Comparative Effectiveness of mHealth-Supported Exercise Compared With Exercise Alone for People With Parkinson Disease: Randomized Controlled Pilot Study

Abstract

Background

Declining physical activity commonly occurs in people with Parkinson disease (PD) and contributes to reduced functional capacity and quality of life.

Objective

The purpose of this study was to explore the preliminary effectiveness, safety, and acceptability of a mobile health (mHealth)–mediated exercise program designed to promote sustained physical activity in people with PD.

Design

This was a 12-month single-blind (assessor), pilot, comparative-effectiveness, randomized controlled study.

Methods

An mHealth-mediated exercise program (walking with a pedometer plus engagement in planned exercise supported by a mobile health application) was compared over 1 year with an active control condition (walking with a pedometer and exercise only). There were 51 participants in a community setting with mild-to-moderately severe (Hoehn and Yahr stages 1–3) idiopathic PD. Daily steps and moderate-intensity minutes were measured using a step activity monitor for 1 week at baseline and again at 12 months. Secondary outcomes included the 6-Minute Walk Test, Parkinson Disease Questionnaire 39 mobility domain, safety, acceptability, and adherence.

Results

Both groups increased daily steps, moderate-intensity minutes, and 6-Minute Walk Test, with no statistically significant between-group differences observed. In the less active subgroup, changes in daily steps and moderate-intensity minutes were clinically meaningful. An improvement in the Parkinson Disease Questionnaire 39 mobility score favored mHealth in the overall comparison and was statistically and clinically meaningful in the less active subgroup.

Limitations

The limitation of the current study was the small sample size.

Conclusions

Both groups improved physical activity compared with expected activity decline over 1 year. The addition of the mHealth app to the exercise intervention appeared to differentially benefit the more sedentary participants. Further study in a larger group of people with low activity at baseline is needed.

The Parkinson disease (PD) population is largely sedentary.¹ Studies have revealed substantial reductions in physical activity in people with PD compared with healthy older adults, even early in the course of the disease.^2,3 Furthermore, data from longitudinal cohort studies indicate that, on average, the number of steps that people with PD walk per day declines by approximately 12% over 1 year.^4,5 The implications of declining activity are alarming, given the known negative consequences for physical capacity of limited engagement in exercise, and the known benefits of physical therapy, exercise, and physical activity on reducing disability and enhancing quality of life in people with PD.^6–8

Behavioral change strategies provide a theoretically based approach for helping people to develop the skills they need (eg, goal setting, self-monitoring) to successfully engage in sustained exercise behavior over the long term.⁹ Behavioral change interventions have resulted in increased physical activity in healthy older adults and in those with other chronic conditions (ie, diabetes, arthritis).¹⁰ Indeed, physical therapist exercise interventions commonly employ rudimentary strategies to promote behavioral change, including goal setting, tailored instruction, and the provision of feedback. More robust strategies, however, would include more frequent opportunities for the provision of feedback, mastery experiences, self-reflection, vicarious experiences, and more extended support from a health care professional. Few studies have investigated the effectiveness of such enhanced behavioral approaches to increase physical activity in people with PD. In a randomized controlled trial (RCT) in PD conducted in The Netherlands, those who received in-person physical therapy that included a behavioral change intervention increased objectively measured physical activity levels over a 2-year period, compared with a standard physical therapy condition.¹¹ However, it might not be feasible for many people with PD to access ongoing in-person physical therapy over an extended period.

Mobile health (mHealth) applications (“apps”) offer a remote mechanism to enhance important behavioral change elements of in-person–delivered exercise interventions, thereby increasing the reach and scalability of physical therapist services.^12–14 The delivery enhancement of mHealth apps includes: (1) ongoing, rather than episodic, goal setting; (2) provision of real-time, rather than delayed, feedback to reinforce positive behavioral change; (3) more frequent adjustments of individually tailored programs; and (4) more continuous connections with professionals through messaging features. For exercise programs, mHealth apps also can provide videos of the exercises, thereby enhancing proper technique with visual instruction. From a patient's perspective, mobile health apps enable physical therapist interventions to be integrated into everyday life, allowing greater flexibility, adaptability, and relevance to the individual.¹² A systematic review of effectiveness of mobile apps among healthy people and in those with chronic conditions (eg, diabetes, chronic lung disease) revealed statistically significant improvements in health outcomes.¹⁵ Effective apps are theoretically grounded and apply evidence-based behavioral-change approaches shown to be effective when delivered in-person, on websites, or using printed materials, with the added advantage of reaching patients remotely at any time and place.^16,17

In this study, we conducted an early phase, pilot RCT to explore the extent to which mHealth technology might enhance outcomes associated with a physical therapist–prescribed intervention designed to increase physical activity in people with PD. Accordingly, we compared the preliminary effectiveness over 1 year of (1) walking with a pedometer plus engagement in planned exercise supported by the mHealth app with (2) walking with a pedometer and exercise with no mobile technology. Program adherence, safety, acceptability, and resulting changes in quality of life and physical function were also compared. We hypothesized that both conditions would slow the anticipated decline in physical activity levels associated with PD. We further hypothesized that physical activity levels would increase among sedentary participants in both conditions, with greater improvement occurring in the mobile technology–mediated condition.

