimum intensity or duration of activity bouts. Walking, important for maintaining mobility and independence, served as the primary form of LPA. The primary behavioral target was to reach and maintain an extra 3000 steps per day above baseline (established during the second week post enrollment). This target represents approximately 30 extra minutes of LPA [70] and is associated with health benefits [29,71]. The secondary behavioral target was to increase the number of active h/d (defined as >250 steps/h in Fitbit) to at least 10 h/d to encourage activity throughout the day and to interrupt and replace sedentary time. Frequent interruption of sitting is associated with better physical function [36,72-74]. To promote gradual and sustained change, participants were asked to increase the number of steps per day during weeks 3-12 (Table 1). Participants were asked to “practice” achieving their new step goal on at least 3 days during the first week and then at least 5 days during the second week. This allowed participants to have one or more low-activity days and still be able to achieve their step goal. In accordance with self-determination theory and MI, participants chose their own strategies for becoming more active and achieving their weekly goals with guidance and (technology) support from their health coach.

A Fitbit activity tracker and mobile app were used to promote awareness and enable self-monitoring of both activity (steps per day) and inactivity (activity tracker vibrates as a reminder to move each hour). During week 2, participants in the intervention group were mailed the Fitbit Inspire 3 (Google) activity tracker, with instructions for installing the app on their smartphone and for using the Fitbit app. The distinguishing feature of this tracker is an activity reminder, in which the tracker gently vibrates if fewer than 250 steps were taken in the previous 50 minutes and provides a suggestion to take more steps in the next 10 minutes (to achieve a goal of ≥250 steps/hour). Data from the tracker were wirelessly synced to the app. Participants were instructed to wear the tracker during waking hours and were encouraged to check their activity by their app at least once a day. The app includes a daily summary of the total steps and number of active hours (hours with ≥250 steps). The health coach assisted with the installation, reviewing instructions for using the tracker and app, setting goals (steps per day, number of active hours), and checking their progress. If needed, a family member or close friend was included in telephone calls to assist the participant.

Bilingual research assistants were trained in MI core processes (engaging, focusing, evoking, and planning) and microskills (open-ended questions, affirmations, reflections, and summaries) to promote behavior change and use in their role as health coaches. MI techniques and microskills were used to guide 5 sessions of telephone-based health coaching. While these calls were designed to occur during weeks 3, 5, 7, 9, and 11 (45 minutes for initial 1-2 calls, then 15-20 minutes each), participants were allowed to reschedule due to illness, etc. Key intervention topics and techniques included establishing rapport; eliciting knowledge about the health benefits of physical activity and interrupting sedentary time; assessing motivation, confidence, and perceived importance and barriers to becoming more active; eliciting change talk; and enhancing participants’ self-efficacy (competence) and motivation to be more active (eg, small, achievable goals). The health coach also provided tech support to the participants for setting up their Fitbit and using their mobile app. The health coaches regularly checked Fitabase (Small Steps Labs LLC) and sent reminders (telephone, email, or text, depending on the participant’s preference) if participants had not synced their Fitbit in 4 or more days or were not wearing their Fitbit (limited or no steps recorded).

Participants were mailed visual aids to serve as a reference for discussions with the health coach. The health coaches followed a semistructured guide that corresponded to the visual aids. The initial call included a discussion about the number of daily steps associated with good health in older adults (7000-9000) [75-77] and the benefits of spreading out activity throughout the day (ie, breaking up sedentary time through at least minimal stepping each hour). Participants were asked to rate the importance of meeting the recommended level of daily steps and to provide reasons why it was important for them to be physically active; examples for participants to consider include reducing stress and anxiety, increasing energy levels, enhancing balance, improving sleep quality, etc. After reviewing the baseline Fitbit data, participants were provided with the suggested goals (daily steps and active hours) for the next 2 weeks (Table 1) and examples for adding 1000 steps per day (one 10-minute or two 5-minute brisk walk, one 15-minute leisurely walk, or adding at least an extra 150 steps for 6 hours). The participants identified strategies and created a plan for achieving these goals. Examples for participants to consider included playing music and dancing, walking the dog, parking further from an entrance, etc. While examples were provided, participants were encouraged to choose their own strategies. The calls ended with a review of how to set the step goal and the reminders to move in the Fitbit app. Subsequent calls reviewed the previous 2 weeks and created plans for meeting subsequent goals. For example, participants who achieved their step goal during the previous week might have chosen the study recommended goal (increase by 500 steps), whereas participants who did not meet the goal (eg, due to illness) might have chosen to maintain their previous step goal. Participants may have kept their same strategies or identified new strategies for achieving their goals and then created a plan for the next 2 weeks. Table 2 provides examples of the use of MI and visual aids by the health coaches in the theoretically-guided approach.

