Rehabilitation Your Way: A Randomized Trial Comparing 2 Home‐Based and Self‐Managed Programs After Stroke

Abstract

Background

Home‐based, self‐managed stroke rehabilitation is needed to supplement limited clinical resources. This study compared a home‐based online self‐managed program, Stroke Rehabilitation Online Guide (STRONG), versus paper exercise program (PEP) for improving upper extremity function after stroke.

Methods

In this randomized trial, 43 participants with stroke were stratified by motor ability and self‐efficacy, then assigned to either STRONG (n=22) or PEP (n=21). A 6‐week intervention phase (>1 hour, 5 days weekly) was followed by 6 weeks of optional training. STRONG included self‐selected upper extremity task practice with optional web‐based video games and guided exercises. PEP included written upper extremity exercises. Primary outcomes were Fugl‐Meyer Upper Extremity Assessment and Wolf Motor Function Test‐Timed. Secondary outcomes included Motor Activity Log amount of use and movement quality, Wolf Motor Function Test‐Function, Stroke Motivation Scale, and Self‐Efficacy Scale.

Results

STRONG did not demonstrate superiority in Fugl‐Meyer Upper Extremity Assessment or Wolf Motor Function Test‐Timed. Within‐group Fugl‐Meyer Upper Extremity Assessment improvements were similar (STRONG: 1.8 points versus PEP: 2.1 points at post training; 3.3 versus 1.9 at follow‐up). However, STRONG participants reported better quality and increased arm and hand use in daily activities (Motor Activity Log‐quality of movement and Motor Activity Log‐amount of use) at post training and follow‐up. Additionally, STRONG participants showed higher self‐efficacy post training, though not maintained at follow‐up.

Conclusion

Both programs produced modest motor improvements. Although STRONG did not outperform PEP in primary outcomes, participants reported improved daily arm and hand use and movement quality and higher program continuation. Home‐based, self‐managed rehabilitation focused on daily tasks and optional exercises appears beneficial in individuals with stroke. Larger studies are needed to confirm these results.

Registration

URL: https://www.clinicaltrials.gov; Unique identifier: NCT03484182.

Nonstandard Abbreviations and Acronyms

FMA

Fugl‐Meyer Upper Extremity Assessment

MAL

Motor Activity Log

PEP

paper‐based arm and hand exercise program

STRONG

Stroke Rehabilitation Online Guide

WMFT‐F

Wolf Motor Function Test‐Function

WMFT‐T

Wolf Motor Function Test‐Time

Research Perspective.

What Is New?

• This is the first randomized controlled trial comparing a web‐based, self‐managed stroke rehabilitation program (Stroke Rehabilitation Online Guide) emphasizing meaningful daily tasks versus a traditional paper‐based exercise program for upper extremity recovery.

• Both home‐based, self‐managed rehabilitation approaches produced modest but measurable motor improvements without direct therapist supervision, demonstrating the potential viability of supplementing limited clinical resources with structured home programs.

• Stroke Rehabilitation Online Guide participants reported significantly greater improvements in daily arm use and movement quality compared with the paper‐based exercise program participants, with these benefits sustained at 6‐week follow‐up, suggesting that task‐oriented, technology‐enhanced approaches may better translate to real‐world functional gains.

What Question Should Be Addressed Next?

• Larger, adequately powered randomized controlled trials are needed to confirm these preliminary findings and determine whether the observed differences in functional outcomes translate to clinically meaningful long‐term benefits.

• Studies examining longer follow‐up periods beyond 6 weeks are essential to determine whether the superior engagement and self‐reported functional improvements with technology‐enhanced programs lead to sustained motor recovery and improved quality of life outcomes.

Stroke remains a leading cause of acquired disability worldwide. ¹ Although stroke incidence has declined in high‐income countries, ² factors such as improved survival rates and aging populations contribute to increasing numbers of long‐term survivors. ³ Evidence suggests that the amount of therapy is crucial for recovery. ⁴ However, health care systems struggle to provide sufficient therapy time to optimize motor outcomes.

Although telerehabilitation has emerged as 1 solution, it requires significant costs and effort from both therapists and patients. Self‐managed, home‐based strategies present a promising approach, facilitating rehabilitation with minimal professional interaction. A Cochrane review of 14 studies on self‐managed rehabilitation reported improvements in quality of life and self‐efficacy but found no substantial benefits for motor activity, ⁵ emphasizing the need for better research into self‐managed approaches, including those delivered in home settings.

We explored “LifeCIT” (Life Constraint Induced Therapy), a web‐based, individualized, home‐based, and self‐managed program specifically targeting upper extremity rehabilitation through a task‐oriented approach, initially trialed in England. ⁶ Founded on Self‐Determination Theory, ⁷ LifeCIT emphasizes motivation and self‐efficacy through goal‐setting and self‐monitoring. ⁸ A feasibility trial comparing LifeCIT with usual care demonstrated promise among individuals with stroke. ⁹ Our program, STRONG (Stroke Rehabilitation Online Guide), adapted LifeCIT by incorporating activities for a broader range of stroke impairments, including both bilateral and unilateral training, as well as modifying language and activities for use in the United States.

