Background:

Exercise interventions to prevent walking difficulty in community-dwelling older adults have focused mainly on improving strength and endurance and have overlooked an important component of walking, namely the timing and coordination of movement. Based on previous research and with critical input from providers and older adults, the On the Move group exercise program was developed. The program includes timing and coordination components and focuses on improving walking.

Objectives:

To compare the effectiveness and sustainability of On the Move against a Standard group exercise program consisting of seated strength, endurance, and flexibility exercises in community-dwelling older adults who reside in independent living facilities and senior apartment buildings, and who live elsewhere but regularly attend senior community centers. We will also explore the effectiveness of On the Move compared with a Standard program when taught by staff activity personnel, and the effectiveness of the On the Move program when delivered by staff activity personnel compared with delivery by exercise leaders (when feasible), using a quasi-experimental design. The acceptability and risks associated with exercise participation are also of interest.

Methods:

The study was a cluster randomized, single-blind intervention trial that compared the effects on function, disability, and mobility of a Standard group exercise program and the On the Move group exercise program in community-dwelling older adults. Randomization to intervention was at the facility level. We had planned to examine the sustainability of the program by randomly assigning participants within each facility to either class 1, taught by an exercise leader, or class 2, taught by staff activity personnel. As we could not randomize to instructor in all facilities as planned, instructor assignment should be considered quasi-experimental. Exercise leaders were research staff who were physical therapists, physical therapist assistants, or exercise physiologists. Staff activity personnel were employees of the facilities themselves who were involved in providing services to the residents. They could be fitness staff, activity directors, social workers, outreach coordinators, care coordinators, or other employees with a similar role. At facilities that did not have staff activity personnel available, we identified (an) older adult(s) from the facility to be trained as a peer leader. Exercise classes were held twice weekly for 12 weeks and were delivered by study exercise leaders (class 1) or staff activity personnel (class 2). The Standard program consisted of warm-up, aerobic, strengthening, and stretching exercises all done while seated. The On the Move program consisted of warm-up, timing and coordination (stepping and walking patterns), strengthening, and stretching exercises completed primarily while standing. The primary outcome of function and disability were the Late Life Function and Disability Instrument (LLFDI) overall function and disability frequency scores, and the primary outcomes of mobility were the 6-minute walk test (6MWT) and gait speed. Outcomes were assessed preintervention and postintervention. Thirty-two facilities were randomized, encompassing 424 individuals.

Results:

The mean ± SD age of the participants was 80.7 ± 7.8 years. The On the Move group had greater improvements than the Standard group in 6-minute walk distance (18.3 ± 60.5 vs 1.9 ± 55.8 m; adjusted difference = 15.3 ± 6.7; P = .0228). There were no significant differences between groups in gait speed (the other primary measure of walking ability), self-reported function, and disability primary outcome or any of the secondary outcomes. When taught by an exercise leader, the On the Move group had greater improvements than the Standard group in the primary measures of mobility, the 6MWT (20.6 ± 57.1 vs 4.1 ± 55.6 m; adjusted difference = 16.7 ± 7.4; P = .0262), and gait speed (0.05 ± 0.13 vs −0.01 ± 0.11 m/s; adjusted difference = 0.05 ± 0.02; P = .0008). The between-group differences were adjusted for the baseline value of the outcome and represent a small but meaningful difference. There were no significant differences in self-reported function and disability as measured by the LLFDI scores. Of the 32 facilities included in the study, we could identify staff activity personnel to train to lead the exercise class at fewer than half of the facilities (15/32 [46.9%]). We could not recruit suitable facility staff at all facilities as planned, thus making the sustainability model—the ability of the facility to continue the program once the research staff was gone—infeasible and the aim exploratory and quasi-experimental rather than randomized. When taught by staff activity personnel, only when we could identify and train someone at the facility, there were no greater gains in any of the primary or secondary outcomes from On the Move (all P > .10). In both programs, attendance (≥20 classes) was greater in the classes taught by the exercise leader compared with the staff activity personnel (65.1% vs 52.0%; 50.0% vs 24.5%). Overall satisfaction was greater in On the Move classes taught by exercise leaders than in those taught by staff activity personnel, as measured by: reporting benefit from class (68.4% vs 42.9%, OR, 2.29 [95% CI, 1.09-4.82]; P = .0294); sufficient individualized instruction (84.2% vs 51.0%, OR, 11.55 [95% CI, 2.17-61.63]; P = .0042); satisfaction with the class (84.2% vs 53.1%, OR, 9.62 [95% CI, 4.05-22.88]; P < .0001); and likelihood of continuing the class if it were to be offered in the future (74.3% vs 53.1%, OR, 1.84 [95% CI, 1.29-2.61]; P = .0007).

Conclusions:

The On the Move group exercise program elicited greater improvements in mobility, as measured by the 6MWT, than the Standard group exercise program when both instructor types were considered together. When taught by exercise leaders, the On the Move group exercise program was more effective at improving mobility than the Standard group exercise program, more safe, and well-liked by community-dwelling older adults, but differences between groups should be interpreted cautiously because we did not a priori plan or statistically power for testing for instructor type × intervention interaction effects and because assignment to an exercise leader or staff activity personnel was not randomized. The On the Move group exercise program did not improve self-reported function or disability. The small number of staff activity personnel, recruited and trained, were unable to sustain a similar level of effectiveness. Therefore, given the difficulty of identifying and training staff activity personnel to deliver the On the Move program and the lack of effectiveness when delivered by staff activity personnel, we believe On the Move is best delivered by an exercise leader. Other modalities of recruiting and training community personnel need to be considered and evaluated for wider dissemination, implementation, and sustainability of On the Move.

Stakeholder Involvement in the Design and Implementation of the Study [PC-1]

Patient and provider stakeholders played a critical role in preparing and conducting this study. We included 2 main types of traditionally nonresearcher stakeholder investigators in our research: participants and providers. We had 2 provider stakeholders who were coinvestigators on the project: (1) a representative of senior management from University of Pittsburgh Medical Center health system (UPMC) Senior Communities department and (2) a lead geriatric outreach nurse also from UPMC Senior Communities. Our participant stakeholders were community-dwelling older adults who reside in independent living facilities or senior housing buildings or those who live in private residences of the community and regularly attend senior community centers. Participant stakeholders also contributed to our pilot studies to develop the intervention²² and were members of our community advisory boards (CABs). The CABs provided ongoing engagement of our participant and provider stakeholders. We created 2 CABs—1 representing the independent living facilities and 1 representing the senior housing buildings and senior community centers. We attempted to have a diverse representation on the CABs with the goal of including individuals' representative of different genders, races, types of facilities, participants, and providers. The CABs met twice a year throughout the study. The meetings were a means for 2-way engagement between investigators and stakeholders to provide ongoing input into the execution and translation phases as well as to be briefed on progress and challenges. Our stakeholders had significant input into the study aims, design, sample, intervention, outcomes, and operational considerations, which are described in detail in our recent manuscript.²³ Below we highlight a sample of our stakeholder involvement.

