Use of a Home-Based, Commercial Exercise Platform to Remotely Monitor Aerobic Exercise Adherence and Intensity in People With Parkinson Disease

Abstract

Objective

Physical therapists are well-positioned to prescribe exercise outside of a clinical setting to promote positive health behaviors in people with Parkinson disease (PD). Traditionally, a barrier to precise exercise prescription has been reliance on participant self-reported exercise adherence and intensity. Home-based, commercially available exercise platforms offer an opportunity to remotely monitor exercise behavior and facilitate adherence based on objective performance metrics. The primary aim of this project was to characterize the feasibility and processes of remote aerobic exercise data monitoring from a home-based, commercially available platform in individuals participating in the 12-month Cyclical Lower Extremity Exercise for PD II (CYCLE-II) randomized clinical trial. Secondary aims focused on using exercise behavior to classify the cohort into exercise archetypes and describing a shared decision-making process to facilitate exercise adherence.

Methods

Data from each exercise session were extracted, visualized, and filtered to ensure ride integrity. Weekly exercise frequency was used to determine exercise archetypes: Adherent (2–4 exercise sessions per week), Over-adherent (>4 exercise sessions per week), and Under-adherent (<2 exercise sessions per week).

Results

A total of 123 people with PD completed 22,000+ exercise sessions. Analysis of exercise frequency indicated that 79% of participants were adherent; 8% were over-adherent; and 13% were under-adherent. Three case reports illustrate how shared decision-making with the use of exercise performance data points guided exercise prescription.

Conclusions

The number of exercise sessions and completeness of the data indicate that people with PD were able to utilize a commercial, home-based exercise platform to successfully engage in long-term aerobic exercise. Physical therapists can use objective data as a part of a shared decision-making process to facilitate exercise adherence.

Impact

Commercially available exercise platforms offer a unique approach for physical therapists to monitor exercise behavior outside of a clinical setting. The methods used in this project can serve as a roadmap to utilizing data from consumer-based platforms.

Introduction

Physical inactivity has been described as the biggest public health problem of the 21st century¹ and is estimated to contribute to 5.3 million deaths annually worldwide.² Despite the known benefits of regular physical activity and detrimental effects of inactivity,^3,4 it is estimated that less than 10% of adults in the United States achieve the 150 minutes per week of moderate to vigorous exercise⁵ recommended in the Physical Activity Guidelines for Americans set forth by the United States Department of Health and Human Services.^6,7

For people with Parkinson disease (PD), aerobic exercise has been proposed as a universal prescription.⁸ The global benefits of aerobic exercise (eg, enhanced cardiorespiratory fitness), coupled with a reduction in PD-specific symptoms and improved functional mobility, make it an important element of comprehensive disease management.⁹ The importance of using exercise as an adjunct therapy in treating PD was highlighted in 2020 when the Parkinson's Foundation and the American College of Sports Medicine (ACSM) Exercise Guidelines recommended people with PD complete 90–150 minutes of aerobic exercise per week.¹⁰ Despite the known benefits of exercise in PD, people with PD take fewer steps than older adults without the disease (approximately 6500 vs 11,400 steps per day, respectively) and spend a greater amount of time being sedentary (approximately 800 vs 580 minutes per day, respectively).¹¹ While disease severity and physical symptoms explain some of the behavioral differences between people with PD and their peers without PD,¹² nonmotor symptoms such as motivation and apathy may also contribute to reduced activity.¹³ As inactivity impacts older adults and those with PD,¹⁴ it is imperative that the fitness and rehabilitation fields search for innovative approaches to increase exercise activity. A potential solution is to leverage health care providers, including physical therapists, to prescribe and facilitate exercise adherence.¹⁵

Physical therapists are well-positioned to prescribe and promote long-term exercise adherence in clinical settings and beyond. A barrier to the precise prescription and progression of exercise programs is the reliance on participant self-report. In general, self-reported exercise behavior is highly variable and inaccurate,^16–18 and outcome measures that capture self-reported adherence to prescribed exercise programs are lacking.¹⁹ When physical therapists are forced to rely on highly variable, self-reported exercise information, the effectiveness of exercise and potential conclusions drawn about its impact as a treatment intervention are compromised. Precise exercise program prescription and progression can only be achieved if accurate and objective data are available to the provider and patient. Aligning the provider and patient around objective data related to exercise behavior may facilitate a shared decision-making process in rehabilitation.²⁰ Shared decision-making in rehabilitation is a collaborative, interactive process between a patient and their rehabilitation specialist that involves making health-related decisions after an interactive discussion about the risks, benefits, and patient preferences of each option.^20,21 In conjunction with the patient's needs and preferences, shared data may serve as a decision-support tool²² to provide both parties with common data points to facilitate discussion and ultimately the best therapeutic direction.

