Few efforts to increase PA in this population have been grounded in theory, yet the literature on PA promotion in the general population has yielded rich insights from theoretical models of health-related behavior change. The ingredients used in the current intervention were derived from a highly influential model: the COM-B framework of Michie and colleagues (27). This framework, which has been called a “meta-theory (28),” has been rigorously developed and validated in extensive research on health behavior (29–33) and was strongly supported in a review of major theoretical approaches to PA promotion (34). According to this model, voluntary behavior and changes in behavior are a function of three factors: Capability, Opportunity, and Motivation. Capability refers to both cognitive/ psychological capacity, e.g., knowledge about why and how to perform a desired behavior and remembering to do it, and the physical ability to perform it. Opportunity means access to necessary social and physical/ logistic resources, including space, time, and social support. Motivation refers both to conscious/ reflective factors affecting drive and reward, and reflexive/ automatic motivation (e.g., priming) that may enhance habit formation (27).

Within this theoretical framework, we designed the intervention to accommodate a wide range of individual choices and physical capabilities. In this regard we were guided by the work of Clanchy et al. (24, 25) as well as research showing a dose-response relationship between PA increase and health benefit—meaning that it is not necessary to meet the standard recommendations for PA levels to experience improvements in mood, energy, and quality of life, as well as health risk reduction (35).

Studies have revealed that personalized PA programs can successfully be delivered via remote means including phone, web, and text messages (36, 37). A recent review focused on smartphone-delivered PA promotion for mental health emphasized the importance of including education about the value of PA and personalized coaching within the intervention (38). There is a burgeoning literature showing that persons with msTBI, even those with significant neurocognitive impairment, can successfully engage in, and benefit from, interventions delivered via phone (39–41) and other remote methods. Mobile health (mHealth) interventions, such as smartphone apps and interventions delivered by text, are increasingly being used in TBI rehabilitation and research (42, 43). As described below, the current study provides both remote therapist contact and mHealth support from a chatbot called RehaBot, which was developed in our laboratory specifically for the needs of participants with TBI. In a feasibility study, participants with msTBI found RehaBot both easy to use and enjoyable (44).

There is considerable evidence that self-report measures of PA are less reliable than objective measurement using accelerometers worn on the body (45); the latter method has been shown to be reliable in persons with TBI (46). In this study, our primary outcome is increase in PA from pre- to post-treatment, measured with a wrist-worn accelerometer that has been successfully used with the msTBI population (47).

As described below, we include a follow-through phase to assess the effects of the intervention after therapist support is withdrawn, with randomization either to no continued support, or continued use of RehaBot (without therapist intervention). We also include treatment ingredients that have been shown to help sustain new PA habits, and less sedentary behavior, as part of daily life (36, 48).

Throughout intervention development, consumer input was obtained through multiple structured methods. Two paid consumer consultants with lived experience of msTBI and self-reported difficulties with PA were engaged in this process. Input was collected through facilitated focus group sessions where consultants reviewed and provided feedback on the treatment rationale and session structure; visual materials and participant handouts; and the use of RehaBot to deliver just-in-time support for PA in participants' daily environments. Additionally, consumer consultants participated in pilot testing of treatment sessions, providing feedback on feasibility, acceptability, and comprehensibility of intervention components. This iterative feedback was incorporated into the final treatment manual. Consultants also selected the name of the intervention: GetUp&Go.

The aim of this study is to examine the efficacy of the GetUp&Go intervention for increasing PA and enhancing associated functional outcomes in msTBI. We propose the following hypotheses:

1a. Participants who receive the GetUp&Go intervention directly after randomization will show significantly more increase in PA, as assessed by average accelerometer activity counts/ day, from pre-randomization baseline to post-treatment 10 weeks later, compared to those waitlisted for 10 weeks.

1b. Participants randomized to immediate treatment (IT) will also show significantly more improvement compared to waitlisted participants (WL) on secondary outcome measures including intensity of PA (derived from accelerometer readings), self-reported PA, emotional function, fatigue, sleep, pain, and health-related quality of life.

2a. Participants randomized to continued use of RehaBot for 10 weeks after treatment (RB) will demonstrate significantly less diminution in accelerometer-based activity counts from post-treatment to 10-week follow-up compared to those for whom RehaBot is withdrawn (No Tx), adjusted for the degree of change in the initial treatment phase.

