Introduction
Stroke remains a leading cause of long-term adult disability in the United States, with persistent gait and balance impairments impacting independence and quality of life.1 Early, intensive task-specific gait rehabilitation promotes neuroplasticity and functional recovery, but inpatient delivery is often constrained by therapist availability, safety concerns, and patient endurance.1–3 Robotic-assisted gait training (RAGT) addresses these barriers by enabling safe, high-repetition, progressively adaptive walking practice.4,5 Meta-analyses and clinical guidelines support RAGT’s effectiveness in improving walking speed, endurance, independence, and activities of daily living, particularly during early stroke recovery,3–6 through mechanisms that facilitate muscle strengthening and neuroplasticity.7 However, real-world implementation, protocol tolerance and operational feasibility of RAGT within routine, acute inpatient rehabilitation settings remain understudied.
Compared with rigid, joint-driven exoskeleton gait systems that attach motorized supports directly around the hip and knee joints, end-effector gait trainers secure only the feet to movable footplates, allowing greater freedom of movement; however, both system types can use programmed gait trajectories, and their effects on muscle activation and gait variability depend on device design, control strategy, assistance level, and patient characteristics.8,9 The Centers for Medicare and Medicaid Services (CMS) require post-acute care providers to report standardized Section GG scores to evaluate patient independence across daily mobility activities during inpatient rehabilitation.10,11 Section GG extends beyond standardized reporting by measuring specific daily mobility activities, including transfers, walking, and stair negotiation. In post-stroke inpatient rehabilitation, Section GG demonstrated responsiveness to functional change and minor admission floor effects, although ceiling effects were evident at 90-day follow-up.10,12 Interventions aimed at enhancing Section GG outcomes, such as RAGT, directly support meaningful functional recovery and post-stroke independence.7,13 Despite promising outcomes, the integration of RAGT into routine inpatient rehabilitation is often hindered by high resource demands, operational complexity, and a lack of evidence regarding real-world feasibility.3,6
The Rise&Walk® end-effector robotic gait trainer (Healing Innovations, Inc., Nashville, TN, USA) incorporates adjustable body-weight support, adjustable handguides, and variable assistance modes ranging from full support to resistance training (Figure 1). In addition, the rating of perceived exertion (RPE)-based training structure provides a framework for real-time dosing and intensity titration. To date, there are no peer-reviewed data evaluating the operational feasibility, protocol tolerance, safety profile, and clinical outcomes of this system during inpatient stroke rehabilitation.
Therefore, the main objective of this retrospective pilot study was to evaluate the clinical feasibility and safety surveillance of integrating end-effector RAGT (E-RAGT) with the study device into routine inpatient stroke rehabilitation, defined as successful session completion, protocol tolerance, and the absence of serious adverse events. A secondary objective was to describe within-stay functional mobility changes measured by standardized Section GG scores. Figure 1The End-Effector Robotic Gait Training System and Structural Components.
Photograph of the end-effector system in clinical operation, illustrating key structural features and operational components.An end-effector robotic gait training system with labeled components and a patient and a physical therapist interacting.
Materials and Methods
Study Design and Participants
This study was approved by the Methodist Healthcare Institutional Review Board and Research Operations Committee, which granted a waiver of informed consent due to the use of de-identified data (IRB Protocol #2313076-1 and 2313076–2). All procedures were conducted in accordance with the Helsinki Declaration.
This retrospective cohort study was conducted at a regional inpatient rehabilitation facility. Clinical data were collected between October 2, 2024, and December 31, 2024. As a feasibility-focused pilot study, the primary objective was to evaluate the integration and potential impact of robotic gait training in a typical inpatient setting, generating preliminary data on clinical trends that may help guide the design of future trials.
Any adult patient who had experienced a stroke, presented with gait and/or imbalance impairments, and was admitted for inpatient stroke rehabilitation was eligible for the study. Exclusion criteria were contraindications to weight bearing, body weight >150 kg, height outside 152–191 cm, open skin lesions/skin breakdown, uncontrolled seizures, circulatory instability, or pregnancy. Regularly scheduled meetings involving the treating clinicians were conducted to review patient progress, and decisions to use E-RAGT for an individual patient were made during these meetings.