Methods

Study Design

We conducted a 12-month single-blind (assessor), comparative-effectiveness RCT comparing a mobile health–mediated exercise program (“mHealth” condition) with an exercise program administered without mobile health technology (“active control” condition) in people with PD. Prescreening was by telephone. Participants who met screening criteria underwent baseline assessments followed by random assignment to the mHealth or active control intervention condition. Computerized stratified randomization procedures were completed by an independent statistician based on gender and Hoehn and Yahr score (1–2 or 2.5–3).¹⁸ Assignments were concealed from all other study staff until participants met screening criteria and completed baseline assessments. The intervention period initially lasted 6 months but was later extended to 12 months to assess the effectiveness of the intervention over a longer time frame. After 6 months, participants were asked if they would like to continue in the program for an additional 6 months.

Participants

Participants were recruited from Boston University Medical Center, Boston University, Center for Neurorehabilitation, and Fox Trial Finder. Participants had idiopathic PD as confirmed by the UK Brain Bank criteria,¹⁹ were older than 18 years, had not been exercising at moderate intensity >3 d/wk for 30 minutes over the past 3 months, had mild-to-moderate disease severity (Hoehn and Yahr score 1–3), were able to walk without assistance or an assistive device for 6 minutes, and had been receiving a stable dose of medication for ≥ 2 weeks. Patients were excluded if they had 2 or more falls in the past month, cognitive impairment as indicated by the Montreal Cognitive Assessment score < 24, significant freezing (>2 on Freezing of Gait Questionnaire, item #3), or other medical conditions that would limit safe participation in an exercise program.

Intervention

Approximately 1 week following baseline assessment, participants in both conditions participated in 1 or 2 initial in-person visits with a licensed physical therapist (T.R.D.) with expertise in PD to develop an individualized exercise and walking program. The exercise component of the program was developed from a predetermined set of exercises based on the American Parkinson Disease Association “Be Active and Beyond” program (Tab. 1) and tailored to the needs and preferences of each participant.²⁰ Strengthening exercises have been shown to improve physical function in people with PD and were therefore included as a main component of the intervention.^21,22 Participants were asked to do 5 to 7 exercises for ≥ 3 d/wk. The walking component of the home program consisted of an individualized recommended range of steps per day (eg, 5000–7500 or 7500–10,000) that was determined from each participant's baseline activity level (using the StepWatch Activity Monitor; Modus Health LLC, Edmonds, WA, USA). A decline in walking is a known marker of disability in PD²³ and hence an important target of intervention. Walking as a form of exercise has been shown to improve functional outcomes and quality of life in people with PD.^24,25

Table 1.

Exercise Intervention

Strengthening
Level	Squat	Lunge	Hip Abduction	Heel Raises	Bridges	Quadruped	Abdominals	Push-Up	Plank
Strength (level 1)	Sit to stand	Reverse step lunge with support	Standing abduction with support	Double leg	Basic lift, hands on floor	Alternating arm lifts	Bent-knee fall outs	Wall or counter push-up	Plank on elbows and knees
Strength (level 2)	Elevated surface or partial squat	Reverse step lunge with no support	Side-lying clamshell (knee bent, heels together, hip abduction)	Single leg	Lift with arms elevated off floor	Alternating leg lifts	Partial sit-up	Modified floor push-up on knees	Plank on elbows and toes
Strength (level 3)	Full squat	Walking lunges	Standing abduction without support/add speed	Single leg with 5-s hold at end range	Lift with alternate single-leg extension	Opposite arm leg lifts, alternating	Oblique exercise	Standard floor push-up on toes	Plank on hands and toes with side bend/knee lifts to side
Stretching
	Hamstring	Hip flexor	Calf	Trunk	Thoracic spine	Hip extensors	–	–	–

Both components of the program included elements designed to promote behavioral change. These included participant goal setting, tailoring the program to participants’ preferences, and the provision of feedback through activity trackers. In addition, reassessments occurred at 3 and 6 months to enhance accountability. Participants subsequently performed the program independently in their respective home and community settings.

mHealth condition

The mHealth condition was designed with cognitive-behavioral elements to enhance the basic behavioral change component of the individualized exercise and walking program and to emphasize participants’ engagement in managing their health condition. Compared with the active control condition, the mHealth approach contained additional key elements shown to be effective in facilitating behavioral change among healthy older adults and those with other chronic conditions; those elements included remote monitoring, more accessible communication, and more frequent program adaptation by a physical therapist. Incremental walking and exercise goals were entered into the mHealth app. Action plans were developed to include what (which exercises, duration of walking), how (appropriate technique), when (time of day, days per week), and where (community, mall) each participant was to engage in exercise. Notifications (ie, automated prompts and reminders) were used to motivate participants to complete their exercise and walking programs. Exercises were adapted remotely over time by the physical therapist in response to improvements or setbacks experienced by participants. Adherence and progress toward goals was graphically displayed, allowing participants to track their own performance over the intervention period.