Commercially available, home-based exercise platforms provide an opportunity for physical therapists to asynchronously monitor exercise behavior outside of a clinical setting. The objective exercise metrics provided by many home-based exercise platforms generate the potential for physical therapists to critically evaluate and refine home exercise prescription by providing an accurate account of the exercise sessions, including exercise frequency, intensity, time, and type. From the perspective of people with PD, the convenience of home-based exercise platforms addresses common barriers to exercise, including time and transportation.^23,24

This manuscript details the use of a commercially available, home-based exercise platform to manage a cohort of people with PD who participated in the Cyclical Lower Extremity Exercise for Parkinson’s Disease II (CYCLE-II) randomized clinical trial.²⁵ The primary aim of the CYCLE-II trial was to determine the disease-modifying capabilities of a 12-month, home-based aerobic exercise program in people with PD. A critical precursor was to ensure the integrity of the data collected from a home-based exercise platform was of sufficient quality to monitor and quantify long-term exercise behavior to appropriately test the hypothesis that regular aerobic exercise can alter the trajectory of PD. This manuscript characterizes the feasibility and processes of remote aerobic exercise data monitoring from a home-based, commercially available platform in people with PD participating in the CYCLE-II trial over a 12-month period. A secondary goal of this manuscript was to demonstrate how exercise adherence data can be used to classify the study population into exercise archetypes and detail how a shared decision-making process was utilized within each archetype.

Methods

As part of a multisite randomized controlled trial (ClinicalTrials.gov; NCT04000360)²⁵ involving the Cleveland Clinic (Cleveland, OH, USA) and the University of Utah (Salt Lake City, UT, USA), people with PD were randomized to a 12-month home-based aerobic exercise intervention. Participants signed a written informed consent approved by the Institutional Review Board at each institution prior to study participation.

Participants were diagnosed with idiopathic PD, using UK Brain Bank Criteria²⁶ with a Hoehn and Yahr Scale (H&Y) stage between I and III²⁷ and demonstrated the ability to safely mount and dismount an upright stationary cycle. The ACSM Preparticipation Health Screen²⁸ was used to stratify for exercise risk; when medical clearance was indicated due to a preexisting condition or active signs and symptoms of cardiovascular, respiratory, or metabolic disease, clearance was obtained through a written letter from the participant’s physician or advanced practice provider clearing the participant for moderate to vigorous exercise. Further details about the screening procedures for CYCLE-II have been published previously.²⁵

Exercise Platform and Remote Monitoring

Cycling was selected as the aerobic intervention based on previous positive results from exercise studies with PD using a cycling intervention.^29–36 Cycling affords several benefits in terms of standardization, safety, and ease of performance in the home environment. A critical factor in selecting a stationary cycle is the high prevalence of postural instability, gait dysfunction, and freezing of gait in people with PD.³⁷ Despite significant gait deficits, the ability to cycle is largely preserved in those with PD.³⁸

Participants received a stationary cycle with an interactive, virtual platform delivered to their home (Peloton Interactive, New York, NY, USA). The Peloton cycle is an upright stationary cycle with a front-mounted computer screen; each bike was fitted with flat pedals with toe cages to ensure proper biomechanical alignment of the lower extremities. The Peloton online library includes thousands of on-demand or live instructor-led classes, scenic routes, game-like environments, and self-directed rides. For the instructor-led classes, the rider can follow along with the instructor by adjusting cadence and resistance per the instructor’s specifications. The instructors generally provide motivational phrases and select music that plays throughout the ride. The instructor-led and scenic rides are of fixed duration (5, 10, 15, 20, 30, 45, and 60 minutes); the self-directed rides are of user-determined duration. Exercise performance variables including class duration (minutes), cadence (revolutions per minute [rpm]), resistance (percentage), distance (miles), output (kJ), heart rate (HR; beats per minute [bpm]), and others are displayed on the screen in real time.

Prior to initiating the home-based cycling program, participants completed an in-laboratory introductory cycling session. The purpose of the 10-minute session was to ensure appropriate seat and handlebar positioning for proper cycling biomechanics and comfort, demonstrate the use of the hardware and software, and inform initial exercise duration and intensity. The platform captured objective performance metrics during the ride, and the therapist recorded the maximal self-reported rate of perceived exertion (RPE) on a 10-point Borg scale.³⁹

Use of the Shared Decision-Making Process in Exercise Prescription and Progression

Principles of the shared decision-making model^20,21 were used to inform monitoring procedures. The physical therapist applied evidence-based information and their clinical knowledge and expertise in exercise prescription, while taking into account participant preferences, values, and circumstances.²⁰ In CYCLE-II, the physical therapist brought a depth of knowledge and expertise related to exercise, PD pathology, and Peloton technology; the participant expressed their exercise facilitators and barriers, their intrinsic motivators and barriers (including the impact of PD in their life), and their social and environmental circumstances. While the goal of the project was clear—exercise 3 times per week at a moderate to high intensity for 12 months—it was achieved through input from and collaboration with both the physical therapist and participant.