2b. Participants randomized to RB in the follow-through phase will demonstrate significantly less diminution/ worsening in other outcome domains listed under Hypothesis 1b, compared to No Tx participants.

In addition to testing these hypotheses, we will explore the predictors of treatment response, using pooled data from all participants from pre- to post-treatment. We will analyze in exploratory fashion the predictive effects of demographic and injury characteristics; baseline cognitive, physical, social, emotional, and personality factors; objective and perceived neighborhood characteristics; intrapersonal variables such as motivation, self-efficacy, outcome expectancies, and identity/ values as related to PA; and process variables such as the amount of interaction with RehaBot during treatment.

This is a randomized waitlist-controlled trial, registered prior to participant recruitment at https://www.clinicaltrials.gov (NCT06028334). As shown in Figure 1, after a baseline assessment, participants are randomized 1:1 into IT (10 weeks of GetUp&Go intervention) or WL (10 weeks on waitlist), after which the primary outcome is measured. WL participants then receive 10 weeks of GetUp&Go. The 10-week treatment period (for all participants) is termed the Acquisition or A phase. Directly after the A phase, participants are re-randomized 1:1 into the Follow-Through (FT) phase: 10 additional weeks, of either RB or No Tx. A final outcome assessment and a debriefing interview are conducted at the end of the FT phase.

Participants will be 70 persons who meet the following inclusion criteria: (1) age ≥18; (2) TBI (open or closed), sustained at least 6 months prior to enrollment, of at least moderate severity as evidenced by: (a) loss or alteration of consciousness ≥30 min and/ or post-traumatic amnesia (PTA) ≥ 24 h, not due to intoxication/ sedation and documented prospectively from the injury (i.e., not retrospectively self-reported); and/ or (b) positive neuroimaging findings consistent with TBI; (3) fully weight bearing on lower limbs and able to walk indoors and outdoors without the assistance of another person (independent use of an assistive device such as a cane is acceptable); (4) cognitively able to participate in treatment, using selected items from the Glasgow Outcome Scale-Extended (49); (5) able to communicate adequately in English for participation in the treatment protocols; (6) informed consent given by participant. Participants are excluded for (1) contraindications to increasing PA as judged by a study physician and/ or study Physical Therapist, using an exam adapted from published screening tools (50); (2) medical or psychiatric instability, including current psychosis or severe uncontrolled substance misuse, as assessed using items from the Mini-International Neuropsychiatric Interview (51), or suicidal ideation with intent or plan, as assessed by the Columbia-Suicide Severity Rating Scale, screening version (52); (3) significant physical or intellectual disability predating the TBI; (4) neurodegenerative disorder, e.g., Parkinson's disease, dementia; (5) insufficiently inactive, i.e., reporting >23 weekly moderate/ vigorous activity units on the Godin Leisure-Time Exercise Questionnaire (53); (6) planned surgery or other hospitalization during the succeeding 9 months; (7) physical or sensory disability (e.g., blindness; severe bimanual incoordination) that prevents use of a smartphone.

The measures collected at each time point are displayed in Table 1. The baseline evaluation (T1) includes psychometric measures and questionnaires as well as demographic and injury-related information collected from medical records and participant interview. Because we have found that primary medical chart information is variable for people with chronic TBI, we measure PTA duration with a structured interview that has been used in other studies and that correlates well with prospectively measured PTA (54). The primary outcome (average activity counts/ minute) and other objective PA outcomes are measured using the wrist-worn Actigraph GT3XP. Activity counts using the GT3XP correlate well with biometric measures of energy expenditure (55). Following a practice prevalent in the PA measurement literature (56, 57), daily activity counts are averaged only from a minimum of 4 full wear days out of each 7-day wear period; if 4 days of data are not recorded, the wear period is repeated or discarded.

This study is approved and overseen by an Institutional Review Board (iRISID-2023-1533) and complies in full with the Helsinki Declaration for the protection of human subjects.