Through the Wellpepper app (Wellpepper, Seattle, WA, USA), the physical therapist remotely monitored data on adherence, pain, and level of difficulty each week. Changes to the exercise program were made via the app approximately 2 to 3 times per month based on the progress of each participant. When changes to the program were deemed appropriate, the physical therapist contacted participants via the text-messaging feature to check in and convey potential changes to the program. This contact ensured collaborative agreement between the physical therapist and participants before changes were made. Beyond this planned contact, other events triggered contact between the physical therapist and participants. These included a change in the pattern of exercise adherence (ie, unexplained lack of engagement > 1 week), a patient-reported acute health condition that warranted a change in the exercise program, or a patient-initiated question about the program.

All participants were provided with an iPad (Apple, Cupertino, CA, USA) with a cellular connection delivered through a data plan (at no cost to participants) for the 12-month study duration. Participants accessed the exercises via the Wellpepper mobile app preloaded onto each iPad (Fig. 1). At the initial visit, participants were instructed in the exercises, and videos were recorded of participants performing the exercises using correct form. The videos were individually tailored to ensure a “just-right level” of challenge. The videos and exercise prescription information (sets, repetitions, and auditory instructions) were added to the participant's Wellpepper account. During the session, participants were also trained in the use of both the iPad and the Wellpepper app. At home, participants accessed their exercises through the app and reported exercise completion, pain, and level of difficulty following each exercise. Participant-reported information was monitored remotely by the physical therapist. The Wellpepper app also enabled communication between the patient and physical therapist via a 2-way text-messaging feature. During the intervention, the physical therapist added or removed exercises based on each participant's response to their exercise program or changes to their health status. Exercises were also progressed to keep the program dynamic and adequately challenging for each participant.

Figure 1.

Wellpepper exercise application. Courtesy of Wellpepper Inc.

mHealth participants were provided with a commercially available activity tracker, the Fitbit Zip (Fitbit Inc., San Francisco, CA, USA), to be worn on their waist to provide feedback on steps taken. The Fitbit Zip has been shown to be more accurate in tracking step counts compared with other commercially available activity trackers in people with PD.²⁶ The tracker collected activity information and synced with a Fitbit app on the iPad via a Bluetooth connection. The data from the Fitbit app automatically synced with the Wellpepper app so that participants could access both exercise and activity information in the same digital location. Progress regarding steps walked per day could be viewed over time (ie, past week or month), thereby enhancing participants’ ability to self-monitor steps taken. Participants were educated on how to safely increase their step totals during the intervention. The physical therapist remotely monitored participants’ daily step totals and made adjustments, as needed, to the target goal during the intervention.

Active control condition

The active control condition used traditional modalities (without mHealth enhancement) to deliver the individualized exercise and walking program and its associated standard behavioral change elements.

All exercises (Tab. 1) were issued in a paper format with photos, general instructions on technique, and specific information regarding prescription (sets, reps, and helpful cues). At the initial 1 to 2 visits, participants received exercise instruction, developed goals and action plans, and were provided with recommendations for progression of each exercise over the study period. Participants were given paper calendars to track the days they performed the exercises and instructed to submit them monthly via postal mail to the research team. After a 48-hour follow-up call following the initial 1 or 2 visits, no scheduled or regular contact was initiated by the physical therapist.

Participants were provided with a commercially available pedometer (Omron Hj-113 Pocket Pedometer; Omron Healthcare Inc., Kyoto, Japan) worn on their waist to provide feedback on daily steps taken. Unlike the Fitbit Zip, feedback was not provided graphically over time (ie, past week or month), thereby offering less robust behavioral change enhancements. Waist-worn activity trackers have been shown to be more accurate in tracking step counts compared with wrist-worn devices.²⁶ At the initial visit, participants were advised on how to progress their daily step counts during the intervention. At home, participants logged and submitted their daily steps in the paper calendar as described earlier. The Fitbit Zip and the Pedometer were used only as a means to provide participants with feedback regarding step counts. A more robust, validated, research-grade monitor (StepWatch Activity Monitor) was used to measure changes in physical activity as a study outcome.

Outcome Measures

All study procedures were completed at the Center for Neurorehabilitation, Sargent College, Boston University. Patients were assessed by a trained physical therapist (K.H.) who was blinded to group assignment. Assessments occurred at baseline and at 12 months during the “on” state. A subset of the assessments was conducted at 3 and 6 months.