The purpose of this pragmatic randomized clinical trial was to compare the STRONG program to a paper‐based exercise program (PEP), which represented standard exercise instructions typically given to patients with stroke at rehabilitation discharge. We hypothesized that after 6 weeks, participants assigned to the STRONG group would demonstrate clinically meaningful and significantly greater improvements in arm and hand function (Fugl‐Meyer Upper Extremity Assessment [FMA]) and the streamlined timed Wolf Motor Function Test (WMFT‐T) compared with those in the PEP group. Additionally, we expected STRONG participants to demonstrate higher motivation to continue their exercises and improved confidence in their ability to perform the exercises (self‐efficacy) compared with PEP.

METHODS

The study adhered to the ethical principles approved by the university institutional review board, and written informed consent was obtained from all participants. The study also conformed to Consolidated Standards of Reporting Trials standards for randomized controlled trial conduct and reporting. Participants were recruited if they were in the subacute (3–6 months) or chronic (>6 months) phase poststroke. Initial recruitment was conducted in person at 3 local hospitals located in suburban, urban, and rural areas. Recruitment, eligibility assessment, consent, and evaluations shifted online during the COVID‐19 pandemic, expanding to include chronic participants nationwide through organizations for survivors of stroke and via online research platforms. The data that support the findings of this study are available from the corresponding author upon reasonable request.

Inclusion criteria were age ≥30 years; diagnosis with ischemic or hemorrhagic stroke ≥3 months before study initiation; discharge from all upper extremity rehabilitation; observable asymmetrical movement patterns between affected and less affected upper extremity (defined as visually apparent differences in movement initiation, trajectory, or quality when performing reaching or grasping tasks during initial screening); ability to move hand forward ≥3 inches from a position of 90° elbow flexion and neutral shoulder; computer/internet access with a video camera for virtual communication and availability of a second individual to assist with camera angles during the 3 assessment sessions (after March 2020); and being capable of following instructions provided during the initial screening session. Exclusion criteria included medical, orthopedic, neurological, or cardiovascular conditions that could jeopardize participation; cerebellar stroke; and recent (<3 months) upper extremity botulinum toxin injection.

Sample size calculations were based on the 2‐arm parallel group pilot randomized controlled trial (comparing LifeCIT with usual care) ¹⁰ , ¹¹ using the between‐group comparison of WMFT‐function (WMFT‐F) assuming a SD of 0.885 and clinically important effect size of 0.3. Power was set to 80% and the significance level to 5%. These calculations indicated that a total sample size of n=134 (67 per arm) would be needed.

The randomization schedule, designed by the study statistician (L.M.), who was independent of all participant assessment and intervention procedures, used computer‐generated pseudo‐random numbers with variable block sizes and a 1:1 allocation ratio, stratifying by motor performance and self‐efficacy. High motor performance was defined by the ability to independently lift a can with the paretic hand, and high self‐efficacy was determined based on an above average score on the Short Self‐Efficacy for Exercise Test for Stroke. ¹² , ¹³ Participants were randomized after baseline assessment. Research assistants conducting the assessments remained blinded during post training and follow‐up testing.

Both STRONG and PEP groups participated in 6 weeks of training (minimum 1 hour daily, 5 days weekly), with flexible participant‐set scheduling options. Participants tracked their training through daily log sheets and received brief (<5 minutes) weekly check‐in phone calls. After post training evaluation, participants could maintain their respective computer‐based or paper‐based training activities if they chose to do so for 6 additional weeks with 1 midperiod check‐in phone call.

The STRONG program used a web‐based computer interface to guide training activities throughout the intervention. Participants first answered questions regarding their abilities on 10 common tasks to determine their functional level. Level 1 participants used their whole paretic arm as a stabilizer during unilateral tasks (eg, holding newspaper under arm) or bilateral tasks (eg, holding paper still while writing with the nonparetic arm). Level 2 participants could stabilize objects using their paretic hand during bilateral tasks (eg, holding a jar or toothpaste tube) but could not use it as a manipulator. Level 3 participants could manipulate objects with their paretic hand for bilateral tasks (eg, taking a lid from a jar) and unilateral tasks (eg, lifting and drinking from a can). Level 4 participants could perform fine‐motor activities (eg, texting and writing) or higher function unilateral tasks (eg, throwing and catching).