Our provider stakeholder was instrumental in developing the aims for the study. Initially, aims were limited to examining the effectiveness of the On the Move program using a simple 2-arm study design. In our pilot work, although the On the Move program was generally well received, a consistent complaint at the end of the pilot testing was that facilities personnel and participants were sad to see the On the Move program end. Based on this feedback, our provider stakeholders expressed interest in finding a way to sustain the On the Move program once the research was finished. The provider stakeholder suggested we train someone at each of the facilities to deliver the exercise program. Valuing the input of our stakeholders, we modified our study design and aims to include examining the sustainability of the program over time with facility staff and peer leaders.

Our participant stakeholders had significant input in participant selection for the study [PC-2]. Our initial pilot work was conducted in independent living facilities. The residents of the facilities were mostly older (mean age 85), white, and well educated, and they did not need to leave the building in which they lived to attend the exercise class. During our focus groups at the independent living facilities, the participant stakeholders encouraged us to include other participants from settings that were different from these facilities. Such input reinforced our belief that we needed broader inclusion in our sample, and with the help of our provider stakeholder (ie, lead geriatric outreach nurse), we decided to include participants from low-income housing buildings (greater diversity) and senior community centers (greater diversity and people need to travel to the center). Our provider stakeholder (ie, lead geriatric outreach nurse) had an established rapport with people in these settings and was able to facilitate access to them.

Participant stakeholders also provided valuable input into the intervention protocol.²² For example, the study team had initially selected the music to be played during exercise. The participants expressed a strong preference for music of their generation, and we incorporated music from the 1950s and 1960s into our intervention protocol. We believe such seemingly minor changes potentially contributed to improved exercise adherence and participant retention [MD-1]. Also, we worked with the provider stakeholders and the various facilities to define and operationalize the Standard group exercise program so it would be consistent among facilities and be based on current practice.

Our participant stakeholders identified maintaining independence as an important outcome [PC-3]. Given the time constraints of study funding (3 years), it was not feasible to use an outcome of loss of independence due to much larger sample size implications. Therefore, we focused on intermediate markers of independence inherent in our primary outcomes of function, disability, and mobility. The primary outcomes of function, disability, and walking ability are highly associated with independence and are extremely important to the older adult [RQ-6; PC-3].^4,7

Methodology Standards

The PCORI Methodological Standards²⁴ represent requirements for sound scientific investigation, and we adhered to all relevant standards throughout the process of planning and conducting our study. We will indicate specific standard numbers within [brackets] throughout the present report and in the Appendix Table. Briefly, gaps in evidence have been described in the Background section with the supporting literature [RQ-1], as were the initial identification of target population as community-dwelling elderly [RQ-3], On the Move as the experimental intervention [RQ-5] and subgroup and heterogeneity of treatment effect in aim 4 as exploratory [RQ-4; HT-1]. A formal study protocol was developed a priori [RQ-2] that includes an analysis plan to address the aims [IR-3]. The protocol was documented by registering it in ClinicalTrials.gov (NCT01986647) and is described below. Sustainability aim 2 was in fact designed with a view toward implementation [PC-4]. We will publish the study results in professional journals and present at conferences for dissemination to the scientific community [PC-4]. In addition, the supporting dissemination and implementation of study results to the wider community and associated challenges are mentioned in the Conclusion section [PC-4]. Other standards for hypothesis-driven (confirmatory) heterogeneity of treatment effects [HT-2; HT-3; HT-4] may be partially applicable, and those pertaining to data registries [DR-x], data networks and research-facilitating structures [DN-x], causal inference methods [CI-x], adaptive and Bayesian trials [AT-x], diagnostic tests [DT-x], and systematic reviews [SR-1] are not applicable to our study.

Study Design Overview

Figure 1 shows how the study was designed to address its aims. The study was a cluster randomized, single-blind intervention trial to compare the effects of a Standard group exercise program against the On the Move group exercise program on function, disability, and mobility in community-dwelling older adults who reside in independent living facilities and senior apartment buildings, and who live elsewhere but regularly attend senior community centers. Randomization to interventions was at the facility level, as necessitated by the group nature of the intervention, stratified by facility type. Independent living facilities were known in advance and stratified by socioeconomic status, affiliation, and county. Each of the other 2 facility types were randomized in the order of agreeing to participate. Group exercise classes were held twice weekly for 12 weeks and were delivered by study exercise leaders and facility staff activity personnel when available. Function, disability, and mobility were assessed preintervention and postintervention.

Figure 1

Study Design.

We planned to examine the sustainability of the program by randomly assigning participants within each facility to either class 1, taught by a study exercise leader, or class 2, taught by staff activity personnel. Study exercise leaders were research staff with training and experience in administering the exercise programs, and were exercise physiologists, physical therapists, physical therapy assistants, or from a similar background. Facility staff activity personnel were employees of the facilities themselves who were involved in providing services to the residents. They could be fitness staff, activity directors, social workers, outreach coordinators, care coordinators, or other employees with a similar role. At facilities that did not have staff activity personnel willing or able to be trained, we identified (an) older adult(s) from the facility to be trained as a peer leader. Partway through the trial, we realized that at some facilities we were unable to identify anyone (staff activity personnel or peer leader) to teach the class. In these circumstances, the class was then taught by a study exercise leader. As we could not randomize to instructor at all facilities as planned, instructor assignment should be considered quasi-experimental. Individuals randomized to class 1 exercised for the first 12 weeks at the facility with the exercise leader. During class 1, the exercise leader trained the staff activity personnel or peer leader, who then taught class 2 at the facility. At facilities with sufficient numbers of participants for 3 classes, 2 sections of class 1 were conducted. The study protocol was approved by the University of Pittsburgh IRB and signed informed consent was obtained from all participants. The study was registered in ClinicalTrials.gov (NCT01986647).

Participants [RQ-3; PC-2]

Participants were enrolled between April 2014 and January 2016 and were recruited through informational sessions provided by the research staff at the participating facilities around the greater Pittsburgh, Pennsylvania, area. To be eligible for the study, participants were required to be (1) 65 years of age or older, (2) a resident/member of the participating facility, and (3) able to ambulate independently (with or without a straight cane) for household distances with a gait speed ≥0.60 m/s. Those who (1) were non-English speaking, (2) had impaired cognition (unable to follow a 2-step command or understand the informed consent process), (3) planned to leave the area for an extended period during the study period, (4) had a progressive neuromuscular disorder, (5) had any acute medical condition or illness that was not stable, or (6) had an inappropriate physiologic response to the 6MWT (exercise heart rate ≥120 beats per minute, exercise systolic blood pressure ≥220, or a drop in systolic blood pressure > 10 mmHg or diastolic blood pressure ≥110 mm Hg) were excluded. We attempted to make the inclusion criteria broad and exclusion criteria minimal, but we did need to consider the safety of the participants. Given the class format and content and the instructor-to-participant ratio of 1:10, we did not feel it was safe to include individuals with greatly impaired mobility. Participants were compensated $50 for each testing session they completed to meet recruitment goals, improve adherence, and reduce dropout, thereby preventing missing data [MD-1].

