The influence of upper extremity improvement on daily function in individuals with chronic stroke following task-oriented interventions: a structural equation modeling approach

Abstract

Background

Stroke survivors often experience persistent upper extremity (UE) motor impairments, which significantly hinder their ability to perform daily functions. While various components of UE motor abilities—such as impairment severity, muscle strength, and motor function—have been linked to daily living performance, the structural relationships among these factors remain unclear. The objective of this study was to investigate the structural relationships among changes in UE motor impairment, muscle strength, motor function, and daily living activities in individuals with chronic stroke before and after intervention.

Methods

The study included 131 individuals with chronic stroke in outpatient rehabilitation settings. Pretest and posttest measures were used, with the Fugl-Meyer Assessment evaluating UE motor impairment and the Medical Research Council scale measuring muscle strength. Motor function quality and speed were assessed using the Wolf Motor Function Test, while daily living function was gauged with the Nottingham Extended Activities of Daily Living Scale.

Results

Model fitting indices indicated an excellent fit. Results showed that changes in UE motor impairment negatively impacted daily living function (β = −0.188, p = .024) and motor function speed (β = −0.270, p = .001). Changes in muscle strength and motor function quality marginally influenced motor function speed (β = −0.165, p = .055 for each).

Conclusions

Reductions in UE motor impairment directly diminished improvements in daily living function. Enhancements in muscle strength and motor function quality indirectly promoted these functions through improved motor speed, aiding in the design and monitoring of rehabilitation protocols for stroke patients.

Introduction

Stroke patients often face long-term disabilities, including somatosensory [1], motor [2], and cognitive deficits [3]. Notably, 50% to 80% experience upper extremity (UE) motor impairment in the acute phase [4], affecting independence in their activities of daily living (ADL). Therefore, rehabilitating UE deficits through methods such as task-oriented training [5] is crucial for enhancing stroke survivors’ ADL autonomy. Understanding the connection between UE motor deficits and ADL performance is essential for developing effective rehabilitation strategies. Among several task-oriented approaches including mirror therapy (MT), recent trials highlight technology-augmented rehabilitation—e.g., wearable-sensor–based contextual training [6] and multisensory brain–computer interface (BCI) [7]—as effective means to boost motor recovery and daily function [6], further underscoring the clinical need to disentangle which UE components most strongly drive ADL/instrumental ADL (IADL) outcomes.

Previous studies have identified several components of motor abilities that are associated with ADL performance. For example, Chen et al. [8] examined the potential predictors for health-related quality of life and found that the UE score of the Fugl-Meyer Assessment (FMA) scale (FMA-UE) measuring motor impairments predicted the ADL/IADL, and Rand [9] also found that an individual’s FMA score was positively correlated with their IADL when they examined the relationship of UE function and daily living on proprioception deficits in chronic stroke. In addition, Bae et al. [10] investigated the relationship between grip and pinch strength and independence in ADL in chronic stroke individuals and found that patients’ UE muscle strength of the affected hand was associated with independence in ADL. Instead of examining the impact of UE motor impairment level and muscle strength, UE motor function was also considered as an important predictor for a stroke individual’s ADL. Complementing these findings, earlier cohort work showed that real-world paretic UE use in the first few weeks’ post-stroke is markedly reduced and covaries with strength (e.g., grip), the Wolf Motor Function Test (WMFT)-Function (positive), and WMFT-Time (negative) [11], linking motor capability to everyday arm use—a behavioral bridge to ADL/IADL.

Nichols-Larsen et al. [12] found that UE motor function assessed by the WMFT-time measuring motor function efficiency was significantly associated with the physical domain of the Stroke Impact Scale (SIS), which contained a subscale of ADL/IADL. Specifically, Franceschini et al. [13] identified that the UE motor function assessed by the Box and Block Test was the most robust predictor of the performance of ADLs after robot-assisted training. Interventional evidence further indicates that strengthening-oriented programs can improve UE strength and function without increasing tone or pain; however, pooled data suggest inconsistent transfer to ADL [14], implying that the path from impairment/strength to daily function may be indirect and mediated by other motor features.

Efficiently coordinated movements are pivotal in enhancing UE muscle strength recovery after stroke [15]. The work of Bütefisch et al. [16] elucidates that the repetitive execution of uniform hand and finger tasks substantially contributes to grip strength augmentation. Consequently, a myriad of therapeutic interventions have been developed with the objective of recuperating functional autonomy in ADL for individuals with UE motor impairments [17]. Recent randomized trials exemplify this trend: a wearable, situationally intelligent training system added to conventional therapy yielded superior gains in FMA (motor function) and Modified Barthel Index (ADL) versus usual care [6]; separately, a multisensory BCI that integrates proprioceptive, tactile, everyday upper-limb action imagery and visual feedback outperformed conventional therapy on UE motor outcomes, accompanied by network-level neuroplastic changes [7]. Notably, neither trial directly quantified muscle strength, underscoring a measurement gap that our model addresses.