Intervention
All patients received standard inpatient rehabilitation physical therapy delivered by physical therapists per the site’s standard protocols; typically including gait training, balance and strength exercises, transfer training, and functional mobility practice. Within this clinical workflow, E-RAGT was integrated into scheduled physical therapy care, intended to substitute for a portion of conventional gait practice rather than providing additional or supplementary therapy time.
Patients received E-RAGT using the study device. Licensed Doctors of Physical Therapy (hereafter referred to as therapists) tailored treatment parameters in real-time, including body-weight support (10–100%), step direction (forward, backward), and gait velocity (0.5–1.8 mph [0.2–0.8 m/s]). Assistance modes were varied between resistance, manual mode, active assist, and full robotic assist.
Clinical applications such as isometric holds and interval training were integrated to align with individual functional goals.
Training intensity was monitored every five minutes using the modified Borg Rating of Perceived Exertion (RPE).14,15 The RPE is a validated, widely used measure of exertion in stroke and cardiac rehabilitation populations for frequent, real-time use at the bedside without additional monitoring equipment. To support consistency and fidelity, the RPE scale and target range were integrated directly into the device’s touchscreen interface, prompting both clinician and participant at each check-in, to minimize measurement variability across sessions and clinicians (Table 1, E-RAGT Session Protocol Progression Table). Therapists aimed to have patients spend the majority of their walking session at moderate-to-vigorous intensity, defined as a score of ≥6 to ≤8/10 on the modified Borg 0–10 scale,14,15 and adjusted device parameters to sustain this target during each session.
E-RAGT sessions occurred 2–8 times per stay, with individual sessions frequency and duration determined by clinical judgment, patient tolerance, and therapist/patient availability.Table 1End-Effector RAGT Session Protocol Progression TableTimeActionDevice ParametersRPE Target0:00Begin stepping. Confirm fit and stepping mechanics. Observe for crouched posture; cue quad activation if needed.
Note baseline RPE at stepping start. Warm-up phase: target 3 minutes, not 5 — advance to intensity protocol as soon as mechanics are confirmed and participant is stableBWS: 70% | Assist: Level 7 | Speed: 0.6 mph (0.3 m/s)RPE monitored continuously. No target — baseline only.1:00Increase speed.
Continue observing gait mechanics. Log RPE on touchscreen.BWS: 70% | Assist: Level 7 | Speed: 1.0 mph (0.5 m/s)RPE logged on touchscreen. Monitor only — no target yet.3:00Check RPE (modified Borg) and log on touchscreen.
If RPE <7, decrease assistance.BWS: 70% | Assist: Level 5 | Speed: 1.0 mph (0.5 m/s)RPE logged on touchscreen. Target: approaching RPE 74:00Check RPE (modified Borg) and log on touchscreen. If still <7, initiate the first interval.BWS: 70% | Assist: Level 4 (30 sec interval)RPE logged on touchscreen.
First intensity challenge.4:30End of first interval — return to Level 5–7. Log RPE (modified Borg) on touchscreen immediately at the transition, before recovery begins.BWS: 70% | Assist: Level 5–7 | Speed: 1.0 mph (0.5 m/s)RPE logged on touchscreen at peak exertion. Allow brief recovery.5:00Log RPE on touchscreen.
Increase speed to max per step length chart. Begin formal interval protocol.BWS: 70% | Assist: Level 5 | Speed: MAX per chartRPE logged on touchscreen. Target: RPE ≥7.5:00–15:00Interval protocol: Level 4 for 1 min, Level 5 for 1 min.
Repeat. RPE (modified Borg) monitored continuously; log on touchscreen every 2 min and check every 5 min.Alternating Level 4/Level 5 intervals | Speed: maxTarget: RPE ≥7 during Level 4 intervals.10:00Log RPE on touchscreen. RPE value at this point drives the BWS decision: reduce BWS by 10% (70% → 60%).
If RPE ≥7 is not yet achieved, prepare to advance titration in the next phase.BWS: 60% | Assist: Level 4/5 intervals | Speed: maxRPE logged on touchscreen. If still <7, proceed with assistance level reduction.10:00–30:00 or > minIf RPE ≥7 still not achieved: Decrease interval assistance level in 1-level increments (Level 4 → 3 → 2 → 1). For each step, perform a 1-min high-intensity interval at the new lower level, then return to Level 5 (minimum assist — powered) for the 1-min recovery.