Randomization

The study statistician used the pseudo-random deviate generator in SAS (SAS Institute, Inc) to randomize facilities to the 2 arms in a 1:1 ratio stratified by facility type. Once the facilities were randomized to an exercise program, we then attempted to randomize participants within each facility to classes run by a study exercise leader or a facility staff activity person. In the spirit of patient-centeredness in the research context [PC-1] and in an effort to improve adherence and prevent missing data [MD-1], couples (married, living together, dependent for transportation) were randomized as 1 to the same class but were treated as separate individuals from there on. Per a midtrial protocol change approved by the stakeholder advisory boards and PCORI, the second-level randomization to exercise leader or facility staff took place only if a facility staff member was available to conduct the class safely. As such, the sustainability aim, or the ability of the facility to continue the program once the research ended, was rendered exploratory, and aspects of the design related to the sustainability aim were rendered quasi-experimental.

Training of Facility Staff

Facility staff identified to lead the second session of classes were encouraged to attend all exercise classes within the first session, with a minimum of 3 recommended (1 each at the beginning, middle, and end of the 12 weeks). Additionally, facility staff were provided with a printed binder that outlined the specific exercise routines, with recommendations for the progression of various components of the program. A member of the research team met with the facility staff twice prior to their exercise session, to review the program and answer any questions that staff might have. The research team assessed the fidelity to the exercise protocols via in-person facility visits or review of videotaped recordings of exercise sessions. Feedback and weekly support was provided to all exercise leaders.

Interventions [RQ-5]

Both exercise programs (On the Move and Standard) were delivered by trained exercise leaders (primary aim) or trained staff activity personnel/peer leaders (sustainability component). The frequency and duration of the programs were identical (50 minutes, twice a week, for 12 weeks) with a maximum of 10 participants in a class. The main difference between On the Move and the Standard group exercise programs was the program content described below.

The experimental intervention On the Move exercise program was based on principles of motor learning that enhance “skill” or smooth and automatic movement control.^25-30 The program contained a warm-up (5 minutes), stepping patterns (15 minutes), walking patterns (15 minutes), strengthening exercises (10 minutes), and cool-down exercises (5 minutes). The warm-up and cool-down contained gentle range-of-motion exercises and stretches for the lower extremities and trunk. The stepping and walking patterns were goal oriented and included progressively more difficult patterns, which promoted the timing and coordination of stepping, integrated with the phases of the gait cycle.^26,27,29,30 The goal of the stepping patterns was to facilitate a shifting of the center of pressure posterolateral and then forward, encouraging hip extension. Stepping patterns consisted of stepping forward and across the midline of the body with 1 foot for several repetitions followed by stepping forward and across the midline of the body with the opposite foot for several repetitions. A similar stepping activity was conducted with backward stepping, stepping backward and across the midline of the body, shifting the body weight in a posterolateral direction. Stepping was progressed from stepping on all 1 side, to alternating left and right steps, and then alternating forward and backward stepping. The goal of the walking patterns was to promote a shift of the center of pressure during medial stance and to promote the timing and interlimb coordination of muscle activations (ie, abductors of the going-to-be-swinging limb with the adductors of the stance limb). Patterns consisted of ovals, spirals, and serpentines that were progressed by changing the amplitude of the pattern (ie, narrower oval), altering the speed of walking, or increasing the complexity of the task (ie, walking past other walkers or object manipulation while walking—bouncing a ball). Only 1 item—amplitude, speed, or complexity—was progressed or changed at a time. The strengthening program was conducted primarily while seated and focused on lower extremity muscle groups such as hip flexors, hip abductors, knee flexors, knee extensors, and ankle planter flexors. Playground balls, the opposite extremity, and body weight were used to provide resistance to the movements. Much of the program was conducted in a standing position (40 minutes), with only a small portion conducted while sitting (10 minutes).

The comparator Standard group exercise program was based on those currently being conducted in the community-based facilities involved in the trial (ie, Standard group exercise). We worked with our provider stakeholders and the various facilities to define and operationalize the Standard group exercise program so it would be consistent among facilities and be based on current practice (ie, usual care). The Standard group program contained a warm-up (5 minutes), upper and lower extremity strength exercises (20 minutes), aerobic activities (20 minutes), and a cool-down (5 minutes). The warm-up and cool-down contained gentle range-of-motion exercises and stretches for the lower extremities and trunk. The strengthening program focused on both upper extremity and lower extremity muscle groups. Playground balls, the opposite extremity, and body weight were used to provide resistance to the movements. Aerobic activities included repeated movements of the lower extremities (marching, tapping, skiing) at various speeds. Upper extremity movements were added to increase the intensity of the activity. The entire program was conducted while sitting (50 minutes).

Measures [RQ-6; IR-4]

The primary outcome of function and disability was measured using the self-reported [PC-3] LLFDI overall function and disability frequency domains, and the primary outcomes of walking ability were the 6MWT and gait speed. Self-reported [PC-3] measures of confidence in walking (Gait Efficacy Scale), walking performance under challenging conditions (challenging gait tasks and figure-8 walk), and gait variability were collected as secondary outcomes of walking ability. All measures were collected at baseline prior to randomization, immediately prior to intervention in those randomized to class 2 (see Figure 1), and immediately following the 12-week intervention by research personnel who were blinded to group assignment. All testing was conducted on site at the facilities. Participants who used a cane for ambulation were permitted to use the cane during the testing.

The LLFDI contains a pair of self-report [PC-3] measures targeted to assess physical function and disability in older adults with acute or chronic problems and is designed to be more sensitive to change than similar measures.^31,32 The 2 components of the LLFDI correspond to the activity (LLFDI–function) and participation (LLFDI–disability) components of the World Health Organization's International Classification of Function, Disability, and Health Model. The LLFDI function component has 32 items in 3 domains: basic lower extremity (BLE), advance lower extremity (ALE), and upper extremity (UE), and the LLFDI disability component has 16 items representing 2 domains: frequency of performance and limitation in performance of life tasks. The LLFDI was selected because the patient stakeholders thought disability was an important outcome, and (1) it measures both function and disability, which are critical components of independence; (2) it includes a wide variety of life tasks in various social areas, thus extending beyond the traditional focus of just activities of daily living; (3) the scale was designed with sufficient breadth of items and increments of rating to minimize ceiling and floor effects and maximize the scale's ability to detect change over time; and (4) it is a continuous outcome, which gives us greater statistical power than a dichotomous outcome to detect change over time and make comparisons thereof. The analyses focus on the LLFDI function and disability dimension scores (ie, BLE function, ALE function, UE function, disability frequency, and disability limitation). The disability domain scores (social role, personal role, instrumental role, and management role) were also examined as secondary outcomes because they may provide insight into the impact of the disability on frequency of performance and perceived limitations.³² The LLFDI function and disability scales have established known groups validity, and the test–retest reliability is moderate to high for the disability component (ICC, 0.68-0.82) and extremely high for the function component (ICC, 0.91-0.98) pursuant to [IR-4]. Scores range from 0 to 100; higher scores represent less difficulty and less disability.