After determining their functional level, STRONG participants selected daily activities from level‐appropriate lists. Activities were organized into 3 categories: personal care, home chores, or recreational activities. Although participants could choose how many activities to practice, the number of activities was self‐selected, and they were initially guided to select 6 (2 from each category). Figure 1 shows examples of home chore activities for each functional level. Beyond the required minimum of 1 hour of daily task practice, 5 times a week, the STRONG program offered optional training components: 5 standard unilateral and bilateral exercises for the affected arm per functional level and 5 computer games. The exercises, developed by a physical therapist (S.M.W.), were presented with images and descriptions tailored to each functional level. Although at least 5 repetitions for each exercise were suggested, participants could choose how many repetitions to perform. The computer games were designed to encourage participants to use their affected arm/hand to control the mouse, with self‐selected practice time. The STRONG program prompted participants to log in twice daily, once to select their daily practice tasks and again in the evening to report completed activities. Weekly reassessments of functional levels allowed for adjustments to daily activities, exercises, and games. This system of daily accountability and weekly assessments provided built‐in feedback, enabling participants to track their progress.

Figure 1. Screenshot of STRONG computer interface for selecting daily practice activities at 3 levels of motor function ability.

Example shown is for home chores. Similar boxes were displayed for personal care and recreational activities. Each day, participants checked the activities they were committed to practicing for a minimum of 1 hour that day within their predetermined functional level (based on a weekly computer program assessment). Following activity selection for personal care, home chores, and recreational activities, participants were able to optionally complete exercises and video games to promote paretic arm and hand use. At the end of each day, participants logged back in and were prompted to report on the activities and time practiced that day. STRONG indicates Stroke Rehabilitation Online Guide.

PEP featured printed handouts with images and descriptions of the same standard exercises per functional level that were included as optional activities within the STRONG program. Within STRONG, participants focused primarily on completing self‐selected goals regarding selected activities of daily living, with an additional optional set of standard exercises and games. In contrast, the PEP approach focused exclusively on the standard exercise set modeled after typical rehabilitation discharge instructions provided to individuals with stroke. Moreover, although STRONG participants completed a structured self‐assessment to determine their functional level, PEP participants were assessed during baseline evaluation and were advised to focus on exercises at their appropriate level although exercises from lower levels were also suggested as appropriate and a warmup. The functional level determination was important for both groups to ensure appropriate challenge and safety. In PEP, progression guidance was provided regarding dosage, which was to increase the number of repetitions, speed, or the distance of each exercise if they felt able to do so. As an incentive to complete the PEP program, participants were offered a crossover opportunity to STRONG following the 6‐week follow‐up assessment.

The fundamental differences being tested between the 2 interventions focused on 3 key aspects: (1) the delivery method (interactive web‐based platform versus traditional paper handouts); (2) the intervention content and approach (STRONG's emphasis on self‐selected meaningful daily tasks combined with optional prescribed exercises and games versus PEP's exclusive focus on prescribed exercises reflecting standard postdischarge care); and (3) the feedback mechanism (STRONG's built‐in tracking and progression system versus PEP's self‐monitoring without structured feedback). Although both groups received identical weekly check‐in calls from research staff and were invited to record a daily log, STRONG participants interacted with an automated system that prompted daily activity selection and completion reporting, creating a different accountability experience.

Evaluations were conducted at baseline, post training, and 6‐week follow‐up. Performance‐based assessments were video recorded by blinded research assistants, and a separate blinded assessor (S.M.W.) reviewed and scored the recordings later. S.M.W. also trained the research assistants to conduct these assessments, and another team member (J.W.) trained them on questionnaire administration. Before March 2020 (pre‐COVID), participants were assessed in person (n=2). The remaining participants were evaluated virtually. For virtual assessments, participants were asked to gather specific household items from a list (with sizes where applicable) and asked to store these items in a box and use the same ones for each testing session. Additionally, a template for the WMFT was sent to each participant to place on a table.

Primary outcomes included the FMA ¹⁴ and the Six‐Item WMFT‐T. ¹⁵ To accommodate virtual testing necessitated by the COVID‐19 pandemic, we carefully adapted our primary outcome measures while maintaining their psychometric integrity. For the FMA, we excluded reflex testing (3 items totaling 6 points) as these could not be reliably assessed virtually. This modification has been validated in previous research, ¹⁶ , ¹⁷ maintaining the construct validity of the test with a reduced scoring range of 0 to 60. For the WMFT‐T, we substituted 2 tasks: elbow extension was performed without weight to ensure standardization across home environments, and turning 3 cards replaced the key‐turning task, as both involve similar forearm supination movements. The card‐turning task is validated in the 15‐item WMFT. ¹⁸ To ensure reliability of video‐assessment, our blinded assessor reevaluated 10 quasi‐randomly selected testing sessions across a variety of impairment levels, achieving 98.7% agreement on FMA item ratings and an average timing discrepancy of only 129 milliseconds for WMFT‐T tasks.