Participants are recruited from a variety of sources including two research registries at the study site, as well as clinical programs at the site and others in the geographical vicinity. Recruitment began in December, 2023 and is expected to continue until December, 2026. Prospective participants are screened for eligibility over the phone after expressing interest in the study and assenting to screening. Eligible participants are invited to an in-person session that includes written consent, administration of T1 measures listed in Table 1, and screening by both the study physician and the study Physical Therapist. These screenings include a review of systems and assessment of vital signs, lower extremity range of motion, and mobility. Exercise tolerance is evaluated using the modified Bruce protocol (74). Following the physical/ medical screening and the T1 evaluation, participants are given the Actigraph and instructed to wear it continuously for 7 days as they go about their normal routines in the home and community. Participants are given postage-paid padded envelopes for returning the devices.

After all baseline data including the Actigraph data are collected, participants are randomized 1:1 to IT or WL. Randomization uses a sequence of permuted blocks randomly selected from sizes 6, 9, and 12. The randomization sequence was created by the study statistician and is kept in a securely housed spreadsheet which kept all but the current treatment assignment obscured from view. To maximize equipoise and minimize attrition, participants are informed that the randomization is to “one baseline evaluation” or “two baseline evaluations 10 weeks apart.” WL participants receive periodic contacts from research staff reminding them of their upcoming activities in the program. After this initial 10-week period, the primary outcome is measured with another 7-day Actigraph wear period, followed by telephone administration of the outcome measures listed in Table 1.

When all post-treatment data are collected (at T2 for IT and T2b for WL), participants are re-randomized 1:1 to 10 weeks of RB or No Tx (see Figure 1). T3 data are collected after that interval and as a final step, the participant receives a 15-minute debriefing interview by phone. This interview includes the Patient Global Impression of Change scale (75), worded as follows: “Since your participation in this study, how would you describe the change (if any) in areas of your life that could be related to physical activity and fitness, such as physical and mental well-being, mood, and overall quality of life?” Responses are on a 7-point scale from “No change (or condition has gotten worse)” to “A great deal better and a considerable improvement that has made all the difference.” Other questions are concerned with features of the treatment and the RehaBot app that the participant particularly liked or disliked.

This trial uses masked outcome assessment, with special precautions taken to prevent inadvertent unmasking. These include making treatment information completely inaccessible to data collection staff, and using a “script” reminding participants not to discuss any of their experiences in the program when contacted by a data collector. Any instances of inadvertent unmasking are recorded for later analysis of their influence, and data collectors are asked to guess the participant's treatment allocation following each assessment.

Adverse events (AEs) are documented systematically throughout the trial. AEs are defined as any of the following: 1) Events that initiate the risk management protocol (i.e., expression of suicidal ideation or intent), 2) Events prompting an unscheduled emergency department visit not resulting in hospitalization, 3) Any unanticipated event that is unexpected and related to the study treatment. Serious Adverse Events (SAEs) are defined as outcomes resulting in death, hospitalization, life-threatening circumstances, person at risk of death at the time of the event, disability and/or incapacity, or events requiring intervention to prevent these outcomes. All hospitalizations are considered SAEs, with the exception of planned hospitalizations for elective procedures unrelated to the study intervention. AEs and SAEs are tracked through multiple surveillance mechanisms: spontaneous reporting by participants, discovery by study staff, identification by clinicians during therapy sessions, or detection during study assessments. When an AE is identified or reported, the therapist/data collector completes an Adverse Event Form immediately. This form documents the adverse event type, duration and timing, event details, and resolution status. AEs are reviewed periodically by a Data Safety Monitoring (DSM) team consisting of a physiatrist with extensive experience in msTBI research, and a statistician with expertise in clinical trials methodology. The DSM team also reviews and makes recommendations on recruitment, attrition, and other data relevant to the conduct of the trial and human subjects protections.

The overall goal of the intervention is to develop and support a personalized plan to increase physical activity and decrease sedentary behavior, using treatment ingredients consonant with the COM-B model (27). The ingredients in each COM-B domain (Capability, Opportunity, Motivation) that are supplied by the therapist and by RehaBot are summarized in Table 2.