Physical activity

Physical activity level was measured during a 7-day assessment period following the baseline, 3, 6, and 12-month assessment sessions using the StepWatch Activity Monitor (SAM). The SAM is the size of a pager, weighs only 38 g, attaches at the ankle with Velcro closures, and has been validated previously in studies of people with PD.^5,27,28 Participants were instructed to wear the SAM on the leg with the least severe motor impairment (based on Movement Disorder Society Unified Parkinson Disease Rating Scale [MDS UDPRS] motor score) during waking hours, except when bathing, showering, or swimming. Monitors were configured by a member of the research team to record stride counts in 1-minute intervals and with optimal sensitivity based on each participant's individual gait pattern. For each participant, the step-counting accuracy of the SAM was verified during the first few minutes of recording by comparing monitor step counts, identified by a flashing indicator light, with visual observation. Step activity monitors were returned in person to study personnel during the initial 1 to 2 visits and via postal mail thereafter. One investigator (J.T.C.) used the SAM manufacturer's software to transform recorded 1-minute stride counts into step counts (step count = stride count × 2). Mean daily values were calculated for (1) the total accumulated number of steps, and (2) the total number of minutes of moderate-intensity stepping, operationalized as the number of 1-minute step counts in excess of 100.²⁹

Health-related quality of life

Mobility-related health-related quality of life was measured using the Parkinson Disease Questionnaire 39 (PDQ-39) mobility domain. Self-reported disability in gait-dependent activities including leisure activities, housework, and carrying shopping bags was assessed. The reliability, validity, and responsiveness to change of the PDQ-39 and the mobility domain have been established in people with PD.^30,31

Walking capacity

Walking capacity was assessed using the 6-Minute Walk Test (6MWT), a measure of the maximum distance a participant can walk in a 6-minute period. Each participant was instructed to cover as much ground as possible. The 6MWT is a safe, valid, reliable, and responsive measure of walking capacity in PD.^11,32

Adherence, safety, and acceptability

Exercise adherence data were collected via daily records of steps taken and exercises performed, using either the mobile health application (mHealth group) or paper calendars (active control group). Safety was assessed by having a blinded research assistant call participants monthly to monitor adverse events and to ask a standard set of questions about recent health events and their relationship to the intervention. Program acceptability was assessed after 12 months by having participants rate their satisfaction using a 1 to 10 Likert scale (0 = not satisfied; 10 = highly satisfied). Participants also were asked at that time if they were interested in continuing the program (yes/no) and if they would recommend it to others with PD (yes/no).

Data Analysis

Baseline characteristics of participants were analyzed according to their randomized assignment (intent-to-treat) and are presented as mean and standard deviations for continuous measurements or number and percent for discrete data. Primary hypotheses were tested using an analysis of covariance (ANCOVA) with an intervention factor (mHealth or active control), adjusted to baseline value. To assess effect of intervention at the end of the study, outcomes were expressed as change from baseline to month 12, and their estimates were presented as mean differences with 95% CIs. Feasibility of the statistical assumptions was explored graphically and quantitatively. Secondary subgroup analyses among relatively sedentary participants were performed for people with an average number of steps < 7500 at baseline.^29,33 The robustness of our findings was asserted by sensitivity analyses conducted with inclusion of all available data from 3- and 6-month time points. To investigate potential modification of treatment effect due to inclusion of midpoint measurements, we used a mixed-model framework with intervention, visit, and intervention-to-visit interaction factors, adjusted to baseline value of outcomes. Data in the bar plot graphs are presented as mean change from baseline and 95% CIs. Adherence, safety, and acceptability data were analyzed descriptively. Based on the effect sizes from our prior longitudinal trial, we anticipated that the sample of 25 participants per group (n = 50 for both arms) would detect, with 80% power, a difference of 1000 steps between arms, (standard deviation [SD] = 1175 steps) assuming a 15% attrition rate over 1-year follow-up and 2-sided α of .05. Statistical analyses were performed using SAS software v.9.3 (SAS Institute, Cary, NC, USA) and R software v.3.2.5 (R Foundation, Vienna, Austria).

Role of the Funding Source

The funder played no role in the design, conduct, or reporting of this study.

Results

We screened 155 volunteers by phone, of whom 60 were eligible for in-person screening. A total of 51 people with PD met inclusion criteria and were randomized. Seven participants withdrew from the study over the course of 12 months (3 in mHealth, 4 in active control; Fig. 2).

Figure 2.

Flow diagram of patient participation in the study. FOGQ = Freezing of Gait Questionnaire; MoCA = Montreal Cognitive Assessment; PD = Parkinson disease.