Secondary motor outcomes included the Six‐Item WMFT‐F and the Motor Activity Log (MAL). WMFT‐F evaluates the quality of movement in the same tasks used for the WMFT‐T. The MAL is a patient‐reported measure of upper limb activity, divided into 2 scales: amount of use and quality of movement. ¹⁹ , ²⁰

Secondary behavioral outcomes included the Stroke Self‐Efficacy Questionnaire ²¹ and the Stroke Rehabilitation Motivation Scale. ²²

Statistical Analysis

For the statistical analysis, outcome distributions were first examined at each time point using graphical methods to gain insights into their location, spread, and skewness. Participant characteristics were compared across intervention groups for all randomized individuals. For each outcome variable, we estimated the mean outcome at each time point for each group using a mixed effects model with a random effect for subject. Our mixed effects models included fixed effects for intervention group, time point, and the group‐by‐time interaction, the binary variables used for the stratified randomization, and a random effect for subject to account for within‐subject correlation. To evaluate the effect of each intervention over time and the differences in these effects, we calculated P values and CIs for selected contrasts. Our primary analyses followed the intention‐to‐treat principle, and subsequent analyses considered compliance. We report the differences along with their corresponding 95% CIs and effect sizes. Effect sizes were defined as the difference between the treatment groups with respect to changes over time divided by the SD of the changes over time.

RESULTS

Figure 2 depicts the study participant flow. From 184 potential participants who expressed interest in the study, 43 were randomized into the STRONG (n=22) or PEP (n=21) groups. Completion group sizes were 16 and 15 for STRONG and PEP, respectively, and 12 (STRONG) and 10 (PEP) at the 6‐week follow‐up testing session. Despite targeting a sample size of 134 participants based on initial power calculations, recruitment challenges limited our final sample to 43 randomized participants. This reduced sample size maintained adequate power for detecting moderate‐to‐large effects but limited our ability to detect smaller but potentially meaningful differences, particularly in our primary outcomes.

Figure 2. Study participant flow chart.

PEP indicates paper‐based arm and hand exercise program.

Table 1 presents the demographic baseline characteristics of participants who commenced the study. There were no significant differences between the groups regarding age, sex, time since stroke, stroke side or dominance, or baseline functional scores.

Table 1.

Baseline Characteristics of the Participants

Variable	Level	All participants (n=43)	STRONG (n=22)	PEP (n=21)
Age group, y	<40	2 (5)	1 (5)	1 (5)
	40–49	7 (16)	3 (14)	4 (19)
	50–59	16 (37)	11 (50)	5 (24)
	60–69	13 (30)	5 (23)	8 (38)
	70+	5 (12)	2 (9)	3 (14)
Sex	Female:male	22:21 (51:49)	12:10 (55:45)	10:11 (48:52)
Years since stroke	<2 y	10 (23)	5 (23)	5 (24)
	2–4 y	11 (26)	7 (32)	4 (19)
	4–6 y	4 (9)	3 (14)	1 (5)
	6+ y	9 (21)	3 (14)	6 (29)
	Missing	9 (21)	4 (18)	5 (24)
Dominant side	Left:right	6:37 (14:86)	3:19 (14:86)	3:18 (14:86)
Stroke dominance side	Motor dominant	12 (28)	9 (41)	3 (14)
	Motor nondominant	31 (72)	13 (59)	18 (86)
Paretic side	Left:right	27:16 (63:37)	12:10 (55:45)	15:6 (71:29)
Fugl‐Meyer Assessment‐Upper Extremity total: mean±SD max=60		33.71±17.67	33.29±17.73	34.14±18.04
WMFT‐Timed, s: mean±SD max=<1 s		38.97±41.95	42.38±41.20	35.56±43.43
WMFT‐Function: mean±SD max=5		2.90±1.41	2.82±1.40	2.98±1.45
MAL‐AU: mean±SD max=5		1.22±0.97	1.05±1.05	1.39±0.87
MAL‐QM: mean±SD max=5		1.38±1.05	1.13±1.11	1.64±0.95
Self‐efficacy: mean±SD max=141		96.67±22.12	97.57±22.50	95.76±22.26
Stroke Rehabilitation Motivation Scale: mean±SD max=140		113.3±8.85	115.7±8.16	111.0±9.09

Data are presented as n(%) and mean±SD. MAL indicates Movement Activity Log; PEP, paper‐based arm and hand exercise program; STRONG, Stroke Rehabilitation Online Guide; and WMFT, Wolf Motor Function Test. All measured items are scored from low to high with high being best except Wolf Time where a low score is a better performance and 120 s is the worst possible score.

Primary Outcomes

At post training, both intervention groups demonstrated improvements in the primary motor outcome of FMA scores (STRONG: 1.8 points,P=0.022, PEP: 2.1 points, P=<0.001), with no significant between‐group differences (P= 0.82; effect size=0.08 in favor of PEP). Improvements persisted at the 6‐week follow‐up assessment (STRONG: 3.3 points, P=0.001, PEP: 1.9 points, P=0.04); however, again, there was no significant difference between the groups at follow‐up (P=0.30, effect size=0.45 in favor of STRONG).

Neither group showed significant within‐ or between‐group WMFT‐T improvements at post training or at follow‐up. The point estimate for the difference between the groups at the post training assessment was 1.2 s in favor of STRONG (an effect size of 0.11), and the point estimate for the difference between the groups at the follow‐up visit was 6.9 seconds in favor of STRONG (an effect size of 0.65) Results are detailed in Table 2.