During the baseline in-person assessment, principles of the shared decision-making model were introduced. The physical therapist stressed that the approach to exercise adherence required collaboration between the participant and physical therapist to achieve the exercise goal. An unstructured discussion on potential facilitators, barriers, and exercise history was conducted. The physical therapist explained how the participant’s exercise data would be used as a decision-support tool²² in conjunction with the participant’s subjective reporting during biweekly phone calls to achieve the target exercise frequency and intensity. The biweekly phone calls, in addition to addressing any physical or technology barriers, were focused on discussing exercise adherence and exercise prescription.

One physical therapist from each site (Cleveland Clinic and University of Utah) was trained on principles of the shared decision-making model prior to study initiation and performed all exercise assessments, exercise prescription, and follow-up phone calls. Both physical therapists had been practicing for 10–15 years, held doctor of physical therapy degrees, and were Board-Certified Clinical Specialists in Neurologic Physical Therapy.

Exercise Prescription

Based on preliminary data indicating consistent, high-cadence cycling mitigated PD symptoms and facilitated cortical and subcortical central nervous system connectivity compared with slower cadence,^29,30,33 exercise frequency (3 times per week) was prioritized as the most important variable, followed by cadence (75+ rpms). The next priority was increasing exercise duration up to 45 minutes. Age-estimated target HR ranges were considered a lower priority metric due to the high prevalence of autonomic dysfunction resulting in blunted HR response in over 50% of people with PD.^40,41

Exercise studies from the mid-to-late 2000s suggested high-cadence cycling (ie, 80+ rpms) may be beneficial to mitigating PD symptoms²⁹ and increasing cortical and subcortical connectivity.^30,33 The initial cycling studies were conducted on a stationary tandem cycle in which a participant without a neurological disease pedaled in the front position on the tandem bike, and a participant with PD pedaled in the second seat; thus, the captain was driving or “forcing” a high cadence.^29,42 Subsequent studies scaled the high cadence intervention by using a motorized cycle.^{31,32,35,36,43} Individuals in the CYCLE-II study were strongly encouraged to cycle at a cadence of 80 rpms; however, it was acknowledged that PD-related bradykinesia may prevent strict adherence to a high cadence, especially without the assistance from another rider or motor. Moreover, a recent community-based study conducted on standard upright cycles reported that a cadence of 76+ rpms resulted in the mitigation of disease symptoms in participants with PD.⁴⁴ If an individual achieved 75+ rpms or reached a cadence plateau, other exercise intensity metrics were progressed.

Participants were provided a validated wireless chest-worn HR monitor (Wahoo Fitness Tickr, Atlanta, GA, USA)⁴⁵ that connected via Bluetooth to their exercise cycle. Participants were progressed to a vigorous HR zone of 60%–80% of their heart rate reserve (%HRR) using the Karvonen formula: Target HR = ([Maximum HR − Resting HR] × [60% − 80%] + Resting HR).⁴⁶ Resting HR was measured after a 5-minute supine resting period. Maximum HR was estimated as (220 − age)⁴⁷ or (164–0.7 × age) if the participant was on a β-blocker.⁴⁸ Using the ACSM classification of exercise intensity based on %HRR, 30%–39% was classified as light, 40%–59% was classified as moderate, and 60%–89% was classified as vigorous.⁴⁹

Exercise Data Extraction and Management

The primary aim of this project was to characterize the feasibility and processes of remote aerobic exercise data monitoring, collection, and analyses from a home-based, commercially available platform. Feasibility was determined by the ability of the study team to collect the data from all study participants using the commercially available website and the integrity of the data management process.

Through the creation of a shared user account, the physical therapist had access to detailed exercise performance data for all exercise sessions for each participant. Each week, summary data with mean values were downloaded from the online dashboard on the company’s user website and were available for the physical therapist to review. As is common for a virtual exercise platform, summary metrics were available for each exercise session. Because the cycling was in-home and unsupervised, exercise performance data underwent a comprehensive data integrity and management process detailed in the following sections.

Identifying Rides of Sufficient Aerobic Exercise Duration

The initial step of ensuring the integrity of the exercise sessions was to evaluate ride duration using 3 ride characteristics: identify sessions in which exercise duration was missing, identify exercise sessions with a duration of 5 minutes or less, and explore abandoned exercises sessions. Figure 1 provides an overview of the process.

Figure 1.

Overview of the data management process for the Cyclical Lower Extremity Exercise for Parkinson Disease II (CYCLE-II) exercise sessions. Over 22,000 rides were recorded and extracted from the Peloton portal from 123 participants with Parkinson disease. Approximately 12.6% of rides were deemed to be of insufficient aerobic duration (red) and were excluded from the final data set. An additional 11.1% of rides were completed by the same rider on the same day in consecutive order and were combined through either summing (time, output) or a weighted average (heart rate, cadence). In total, 17,093 exercise sessions were represented in the final data set. Ex = exercise.