The primary measures of walking ability were the 6MWT and gait speed. The 6MWT measures distance walked (meters) in 6 minutes, including time for rest as needed.³³ The 6MWT is (1) a performance-based measure of walking ability, which is an important component of independence; (2) an indicator of community ambulation (ie, the ability to walk 300 m in 6 minutes)^34,35; (3) a continuous outcome that provides greater statistical power than a dichotomous outcome to detect change over time³⁶; and (4) a widely used measure of mobility that is included in the National Institute of Health (NIH) Patient Reported Outcomes Measurement Information System project to establish measures of clinical assessment [PC-3]. Pursuant to [IR-4], the 6MWT has established psychometric properties, excellent test-retest reliability (Pearson r = .95) in older adults,^37,38 and construct validity for graded exercise testing and functional classification.³⁹ Changes in the 6MWT of 20 m and 50 m are considered small but meaningful and substantial, respectively.⁴⁰

The second main walking ability outcome was gait speed. Gait speed is a strong indicator and predictor of disability, morbidity, and mortality in the older adult.^2,3,5,7 Gait speed was assessed during usual walking using an instrumented walkway. After receiving instructions, participants completed 6 passes at their usual, self-selected walking speed. Gait speed was averaged over the 6 passes. Pursuant to [IR-4], the test–retest reliability of gait speed measured using instrumented walkways is excellent (ICC, 0.98).⁴¹ Changes in gait speed of 0.05 m/s and 0.10 m/s are considered small but meaningful and substantial, respectively.⁴⁰

Additional mobility measures included confidence in walking, walking performance under challenging conditions (narrow, obstacle, and figure-8 walk), and gait variability (stance time, step length, and step width SD). To determine if interventions affected confidence in walking, confidence was assessed using the Gait Efficacy Scale [PC-3].^42-44 The scale items include a range of gait activities such as walking over different surfaces, stepping on and off curbs, and negotiating stairs. Each item is scored on a 10-point Likert scale, with the total score for the 10 items ranging from 0 to 100. A higher score represents greater confidence. The figure-8 walk was designed to measure motor skill in walking.⁴⁵ The test involves walking a figure-8 pattern around 2 markers placed 5 feet apart. Performance was scored based on the time needed to complete the figure-8 walk and the number of steps taken. Challenging gait tasks were used to examine participants' ability to adapt their gait to different environmental conditions.⁴⁶ Subjects completed two 12-meter trials of each challenging condition, obstacle, and narrow path. The time needed to complete each task, averaged over 2 trials, was the summary indicator of gait during challenging tasks. In a published sample of 40 community-dwelling older adults, the 1-week test–retest reliability of the timed measures was excellent (ICC, 0.70-0.94), satisfying [IR-4].⁴⁵ Gait variability, defined as fluctuations in gait characteristics from one step to the next,⁴⁷ is an important indicator of impaired mobility in older adults.⁴⁸ Gait variability was quantified using established measures of temporal and spatial gait characteristics including stance time, step length, and step width. Variability was calculated as the SD of the set of steps/stances recorded over 6 passes on the instrumented walkway (described above). In general, lower variability is better.^48,49

Satisfaction [PC-3], adherence, and adverse events were also measured and documented. Participant satisfaction was assessed using a satisfaction survey and in-depth interviews. The satisfaction survey was administered to all participants at the conclusion of the exercise program. It included 5 items measured on a 5-point Likert scale, 6 items measured on a 3-point scale (changed a lot, changed some, or no change), and 2 open-ended questions. The items assessed degree of satisfaction with various components of the exercise program (ie, the exercises, safety, individualized instruction) and likelihood of continued participation. A series of yes/no questions were used to determine if the program met the participants' expectations, the perception of benefit from the program, and the likelihood of recommending the program to others. In-depth phone interviews were used to assess satisfaction in a random subsample. We systematically selected every fifth participant from a range of settings representing the different intervention arms (On the Move, Standard, exercise leader, activity staff personnel) and facility types (independent living facility, community center, and senior housing). The interview included a mixture of closed-ended and open-ended questions. Topics included in the interviews were perceived benefits and risks; satisfaction with the program, facilities, and instructor; and amount of individualized instruction.

A roster of participants was maintained for each class. At the beginning of each class, the instructor recorded attendance. The instructor also recorded reasons for missed classes, when available. Attendance rate ([number of sessions attended by the participant/total number of classes offered, ie, 24] × 100%) for each participant was the main indicator of adherence.

Adverse events that occurred during testing or intervention were recorded on a clinical event form. Adverse events other than muscle soreness (which was greater than anticipated) included labored breathing, chest pain, fall, or other injury. All adverse events were forwarded to the study consulting physician (NKN) for adjudication and direction.

Data from instrumented walkway testing, onsite physical performance testing, and self-report questionnaires were merged with the randomization and participant tracking databases by participant identification number and time point to create analytic data sets [IR-2]. All data, including outcomes and other participant characteristics, were collected specifically for the present study and thus meet standard [IR-1].

Sample Size Justification

We based the sample size on pilot studies^18-20 and the ability to detect clinically meaningful⁵⁰ or moderate effect sizes (Cohen d = 0.5).⁵¹ The computations were based on 2-tailed α = .05 tests, an attrition rate of 10% [MD-1], an anticipated class size of 10 participants, intracluster correlation of 0.1, published methodologies and commercially available sample size and power software (PASS 2002: Number Cruncher Statistical Systems). We estimated that 90 participants per arm would allow us, with 80% statistical power, to detect statistical significance of a difference as small as 3.1 points in LLFDI overall function change between the 2 interventions; 80 per arm for 3.2 points in LLFDI disability frequency; 140 per arm for 0.1 m/s in gait speed; and 40 per arm for 50 m in 6-minute walk distance. We estimated that 70 per arm receiving the On the Move intervention would allow us, with 95% statistical power, to detect a similar difference in LLFDI overall function change between the 2 types of instructors; 60 per arm for LLFDI disability frequency; 110 per arm for gait speed; and 30 per arm for 6-minute walk distance. We used 95% statistical power (rather than the customary 80%) in study planning for the sustainability hypothesis to reduce the likelihood of a type II error, thereby minimizing the chances of a finding in favor of sustainability because of lack of statistical power rather than because of an actual similarity. Therefore, 140 participants per arm, or a total of 560 participants, were deemed necessary to accommodate all primary outcomes and both aims 1 and 2. We reached participant recruitment targets for exercise leader arms, meeting [IR-1], but not for facility staff arms. With the midstudy approval of stakeholder advisory boards and PCORI, the aims and hypotheses that involved facility staff were deemed exploratory and were based on a smaller number of participants due to safety and feasibility concerns outlined elsewhere in the report.

Data analysis

Aim 1: Effectiveness of On the Move

We first present the analysis results comparing the 2 interventions with both instructor types combined. This analysis was a priori not planned, but reviewers requested we precede with the planned analysis because of the failed randomization to instructor type. Sixteen sites with a total of 201 participants were randomized to receive the On the Move program taught by research or facility staff, and 16 sites with 223 participants were randomized to receive the Standard program taught by research or facility staff (Figures 2 and 3). Of the 201 participants in the On the Move group, 179 individuals (89.1%) completed some or all of the postintervention testing, and 22 did not complete any postintervention testing (8 were lost to follow-up, 11 had health issues, and 3 dropped out of the study). Of the 223 participants in the Standard intervention, 205 individuals (91.9%) completed some or all of the postintervention testing, and 18 did not complete any postintervention testing (5 were lost to follow-up, 7 had health issues, and 6 dropped out) [IR-6; MD-4]. Those who did not complete any postintervention testing had poorer 6MWT results (280.1 vs 245.0 m; P = .0187), and a lower proportion of this group reported excellent/very good balance (32.3% vs 17.5%; P = .0119; Table 1.