Maintain same max speed and established BWS (60%). Continue decreasing intensity level until RPE ≥7 is achieved. RPE (modified Borg) monitored continuously; log on touchscreen every 2 min and check every 5 min.
Document the level at which RPE ≥7 is first achieved and sustained.BWS: 60% | Intervals: Level 4→3→2→1 as needed | Recovery: Level 5 | Speed: maxTarget: RPE ≥7 during intensity intervals. Upper limit: RPE 8.EndCool-down at reduced intensity (5 min standard; shorten to 2–3 min if participant has limited tolerance). Final RPE, step count.
Adverse events. Document: (1) threshold-achieving parameter combination, and (2) total active stepping time achieved.BWS: increase | Assist: increase | Speed: reduceFinal RPE.Abbreviations: BWS, body weight support; mph, miles per hour; RAGT, robotic-assisted gait training, RPE, rating of perceived exertion.
Treating therapists were instructed to report any safety concerns or changes in clinical status associated with the device (eg, serious health events, abnormal physiologic responses, skin irritation/breakdown, orthopedic pain or injury, or other unexpected issues). Safety was also monitored through scheduled virtual meetings with treating therapists, site leadership, and the technology vendor, during which clinicians were asked to report any safety challenges or serious adverse events related to E-RAGT sessions. Any of the following events, if they occurred, were considered a serious adverse event: Any untoward medical occurrence that resulted in death, was life-threatening, required or prolonged hospitalization, resulted in significant or persistent disability, or was otherwise deemed medically significant.
Mild adverse events were any transient, self-limited occurrences that did not require medical intervention or interruption of the intervention (eg, muscle soreness, fatigue, minor skin irritation), which were reported spontaneously rather than systematically logged.
Primary Outcomes
Functional mobility was assessed at admission and discharge using seven mobility items from CMS Section GG0170:11 Sit to stand (Item D), chair/bed-to-chair transfer (Item E), walking 10 feet (3 m) (Item I), walking 50 feet (15 m) with two turns (Item J), walking 150 feet (45 m) (Item K), and ability to ascend and descend 4 steps (Item N) and 12 steps (Item O).11
Section GG scores were collected as part of routine clinical documentation assigned by the treating therapists during usual inpatient care. Treating clinicians were aware of patient device usage, and outcome assessment was unblinded. These assessments are standardized, federally mandated data elements under the IMPACT Act of 2014 for national quality reporting and reimbursement across post-acute-care settings.11 Each task is rated on a 6-point functional scale where higher scores indicate greater independence: 6=Independent; 5=Setup or clean-up assistance; 4=Supervision or touching assistance; 3=Partial/moderate (helper lifts, holds, or supports trunk or limbs, but provides less than half the effort); 2=Substantial/maximal assistance (helper lifts or holds trunk or limbs and provides more than half of the effort); and 1=Dependent (helper does all of the effort).10
Secondary Outcomes
Secondary functional outcomes were assessed using the Berg Balance Scale (BBS) and the 10-Meter Walk Test (10MWT). The BBS is a 14-item measure of static and dynamic balance, with mobility tasks graded on a 5-point ordinal scale (0–4) for a maximum score of 56, where higher scores indicate superior balance and a reduced need for assistive devices.16,17 Walking capacity was quantified using the 10MWT, which measured gait velocity over 10 meters at both a self-selected comfortable pace and a maximal safe pace.18 Participants completed the 10MWT using their customary assistive device and/or bracing (eg, cane, walker, ankle-foot orthosis), consistent with standard practice for gait speed testing in this population. No hands-on physical assistance from staff was provided unless needed for fall prevention during the timed walking test.
The assistive device and bracing used by each participant were held constant between the pre- and post-intervention assessments so that changes in gait speed reflect changes in walking performance rather than a change in the assistive setup.