First, exercise sessions without a duration data point were not included in the final data set due to the inability to confirm the validity of the ride. Second, the virtual exercise platform ride library contains multiple 5-minute warm-up or cool-down sessions. The duration of these sessions do not meet the ACSM definition of physical activity (ie, at least 10 minutes)⁴⁹ and were ultimately not counted in the total exercise time or number of sessions.

Finally, data were evaluated to determine if the ride was completed or abandoned. To facilitate user class selection in the online platform, each ride is labeled with a session duration (eg, “30 min Country Music ride”). On the dashboard, ride duration is automatically populated from the ride title. These data persist even if the ride ended early; hence duration is reported as the number of minutes the class was intended to run, not the number of minutes the participant actually exercised.

To identify abandoned exercise sessions, ride duration was rounded to the nearest 5-minute increment (eg, 10, 15, 20 minutes, etc.); mean work in kJ/min was then calculated for each ride duration. Next, outlier rides, those with kJ/min below 1.5, 2, 2.5, and 3 SD of the mean for each 5-minute increment were examined. Based on the visual assessment of data,⁵⁰ a threshold of 2.5 SD below the mean appeared to conservatively identify abandoned rides. To evaluate the integrity of the 2.5 SD filter, data for each identified session were manually reviewed by study personnel not affiliated with the exercise monitoring or prescription using the second-by-second data visualization tool on the exercise platform website. A total of 360 rides falling below the 2.5 SD (rides excluded) and 290 rides falling within the 2.5 SD (rides included) were inspected. Rides were correctly classified by the filtering process 89.2 and 100% of the time for the below and within 2.5 SD, respectively. Overall, this approach was deemed appropriate, and the 2.5 SD filter was implemented.

Consecutive Rides on a Single Day

Some participants performed 2 consecutive shorter rides in a single day to meet the desired exercise duration for a given day (ie, a 20-minute ride with 1960s music followed by a 20-minute ride with 1970s music). Because the rides were within a single exercise bout, the total output (kJ) was summed. The HR and cadence values were given a weighted average that corresponded with exercise duration. For example, if an individual completed 2 consecutive 20-minute rides, each cadence and HR value was multiplied by 0.5 and then summed for the weighted total average.

Converting Heart Rate Data to Heart Rate Reserve

Mean HR (bpm) was reported for each ride. The Karvonen formula described above was used to calculate % HRR and create an HR metric that was comparable across participants. To ensure Bluetooth data transmission between the HR monitor and cycle was functioning properly, rides where mean exercise HR was less than the participant's resting HR were considered a data transmission error and were removed from the final data set (eg, a mean exercise HR of 39 bpm and a supine resting HR of 72 bpm). In the exercise sessions where HR was missing or invalid, HR was removed from the exercise session; the other metrics were vetted using the aforementioned process and were included in the final data set.

Using Data to Classify Exercise Adherence

To quantify overall adherence, the study team identified exercise archetypes and quantified the number of individuals who fell into these archetypes. Three exercise adherence archetypes were identified: Adherent, Over-adherent, and Under-adherent. The archetypes were based on exercise frequency (prescribed 3 times per week), as that was the prioritized variable in the CYCLE-II study due preliminary data indicating individuals with PD regress to their baseline motor performance following exercise cessation in as few as 4 weeks.²⁹ To account for some level of expected variability over the 12-month period, the Adherent archetype was defined as mean exercise frequency of 2–4 times per week over the 50 week exercise period. The Over-adherent and Under-adherent archetypes were defined as a mean exercise frequency of >4 and < 2 day/week, respectively. A 50-week exercise duration was utilized to account for the plus or minus 10-day visit window for the final assessment.

Acknowledging that medical events occur frequently in older adults (with and without neurological disease),⁵¹ medical events were documented during the 12-month intervention. For this project, a medical event was operationally defined as a reported bodily injury or illness that resulted in a minimum of 2 weeks without a recorded ride. The 2-week duration was based on the adverse impact on body composition, aerobic capacity, muscular strength, and cellular metabolism following 2 weeks of inactivity.^52–55 Medical events included scenarios such as musculoskeletal injury (eg, knee pain), bacterial/viral infection (eg, COVID-19), procedure or surgery (eg, cataract removal), and others. The medical interruption was deemed resolved when the participant completed 2 consecutive weeks with one or more exercise sessions. For those who experienced a medical event, the adherence rate calculation excluded the weeks of the medical interruption.