Table 1

Participant Characteristics of Classes Taught by Exercise Leaders and Facility Staff Stratified by Whether Completed or Dropped Out: Mean ± SD or No. (%).

Overall, participants had a mean ± SD age of 80.7 ± 7.8 years and were mostly female (82.3%) and white (83.1%). Participants reported 2.8 ± 1.4 chronic condition domains on the comorbidity index. Participants walked slowly (gait speed = 0.91 ± 0.20 m/s), and 56.4% walked less than community ambulation distance (ie, 300 m) on the 6MWT. The intervention groups were similar on all baseline measures except facility type (Table 2).

Table 2

Participant Characteristics and Measures at Baseline by Intervention Group, On the Move and Standard Taught by Exercise Leaders and Facility Staff: Mean ± SD or No. (%).

The On the Move group had greater improvements than the Standard group in the 6MWT (18.3 ± 60.5 vs 1.9 ± 55.8 m; adjusted difference = 15.3 ± 6.7; P = .0228). There were no significant differences between groups in gait speed, self-reported function and disability, or any of the secondary outcomes (Tables 3-4).

Table 3

Baseline to Follow-up Change in Participant Primary Outcome Measures and Between-Intervention Differences When Interventions Were Delivered by Exercise Leaders and Facility Staff: Estimate ± SE (P Value).

Table 4

Baseline to Follow-up Change in Participant Secondary Outcome Measures and Between-Intervention Differences When Delivered by Exercise Leaders and Facility Staff: Estimate ± SE (P Value).

We next present the a priori planned primary analysis results using only those taught by exercise leaders. Of the participants, 152 were randomized to receive On the Move taught by exercise leaders, and 146 were randomized to receive Standard taught by research staff (Figures 2 and 3). Of the 152 participants in the On the Move group, 142 individuals (93.4%) completed some or all of the postintervention testing, and 10 did not complete any postintervention testing (4 were lost to follow-up, 5 had health issues, and 1 dropped out of the study). Of the 146 participants in the Standard intervention, 139 individuals (95.2%) completed some or all of the postintervention testing, and 7 did not complete any postintervention testing (1 was lost to follow-up, 3 had health issues, and 3 dropped out) [IR-6; MD-4]. Those who did not complete any postintervention testing had poorer 6MWT results (279.5 vs 205.9 m; P = .0012), gait speed (0.92 vs 0.80 m/s; P = .0290), and LLFDI disability frequency (52.7 vs 48.4; P = .0110; Table 1).

Table 5 shows participant characteristics by intervention group and by whether participants were taught be staff or exercise leaders. Participants had a mean ± SD age of 80.0 ± 8.1 years and were mostly female (84.2%) and white (83.6%). Participants reported 2.8 ± 1.4 chronic condition domains on the comorbidity index. Participants walked slowly (gait speed = 0.91 ± 0.21 m/s), and 56.4% walked less than 300 m on the 6MWT, which has been proposed as a distance compatible with community ambulation.^38,61 The intervention groups were similar on all baseline measures except facility type (Table 5).

Table 5

Participant Characteristics and Measures at Baseline: Mean ± SD or No. (%).

When examining the effectiveness of On the Move when delivered by an exercise leader, the 2 intervention groups had significantly different improvements when all 4 primary outcomes were simultaneously considered in a multivariate Hotelling test (P = .0164). The On the Move group had greater improvements than the Standard group in the primary measures of walking ability, the 6MWT (20.6 ± 57.1 vs 4.1 ± 55.6 m; adjusted difference = 16.7 ± 7.4; P = .0262), and gait speed (0.05 ± 0.13 vs -0.01 ± 0.11 m/s; adjusted difference = 0.05 ± 0.02; P = .0008). In total, 48.5% of On the Move participants showed a small but meaningful improvement in gait speed (0.05+ m/s) compared with only 29.5% in the Standard intervention, resulting in a number-needed-to-treat of 5.2. There were no significant differences between groups in self-reported function and disability primary outcome or any of the secondary outcomes (Tables 6 and 7). Sensitivity analyses that adjusted for facility type as an additional covariate, including an additional parameter in the working correlation structure to account for a potential correlation between members of the same couple; that averaged data by facility and fitted a simpler analysis of covariance model with facility as the unit of analysis; that used an immediate preintervention measurement instead of the pre-randomization assessment; or that ignored missing data instead of multiple imputation, did not materially change the results [IR-5; MD-5].

Table 6

Baseline to Follow-up Change in Primary Outcomes and Between-Intervention Differences: Estimate ± SE (P Value).

Table 7

Baseline to Follow-up Change in Participant Secondary Measures and Between-Intervention Differences: Estimate ± SE (P Value).

Aim 2: Exploration of Sustainability of On the Move When Facility Staff Are Available to Teach

Of the 32 facilities included in the study, we could identify a member of the staff or a peer to train to lead the exercise class at fewer than half of the facilities (15/32 [46.9%]). As we could not randomize to instructor as planned in all facilities, instructor assignment should be considered quasi-experimental and results treated as exploratory. We were more likely to identify someone to train at the independent living facilities (9/10 [90%]) than at the community facilities (6/22 [27.3%]). Forty-nine individuals were randomized to receive On the Move taught by facility staff and 77 individuals to receive the Standard intervention delivered by facility staff (Figures 2 and 3). Of the 49 randomized to receive On the Move, 37 people completed some or all of the postintervention testing, and 12 people did not complete any postintervention testing (4 were lost to follow-up, 6 had health issues, and 2 dropped out). Of the 77 randomized to receive the Standard intervention taught by facility staff, 66 people completed some or all of the postintervention testing and 11 individuals did not complete any postintervention testing (4 were lost to follow-up, 4 had health issues, and 3 dropped out) [IR-6; MD-4].

The 4 groups defined by the intervention × instructor combinations were similar at baseline on most demographic, health status, function, disability, and mobility measures (Table 5). The groups differed on facility type, presence of diabetes, and results of digit-symbol substitution testing (P < .05).

When taught by facility staff personnel, no greater gains were attributable to On the Move than to the Standard program in any of the primary or secondary outcomes of function, disability, and mobility (all P > .10; Tables 6 and 7). However, the sample size used in these comparisons was limited. Sustainability of the On the Move program was also examined by comparing outcomes of the On the Move program when taught by an exercise leader with those when taught by facility staff personnel. Some evidence showed that facility staff elicited smaller gains with On the Move than exercise leader in LLFDI overall function (AD = 1.89 ± 1.02; P = .0647), figure-8 walk number of steps (adjusted difference = AD = −1.25 ± 0.55; P = .0224), gait efficacy scale (AD = 4.5 ± 2.1; P = .0314), LLFDI advanced lower extremity function (AD = 3.88 ± 1.96; P = .0484), LLFDI disability frequency (AD = 1.87 ± 0.89; P = .0370), LLFDI social role (AD = 2.30 ± 1.21; P = .0576), LLFDI management role (AD = 4.18 ± 2.03; P = .0402), gait speed (AD = 0.05 ± 0.03; P = .0578), and step length variability (AD = −0.46 ± 0.19; P = .0179). However, given the difference in sample size between the groups, these results should be interrupted cautiously. We found no such differences in the other measures (Tables 6 and 7).