We used evidence-based MI training procedures identified in previous studies [78,79]. The health coaches received an initial 1.5 days (12 hours) of in-person MI training by an experienced MI trainer (STW). Before interacting with participants, the coach completed 4 practice sessions and received feedback through 4 phone supervision sessions. These sessions were spread over 6-8 weeks to allow for further review of materials between feedback sessions. STW coded the sessions for MI fidelity by using the MI Treatment Integrity Coding System (version 4.1) [80] and used this information during supervision sessions. During the study period (15 months), each health coach submitted 1 audio recording and participated in 1 phone supervision per month (months 1-3), then every other month (months 4-15). Health coaches were provided with training on the intervention design, measures, and outcomes. Fidelity to the intervention delivery was ensured through the use of the operations manual, spreadsheet software for tracking day-to-day activities, and weekly team meetings.

A waitlist control was selected to help boost retention [81] and balance validity, cost, and time. Participants were asked to maintain their usual lifestyle during the initial 12 weeks. Upon completion of the 13-week assessment, waitlisted participants were mailed a Fitbit Inspire 3, instructions, and visual aids and received the full intervention (Table 1).

To compensate participants for their time, a US $50 merchandise card was provided for each outcome assessment (US $150 total). In addition, both groups were permitted to keep the Fitbit activity tracker after the study ended to maintain activity (at the time of the intervention, a US $100 value).

Data were collected by mailed packets (activity monitor and questionnaires) and videoconferencing (physical performance tests) at baseline, immediately after the intervention, and at longer-term follow-up (3 months after intervention). The assessment packet included an activity monitor (for objective measurement of steps, LPA, and sedentary behavior), questionnaires, and print instructions for attaching the activity monitor and for completing the remote assessment of physical function. In addition, participants were emailed links to the YouTube videos of the instructions. Participants were instructed to wear the activPAL (PAL Technologies Ltd; using 24-hour protocol) activity monitors [82] for 1 week, consistent with other studies [40,83,84]. At the end of the week, participants completed questionnaires on both sedentary behavior and physical activity to provide context to the objective data. Instructions were provided for returning (through self-addressed stamped mailer) the monitor and questionnaires to study staff. The health coach called to verify receipt of the assessment packet, review the instructions, if necessary, answer any questions, and schedule the remote assessment of physical performance. Participants were provided an option to complete the study surveys online through a REDCap database. Study data were collected and managed using REDCap electronic data capture tools hosted at the University of New Mexico [85]. REDCap is a secure, web-based application designed to support data capture for research studies, providing (1) an intuitive interface for validated data entry, (2) audit trails for tracking data manipulation and export procedures, (3) automated export procedures for seamless data downloads to common statistical packages, and (4) procedures for importing data from external sources [85].

Fitabase (Small Steps Labs LLC) was used to collect data from participants’ Fitbit trackers. The data collected by Fitabase included total daily steps, number of steps per minute, estimated energy expenditure, distance moved, minutes of vigorous, moderate, and light activity, minutes of sedentary time, sleep length and quality (if Fitbit worn overnight), heart rate, and manually entered and automatically detected physical activities such as walking or running. Fitabase also indicated the last day and time each study Fitbit tracker was synced. Each study participant was deidentified with a participant ID.

Based on our earlier work [58], physical performance tests were conducted in the home setting and remotely assessed using videoconferencing software. Participants used their smartphone, tablet, or camera-enabled computer for the remote assessment. If a participant did not have a suitable device, the study provided a smartphone to capture the physical performance tests. Both written and video instructions were provided that demonstrated the performance tests, safety measures, and how to prepare for the videoconference session with the study investigator. The investigator (TB) performed the assessment from a remote location using videoconferencing software (Zoom [Zoom Communications, Inc] [86] or Skype [Microsoft] [87]; participant’s choice). During the videoconference, the investigator reviewed the instructions for the tests and timed the tests. A separate software (Snagit [TechSmith]) run on a study computer was used to record the assessments for later viewing for quality control purposes.