The 10-week GetUp&Go intervention is delivered entirely remotely. The first 2 sessions last 60–90 min each and are delivered within the first week over Zoom for Healthcare (or smartphone if needed). The following 3 sessions (in weeks 3, 5, and 8) are delivered by phone and last 30–45 min. The RehaBot app is installed on the participant's phone in Session 2 followed by a guided overview and brief training that includes hands-on practice, troubleshooting, and opportunities to ask questions, until the participant expresses confidence in using the app. Participants who do not own a smartphone or do not have access to a computer/ tablet are issued a phone with unlimited texting and/ or an iPad for the duration of the trial. To maximize scheduling flexibility, participants are assigned to one of 4 therapists: Two doctoral level neuropsychologists, one Masters-trained clinician with extensive experience in neuropsychological approaches to the treatment of msTBI, and one doctoral-level Physical Therapist. All therapists are experienced clinicians with specific expertise in treating persons with msTBI. To ensure fidelity, a detailed manual is used for all treatment sessions. In addition, two therapists were present during treatment sessions for the first four participants in the trial in order to achieve proficiency and consistency in delivering the intervention. All therapists continue to meet weekly to review cases to ensure that their approaches and procedures are comparable.

In Session 1, the therapist provides education on the benefits of PA, emphasizing the dose-response relationship (i.e., the fact that any degree of PA increase is potentially beneficial). In discussion with the participant, the therapist explores in more depth the factors revealed in screening, and they begin to brainstorm a list of possible activities and routines to increase PA/ decrease sedentary behavior within the participant's preferences and opportunities in the home and community environment. The therapist promotes consideration of new activities, existing activities that might be modified or increased, and ways to “sneak in” PA, such as taking extra steps to reach a destination. The therapist also elicits overall goals and specific intentions regarding increased PA and reduction of sedentary time (64, 76), expected outcomes (64), and other statements for later use as personalized motivational reminders, delivered by RehaBot. The capabilities of RehaBot are explained and demonstrated in Session 1 using sample screen shots, to help participants begin to consider how they might benefit from using the app. See Figure 2 for sample screenshots.

As the list of planned activities takes shape, the therapist engages the participant in action planning, i.e., specific plans for when, where, how often, and (if applicable) with whom activities will occur during the week (64, 76–78). Additional discussion involves potential barriers to activity completion, and possible ways of overcoming them. Obstacles may be overcome with various types of coping plans for one-time solutions (e.g., buying suitable shoes for walking or jogging; joining the nearby fitness center), or via just-in-time reminders framed as implementation intentions (76, 79, 80) (e.g., “If I don't feel like taking my walk, I’ll remind myself that….”) that may be accessed through RehaBot.

In Session 2, RehaBot is installed on the participant's phone, the participant is trained on all of its features, and the therapist and participant collaborate on personalizing the app. RehaBot is accessed via regular text-messaging, and engages in bi-directional communication with the user (see Figure 2). Therapists use a portal to enter participants' action plans, coping plans, goals, and other relevant data into the RehaBot app. Message timing is fully customizable. Participants select their preferred times for daily “good morning” and “good evening” check-in messages. They also determine how many activity reminder messages they wish to receive throughout the day and when those reminders should be delivered. Scheduled contacts come in the form of text message reminders to check your plan for the day, carry out a planned activity at a scheduled time, and report on your accomplishments. However, RehaBot is also available on-demand to serve as a “coach” to provide reminders, suggestions, feedback, reinforcement, coping plans, and personalized motivational messages. Activity reminder messages are worded using the participant's own language and selected activities. Participants also personalize the content of activity suggestions and the language and content of their coping plans. When an activity is logged as complete, RehaBot sends encouraging messages that reinforce the effort and tie the participant's accomplishments to progress towards a specific personal goal. Self-efficacy, thought to be a strong mediator of positive behavior change, is a particular target of intervention via specific positive feedback from both therapist and chatbot, and reminders of task mastery and overall progress (81).

In the phone sessions in weeks 3, 5, and 8, the participant's plan and their reactions to it are reviewed. Progress is reinforced, problems and obstacles are addressed, and any desired changes to the plan are instituted in the RehaBot portal. In the final telephone session, the therapist guides the participant in reflecting on their overall progress and their ability to overcome barriers to physical activity, while offering feedback on positive changes observed throughout the intervention. The participant continues to use RehaBot independently until Week 10, when they receive a letter reminding them of the upcoming telephone data collection and Actigraph wear period.

Participants randomized to RB for the FT phase receive a call notifying them that their RehaBot plan will remain in place for another 10 weeks. These participants are encouraged to use RehaBot to continue with their activity plans until the next evaluation period. Those randomized to No Tx are reminded of the final evaluation in 10 weeks and encouraged to keep up their PA plans. Neither FT condition includes any further contact from the therapist, although we would respond to participants who contacted us regarding a technical problem with the app.