Sensitivity analyses that adjusted for facility type, presence of diabetes, and digit symbol substitution test performance as additional covariates, using an immediate preintervention measurement instead of the prerandomization assessment or ignoring missing data instead of multiple imputation, did not materially change the results [IR-5; MD-5].

Aim 3: Acceptability and Risks of the 2 Programs When Delivered by Exercise Leaders and Staff Activity Personnel

As requested by the reviewers, we examined the acceptability and the risks of the 2 programs regardless of instructor type. Individuals in On the Move were less likely than those in the Standard class to attend ≥20 classes (43.8% vs 60.5%; OR, 0.51 [95% CI, 0.33-0.78]; P = .0021; Table 8). More than half of the individuals in both programs (On the Move and Standard) felt (1) they benefited from the class, (2) the classes were at least somewhat challenging, (3) they received just enough or more personalized instruction, and (4) safe or very safe while doing the exercises. In both groups, more than half of the participants (On the Move = 76.6% and Standard = 82.1%) were satisfied or very satisfied with the programs.

Table 8

Comparisons of Follow-up Measures of Exercise Attendance and Satisfaction: No. (%) and Odds Ratio [95% CI] (P Value).

When taught by the exercise leaders, individuals in On the Move were less likely than those in the Standard class to attend ≥20 classes (50.0% vs 65.1%; OR, 0.55 [95% CI, 0.30-0.90]; P = .0418); Table 8). When taught by an exercise leader, the vast majority of individuals in both programs (On the Move and Standard) felt (1) they benefited from the class, (2) the classes were at least somewhat challenging, (3) they received just enough or more personalized instruction, and (4) safe or very safe while doing the exercises. In both groups, the majority of the participants (On the Move = 84.2% and Standard = 86.3%) were satisfied or very satisfied with the program and the majority said they would likely continue with the same program if it were offered in the future (74.3% vs 74.0%; OR, 1.16 [95% CI, 0.65-2.05]; P = .6167).

When the On the Move and the Standard programs were taught by staff activity personnel, individuals in On the Move were less likely to attend ≥20 classes than those in the Standard program (24.5% vs 52.0%; OR, 0.28 [95% CI, 0.08-0.93]; P = .0384; Table 8). Approximately half of the participants in each group felt they benefited from the exercise classes taught by staff activity personnel (42.9% vs 53.3%; OR, 0.83 [95% CI, 0.42-1.65]; P = .5915). However, compared with participants in the Standard program taught by staff activity personnel, participants in On the Move taught by staff activity personnel were less likely to report that they felt safe (55.1% vs 75.3%; OR, 0.27 [95% CI, 0.07-1.04]; P = .0574), that they were satisfied with the program (53.1% vs 74.0%; OR, 0.29 [95% CI, 0.11-0.79]; P = .0153), or that they would continue in the same program if it was offered in the future (53.1% vs 72.7%; OR, 0.31 [95% CI, 0.13-0.72]; P = .0066).

Last, we compared adherence to and satisfaction with On the Move when it was taught by exercise leaders and staff activity personnel. Attendance (20+ classes) was greater in the classes taught by the exercise leader compared with the facility staff personnel (50.0% vs 24.5%; OR, 3.38 [95% CI, 1.10-10.38]; P = .0333; Table 8). Individuals in the classes taught by exercise leaders attended on average 16.0 ± 7.9 classes, while those in the classes taught by facility staff personnel attended on average 11.7 ± 9.0 classes. Overall satisfaction was greater in On the Move classes taught by exercise leaders than in those by staff activity personnel. Participants reported benefit from class (68.4% vs 42.9%; OR, 2.29 [95% CI, 1.09-4.82]; P = .0294), sufficient individualized instruction (84.2% vs 51.0%; OR, 11.55 [95% CI, 2.17-61.63]; P = .0042), satisfaction with the class (84.2% vs 53.1%; OR, 9.62 [95% CI, 4.05-22.88]; P < .0001), and likelihood of continuing the class if it were to be offered in the future (74.3% vs 53.1%; OR, 1.84 [95% CI, 1.29-2.61]; P = .0007). Finally, individuals in On the Move classes taught by exercise leaders were much more likely to report feeling safe during the exercise class than those taught by staff activity personnel (86.2% vs 55.1%; OR, 33.42 [95% CI, 4.05-275.80]; P = .0011).

Few adverse events occurred, indicating that On the Move as currently designed was well tolerated. Throughout the study, 7 reported adverse events occurred during the intervention. The majority of the events occurred during the On the Move classes (n = 6; 3 falls, 2 fatigue, and 1 pain), with only 1 event (chest pain) occurring during the Standard class. Of the 6 events that occurred during On the Move classes, 4 occurred during classes taught by exercise leaders and 2 occurred during classes led by staff activity personnel. Of the events, only the chest pain event was classified as serious. The participant who experienced chest pain was referred for medical attention, had a stent placed, and eventually returned to the exercise program without further incident.

Aim 3: In-Depth Phone Interviews

A complete report of the in-depth phone interviews can be found in the Appendix. Fifty-one individuals completed the phone interviews. Distribution by study arm was as follows: On the Move taught by exercise leader, n = 10; Standard taught by exercise leader, n = 21; On the Move taught by staff activity personnel, n = 9; and Standard taught by staff activity personnel, n = 11. The feedback received by the phone surveys was in general very positive. The findings from the interviews mirrored the quantitative findings. Briefly, all participants who were interviewed stated they felt safe while attending the exercise classes. The majority of the participants reported that they experienced a small to a great change in their walking ability. Some of the changes reported were improved walking; walking better, faster, farther, and for longer periods of time; and less reliance on a cane for walking. We found no appreciable differences in the responses given by the participants in the On the Move compared with the Standard program. However, participants in exercise leader–led classes tended to mention more positive outcomes than the staff activity personnel–led classes. An unexpected benefit of the exercise program mentioned by several participants was the positive nature of the group experience. Several mentioned the social benefits and the camaraderie of the group as a benefit of the group exercise program. In fact, more than two-thirds mentioned the social aspects of participating in a group activity as what they liked most about the exercise class, namely the feeling of community and companionship.

“I guess one thing that I really enjoyed about it was the feeling of community that we had. We really enjoyed each other as well as enjoying the exercise together. That was a real plus for me.” [WF/ILF/OTM/SAP]

Another interesting finding from the surveys was that the Standard program was likely a much better exercise program than the normally offered standard seated program. Some of the comments included, “Even if it was a seated program it was quite a workout”; “it was not the same old, same old program”; and “it was in general more demanding and organized than the seated programs at their facility.”

A few of the participants who were interviewed expressed concern with the staff activity personnel. Only 3 participants (2 in On the Move and 1 in Standard, all taught by staff activity personnel) expressed some reservations about the way their exercise instructor conducted the class.

“the instructor. She was the most disappointing. The instructor was not well qualified, could not do some of the exercises and show you.” [WF/SC/OTM/FS]

Several of those interviewed found the On the Move program either too easy or too challenging. Participants suggested offering different levels of the On the Move program, such as a beginner and an advanced class, so that individual differences and capabilities could be addressed.