Therefore, the exploration of the intricate causal dynamics between various facets of motor functionality and ADL proficiency is paramount. Such insights would empower rehabilitation professionals to formulate tailored and efficacious recovery schemes, thereby expediting the restoration of stroke survivors’ functional independence. In addition, a prognostic modeling study of post-stroke rehabilitation prioritize age, balance, mobility, and confidence to predict ADL gains, with some omitting muscle strength from the final parsimonious model despite in-patient strength improvements [18]—again suggesting complex, potentially mediated pathways from impairment and strength to functional independence.

Although existing research has explored the link between motor abilities and ADL/IADL performance, the detailed structural relationship between UE performance and ADL remains unexplained. Using a covariance-based structural equation modeling (SEM) approach could significantly contribute to understanding these structural relationships. SEM offers a framework for defining and estimating the linear interconnections among diverse measures, positing that these measures are interconnected within a theoretical structure [19]. This structural framework suggests statistical, and frequently causal, associations among the measures, essentially proposing a hypothesized pattern of linear relationships within the context of UE performance and ADL. Given mixed evidence on the direct ADL impact of strength training [14] and growing demonstrations of technology-enhanced therapies [6] improving motor scores and ADL, SEM provides a principled way to test whether strength and impairment act directly on IADL or indirectly via motor function quality and speed.

This study conducts a secondary (post hoc) analysis of pre- and post-intervention datasets from previously completed motor rehabilitation trials, applying covariance-based SEM on change scores (Δ post–pre) to model direct and indirect pathways among UE motor impairment, muscle strength, motor function, and ADL Notably, although individuals recovering from stroke may regain independence in basic ADLs, such as eating, bathing, and toileting [20], challenges often persist in more complex IADLs such as shopping, housekeeping, financial management, meal preparation, and transportation when reintegrating into work or leisure pursuits [21]. Accordingly, this research adopts the IADL measurement as a proxy for functional independence in daily living among stroke survivors. This choice aligns with evidence that early deficits in real-world paretic limb use predict poorer Activity performance (speed/efficiency on the WMFT), with downstream implications for complex daily functional activities [11], while strength changes alone may be insufficient to capture the breadth of functional independence [14].

Given previous studies showing that the daily living function of stroke individuals is influenced by UE motor impairment [8, 9], muscle strength [15, 17], and motor function [13], we hypothesized that the changes in these three motor abilities directly influence the change in daily living function. Additionally, the FMA is commonly used to evaluate the recovery of motor abilities in stroke patients [22] and can predict changes in muscle strength, such as grip strength [23]. We hypothesized that changes in the UE motor impairment level would directly impact changes in muscle strength.

Considering that motor function is typically evaluated by its quality and speed [12] and that previous research indicates that FMA-UE scores can predict changes in motor function [24] in stroke survivors, we should consider the direct effects of changes in UE motor impairment levels on changes in motor functional quality and speed. From the perspective of motor learning, motor function quality significantly impacts execution speed through mechanisms such as motor control efficiency [25], skill acquisition [26], and biomechanical optimization [27]. Improved movement quality, characterized by precise and coordinated actions, facilitates faster task execution by enhancing motor control efficiency, reducing energy expenditure, and optimizing force application. We therefore hypothesize that changes in UE motor impairment directly affect changes in motor functional quality and indirectly influence changes in motor functional speed, mediated by changes in motor functional quality. This hypothesis aligns with the understanding that comprehensive assessments of motor recovery, including FMA-UE, provide essential insights into the rehabilitation needs and progress of stroke patients.

From the perspective of exercise physiology and motor control mechanisms, muscle strength plays a pivotal role in influencing both the quality of motor function and the execution speed of movements. Defined as the maximum amount of force that a muscle can exert in a single effort, muscle strength is essential for the efficient execution of various tasks and activities. The relationship between muscle strength and motor function execution speed is underpinned by the ability of strong muscles to generate higher force outputs quickly. This rapid force generation capability allows for quicker initiation and completion of movements, leading to increased movement speed [28]. Stronger muscles can also endure greater loads and perform repetitive tasks with reduced fatigue, crucial for maintaining high-speed movements over time [29]. Therefore, strength training is important in rehabilitation protocols to improve motor function execution speed in individuals with muscle weakness or motor control impairments [30]. Considering the impact of motor function quality and muscle strength on motor execution speed, we hypothesize that the changes in motor functional quality and muscle strength both have a direct impact on the change in motor execution speed. Nevertheless, meta-analytic evidence suggests that strength gains do not always translate to ADL improvements [14], reinforcing our use of SEM to test direct versus mediated pathways from strength to daily function.