Table 2.

Model‐Based Estimates of Mean Outcomes by Group at Each Time Point Based on a Mixed Effects Longitudinal Regression Model

Outcome	Group	Baseline mean±SE	Post training mean±SE	Mean change from baseline (95% CI)	P value within group	Mean difference between groups in change (95% CI)	P value between groups	Follow‐up mean±SE	Change from baseline	P value within group	Difference (95% CI)	P value between group
Fugl‐Meyer Assessment‐Upper Extremity	STRONG	32.8 ±1.8	34.6 ±1.8	1.8 (0.3 to 3.3)	0.022*	−0.3 (−2.6 to 2.0)	0.82	36.0 ±1.8	3.2 (1.4 to 5.1)	<0.001*	1.4 (−1.2 to 4.0)	0.30
	PEP		34.8 ±1.8	2.1 (0.4 to 3.8)	<0.001*			34.7 ±1.9	1.9 (0.0 to 3.7)	0.049*
WMFT‐Timed	STRONG	41.2 ±4.1	40.1 ±4.6	−1.1 (−6.4 to 4.2)	0.68	−1.2 (−9.0 to 6.7)	0.77	36.3 ±4.9	−4.9 (−11.3 to 1.6)	0.13	−6.9 (−16.0 to 2.1)	0.13
	PEP		41.3 ±4.7	0.1 (−5.8 to 5.9)	0.98			43.3 ±4.9	2.1 (−4.4 to 8.6)	0.52
WMFT‐ Function	STRONG	2.8 ±0.1	3.0 ±0.2	0.2 (0.0 to 0.4)	0.019*	0.1 (−0.1 to 0.4)	0.31	2.9 ±0.2	0.1 (−0.1 to 0.3)	0.31	−0.1 (−0.4 to 0.2)	0.40
	PEP		2.9 ±0.2	0.1 (−0.1 to 0.3)	0.41			3.0 ±0.2	0.2 (0.0 to 0.4)	<0.001*
MAL‐amount of use	STRONG	1.2 ±0.1	2.3 ±0.2	1.1 (0.7 to 1.4)	<0.001*	0.6 (0.1 to 1.1)	0.013*	2.3 ±0.2	1.1 (0.7 to 1.4)	<0.001*	0.7 (0.1 to 1.2)	0.023*
	PEP		1.7 ±0.2	0.5 (0.1 to 0.8)	0.012*			1.6 ±0.2	0.4 (0.0 to 0.8)	0.036*
MAL‐quality of movement	STRONG	1.4 ±0.1	2.3 ±0.2	1.0 (0.6 to 1.3)	<0.001*	0.6 (0.2 to 1.1)	0.012*	2.3 ±0.2	1.0 (0.6 to 1.3)	<0.001*	0.6 (0.1 to 1.2)	0.024*
	PEP		1.7 ±0.2	0.3 (−0.0 to 0.7)	0.088			1.7 ±0.2	0.3 (−0.1 to 0.7)	0.14
Self‐efficacy	STRONG	96.2 ±3.2	102.3 ±3.9	6.1 (0.1 to 12.1)	0.047*	7.6 (−0.8 to 16.1)	0.076	101.6 ±4.2	4.4 (2.3 to 11.1)	0.20	6.3 (−3.4 to 16.1)	0.20
	PEP		94.7 ±4.0	−1.6 (−7.7 to 4.6)	0.62			94.3 ±4.4	−2.0 (−9.2 to 5.3)	0.59
Stroke Rehabilitation Motivation Scale	STRONG	113.3 ±1.4	109.0 ±2.0	−4.3 (−8.0 to −0.6)	0.024*	0.3 (−4.9 to 5.4)	0.92	111.3 ±2.2	−2.0 (−6.1 to 2.2)	0.35	2.4 (−3.6 to 8.4)	0.42
	PEP		108.7 ±2.0	−4.6 (−8.4 to −0.7)	0.020*			108.9 ±2.4	−4.4 (−8.9 to 0.2)	0.059

MAL indicates Movement Activity Log; PEP, paper‐based arm and hand exercise program; STRONG, Stroke Rehabilitation Online Guide; and WMFT‐T=Wolf Motor Function Test‐Timed.

^* P<0.05.

Secondary Motor Outcomes

At post training, the STRONG group demonstrated significantly improved WMFT‐F, but the difference between the groups with respect to improvement was not statistically significant (P=0.31, effect size in favor of STRONG of 0.22). At the follow‐up assessment, improvements were greater in the PEP group, but again, group differences were not statistically significant (P=0.4, effect size=0.36).

For the self‐reported measures, both groups demonstrated improvements in daily arm use (MAL‐amount of use), but only STRONG demonstrated improvements in movement quality (MAL‐QM) from baseline to post training, as well as significantly greater improvements compared with PEP (MAL‐amount of use: 1.1 versus 0.5 points, P=0.013, effect size=0.87 and MAL‐quality of movement: 1.0 versus 0.3 points, P=0.012, effect size=0.89). These between‐group differences in favor of STRONG were maintained at follow‐up. Results are detailed in Table 2.