Aim 4: Who Is Likely to Benefit From On the Move Taught by Exercise Leaders?

Post hoc exploratory subgroup analyses using only those taught by exercise leaders showed significant treatment by class attendance (P = .0112) and facility type (P = .0443) interactions with respect to the outcome of the 6MWD. From a hypothesis-generating perspective for subsequent studies, we feel those attending ≥20 classes (6MWD improvement 31.8 m; P = .0016) or those attending community centers (48.9 m; P = .0051) may derive greater benefits from the On the Move program than from the Standard program.

Our responder vs nonresponder comparisons are based on a smaller number of participants due to the unanticipated staff recruitment difficulties and significant differences between exercise leaders and facility staff personnel. As such, the results should be considered preliminary and hypothesis generating, rather than confirmatory. Comparisons of baseline characteristics between responders (gait speed gain of 0.05+ m/s) and nonresponders to On the Move when taught by an exercise leader (Table 9) showed that responders tended to have a lower body mass index (BMI) (27.7 ± 5.1 vs 30.6 ± 9.4; P = .0130) and greater scores on LLFDI disability frequency (55.1 ± 6.5 vs 52.4 ± 5.4; P = .0133) and social role (50.6 ± 8.2 vs 46.5 ± 7.0; P = .0037). Further, those with a lower BMI (OR, 0.94 [95% CI, 0.90-0.98]; P = .0130) and greater LLFDI scores for disability frequency (OR, 1.08 [95% CI, 1.02-1.15]; P = .0051) and social role (OR, 1.08 [95% CI, 1.03-1.12]; P = .0007) were more likely to respond to On the Move when taught by an exercise leader. We attempted to find a parsimonious set of independent baseline predictors through forward selection (Table 10); we found that having had cancer (OR, 0.20 [95% CI, 0.07-0.56]; P = .0022), a lower BMI (OR, 0.91 [95% CI, 0.82-1.00]; P = .0527), a greater score on the LLFDI social role (OR, 1.15 [95% CI, 1.07-1.24]; P = .0002), and a faster gait speed (OR, 0.50 [95% CI, 0.33-0.76]; P = .0011) were associated with a likelihood to respond to On the Move. The random forest and classification tree analysis (Figure 4) showed largely similar results, splitting on performance measures of walking ability, BMI, and LLFDI disability frequency and social role as important baseline predictors with an AUROC of 0.869 in the analyzed sample and 0.738 in internal cross-validation.

Table 9

Comparisons of Baseline Characteristics Between Those Responding to On the Move Intervention Taught by Exercise Leaders and One-at-a-Time Baseline Predictors of a 0.05+ m/s Gain in Gait Speed: No. (%), Mean ± SD or Odds Ratio (95% CI) [P Value]. (more...)

Table 10

Independent Baseline Predictors of a 0.05+ m/s Gait Speed Gain From On the Move Intervention Taught by Exercise Leaders, Identified Through Forward Selection.

Figure 4

Classification Tree Analysis for Identifying Potential Responders (as defined by a 0.05+ m/s gain in gait speed) to On the Move Intervention Taught by Exercise Leaders.

Decisional context

The On the Move group exercise program elicited greater improvements in mobility as measured by the 6MWT than the Standard group exercise program when the exercise leaders and facility staff arms were combined. Combining the 2 instructor types was deemed necessary by the reviewers as we could not randomize participants to the instructor as we had originally planned, due to the difficulty of identifying and training staff activity personnel to deliver the On the Move program. Therefore, we believe On the Move is best delivered by an exercise leader. When taught by an exercise leader, the On the Move group exercise program, which was designed to target the timing and coordination of movement important for walking, elicited greater improvements in mobility than the Standard group exercise program. The gains in mobility were both statistically significant and clinically meaningful.⁵⁰ The adjusted difference of gains between groups in the 6-minute walk distance of 20 m and the adjusted group difference between groups in gait speed of 0.05 m/s have been described as small but meaningful changes in mobility.⁵⁰ The results translate to a very favorable number-needed-to-treat of only 5.7. Interestingly, the greater improvements in mobility in On the Move compared with the Standard group occurred even though adherence to the exercise program was lower in On the Move than in the Standard program. Our findings support the idea that timing and coordination exercises should be included in exercise programs to improve mobility in older adults.

Study Results in Context

Group exercise programs for older adults that have included standing and walking activities have had conflicting findings about their impact on mobility.^11-13 Often these interventions were compared with a nonexercise control group and were conducted in “young” older adults.^11-13 The 1 group exercise program that did improve mobility consisted of a very high dose of exercise (65 minutes a day, 5 days a week, for 24 weeks), which may not be acceptable to all older adults.¹¹ In addition, many programs exclude an important component of exercise that is critical to walking, namely the timing and coordination of movement.^14-16

In our study, the On the Move group exercise program, which was designed to target the timing and coordination of movement important for walking, elicited greater improvements in mobility than the Standard group exercise program. We compared the On the Move program with the Standard group exercise program instead of comparing it with a nonintervention control. Many of the previous research reports on community-based group exercise programs have compared a group exercise program with a nonexercise control.^11-13 We compared the effectiveness of On the Move against a more challenging Standard program [RQ-5]. The comparison group received an exercise program that was not only active but also well taught by trained exercise professionals. Despite being held to a more challenging Standard program, the On the Move exercise program participants demonstrated a greater improvement in mobility. Also, the participants in this study sample constituted a somewhat frail group of adults of advanced age who are not usually included in exercise intervention studies. The mean age of these participants was greater than 80 years and they had multiple chronic conditions, impaired mobility, fear of falling, and a history of falls. However, it is important to note that, to participate in the study, the older adults had to have a certain level of mobility (ie, gait speed ≥0.60 m/s) and be able to ambulate with no more than a straight cane. Therefore, the findings cannot be generalized to older adults with greater mobility limitations. Finally, the On the Move program consisted of a lower dose of exercise (ie, 50 minutes twice a week for 12 weeks) than the previous study (65 minutes a day, 5 days a week, for 24 weeks), which was shown to improve mobility.

Implementation of Study Results

We demonstrated the effectiveness of the On the Move program on site in 3 different settings—independent living facilities, community centers, and apartment buildings. All testing and interventions were delivered at the facilities, thus indicating the program can be conducted in various community locations and supports the implementation in a variety of settings.

To examine the facility's ability to maintain the program after the research project had ended, we attempted to identify and train staff at the various facilities to lead the exercise programs. We had great difficulty identifying staff activity personnel to teach the group exercise programs, especially at the senior apartment buildings and community centers, where we were able to identify and train someone only at less than 30% of the facilities. Therefore, in a way, the question about the facility's ability to maintain the exercise program, as asked, was mostly answered through the feasibility aspect during the process of conducting the study. Because the randomized assignment to instructor could not be accomplished as planned, the analytic results for our sustainability aim should be considered to be of the quality of a quasi-experimental study, and lesser than those from a fully randomized trial. Subject to this limitation, the individuals who were trained to lead the exercise programs were unable to get the same improvements in participants' mobility with the On the Move program as the exercise leaders. In fact, participants were more satisfied when the staff activity personnel taught the Standard program.