Results

A total of 129 individuals were randomized to the exercise group. Three individuals withdrew from the study prior to receiving their home-based cycle due to COVID-related concerns (n = 2) and an unrelated illness (n = 1). Three additional individuals withdrew from the study after receiving their home-based bicycle due to unrelated illness (n = 2) and lost to follow up (n = 1). A total of 123 individuals were included in the analysis with the sample evenly distributed between the Cleveland Clinic and the University of Utah (61 and 62 participants, respectively). Participants were 62.7 (8.1) years old, 65.9% male, 95.1% White, average disease duration of 4.2 (3.5) years, and average Movement-Disorder Unified Parkinson’s Disease Rating Scale,⁵⁶ motor portion (MDS-UPDRS III) of 37.7 (15.0) points in the off-medication state. In terms of disease classification, 11 (8.9%) participants were classified as an H&Y I (unilateral involvement), 72 (58.5%) were classified as H&Y II (bilateral involvement without balance impairment), 37 (30.1%) were classified as H&Y III (mild–moderate involvement with balance impairment), and 3 (2.4%) were classified as H&Y IV (severe disability and needs assistance with standing and walking) in the off-medication state.

Data Management & Integrity Process

Figure 1 illustrates the data management and integrity process implemented for the 22,412 recorded rides, of which nearly 88% met the study criteria for exercise sessions of sufficient aerobic exercise duration. Heart rate data were missing in 1125 (5%) of the exercise sessions; average exercise HR values less than the participant’s recorded supine resting HR were present in 162 (0.7%) of the sessions.

Exercise Archetype Classification

The Table details the number of participants in each exercise archetype. Of the 123 participants, 78.9% were classified as Adherent, 8.1% as Over-adherent, and 13% as Under-adherent. Approximately 28% (35 out of 123 participants) reported a significant medical event that rendered them unable to cycle for at least 2 weeks. Median (IQR) length of the medical interruption was 5 (2.5, 11) weeks. In the Under-adherent archetype, 12 out of the 16 participants experienced a medical interruption.

Table.

Description and Classification of Three Exercise Archetypes^a

Exercise Archetype	Description	Total Number of Participants (N = 123)	No Significant Medical Interruption (N = 88)	Significant Medical Interruption Reportedb (N = 35)
Adherent archetype	Mean exercise frequency of 2–4x/wk over the 50-week exercise period	97 (78.9%)	74 (84.1%)	23 (65.7%)
Over-adherent	Mean exercise frequency of >4 d/wk over the 50-week exercise period	10 (8.1%)	10 (11.4%)	0 (0%)
Under-adherent	Mean exercise frequency of <2 d/wk over the 50-week exercise period	16 (13.0%)	4 (4.5%)	12 (34.3%)

^a Data presented as N (%).

^b If significant medical interruption reported, the period of interruption was deducted from the total number of weeks.

Case Reports

Adherent Archetype

A representative participant meeting the Adherent archetype (Fig. 2A) was a 68-year-old retired female with a 2-year history of PD and a self-reported most bothersome symptom of upper extremity tremor; her baseline MDS-UPDRS III score off-medication was 40, H&Y II. Relevant past medical history included osteopenia. She participated in a Silver Sneakers program at her local YMCA and enjoyed walking outdoors with friends when the weather was warm; these activities were self-reported at low to moderate intensity. Facilitators of exercise included a supportive spouse; barriers included regular travel to visit her family several hours away. At her initial assessment, she completed a 10-minute supervised cycling session with an average cadence of 56 rpms, average %HHR of 56%, and an RPE of 7. As part of the shared decision-making process, the participant reported that she responded well to physical activity goals that were challenging but achievable and did admit to becoming discouraged if she was unable to meet a goal. The participant/therapist team decided to set frequent, incremental goals in line with the participant’s preferences.

Figure 2.

Case series cycling data. The shaded region in each figure represents the recommended weekly exercise duration for people with Parkinson disease per the Parkinson’s Foundation and American College of Sports Medicine (90–150 m/wk). For each week, the aggregate length of workouts (min) is represented by the height of the vertical line (left y-axis), and the mean cadence over the course of those minutes is represented by the corresponding circle (left y-axis). (A) Adherent archetype: the participant was exercising 3x/week and progressing well with exercise duration and cadence until experiencing a significant injury at week 41 that resulted in a 4-week period of biking cessation. Following the imposed rest, the participant slowly progressed back from injury to resume the exercise routine and ultimately completed 133 exercise sessions or 2.9 rides per week (calculated out of 46 weeks due to her medical interruption). (B) Over-adherent archetype: the participant was over-adherent to the exercise program due to cycling 5–7 d/wk with a cadence of 80+ rpms and completed a total of 288 exercise sessions or 5.8 rides per week. (C) Under-adherent: the participant was under-adherent to the exercise program with irregular cycling behaviors and eventually cessation of the program after completing a total of 41 exercise sessions or 0.9 rides per week (calculated out of 46 weeks due to a medical procedure). m = minutes; RPM = revolutions per minute.