Both the BBS and 10MWT are well-validated in stroke populations to predict functional ambulation, fall risk, and community mobility.17–19 These outcomes were chosen to reflect high-priority stroke rehabilitation goals: Safe mobility and return to meaningful walking. The BBS provides a practical snapshot of post-stroke balance control during everyday tasks and is clinically useful as balance impairments are closely tied to fall risk and the level of supervision a patient requires during mobility.17,19 The 10MWT captures gait speed, an indicator of walking capacity that relates to a patient’s ability to progress from household ambulation toward community mobility.18,20 Together, these measures complement Section GG by helping interpret whether functional gains are accompanied by improvements in balance safety and walking performance that matter to patients during and after inpatient rehabilitation.17,18
Data & Analysis
Outcome data were collected using a combination of the Uniform Data System for Medical Rehabilitation (UDS Pro®) and paper medical records. All analyses were performed by an independent data analyst, unaffiliated with the clinical study team or the manufacturer.
Continuous variables were summarized using mean and standard deviation (SD) for approximately normally distributed data, and median with interquartile range (IQR) for skewed or ordinal data. Ranges (minimum–maximum) were reported where appropriate. Categorical variables were summarized using counts and percentages.
Ninety-five percent confidence intervals (95% CI) were reported for Section GG scores based on 10,000 bootstrap resamples to estimate uncertainty around the median change from baseline.
Missing data were present for the secondary BBS and 10MWT outcomes because these assessments were not administered to all patients and were available for only a subset of five patients. These outcomes were analyzed descriptively on a complete-case basis without imputation. Missingness was assumed not to be systematically related to the outcome; however, because patient identifiers were not separately recorded on these assessment forms, the relationship between missingness and patient characteristics or outcomes could not be formally evaluated.
Exploratory pairwise Spearman correlations were conducted to evaluate associations between length of stay (LOS) and Section GG mobility scores at admission and change from admission to discharge. Associations between LOS and admission scores may reflect differences in baseline functional status or clinical severity, whereas associations between LOS and change scores may reflect greater opportunity for rehabilitation during longer hospitalizations.
No formal hypothesis testing was prespecified; all analyses were descriptive and exploratory in nature. Analyses were performed in R version 4. 6.
1 using RStudio (2026. 05.1 Build 225).
Results
Participant Flow and Analytic Sets
This study included 13 consecutive adult stroke inpatients meeting eligibility criteria admitted following device implementation (October-December 2024) who underwent E-RAGT during routine clinical care. Primary Section GG outcome data were available and analyzed for all 13 participants, and secondary balance and gait-speed measures (BBS and 10MWT) were available for a subset of n = 5, due to routine clinical staffing constraints. No patients were lost to follow-up or excluded after enrollment; all available cases were analyzed.
Baseline and Intervention Characteristics
Baseline characteristics are summarized in Table 2. The median age was 66 years (IQR: 53–72), and 9 patients were male (69%). Patients completed a mean of 4.2 E-RAGT sessions (range: 2–8) per stay, and individual session durations ranged from 7 to 35 minutes.
All patients demonstrated substantial baseline functional impairment. Section GG mobility scores reflected most walking and stair-related items clustered at or near complete dependence (median score of 1 for all related tasks). Transfers from chair-to-bed and sit-to-stand reflected median scores of 2 and 3, respectively, consistent with substantial to moderate assistance required.Table 2Baseline Characteristics of the End-Effector RAGT Cohort at AdmissionVariableEnd-Effector RAGT (N=13)Age, years66 (53–72) [47–82]Male sex, n/N (%)9/13 (69%)Section GG mobility items Chair-to-Bed2 (2–3) Sit-to-Stand3 (2–3) 4 Steps1 (1–1) 12 Steps1 (1–1) Walk 10 ft1 (1–2) Walk 50 ft + 2 Turns1 (1–1) Walk 150 ft1 (1–1)Berg Balance Scale (n=5)23.0 ±17.8 [3–42]10MWT Comfortable, m/s (n=5)0.25 ± 0.24 [0–0.51]10MWT Fast, m/s (n=5)0.35 ± 0.32 [0–0.64]Notes: Data are presented as n/N (%) for categorical variables, median (IQR) [min–max] for non-normally distributed numeric or ordinal variables and mean ± SD [min–max] for normally distributed variables.Abbreviation: RAGT, robotic-assisted gait training; 10MWT, 10-Meter Walk Test.