Secondary Behavioral Outcomes

At post training, only the STRONG group demonstrated a greater improvement in self‐efficacy (P=0.076, effect size=0.62), although this improvement was attenuated at the 6‐week follow‐up. Regarding the Motivation Scale, both groups experienced a decline in motivation at the post training assessment, with no significant difference between the groups. Scores for both groups returned to baseline at the follow‐up assessment. Results are detailed in Table 2.

Adherence to the goal of 30 hours of training over 6 weeks was generally good among participants who reported their training times (n=9 STRONG; n=7 PEP). On average, STRONG participants logged 33.4 hours (SD, 14.8; range, 17–58.3), versus 26.4 hours for PEP participants (SD, 7.8; range, 14.8–37). Adherence varied, with only 1 PEP participant exceeding the target compared with 3 from STRONG. Two caveats must be noted: first, daily logs were introduced for participants starting with participant #14, resulting in 5 prior participants (3 PEP and 2 STRONG) not having logs to complete. Second, 5 participants from each group did not return their record sheets despite reminders, leading to an overall log return rate of 62%.

Willingness to continue training during the follow‐up period was notably higher among STRONG participants, with 5 out of 7 reporting completers continuing training compared with none from PEP. Additionally, interest in the STRONG program was evident, as 5 out of 10 PEP participants who completed the study expressed a desire to cross over to STRONG, whereas only 1 of 12 STRONG participants chose to transition to PEP.

There were no serious adverse events reported among participants in either the STRONG or PEP group. Adverse events were systematically monitored through weekly telephone check‐ins with all participants and a specific query regarding any difficulties or injuries experienced during practice sessions.

DISCUSSION

In this randomized superiority trial, we compared STRONG, a web‐based, home‐based, and self‐managed upper extremity rehabilitation approach emphasizing self‐selected meaningful tasks, motivation, and self‐efficacy, to a home‐based, self‐managed control group who received a standard set of PEP without task‐oriented or feedback components and more similar to standard practice. Our primary hypothesis was that STRONG would outperform PEP at the primary end point after 6 weeks training. Instead, both approaches produced modest motor impairment improvements. However, STRONG participants reported greater daily engagement of the paretic arm and improved movement quality compared with the PEP group. Only STRONG participants exhibited improvements in self‐efficacy following 6 weeks training. Participants were allowed to continue with their intervention for 6 extra weeks after the first assessment. This pragmatic, clinically meaningful option generated useful data on acceptance of each intervention. Both interventions showed sustained within‐group FMA improvements at 12 weeks. STRONG participants sustained their report of greater daily engagement of the paretic arm and improved movement quality compared with the PEP group.

The findings for FMA align with previous studies showing comparable FMA improvements in both experimental and control groups with equivalent training doses, as reported in a comprehensive systematic review with similar results in 69% of the analyzed studies. ²³ Importantly, our results suggest that modest improvements in FMA can be achieved even with a home‐based, self‐managed intervention consisting solely of written exercises. Although the improvements observed in both groups fell short of the minimally clinical important difference of 8.42 points previously reported for subacute (<3 months) poststroke and 4.25 to 7.25 points reported for chronic (>6 months) stroke, ²⁴ , ²⁵ the effect sizes showed a small advantage for PEP at 6 weeks and a moderate advantage for STRONG by 6 weeks. These findings suggest that although both interventions produced measurable motor function improvements, more intensive or different approaches may be necessary to achieve clinically meaningful improvements in upper extremity function after stroke with a larger advantage for STRONG over time.

The absence of significant changes in the WMFT‐T was unexpected, given its prior sensitivity to change resulting from laboratory‐based upper extremity rehabilitation. ²⁶ , ²⁷ , ²⁸ , ²⁹ A contributing factor to the overall lack of improvement in WMFT‐T may be that neither the STRONG nor PEP groups were specifically instructed to prioritize speed during their training. Our focus was on ensuring participants engaged in the activities for a set duration. Although we did not find statistical differences between groups in this outcome, the CI for the sustained reduction in WMFT‐T suggested the potential for a greater reduction of up to 16.4 seconds in the STRONG group and a medium to high effect size in favor of STRONG.

Encouragingly, STRONG is superior to PEP based on secondary motor and behavioral outcomes. STRONG participants reported greater engagement in activities of daily living and improved movement quality compared with PEP participants, with these gains sustained after training became optional. Notably, improvements in the STRONG group reached the minimal clinically important difference for the MAL in acute/subacute population ³⁰ and resulted in large effect sizes. STRONG's effectiveness likely stemmed from its emphasis on meaningful activities, built‐in weekly self‐assessments, and quantitative feedback system. The program's focus on daily living tasks over generic exercises enhanced MAL scores while promoting participant autonomy and engagement. Moreover, only after STRONG training did we observe enhancements in the functional scores of the WMFT‐F, supporting the findings of LifeCIT in a higher functioning group of mixed chronicity individuals after stroke ⁹ that reported a clinically important difference after 3 weeks of similar web‐based training. ³¹ Although these improvements were not maintained at the 6‐week follow‐up, they suggest that STRONG may be more effective in inducing changes in motor function compared with PEP.