The initial intervention was prescribed as 3 cycling sessions per week, each for 15 minutes, with an average cadence of 60 rpms due to the high RPE reported during the 10-minute ride. Data from the initial 2 weeks of riding indicated that the participant was successful in achieving the frequency, duration, and cadence goals, and 20-minute rides were gradually incorporated at week 3, starting at one 20-minute ride per week interspersed with the 15-minute rides and gradually progressing to three 20-minute rides. Over the next 6 weeks, the participant adhered to her exercise program, and her average cadence goal increased by 1–2 rpms per week until she reached 65 rpms. The incremental increase in cadence was consistent with her personal preference of setting frequent, achievable goals.

At week 10, 30-minute rides were introduced, starting with 1 per week and increasing to 3 per week. By week 24, the participant was experiencing a plateau with her cadence, which was hovering at ~65 rpms. The participant was frustrated by the stagnation. The physical therapist and participant decided to strategically reduce the exercise duration to 20 minutes and have the participant focus primarily on increasing the average cadence to 70 rpms for 4 weeks. The strategy was successful, and the participant maintained a cadence of ~70 rpms and was able to increase her ride duration to 30 minutes while maintaining that cadence. She was encouraged by this progress which boosted her exercise confidence, and she began incorporating 45-minute rides into her workouts at week 35. The primary reason for gaps in exercise frequency was traveling to visit family.

Unfortunately, during week 41, she experienced a fall at home, unrelated to the study intervention, that resulted in rib bruising and a nondisplaced rib fracture. The subsequent pain prevented her from cycling for 4 weeks. The therapist and participant communicated weekly and discussed a return-to-exercise plan. Once the pain was tolerable and the participant could complete household activities without a significant increase in pain (≤3 on a 10-point pain RPE scale), the participant began with three 10-minute rides per week with 24 hours of rest between rides to monitor postexercise pain with the handlebars of the bike raised to maximum height for upright trunk posture. She increased exercise duration by 5 minutes every week without an increase in pain and eventually resumed 30-minute rides. The postinjury average cadence goal was lowered to 65 rpms, which was very challenging for her to achieve. Data from the virtual platform indicated that her mean cadence varied between about 58 and 68 rpms. Despite her injury resulting in 4 weeks of inactivity, the participant completed a total of 133 exercise sessions or 2.9 rides per week (calculated out of 46 weeks due to her medical interruption).

Over-Adherent Archetype

A representative participant meeting the Over-adherent archetype (Fig. 2B) was a 70-year-old retired male with a 2-year history of PD and a self-reported primary symptom of upper and lower extremity tremor; his baseline MDS-UPDRS III score off-medication was 38, H&Y II. Upon enrollment, the participant reported exercising several days per week on his home elliptical or participating in a local boxing class for people with PD at a moderate intensity. Relevant past medical history was significant for coronary artery disease (cardiac stent placed 10 years ago, currently on a β blocker) and Type II diabetes. Facilitators of exercise included high motivation and time; no substantial barriers to exercise were reported. At his initial assessment, he completed a 10-minute supervised cycling session with an average cadence of 88 rpms, average %HRR of 42% (adjusted for β-blocker medication), and a RPE of 3. As part of the shared decision-making process, the participant reported that exercise consistency and routine were important to him. He felt that if he maintained a disciplined, repetitive routine, he would have the greatest chance of success.

The participant was active at moderate intensities at baseline and achieved a high average cadence during his supervised ride; however, he did endorse discomfort with the bicycle seat and stated that a vigorous cadence coupled with the bike seat may be discouraging. To allow him to acclimate to the seat, he was prescribed alternating 20- and 30-minute rides, 3 times per week, with a target cadence of 77+ rpms. The exercise performance data revealed he was meeting his duration goal and exceeding his frequency goal by cycling 5-7 times per week; however, the cadence goal was challenging, with the participant meeting the target ~50% of the time. During weeks 4 and 5, the physical therapist instructed him to increase his ride duration to 30–40 minutes while maintaining his cadence goal. Again, the data indicated that he was successful with these goals while meeting his cadence goal ~75% of rides. During week 13, the participant and therapist agreed that it was time to increase the ride duration to 45-minute with a cadence goal of 80+ rpms. The data demonstrated that he was able to meet his cadence goal ~90% of the time.

Approximately 6 months into the study, the participant reported signs of exercise burnout (ie, decreased motivation to ride the bicycle, decreased focus and attention during the exercise session, and increased fatigue) during the biweekly phone calls. He expressed a desire to continue his cycling routine most days of the week because he felt a personal benefit from a general health and PD symptom-mitigation perspective. The therapist and participant decided on an exercise schedule that included 1 day of rest (ie, no formal exercise) per week. On the remaining days, the participant would aim for three 45-minute rides per week interspersed with 20-minute rides. Following the change in routine, the participant endorsed less fatigue, and the program was sustainable throughout the 12 months. The participant completed a total of 288 exercise sessions or 5.8 rides per week.