Participants were not as satisfied and did not feel as safe when the staff activity personnel taught the On the Move program compared with the Standard program. Therefore, we currently believe, given the difficulty in identifying and training staff activity personnel to lead the exercise classes and the present preliminary findings about effectiveness, that the On the Move program is best taught by an exercise leader who is a health professional.

It is important to note that the On the Move program was designed based on a task-specific motor skill program that is delivered 1-on-1 by a physical therapist. Although the improvements were smaller for the On the Move group class than for the 1-on-1 task-specific exercise program, the improvements are still significant and clinically meaningful. The group exercise class format is a more efficient way to deliver the exercise and could be a valuable option for community-based health and wellness programming.¹² However, trained staff activity personnel who were primarily lay individuals without an exercise or physical therapy background were unable to get the same improvements in mobility when they taught the On the Move class, thus indicating that the On the Move program should be taught by someone with physical therapy or exercise physiology training. Some potential explanations for the lack of success with staff activity personnel delivering the On the Move program include (1) inadequate training, (2) lack of interest in teaching the program, (3) lack of baseline knowledge of exercise principles, and (4) participants' unwillingness to recognize the staff activity personnel as having exercise expertise.

As stated earlier, adherence to the On the Move program was lower than it was for the Standard exercise program when it was taught by exercise leaders (median number of classes attended was 19.5 vs 21.0). Participants were equally satisfied with either program (84% vs 86% satisfied or very satisfied), and they felt safe in both programs (86% vs 87%) when taught by exercise leaders. The On the Move program is conducted primarily in the standing position and was designed to be a more challenging program than the Standard exercise class. One potential explanation for the lower adherence could be that individuals who had a change in health status during the 12-week program may have been more likely to be able to continue in the Standard program, which is performed primarily while sitting.

Generalizability

As our sample was representative of the Pittsburgh area (more than 80% white), it was fairly racially homogeneous and did not represent the entire spectrum of the older adult population. However, our study participants did represent a wide range of living settings—such as independent living facilities, senior high-rise apartments, and those residing in private homes and attending community centers—and a range of physical function abilities.

Subpopulation Considerations

To facilitate informed patient decision-making, we explored heterogeneity of treatment effect in post hoc [HT-1] analyses and attempted to generate hypotheses about baseline predictors of who would benefit the most from participation in the On the Move program. Given the exploratory nature of these analyses, the findings should be interpreted with caution. Based on these exploratory analyses, we speculate that those able to be adherent and those with some level of intact walking speed, endurance, and cognitive function will benefit more from On the Move than from the Standard program. And among those in On the Move, those with somewhat lower levels of physical performance of mobility, which signals room for improvement, and greater levels of social function and desire for interaction, are more likely to benefit. In subgroup analyses, individuals with lesser mobility impairment, lower comorbid burden, and better processing speed had greater improvements in mobility with the On the Move program, suggesting that individuals need a certain level of mobility and health to reap greater benefits from the On the Move program compared with a Standard program. Individuals with impaired mobility demonstrated similar improvements in walking endurance with the 2 programs. Individuals with impaired mobility may benefit from first participating in a Standard program to improve endurance so that they could later elicit greater benefits from the On the Move program.

Study Limitations

When interpreting the results, some limitations should be considered. First, we had difficulty identifying staff activity personnel to train to lead the exercise class. Therefore, we could not randomize participants to instructors at all sites as we had originally planned, leading the reviewers to request analyses combining both instructor types. The staff activity personnel whom we did identify had various backgrounds and levels of training and exercise experience. We did not attempt to measure or quantify the staff activity personnel's level of education or exercise experience. Second, participant attrition was greater in the facility staff study arm than in the exercise leader study arm (13.5% vs 5.7%). Given the study design, it is difficult to determine if the difference in attrition was due to a lack of satisfaction with the intervention arm or to changes in participants' health status. The 2 exercise session study arms (exercise leader and staff activity personnel) were not taught at the same time but in sequence. The study was so designed that during the first session, which was taught by an exercise leader, we could train the staff activity personnel to lead the second exercise session. However, it also meant that participants randomized to the staff activity personnel class had to wait an extra 12 weeks before starting their exercise session. The additional waiting time in the study could potentially expose the participants to additional adverse health events and lead to potential disappointment later upon learning that they would be taught by someone other than the professionally trained exercise leader. When we examined reasons for dropping out of the study, the percentage of those who dropped out due to health concerns was indeed greater in the staff activity personnel arm (staff activity personnel 10/126 or 7.9% and exercise leader 8/298 or 2.7% dropped out for health reasons). Third, in this comparative effectiveness trial, we compared the On the Move program with a Standard program that was similar to usual care for the facilities included in the study, based on input from our participant stakeholders. The Standard program that was completed primarily in a seated position would, in theory, be less likely to affect mobility outcomes such as gait speed than a program that included walking activities. However, in other high-profile studies—such as the LIFE-P, which included a physical activity intervention with a goal of completing 150 minutes of walking a week—the change in gait speed was minimal (ie, 0.01 m/s),⁹ thus indicating that interventions that include walking do not necessarily improve walking speed. Fourth, we did not exclude participants based on their cognitive status. To participate in the trial, participants had to be able to follow a simple 2-step command. It is possible that some participants had mild cognitive impairment, which may have affected the self-reported outcomes. Last, we measured outcomes only immediately following the intervention, so we cannot comment on the long-term sustainability of mobility improvements.

Regarding other methodological limitations, when planning the study we did not expect to find differences in gains elicited by exercise leaders and facility staff with On the Move and had proposed to combine them in exploring baseline participant characteristics predictive of likelihood of benefit. Because of the unanticipated differences we found in aim 2, and the smaller numbers taught by facility staff due to feasibility concerns, we were able to conduct the said exploration only with those taught by the exercise leader. The resulting reduced sample size thus limited our analyses to definitively identify participants most likely to benefit from On the Move.

Future Research Including Dissemination and Implementation

We plan to prepare and submit an Aging and Disability Evidence-Based Programs and Practices (ADEPP) application to the Administration for Community Living, US Department of Health and Human Services, to have the On the Move program evaluated and designated as an evidence-based program. The ADEPP's goal is to improve access to information on evidence-based interventions, with the ultimate goal of reducing the lag time between the creation of scientific knowledge and its practical application in the field. Interventions that have undergone an ADEPP review have met the highest-level criteria for Title IIID funding of the Older Americans Act. Many community sites are looking for evidence-based programs to offer at their facilities. Obtaining the designation of an evidence-based program is an important first step in translating On the Move into widespread use in the community.

We also plan to develop a website to house information about the On the Move program. The website will be the comprehensive resource that contains all information about the On the Move program. The following information may be included on the website: a description of the program, evidence to support the program, research opportunities, training materials, resources for instructors, and participant testimonials. We will also meet with local stakeholders, such as the UPMC Senior Communities Area Agency on Aging and various other agencies that provide resources for older adults (Life Span Resources and Services for Seniors, Senior Care Network, The Hill House Association, Presbyterian Senior Care), to finalize the research questions and develop a research plan to study the implementation and sustainability of the On the Move program in the real world.

Appendix Table. PCORI Methodology Standard Adherence (PDF, 128K)

Appendix Report (PDF, 420K)