Baseline Characteristics

Participants were aged 64.1 ± 9.5 years at baseline with average disease duration of 4.8 ± 3.1 years. Fifty-five percent of participants were male. Participants in the mHealth and active control groups had similar demographic characteristics (Tab. 2). Mean numbers of daily steps were 8478 ± 3699 versus 8902 ± 2967 for mHealth and active control cohorts, respectively. The moderate-intensity stepping metric (ie, number of minutes containing >100 steps), PDQ-39 mobility domain, and 6MWT were also similar between groups at baseline.

Table 2.

Baseline Characteristics by Subgroup and Combined^a

Baseline Characteristics	Combined (n = 51)	mHealth (n = 26)	Active Control (n = 25)
Age (y)	64.1 (9.5)	64.8 (8.5)	63.3 (10.6)
Disease duration and median (y) (IQR)b	4.8 (3.1)	5.9 (3.5)	3.7 (2.1)
	5 (2, 7)	6 (3, 9)	4 (2, 5)
Total no. of comorbidities	1.8 (1.4)	1.8 (1.5)	1.7 (1.4)
Hoehn and Yahr stage, mean (SD), no. (%)	2.1 (0.3)	2.0 (0.3)	2.2 (0.4)
1	1 (2.0)	1 (3.9)	0 (0)
1.5	2 (3.9)	1 (3.9)	1 (4.0)
2	38 (74.5)	20 (76.9)	18 (72.0)
2.5	7 (13.7)	4 (15.4)	3 (12.0)
3	3 (5.9)	0 (0)	3 (12.0)
Male sex, no. (% male)	28 (54.9)	15 (57.7)	13 (52.0)
Living alone	2 (3.9)	1 (3.8)	1 (4.0)
Living with others	49 (96.1)	25 (96.2)	24 (96.0)
High school/some college	7 (14.0)	4 (15.4)	3 (12.5)
College graduate or higher	43 (86.0)	22 (84.6)	21 (87.5)
Retired	24 (47.1)	11 (42.3)	13 (52.0)
Race (white)	51 (100)	26 (100)	25 (100)
Parkinson Disease Questionnaire 39 (PDQ-39) index	11.5 (6.1)	10.8 (4.9)	12.2 (7.2)
PDQ-39 mobility score	8.7 (8.5)	9.2 (8.5)	8.1 (8.7)
6-Minute Walk Test (m)	521.5 (88.0)	505.4 (72.0)	538.2 (100.7)
Movement Disorders Society: Unified Parkinson Disease Rating Scale motor score	29.6 (10.0)	31.6 (10.7)	27.6 (9.1)
Montreal Cognitive Assessment score	27.5 (1.7)	27.4 (1.6)	27.7 (1.7)
Geriatric Depression Scale score	5.0 (3.6)	4.9 (3.7)	5.1 (3.6)
Step Activity Monitor (steps/d)	8686 (3334)	8478 (3699)	8902 (2967)
Activity monitor worn, mean (d) (SD)	7.3 (1.5)	7.3 (1.3)	7.3 (1.6)

a

Data are reported as mean (standard deviation) for continuous measurements and as number and percent for discrete data.

b

For duration of disease, medians and interquartile ranges are reported below mean and SD. IQR, interquartile range.

Physical Activity

Among 44 participants who completed the study, the estimated mean change in daily steps at 12 months was 102.6 steps (95% CI = −888 to 1093) in the mHealth group compared with 159 steps (95% CI = −878 to 1195) in the active control group (Fig. 3; Tab. 3; eTab. 1, available at https://academic.oup.com/ptj). The difference between groups at end of the study was not statistically significant (−56 steps, 95% CI = −1494 to 1382; P = .94). Estimated mean change in the number of moderate-intensity minutes per day at 12 months was 17.4 minutes (95% CI = −17.2 to 52.0) and 12.3 minutes (95% CI = −23.9 to 48.5) for the mHealth and active control groups, respectively. The difference between groups was not statistically significant (−5.2 min, 95% CI = −55.3 to 45.0; P = .84).

Figure 3.

Estimated changes from baseline for daily steps and number of minutes of moderate-intensity activity (results for entire cohort and subgroup of participants with < 7500 steps/d at baseline). Bar plots and error bars represent estimated mean changes from baseline and 95% CIs by intervention arm, extracted from ANCOVA model. P values for the between-group effect for outcomes defined as a change from baseline. Asterisk (*) indicates significant P value (< .05) for within-group change from baseline.

Table 3.