Participant feedback collected during weekly check‐in calls revealed that STRONG's built‐in feedback mechanisms played a crucial role in maintaining engagement. Several participants specifically valued the daily activity logging feature, which allowed them to visualize their progress over time. One participant noted, ‘Seeing what I accomplished each day motivated me to keep going.’ The weekly self‐assessment component, which could trigger progression to more challenging activities, provided participants with concrete evidence of improvement that many found rewarding. In contrast, PEP participants, while generally adherent to the exercise regimen, commented that they would like more feedback and advice regarding their progression. The difference in feedback mechanisms and self‐choice opportunities likely contributed to STRONG participants' higher self‐efficacy scores and willingness to continue training during the follow‐up period.

The observation that PEP participants showed within‐group sustained improvement in the quantity, although not quality, of their activities of daily living, relative to baseline measurements is noteworthy, given that their training was not task specific. This outcome suggests a successful transfer of functional skills to tasks beyond those practiced. It is possible that the exercises available/selected for both groups facilitated impairment‐based improvements, as evidenced by the FMA results, which were generalizable to tasks of daily living. Given that the STRONG protocol allows for the practice of exercises alongside specific tasks, we tentatively propose that this combined approach, with a greater emphasis on task practice, may be more beneficial than focusing solely on either tasks or exercises.

Overall, our participants demonstrated high motivation at baseline, likely reflecting their commitment to a 12‐week study without compensation, which necessitated considerable self‐discipline to manage their interventions. However, both groups exhibited a decline in motivation scores over the training period. The paradoxical finding of good adherence despite declining motivation scores warrants deeper examination. We propose several potential explanations for this phenomenon. First, the Stroke Rehabilitation Motivation Scale used in our study was designed to assess intrinsic motivation specifically related to rehabilitation exercises themselves. The scale includes items such as “Do you find rehabilitation to be fun?” and “Does rehabilitation help you feel like you're achieving something.” As participants progressed through the 6‐week program, their initial enthusiasm for the tasks and exercises themselves may have naturally waned although their commitment to the overall goal remained strong. Motivation theory distinguishes between different types of motivation, intrinsic (enjoyment based) versus more regulated forms (commitment based). Our scale likely captured the decline in intrinsic enjoyment of repetitive activities over time, whereas participants maintained extrinsic motivation to complete the study protocols. Second, the Stroke Rehabilitation Motivation Scale was originally validated with inpatient populations in the early poststroke phase when rehabilitation gains are typically more rapid and noticeable, potentially making it less sensitive to the motivational nuances of participants with chronic stroke engaged in home‐based rehabilitation.

To maintain engagement, we used several strategies: weekly personalized check‐in calls providing encouragement and addressing technical challenges; allowing participants flexibility in scheduling their practice sessions; and for STRONG participants specifically, the opportunity to modify activity selections weekly based on self‐assessments. Notably, despite similar declines in motivation scores, STRONG participants demonstrated significantly greater willingness to continue training during the follow‐up period (5/7 STRONG versus 0/7 PEP), suggesting that the interactive, task‐oriented and self‐choice nature of STRONG fostered a more sustainable engagement model. The self‐assessment aspect of functional level determination in STRONG as well as the self‐selected choice of activities may have contributed to greater participant engagement and motivation compared with the predetermined exercise levels in PEP, along with no external feedback except for weekly encouragement. STRONG was aligned with self‐determination theory principles of autonomy and competence. ³² Future iterations of both programs could and should incorporate additional motivational elements, such as social connectivity features or gamification elements, to address this decline in motivation.

Recruitment for this study proved challenging for several reasons. Initially, we targeted only a subacute population (3–6 months poststroke) discharged from 3 local rehabilitation facilities. When COVID‐19 emerged in March 2020, in‐person recruitment became impossible, and many hospital‐based research activities were suspended. Our pivot to nationwide virtual recruitment expanded our geographical reach but introduced additional barriers: participants needed adequate technology skills, computer access, and an individual available to assist with camera positioning during virtual assessments. Many interested participants lacked this support. Additionally, the commitment to a 12‐week unpaid study requiring at least 5 hours of weekly self‐directed practice deterred some potential participants, particularly when paired with the technical requirements.