Section GG Mobility Items
Patients demonstrated median improvements ranging from 3 to 5 points across all Section GG mobility items from admission to discharge (Table 3). The largest improvement was observed in walking 50 feet with two turns, whereas the smallest improvement was 12-step stair negotiation. A visualization of change scores from admission to discharge is provided in Figure 2.Table 3Change in Section GG Mobility Scores, Balance, and Gait Speed Outcomes from Admission to DischargeOutcomeAdmissionDischargeChange95% CI for ChangeTotal Cohort (N=13)Chair to Bed2 (2–3)6 (6–6)+3 (3–4)3–4Sit to Stand3 (2–3)6 (6–6)+3 (3–4)3–44 Steps1 (1–1)4 (4–6)+3 (2–3)2–312 Steps1 (1–1)4 (4–6)+3 (3–3)3–3Walk 10 ft1 (1–2)6 (6–6)+4 (3–5)3–5Walk 50 ft + 2 Turns1 (1–1)6 (6–6)+5 (3–5)3–5Walk 150 ft1 (1–1)6 (4–6)+3 (3–5)3–5Patient Subset (N=5)Berg Balance Scale (BBS)23.0 ± 17.847.4 ± 5.624.4 ± 14.9–10MWT Comfortable (m/s)0.25 ± 0.240.60 ± 0.340.35 ± 0.28–10MWT Fast (m/s)0.35 ± 0.320.90 ± 0.390.54 ± 0.27–Notes: Unless otherwise specified, data show the median (IQR, Q1–Q3) for each Section GG mobility item at admission and discharge, as well as the median change from admission to discharge.
Bootstrap 95% CIs for median change were calculated using 10,000 resamples. BBS and 10MWT values are presented as mean ± standard deviation; 95% CIs were not calculated for these outcomes because only 5 patients contributed data. Change scores represent discharge minus admission values.Abbreviations: BBS, Berg Balance Scale; CI; Confidence Interval; 10MWT, 10-Meter Walk Test.Figure 2Individual Section GG Mobility Scores at Admission and Discharge.
Each row in the heatmap corresponds to a patient, and each cell represents a Section GG score for a specific functional domain. Scores are shown at admission and discharge. Patients are ordered from top to bottom by descending total discharge score.A) Heatmap titled, Admission.
X-axis label, Functional Domain. Y-axis label, Patient number. Rows: Patient1, Patient3, Patient6, Patient8, Patient4, Patient7, Patient11, Patient13, Patient9, Patient12, Patient10, Patient5, Patient2.
Columns: 4 Steps, 12 Steps, Chair to Bed, Sit to Stand, Walk 10ft, Walk 50ft plus 2 Turns, Walk 150ft. Legend titled, Section GG Score: 6 equals Independent; 4 equals Supervision touching assistance; 3 equals Partial moderate assistance; 2 equals Substantial maximal assistance; 1 equals Dependent. Cells show many 1 and 2 scores across domains, with some 3 and 4 scores in Chair to Bed and Sit to Stand.
B) Heatmap titled, Discharge. Same axes, rows, columns and legend as A). Cells show many 4 and 6 scores across domains, with lower scores appearing most often in 12 Steps and Walk 150ft for some patients.Heatmap of Section GG mobility scores by patient and functional domain at admission and discharge, with cell colors representing scores 1-6.Note: Score 5 (setup or clean-up assistance) is not included in the legend because no patients had this score.
At discharge, four patients were completely independent across all individual mobility tasks. Three patients remained dependent on at least one mobility item; two of these patients had dependence only for stair negotiation, while the third demonstrated functional limitations across multiple mobility tasks in addition to stair dependence. All three were discharged to skilled nursing facilities, while the remaining patients were discharged home.
Balance & Gait Speed
Secondary outcome data were incomplete due to therapy time constraints that precluded testing in some cases. Balance (BBS) and gait speed (10MWT) data were available for five patients. All five patients demonstrated improvement in each assessed outcome from admission to discharge.
A summary of balance and gait speed outcomes is provided in Table 3 and visualized in Figure 3. Figure 3Berg Balance Scale and 10-meter walk tests at admission and discharge (n=5). The dots represent individual measurements at admission and discharge for each outcome, with connecting lines showing trajectories of improvement for each patient.A) Dot plot titled, BERG Balance Scale.