Although primary enrollment and data capture activities have been completed, final analyses have not yet been completed. For this feasibility study, our goal was to meet the accrual target of 64 inactive, older cancer survivors within 18 months, retain 80% of the sample, and achieve 80% adherence and fidelity to the intervention while minimizing the number of adverse events. Retention will be calculated as the percentage of participants who completed the final assessment. Adherence to the intervention will be calculated as (1) the percentage of participants who complete all 5 health coaching calls and (2) the percentage of days that participants wore their Fitbit during the intervention. Fidelity to intervention delivery will be calculated as the average score on global ratings for MI (scale of 1 to 5; higher scores are better), which includes cultivating change talk, softening sustain talk, partnership, and empathy. The number of adverse events that were attributable, possibly attributable, or not attributable to the intervention were tracked and reported. Satisfaction with the intervention will be evaluated by a 13-item survey that we developed to assess the level of agreement or disagreement (5-point Likert scale, ranging from strongly disagree=1 to strongly agree=5) on questions about the Fitbit activity tracker, the Fitbit app, and the health coach. Example questions include “The Fitbit activity tracker made me more aware of my activity and inactivity during the week,” “The Fitbit mobile phone app was easy to use,” and “Talking to a Health Coach helped me to identify strategies to be more physically active (get more steps).” Satisfaction with the intervention will be calculated as the average score.

The measures and a brief description for each primary outcome, secondary outcome, safety assessments, covariates, and theoretical constructs are provided in Table 3.

Using a 2-sided, 2-sample t test, with a significance level of .05 and 25 people per arm, we estimated that we would have 80% power to detect a mean difference of 8.1 (SD 10) in physical function scores (a large effect size). To allow for 20% attrition, we planned to enroll 32 participants per arm. This feasibility trial will provide estimates of within-group SDs and between-group differences of physical function scores for the target population. These results will inform the design (sample size, effect size) of a larger trial powered to examine the efficacy of the intervention.

Outcome variables violating the normality assumption will be transformed before analyses. The pattern of missing data will be evaluated [114], and the appropriate procedures for multiple imputation will be used. Primary analyses will be performed according to the intent-to-treat principle; as this is a feasibility study, complete case analyses will also be explored. activPAL data will be downloaded using PAL technologies software (version 8) [115] and analyzed using SAS (version 9.4, SAS Institute) and R software (version 3.4.3; R Foundation for Statistical Computing). Raw and event data (start and stop time for sitting or lying, standing, and stepping) will be processed using standard methods, for example, determining valid wear time [116-118], threshold values for activity intensity [119,120], etc. A valid day is defined as 10 or more hours of wear during awake time. Consistent with the intention-to-treat principle and similar to other recent trials [53,54,121], participants with at least one valid day of activPAL data at baseline will be included in the analyses. If needed, a sensitivity analysis will be conducted that will exclude participants with fewer than four valid days of activPAL data. Quality controls (daily logs and visual examination of heat maps) will be used to check the classification of accelerometer data. Accelerometer methods will be published according to guidelines [122-124].

We hypothesize that physical function scores will improve in the intervention group compared with the control group. We will calculate means and precision estimates for between-arm differences in physical function scores at each time point. A linear mixed-effects model will be used with group, time, and group-by-time interaction, adjusted for ethnicity, rural or urban status, and baseline values as needed. The appropriate variance-covariance structure will be selected to account for the potential correlation among repeated measurements from the same participant. The primary test of interest is the group-by-time interaction, indicating whether there was a difference between groups in the primary outcome over time. Similar analyses will be conducted for secondary outcomes.

Acceptability and satisfaction related to the intervention will be evaluated by a mailed survey. Process data and other study-related outcomes will be reviewed to identify the next steps and strategies to ensure the most appropriate design and methods for the future, larger RCT, with a potential for broad dissemination and implementation.