Clinical Implications

The findings from this study have useful implications for clinical practice and the future development of rehabilitation programs. First, our results demonstrate that home‐based, self‐managed rehabilitation can produce measurable improvements in upper extremity function after stroke, even without direct therapist supervision. This is particularly relevant given health care system constraints and the growing emphasis on telehealth and remote care delivery models. The modest but consistent improvements observed in both groups support the viability of supplementing limited in‐person therapy with structured home programs. Second, STRONG's superior performance in promoting actual use of the affected arm in daily activities highlights the importance of designing rehabilitation programs that emphasize functional, meaningful tasks rather than isolated exercises alone. Any paper‐based rehabilitation programs should consider adding self‐selected choice of tasks as well as a progression of exercises. Third, our findings regarding participant engagement and program continuation suggest that web‐based platforms with built‐in accountability and feedback mechanisms may offer advantages for long‐term adherence. Health care systems considering implementation of home‐based rehabilitation programs should prioritize incorporating feedback mechanisms, such as self‐assessment questions, even if using simpler technological solutions.

From a scalability perspective, STRONG shows considerable promise. The web‐based platform could be expanded to serve larger populations, with minimal increases in administrative costs. Future iterations could incorporate automated check‐ins and machine learning algorithms to adjust exercise recommendations based on reported progress, further reducing the need for direct therapist involvement while maintaining personalization. Although initial development costs are higher than paper‐based alternatives, the potential for improved outcomes and sustained engagement suggests favorable cost‐effectiveness in the long term. Long‐term effects merit further investigation, as our follow‐up period was limited to 6 weeks. However, the continued voluntary use of STRONG by most participants suggests potential for sustained benefits. Future research should examine whether the improvements in daily arm use translate to longer‐term functional independence and quality of life outcomes. Additionally, investigating whether early intervention with STRONG during the subacute phase might yield more substantial motor improvements represents an important direction for future study.

The study design possessed several strengths, including a pragmatic approach that allowed all participants to choose their practice times and intensity beyond the suggested minimum. Additionally, the provision of a control group with similar time‐on‐task requirements and the inclusion of a follow‐up assessment enhanced the study's robustness. Both groups received equal amounts of information and interaction from the research team, thereby increasing the internal validity of the comparisons.

Study Limitations

The key limitation is the reduced sample size. Our reduced sample size for analysis provided sufficient power to detect differences in some secondary outcomes, but it likely limited our ability to detect smaller but potentially meaningful differences in our primary outcomes. For the WMFT‐T specifically, we observed nonsignificant improvement in the STRONG group (reduction from 39.0 seconds to 33.9 seconds) compared with a slight increase in the PEP group (39.0 seconds to 41.1 seconds) at follow‐up. The wide CIs (−16.4 to 11.7 seconds) surrounding this between‐group difference and the high effect size, suggest that with a study of this size, we could not rule out substantial differences between the groups.

We acknowledge, too, that the small sample sizes for adherence data also limit data interpretation. It is plausible that the higher adherence from the STRONG group both during and after the 6‐week training was related to, for example, enjoyment of STRONG versus enjoyment of PEP, but we cannot rule out other factors with the small sample for adherence. We did learn, however, that self‐reported compliance data are best collected on a daily or weekly basis and not at the end of a study. Relatedly, the relatively large dropout rates, though similar in amount and causes across the 2 groups, could lead to biased estimates of change because those who were evaluated may have the most adherence. The limited adherence data we do have do not necessarily support this idea since adherence may or may not be spread evenly over the 2 groups.

A second limitation is the heterogenous population of subacute and chronic individuals. We failed to restrict participants to those recently discharged from outpatient rehabilitation (the subacute population), which was the original aim. As a result, the generalizability of study findings may have been affected regarding applicability to this early phase of stroke chronicity. A more homogeneous subacute cohort would have minimized variability due to motivation and proximity to the stroke, thereby enhancing the internal validity of the study. On the other hand, although 4 of the original 6 who consented in the subacute phase were randomized, only 3/31 were analyzed at posttest (2 in STRONG and 1 in PEP) and 1/22 at retention (from STRONG). Therefore, the results do more generally apply to the chronic population, beyond the more “ideal” subacute population where most recovery is expected.

Overall, the promising results from both (particularly STRONG) self‐managed, home‐based programs suggest potential effectiveness for poststroke upper extremity rehabilitation. However, the findings from this preliminary study should be interpreted with appropriate caution given the reduced sample size. They require confirmation in larger, adequately powered trials before definitive clinical recommendations can be made. Next iterations of STRONG will require technological enhancements to improve compliance assessment. Future steps include developing a comprehensive level 3 randomized controlled trial and incorporating additional methods to enhance patient engagement.

CONCLUSIONS

Both STRONG and PEP modestly improved motor function, with STRONG showing sustained clinically important advantages in self‐reported activity amount and quality. STRONG uniquely enhanced self‐efficacy and demonstrated greater participant retention of ongoing use of the STRONG program even when exercising became optional. These preliminary findings support the potential for home‐based, self‐managed rehabilitation and suggest incorporating web‐based interfaces, participant choice, augmented feedback, and the combining of daily task activities with optional exercises and games.

Disclosures

Authors declare no conflict of interest.

Source of Funding

The study was supported by National Institute on Disability, Independent Living, and Rehabilitation Research Field Initiated Grant 90IFRE0011.