The x axis categories are Admission and Discharge. The y axis label is Score, with labeled ticks at 10, 20, 30, 40, 50. Five paired dots are connected from Admission to Discharge, representing indvidual patient data.
Admission values are approximately 3, 5, 32, 33, 42. Discharge values are approximately 42, 44, 45, 50, 56. B) Dot plot titled, 10MWT (Comfortable).
The x axis categories are Admission and Discharge. The y axis label is meters per second, with labeled ticks at 0.0, 0.3, 0.6, 0.9. Five paired dots are connected.
Admission values are approximately 0.0, 0.32, 0.42, 0.50, 0.52. Discharge values are approximately 0.20, 0.45, 0.58, 0.67, 1.05. C) Dot plot titled, 10MWT (Fast).
The x axis label categories are Admission and Discharge. The y axis label is meters per second, with labeled ticks at 0.0, 0.5, 1.0, 1.5. Five paired dots are connected.A set of three dot plots showing individual Berg Balance Scale and 10 meter walk test scores (at comfortable and fast pace) at admission and discharge.
Length of Stay
Mean LOS was 16.7 ± 5.2 days (range: 7–23). Exploratory Spearman correlations examined the relationships between LOS and Section GG scores at admission and changes from admission to discharge. Correlations between LOS and baseline scores were consistently inverse across all Section GG items, indicating that longer LOS was associated with poorer baseline functional performance.
Correlations between LOS and change scores varied in direction across items, with no consistent pattern indicating that longer LOS was associated with greater functional improvement. Given the small sample size and exploratory nature of these analyses, findings should be interpreted descriptively and do not establish the absence of an association between LOS and functional improvement. The correlations were also associated with wide 95% CIs, reflecting substantial uncertainty in the estimated associations.
Detailed results are provided in Supplementary Figure 1.
Adverse Events
No serious adverse events were reported through clinician reporting or scheduled safety monitoring meetings. Mild adverse events included muscle soreness and activity-related fatigue reported by participants following sessions.
Discussion
This feasibility study provides preliminary descriptive evidence on the use of E-RAGT with the study device in patients with recent stroke undergoing inpatient rehabilitation. Improvements in functional mobility were observed across the measured transfer and walking tasks, which suggests that the study device may warrant further evaluation as a rehabilitation approach in a controlled setting. These findings do not establish a causal effect of E-RAGT or determine whether the observed improvements exceed those expected with standard inpatient rehabilitation.
Similarly, the absence of device-related adverse events during the study period supports the feasibility of implementing the device in an inpatient rehabilitation setting but does not by itself establish its safety profile. Taken together, these observations provide a basis for hypothesis generation and for prospective controlled studies evaluating whether E-RAGT provides additional benefit beyond standard-care physical therapy.
Meta-analytic evidence has shown that RAGT significantly increases the odds of achieving independent walking and improved walking speed, particularly when initiated during the first three months of recovery.3 Recent literature suggests that end-effector systems yield more consistent improvements in gait velocity and balance than exoskeletons in subacute populations, likely due in part to their ability to provide naturalistic gait trajectories and plantar sensory stimulations.6,21 By offering variable assistance and resistance modes, these systems facilitate structured, high-repetition practice and active patient engagement, aligning with established principles of motor learning in stroke care.2,22 Within this broader context, the present findings describe the operational implementation and functional trajectories observed when integrating an end-effector platform into routine inpatient stroke rehabilitation.
Standardized functional reporting under the IMPACT Act mandates tracking patient mobility across post-acute settings;10 evaluating individual Section GG mobility items provides a consistent, patient-centered method for describing functional trajectories during inpatient rehabilitation. Within this context, integrating RAGT with a technology-driven RPE model offers a practical framework for delivering high-intensity, repetitive gait practice during inpatient stroke care.2,22 Although clinical workflow parameters, therapist workload, and hospital resource utilization were not directly evaluated in this pilot study, technology assisted gait training may offer a practical approach for supporting intensive training in routine care.2
Improvements across individual Section GG mobility items were consistently observed across all assessed transfer and ambulation tasks in this cohort. Because Section GG data were available for all participants (n=13), these findings provide the most robust estimate of within-stay functional change in this study. Secondary outcomes were only available for an exploratory subset (n=5), among whom descriptive improvements were observed in comfortable gait speed (mean increase of 0.35 m/s) and balance performance (mean BBS scores increase of 24.4 points).