Summary of Results for Primary and Secondary Outcomes in Each Treatment Arm

		mHealtha		Active Controla
Sample	Variable or Test	Baseline Mean (SD)	12-mo Mean (SD)	Baseline Mean (SD)	12-mo Mean (SD)	Mean Adjusted Difference Between Groupsb (95% CI)
Entire sample	Step activity monitor, steps/d	8478 (3699)	8457 (3184)	8902 (2967)	9028 (3366)	−56.2 (−1494.3 to 1382.0)
	No. of minutes containing ≥ 100 steps	12.3 (13.4)	14.2 (14.8)	10.4 (7.9)	12.6 (12.9)	−5.2 (−55.3 to 45.0)
	Parkinson Disease Questionnaire (PDQ-39) index	10.8 (4.9)	11.4 (5.9)	12.2 (7.2)	13.4 (8.1)	−1.3 (−4.7 to 2.1)
	PDQ-39 mobility score	9.2 (8.5)	7.1 (6.6)	8.1 (8.7)	9.4 (10.6)	−3.8 (−7.8 to 0.20)
	6-Minute Walk Test	505.4 (72.0)	536.0 (92.4)	538.2 (100.7)	546.9 (105.5)	28.5 (−14.4 to 71.5)
Participants with < 7500 steps/d reported at baseline	Step activity monitor, steps/d	6028 (1046)	6918 (1900)	6330 (560)	6788 (1636)	322.0 (−1326.2 to 1970.2)
	No. of minutes containing ≥ 100 steps	4.1 (4.1)	12.7 (10.0)	5.1 (4.3)	7.3 (7.2)	−40.3 (−101.9 to 21.4)
	PDQ-39 index	10.5 (4.5)	11.6 (6.6)	13.3 (7.6)	16.4 (9.5)	−3.9 (−10.3 to 2.4)
	PDQ-39 mobility score	9.6 (6.4)	8.8 (6.7)	6.3 (6.9)	13.5 (14.2)	−8.2 (−15.4 to −0.93)
	6-Minute Walk Test	497.6 (60.2)	528.9 (99.2)	530.3 (115.7)	521.3 (99.6)	23.4 (−49.4 to 96.3)

a

Mean (standard deviation, SD) by intervention arm at baseline and at the end of the study.

b

Difference, adjusted to baseline value, extracted from ANCOVA model.

Among less active participants (ie, patients with < 7500 steps/d at baseline), the estimated mean change in daily steps in the mHealth group was 763 steps (95% CI: −273 to 1800) compared with 441 steps (95% CI = −831 to 1713) in the active control group (Fig. 3; Tab. 3; eTab. 2, available at https://academic.oup.com/ptj). The between-group difference was not statistically significant (322 steps, 95% CI = −1326 to 1970; P = .69). The increase in daily steps in the mHealth condition approximates the minimally clinically important difference (MCID) value of 779 daily steps associated with small, clinically meaningful changes in ambulation reported in people with multiple sclerosis.³⁴ Participants in the mHealth group displayed a greater and statistically significant increase (55.7 min, 95% CI = 16.8–94.5; P = .01) in weekly moderate-intensity minutes of walking compared with the active control group (15.4 min, 95% CI = −32.2 to 63.0; P = .50); however, the difference between groups was not statistically significant (P = .19).

Mobility-Related Quality of Life

PDQ-39 mobility domain scores for the mHealth group declined from baseline to 12 months, which reflected improved mobility (−1.7 points, 95% CI = −4.4 to 1.1; lower score indicates better mobility). During the same period, the active control group scores increased (ie, had worse mobility) by 2.1 points (95% CI = −0.76 to 5.0) (Fig. 4; Tab. 3; eTabs 1–3, available at https://academic.oup.com/ptj). The between-group difference did not cross the threshold level of .05 α (estimated mean change: −3.8 points, 95% CI = −7.8 to 0.2; P = .06); however, the magnitude of the difference was clinically meaningful.³¹ Participants with lower activity at baseline reported better mobility-related quality of life in the mHealth condition compared with the active control condition, with a statistically significant and clinically meaningful difference in the change in PDQ-39 mobility over 12 months between groups (−8.2 points, 95% CI = −15.4 to −0.9; P = .03). An increase of 3.2 in the PDQ-39 mobility score has been associated with clinically meaningful worsening of mobility-related quality of life.³¹

Figure 4.

Estimated changes from baseline for Parkinson Disease Questionnaire 39 (PDQ-39) mobility domain and 6-Minute Walk Test (results for entire cohort and subgroup of participants with < 7500 steps/d at baseline). Bar plots and error bars represent estimated mean changes from baseline and 95% CIs by intervention arm, extracted from ANCOVA model. P values for the between-group effect for outcomes defined as a change from baseline. Asterisk (*) indicates significant P value (< .05) for within-group change from baseline.