Under-Adherent Archetype

A representative participant meeting the Under-adherent archetype (Fig. 2C) was a 54-year-old male with a 3-year history of PD and a self-reported primary symptom of upper extremity tremor; his baseline MDS-UPDRS III score off-medication was 22, H&Y II. At the time of enrollment, he was working full-time, and his job included frequent traveling; however, due to the COVID-19 pandemic, he was working from home without travel. For exercise, he walked on the treadmill at low intensity 1–2 times per week. Other than PD, he had no relevant past medical history. The facilitators of exercise were a change in his work schedule that allowed him to be at home daily; barriers included long hours of work each day. During the 10-minute supervised cycling session, his average cadence was 84 rpms, average %HRR was 54%, and RPE was 4.

Due to his low baseline activity level, he was prescribed 20-minute rides with a cadence of 75 rpms at a frequency of 3 times per week. He initially expressed high motivation and met his exercise goals for the first 4 weeks. As the COVID-19 pandemic restrictions lifted, the participant began traveling again for work which took him away from home for 1–2 weeks of the month. The shared decision-making process focused on ways to incorporate exercise into his work schedule during the weeks he was at home. In response to the participant feeling overwhelmed with work, family, and study duties, the therapist and participant agreed to decrease the exercise frequency goal to 2 times per week on the weeks he was not traveling to increase the likelihood of success. The shared goal setting was effective for brief periods, but eventually, the participant stopped exercising completely. The participant completed 41 exercise sessions over the 50-week period or 0.9 rides per week (calculated out of 46 weeks due to a medical procedure).

Role of the Funding Source

The protocol described in this publication is supported by the National Institute of Neurological Disorders and Stroke of the National Institutes of Health under award number 2R01NS073717. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Discussion

Remote monitoring of exercise behaviors in people with PD using data from a home-based commercially available exercise platform was feasible and was used as a decision-support tool in a shared decision-making process. More than 22,000 exercise sessions from 123 individuals with PD exercising throughout the United States were successfully monitored. The number of rides and overall completeness of data suggest that participants with PD were able to successfully navigate the exercise platform to complete their prescribed exercise sessions. Data from the exercise session went through a rigorous data cleaning process to ensure that each participant's exercise behaviors were accurately reflected in the final data set. The multistep processing was a key component of accurately quantifying the adherence to the exercise protocol of CYCLE-II and ultimately will facilitate the evaluation of the impact of regular aerobic exercise on PD progression.

Commercially available exercise platforms should be considered by researchers and clinicians as an approach to monitoring exercise behavior outside the clinical environment. Data from this project were collected from 2019 to 2023; the remote monitoring platform was critical to the continuation of the project during the COVID-19 pandemic. The use of remote monitoring, coupled with asynchronous supervision from a physical therapist, rooted in principles of the shared decision-making model resulted in exercise frequency adherence or over-adherence in 87% of participants, suggesting that the strategy facilitated consistent exercise behaviors. Considering the physical therapy plan of care for nearly all conditions includes a home-exercise component, approaches that facilitate patient adherence are helpful in appropriately iterating on that plan of care to facilitate the best outcome. Furthermore, the model of data filtering and its integration into patient–therapist interaction is universally helpful to the profession. The shared decision-making process may have been a key component of the adherence in this project by placing the participant at the center of clinical decisions.²⁰ While a majority of physical therapists see the value of shared decision-making,⁵⁷ implementation is low, and physical therapists often make unilateral decisions with little input from the patient.^58,59 One reason for the low implementation may be the lack of objective performance feedback, especially with home exercise performance. The impact of shared, objective data has the potential to enhance the decision-making process by maximizing the quantity and quality of relevant input to both clinicians and patients/participants.²² Easily interpretable data points are of tremendous value to ensure that therapists and clients are “speaking the same language” in terms of goal setting and evaluating progress. As highlighted by the case reports, the therapist had a comprehensive understanding of what the participant was doing at home through objective data provided by the exercise platform and biweekly interaction with the participant. The shared data fostered trust between the patient and therapist: there was no need for participant to recall or report exercise variables. Subjective feedback from the participant augmented objective outcomes and facilitated open discussions.

The role of the physical therapist is evolving beyond traditional health care delivery environments to models of care that work to increase participation and improve overall health and well-being in people with neurological disease.^60,61 It is neither feasible, nor is it necessary for the vast majority of patients, for physical therapists to be physically present during all aerobic exercise sessions. Until recently, remote monitoring technology has been technologically feasible, but clinically impractical due to a lack of insurance reimbursement. In 2022, updates to the Medicare Physician Fee Schedule added 5 current procedure terminology (CPT) codes related to remote therapeutic monitoring,⁶² creating huge potential for physical therapists to incorporate in-home exercise prescription into their plan of care. Many home-based, commercial platforms offer remote monitoring capabilities that may comply with the requirements set forth by Medicare by providing data points such as exercise frequency, time, and intensity. While the impact of reimbursable remote therapeutic monitoring in the field of physical therapy is new and the full impact has yet to be realized, the methodology described in this manuscript provides a road map for clinicians, researchers, and health care systems interested in the data extraction and management process.