In this subset, observed gait speed changes reflected progression toward community-level walking (≥0.4 m/s), which is generally associated with increased independence and social participation.20 Similarly, improvements in postural stability reflect enhanced balance control during daily mobility tasks.19 However, because these secondary measures were collected in a small, uncontrolled subset, these findings remain strictly exploratory and do not imply device-specific efficacy or population-level clinical meaningfulness.
Study Limitations
This study is limited by its retrospective, single-site design and small cohort size, lack of a control group, and no fixed selection algorithm for the decision to use the device. These limitations introduce possible selection bias, limit generalizability, and restrict causal inferences. Because participants received standard care physical therapy concurrently with E-RAGT, gains in functional outcomes cannot be fully attributed to the E-RAGT sessions alone.
Intervention fidelity varied in session frequency and duration, and de-identification prevented linking specific session counts to outcomes, precluding formal dose-response analysis. Although E-RAGT was provided in combination with conventional physical therapy, participant-level measures of conventional rehabilitation exposure, including total PT minutes or gait-training time, were not available.
LOS and rehabilitation exposure varied across patients and may influence observed functional improvement. Longer LOS may provide more opportunity for rehabilitation while also reflecting lower baseline function or slower recovery. No consistent correlation was observed between LOS and Section GG improvements, although baseline scores were generally inversely correlated with LOS.
These analyses were exploratory and were used to identify potential patterns that might inform interpretation of the observed functional changes. Because LOS may reflect both rehabilitation exposure and factors related to recovery and discharge readiness, its relationship with outcomes is complex; conditioning on LOS could distort the treatment-outcome relationship.
The characteristics and outcomes reported here represent the complete set of patient-level information collected and available for analysis. More detailed clinical characterization, including stroke type, stroke chronicity, laterality, neurological severity, motor impairment, cognition/aphasia, pre-stroke ambulation, and relevant comorbidities, were not available. Similarly, participant-level E-RAGT dose and protocol-fidelity measures, including session number, duration, stepping time, total step count, body-weight support, velocity, and assistance level, were not available.
Consequently, formal assessment of treatment dose, protocol fidelity, or dose-response relationships was not possible.
RPE (modified Borg) was the only measure of intensity tracked, and was not paired with an objective measure such as heart rate for several practical and clinical reasons: Stroke patients commonly take beta-blockers, other rate-controlling antihypertensives, or have post-stroke autonomic or cardiac conduction changes that can blunt or distort the heart-rate response to exertion, which would have made a fixed percentage of heart-rate maximum an unreliable titration target for a meaningful subset of the sample. We recommend that future studies pair RPE with continuous heart-rate monitoring where feasible, particularly for participants without rate-altering medications, to corroborate the subjective measure.
Secondary outcomes (BBS and 10MWT) were available for only a small subset of patients due to clinical constraints. Section GG functional scores were assigned by treating therapists during routine clinical documentation without assessor blinding, introducing the potential for detection bias. Safety surveillance relied on routine clinical reporting and periodic administrative reviews rather than standardized per-session adverse event forms, and mild events (such as transient muscle soreness or fatigue) were reported informally rather than systematically quantified.
Additionally, the lack of a control group limits attribution of observed improvements to the E-RAGT sessions alone, as changes may reflect concurrent standard rehabilitation or spontaneous recovery. The absence of multiple follow-up time points further limits assessment of the durability and trajectory of observed functional gains beyond discharge. Finally, while no serious adverse events occurred, mild adverse events occurred in some patients but were not formally assessed.
Conclusion
E-RAGT represents a feasible and practical tool during inpatient stroke rehabilitation. In this pilot study, participants demonstrated descriptive within-stay improvements in functional mobility, balance, and gait speed alongside standard physical therapy, with high session completion and no serious adverse events reported. Future research should prioritize adequately powered, prospective, controlled trials to confirm these findings, establish standardized strategies for broader implementation, and evaluate long-term outcomes such as community ambulation and healthcare utilization relative to standard of care physical therapy.