Walking Capacity

The change in 6MWT from baseline to 1 year (33.8 m, 95% CI = 5.1–62.5) was statistically significant (P = .02) and could be considered clinically meaningful for the mHealth group but not the Active control group (5.3 m, 95% CI = −25.6 to 36.2) (Fig. 4; Tab. 3; eTabs 1–3, available at https://academic.oup.com/ptj). However, the difference in the change scores between groups was not statistically significant (28.5 m, 95% CI = −14.4 to 71.5; P = .19). Among less active participants, there was a nonsignificant between-group difference in improved distance walked (23.4 m, 95% CI = −49.4 to 96.3; P = .51); however, within-group changes in the mHealth group were similar to those in the whole sample (29.1 m, 95% CI = −16.8 to 75.0; P = .20) although this improvement was not significant. Clinically meaningful changes in the 6MWT have not been established; however, changes of 20 and 50 m have been associated with small and substantial meaningful changes, respectively, in community-dwelling older adults and in people who have had a stroke.³⁵

Adherence, Safety, and Acceptability

Adherence to the exercise program over the 48-week study period was similar between groups. Average number of days per week when participants used the application was 3.2 in the mHealth cohort. In the active control group, based on records reported on calendars, participants on average spent 3.5 d/wk on the exercise program. The number of adverse events per condition are included in Supplemental Materials (eTab. 3). There were no differences in adverse events between conditions and no serious adverse events related to the exercise programs. Nine of the minor musculoskeletal adverse events could have been related to the exercise program (3 in the mHealth condition; 6 in the active control condition); all resolved without medical attention. No falls occurred while exercises were being performed. Participants in both groups were relatively satisfied with the program after 12 months—mean ratings = 8.7 points (mHealth), 8.5 points (active control). In the mHealth group, 82% of participants wanted to continue the program, and 100% would have recommended it to others. In the active control group, 70% of participants wanted to continue the program and all but 1 participant would have recommended it to others.

Sensitivity Analyses

Analyses performed with inclusion of 3- and 6-month records showed similar results for all analyzed outcomes. The estimated mean between-group difference from baseline at 1 year, extracted from a mixed-model framework, for daily steps was −47.3 steps (95% CI = −1477 to 1383; P = .95). The moderate-intensity minutes metric (0.64 min, 95% CI = −6.5 to 7.8; P = .86), PDQ-39 mobility domain (−3.7 points, 95% CI = −7.7 to 0.32; P = .07), and 6MWT (18.3 m, 95% CI = −24.1 to 60.7; P = .39) also did not show any effect modification due to inclusion of midpoint measurements.

Discussion

To our knowledge, this is the first study to evaluate the long-term (ie, 1-year) comparative effectiveness of a structured, individually tailored home exercise and walking program enhanced with mHealth technology for people with PD compared with a similar program delivered without mHealth technology. As hypothesized, participation in either home program appeared to prevent the expected natural decline in daily physical activity in a sample of people with PD over a 1-year period.^4,5 Although the exercise-induced improvements in daily steps and minutes of moderate-intensity stepping were not statistically significant across the “somewhat active” sample as a whole (ie, 7500–9999 baseline daily steps³³), they represented a reduction in the rate at which disability evolves.^4,5 This achievement is an important long-term goal of physical therapist intervention for people with neurodegenerative disease.

Both interventions were well tolerated and acceptable to participants, with no serious adverse intervention-related events. Similar levels of adherence were observed in each. Both interventions received favorable satisfaction ratings. However, the enhanced behavioral change elements contained in the mHealth app appeared to benefit fewer active participants in particular. Compared with more active participants, fewer active participants displayed an approximately clinically meaningful improvement in daily steps,³⁴ a greater increase in minutes of moderate-intensity stepping, significant improvements in mobility-related quality of life, and clinically meaningful increase in walking capacity. The reasons for this differential result are unclear. We speculate that more active participants, many of whom already exceeded 7500 steps per day at baseline, had less need for behavioral change as they were already meeting recommended physical activity targets for walking.^33,36

Reducing disability and optimizing health-related quality of life early in the course of the PD should be a major focus of physical therapist intervention. Numerous exercise studies reveal significant benefits at the behavioral level (ie, improved function, less disability), with a recent trial revealing symptom-modifying effects of early, intense exercise in patients with de novo PD.³⁷ Given these benefits, evidence-based behavioral targets are needed to help people with PD sustain exercise and physical activity over the long term. For example, physical activity guidelines for adults recommend 10,000 daily steps or 150 minutes of moderate-intensity exercise each week.^33,36 A retrospective study of 2200 people with PD revealed that those engaged in exercise >150 min/wk had less disability, better quality of life, and slower disease progression 1 year later compared with those who were less active.⁸ Importantly, given that approximately 5000 of one's daily steps come from habitual activity,³³ some daily walking must be deliberate, planned, and structured—in the form of exercise—to meet activity guidelines. Moderate-intensity steps, rather than the total number of steps, might be a better indicator of planned, structured steps that are more likely to bring health benefits.