Nearly 30% of participants in this project experienced a medical interruption of 2 weeks or more. Data from the Lifestyle Interventions and Independence for Elders (LIFE)⁶³ project highlighted the impact of medical interruptions in an older adult population. The LIFE study, which randomized over 1600 older adults into a physical activity program or a health education program and followed them for an average of 2.6 years, reported that over 58% of participants in the physical activity program experienced a “medical leave” during the course of the study due to a health condition; the median (IQR) duration of medical leave was 49 (21–140) days.⁵¹ The high rate of medical interruptions in older adults presented by our project and the LIFE project, in conjunction with our data indicating that those who experienced a medical interruption were more likely to be under-adherent to the intervention, underscores the role of a physical therapist in guiding these individuals back to physical activity and exercise. The Adherent archetype details how objective metrics (eg, exercise frequency, duration, and cadence) were coupled with participant feedback (eg, pain scale and self-report) to resume exercise in a safe and efficient manner.

The randomized controlled clinical trial, CYCLE-II, aimed at evaluating the disease-modifying capability of aerobic exercise, utilized the described data monitoring and filtering techniques to precisely characterize and report exercise behavior in people with PD.²⁵ To draw accurate conclusions about the impact of aerobic exercise, we must first confirm that the exercise data can be trusted. This manuscript provides the road map to ensuring the integrity of the exercise performance metrics for the CYCLE-II trial and provides guidance for other large exercise trials.

Considerations and Limitations

While this report summarizes the feasibility of remote aerobic exercise monitoring in people with PD, several things should be considered when looking to expand the prescription of aerobic exercise in a home environment. There is a cost associated with home-based fitness platforms. The Peloton indoor cycle costs about $1250 and $49 per month for the subscription—a price that may serve as a barrier. The data collected were derived from a single type of commercially available stationary bike and virtual platform (Peloton Interactive, New York, NY, USA). However, the data management and cleaning techniques utilized were performed on common summary exercise metrics that are typically provided by multiple aerobic exercise platforms and are therefore applicable beyond the platform utilized in this project. Potential modifications to the methods described in this manuscript may be necessary based on specific exercise goals, mode of exercise, and exercise platform.

This project did not place a large emphasis on HR as a measure of intensity since a maximal exercise stress test was not performed, and thus maximal HR was age-estimated. We cautiously interpret the %HRR metric in this population, as over 50% of people with PD experience a blunted HR response to exercise.^40,41 Thus, the %HRR calculation likely under-estimated the level of intensity at which the participants were exercising.

While the in-home exercise model coupled with the shared decision-making approach was successful in facilitating adherence in the majority of participants, ~13% of our sample was under-adherent to the exercise frequency prescription. Continued investigation of ways to promote exercise adherence in people with PD and in the general population who do not respond to a shared decision-making approach should occur.

Conclusions

By embracing remote exercise data monitoring, physical therapists have the potential to shape the way exercise prescription and management is conducted outside of a clinical setting. The exercise and rehabilitation field can evolve beyond heavy reliance on self-reported exercise recall to a partnership between a physical therapist and patient that revolves around easily interpretable shared data points. Considering the limited amount of time an individual spends with a physical therapist, commercially available platforms offer unprecedented insight into real-world exercise behaviors. The ability to precisely monitor exercise behavior both in and out of the clinic may allow rehabilitation specialists to offer more precise and individualized exercise prescription.

Author Contributions

Anson Rosenfeldt (Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Writing—original draft), Cielita Lopez-Lennon (Data curation, Project administration, Writing—review & editing), Erin Suttman (Data curation, Project administration, Writing—review & editing), A. Jansen (Data curation, Project administration, Writing—review & editing), Kelsey Owen (Formal analysis, Software, Visualization), Leland Dibble (Data curation, Funding acquisition, Investigation, Methodology, Project administration, Writing—review & editing), and Jay Alberts (Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Writing—review & editing)

Ethics Approval

This study received approval from the Institutional Review Boards of the Cleveland Clinic (IRB of record) and the University of Utah.

Funding

This study was supported by a grant from the National Institute of Neurological Disorders and Stroke of the National Institutes of Health (2R01NS073717-06A1).

Clinical Trial Registration

This study was registered on ClinicalTrials.gov (NCT04000360).

Disclosure

The authors completed the ICMJE Form for Disclosure of Potential Conflicts of Interest and reported no conflicts of interest.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.