For the primary endpoint (Hypothesis 1a: primary hypothesis) and each of the secondary endpoints (Hypothesis 1b), the mean change from T1 to T2 in the IT group will be compared to the mean change from T1 to T2 in the WL group using the two-sided two-sample t-test with alpha = 0.05. The mean change in each group and the mean difference between changes in IT vs. WL group will be estimated with the corresponding 95% confidence interval. If the normal distribution assumption is not appropriate for patient-specific changes in one or both groups, then the two-sided two-sample Wilcoxon test will be used instead of the t-test. We will employ intent-to-treat for the primary analysis, and as such, we will make every attempt to collect T2 outcome data on all randomized participants regardless of treatment received. Per current recommendations (82–84), we will use multiple imputation to account for missing T2 data, and evaluate missingness assumptions (by comparing available data for participants with complete vs. missing T2). If the assumption of the data being missing at random is not appropriate, the imputations will be performed using the methods specifically developed for data missing not at random (85). As a secondary analysis we will perform a complete-case analysis with dropouts excluded. With the proposed sample size of 35 subjects per arm and assuming up to 15% loss to follow-up, we expect to have at least 30 subjects per arm available for analysis. The sample size of 30 subjects per arm provides 81% power to detect the mean difference between changes in IT vs. WL arms corresponding to the effect size of 0.75 (effect size for the primary endpoint as reported in Clanchy et al. (24) using the two-sided two-sample t-test with alpha 0.05.

For Hypothesis 2a, the primary endpoint is the change in average accelerometer activity counts/ day from post-intervention timepoint (T2 for IT participants, T2b for WL) to the follow-up endpoint, T3, 10 weeks later. The subject-specific T2/T2b-T3 changes will be modeled in an ANCOVA model as dependent on group (RB vs. No Tx) and the average accelerometer activity counts/day during the A phase (from timepoint T1 to post-intervention timepoint T2/T2b). The interaction between group and T1-T2(b) changes will be considered and retained in the model, if significant. If the interaction is not significant, Hypothesis 2a will be tested by comparing the mean change in RB vs. No Tx group using the model-based two-sided two-sample t-test with alpha 0.05. In the case of the main effects model (no significant interaction), this is equivalent to testing the main effect of RB (vs. No Tx) as the difference between intercepts of the regression lines with T1-T2(b) change as a predictor. A significant interaction would imply that the effect of RB (vs. No Tx) depends on the amount of change achieved during the A phase, and the test of interaction corresponds to comparing the slopes of the T1-T2(b) change regression lines between RB and No Tx groups. In the case of a significant interaction, the difference between RB and No Tx groups will be quantified in terms of the difference between the T1-T2(b) change regression slopes, and the significance (p-value) for that difference is the same as the significance (p-value) of the interaction. It is not expected that this interaction is important, and the study is not powered to detect such interaction. The secondary outcome measures (Hypothesis 2b) will be analyzed using the same approach.

We will report AEs and SAEs descriptively by treatment arm, including the number and percentage of participants experiencing at least one AE/SAE, as well as the total number of events by category. All AEs will be categorized by severity, relatedness to the intervention, and resolution status.

With regard to power, in the study of Clanchy et al. (24), participants in the active intervention group lost a mean of 1.7 standard deviations in average activity counts per day between their post-treatment and follow-up assessments, without any treatment during this time (comparable to our No Tx condition). We hypothesize that the loss in activity in the RB (treated) group in our study will be reduced by at least 50% of this value, an expected difference in effect size between RB and No Tx groups of 0.85. We make the conservative assumption of up to 15% additional attrition during the FT phase, leaving N = 46. With this sample size we will have 80% power to detect a difference in effect size of at least.85 between groups, using the two-sided two-sample t-test with alpha 0.05.

The analyses of predictors of treatment response will be exploratory, as the large number of predictors of possible interest will preclude hypothesis-driven methods. We will consider the use of analysis techniques such as model averaging (86, 87), a method that allows for more robust conclusions about important candidate predictors when the possible predictors are numerous and the sample is relatively small, and when different variable selection methods may result in different and possibly suboptimal models.

The sample population for this trial is limited to persons with chronic msTBI who are living in the community, and who are able to walk without the assistance of another person. Our findings may not generalize to those with mild or acute TBI, institutionalized people, or those who are non-ambulatory.