The design and results of our study highlighted several important aspects of physical therapist service delivery. First, if the progression of disability can be mitigated, physical therapy intervention early in the course of the disease appears warranted. Second, behavioral change elements can be embedded in a home exercise prescription across a continuum, from rudimentary (eg, goal setting, tailored instruction, feedback) to enhanced (eg, ongoing goal setting and feedback, mastery experiences, self-reflection, and more extended support from a health care professional). Third, a therapist's prescription of behavioral change elements should consider the baseline activity level of a given patient, because less active individuals could differentially benefit from mHealth-enhanced behavioral change elements. Fourth, regardless of an individual patient's circumstances, program outcomes could be further enhanced by actively involving patients in (1) tracking their progress and outcomes, either on paper or via technology, and (2) participating in planned, periodic physical therapist follow-up assessments to increase accountability, sustain engagement, allow close monitoring of changes, and ensure adequate physical activity intensity levels, especially in less active individuals.

The study limitations had implications for future investigations designed to examine the benefits of mobile health technology to increase physical activity and engagement in exercise in larger cohorts of participants with PD. Our sample was highly educated, lacked racial diversity, and was more active than anticipated. People who are relatively active are more likely to volunteer to participate in an exercise study. Consecutive sampling of patients attending regular neurology visits from a greater variety of clinical sites could enhance enrollment of a more diverse and less active sample. Moreover, our screening methods, using self-report exercise, might not have been adequate to reflect physical activity levels. Using baseline SAM data may be a more accurate method to screen, identify less active candidates, or to stratify participants and examine outcomes based on initial physical activity profiles. Nonetheless, the distribution of active and inactive participants allowed us to examine the similarities and differences in the results between less and more active participants. Treatment cohorts used different methods to collect adherence data, which limited the validity of direct comparisons of these data. In the mHealth group, adherence data were collected directly when participants logged into the app, whereas in the control condition self-reported adherence to the exercise program was recorded on paper calendars. An automated approach to collecting adherence data across both groups is needed. Future studies could also consider additional in-person visits. We limited the in-person physical therapist visits to 1 to 2 visits, which is considerably less than the number provided under usual care. Other studies applying behavioral change strategies typically rely on more in-person visits to increase uptake and facilitate lifestyle changes.^38,39 It is not known if more in-person visits, augmented with an mHealth program, would lead to more robust outcomes. The study also lacked a no-exercise control condition. Although we compared our findings with the results of a natural history, prospective, longitudinal cohort study in PD,⁴ the sample characteristics might not have been equivalent. The study was not sufficiently powered to detect statistically significant differences between groups in the small subset of less active participants, particularly in the physical activity outcomes. However, the results of this pilot revealed clinically meaningful changes in the mHealth, less active cohort and provide useful data in powering larger trials.

Conclusion

An individually tailored home exercise and walking program for individuals with PD appeared to prevent expected physical activity decline over 1 year. The addition of enhanced, remotely monitored, mobile technology–based, behavioral change elements to the exercise prescription appeared to differentially benefit participants who were less active. Further study in a larger group of sedentary people with PD is needed.

Author Contributions and Acknowledgments

Concept/idea/research design: T. Ellis, J. Cavanaugh, T. DeAngelis, C. Thomas, M. Saint-Hilaire, N. Latham

Writing: T. Ellis, J. Cavanaugh, T. DeAngelis, K. Pencina, N. Latham

Data collection: T. DeAngelis, K. Hendron

Data analysis: T. Ellis, J. Cavanaugh, K. Hendron, K. Pencina, N. Latham

Project management: T. Ellis, K. Hendron, N. Latham

Fund procurement: T. Ellis, N. Latham

Providing participants: T. Ellis, C. Thomas, M. Saint-Hilaire

Providing facilities/equipment: T. Ellis

Providing institutional liaisons: T. Ellis, C. Thomas, M. Saint-Hilaire

Consultation (including review of manuscript before submitting): T. Ellis, T. DeAngelis, C. Thomas, M. Saint-Hilaire

We thank all the people with Parkinson disease who participated in this study. Their important contributions made this study possible. We also thank Diane Dalton for contributing to the development of the exercise program, Nicholas Wendel for assisting with data management, Jon Venne for providing technology support, and Nicole Sullivan for her assistance with project management.

Ethics Approval

The study was approved by the Boston University Institutional Review Board. Patients provided written informed consent.

Funding

This study was funded by the American Parkinson Disease Association (grant no. 55,202,909).

Clinical Trial Registration

This study protocol is registered in the Clinical Trials Registry of the National Institutes of Health (ClinicalTrials.gov identifier NCT01955889).

Disclosures and Presentations

The authors completed the ICJME Form for Disclosure of Potential Conflicts of Interest and reported no conflicts of interest. Part of this manuscript was presented at the World Parkinson Congress in Portland, Oregon, on September 23, 2016; IV Step, The Ohio State University, Columbus, Ohio, in July 2016; and the 20th International Congress of Parkinson's Disease and Movement Disorders in Berlin, Germany, on June 19–23, 2016.