Examining how these motor abilities interact to predict the performance of independence in daily function in stroke individuals would enhance our understanding of the mechanisms between motor capabilities and the patient’s ability to perform daily functions. This might shed light on possible directions for providing personalized and hierarchical treatment protocols for individuals with stroke. Specifically, understanding the predictive relationship of these abilities to an individual’s independence in performing daily function is important for setting realistic treatment goals and identifying aspects that may be strengthened in future trials to enhance the transfer of improved UE motor abilities into daily life. Guided by the rationale synthesized across the preceding literature review, we hypothesize that (H1) changes in UE impairment, muscle strength, and motor function each directly predict change in IADL; (H2) changes in UE impairment directly predict changes in muscle strength; (H3) changes in UE impairment directly predict changes in motor-function quality and indirectly predict changes in motor-function speed via quality; and (H4) changes in motor-function quality and muscle strength each directly predict changes in execution speed. Accordingly, our model integrates impairment (FMA-UE), muscle strength, and motor function (quality/speed) to clarify which changes most strongly predict IADL gains and where technology-augmented therapies might exert their effects [6, 7].

Methods

Participants

A secondary analysis was performed on the data from pre- and posttests obtained from completed trials of stroke rehabilitation therapy. These trials were conducted in various health care settings across Taiwan, including two medical centers, eight regional hospitals, two district hospitals, and one clinic. The focus was on stroke patients who had undergone behavioral assessments before and after participating in the intervention program, with data from 131 participants included for further analysis.

To justify adequate sample size for conducting SEM, we adopted an N: q rule-of-thumb (N = participants; q = free parameters). For SEM using maximum likelihood (ML) estimation, a 5:1 to 10:1 ratio is commonly acceptable [31]; simulations suggest the incremental impact of N: q is modest under typical conditions [32, 33], and N: q ≥ 5 yields acceptable behavior of common fit indices [34]. Our hypothesized model estimates q = 13 free parameters (8 structural paths and 5 variances); with N = 131, N: q is about 10.1, satisfying these benchmarks and meeting the more conservative 10:1 guideline.

Informed written consent was obtained from all participants in these trials, and these consents were approved by the Institutional Review Boards of Chang Gung Memorial Hospital (ethics codes: 102-5696A3, 103-1621A3, 103-3564A3, 201600227A0), Taipei Medical University-Shuang Ho Hospital, Ministry of Health and Welfare (ethics code: N201510035), Cheng Hsin General Hospital (ethics code: (363)102 A-11), China Medical University & Hospital (ethics code: CMUH104-REC3-052), Cathay General Hospital (ethics code: CGH-P101054), Taipei Hospital, Ministry of Health and Welfare (ethics codes: TH-IRB-0014-0021, TH-IRB-0016-0016), and the Buddhist Tzu Chi General Hospital (ethics codes: 03-M01-013, 03-M03-055, 05-M02-010). All protocols used in these trials were performed in accordance with relevant guidelines and regulations.

Criteria for inclusion in this analysis were set to include individuals who had suffered a unilateral stroke at least 3 months prior, showed a modified Ashworth Scale (MAS) score of up to 3, achieved a score of 18 to 57 on the FMA-UE (indicative of mild to moderate motor severity), and were aged 18 years or older. The study excluded those with additional neurological or psychological diagnoses aside from stroke, such as Parkinson disease, individuals lacking tactile or proprioceptive sensation in the affected limb, those who had received Botulinum toxin injections in the 3 months preceding the study, and participants unable to follow instructions or perform tasks as determined by a Mini-Mental Status Examination (MMSE) score of 20 or below. The Institutional Review Boards of the involved clinical trials granted ethical approval for this secondary analysis, and informed consent was secured from all participants.

Task-oriented interventions

Patients received task-oriented intervention from one of the following protocols: (i) MT; (ii) robot-assisted training targeting the proximal upper limb; (iii) robot-assisted training targeting the distal upper limb; or (iv) MT combined with transcranial direct current stimulation (tDCS) [5, 35–37]. All interventions were delivered 90 min per session, 3–5 sessions per week, for a total of 20 sessions. Within each parent trial, participants remained on the same intervention throughout (no within-participant switching), and session frequency, duration, and total dose were prespecified in the trial manuals to ensure cross-site standardization and treatment fidelity. Interventions were implemented by licensed occupational therapists who had undergone protocol-specific training. Only staff who completed protocol training and passed competency checks by the principal investigator were authorized to deliver the interventions.

• I.MT A mirror was placed in the midsagittal plane so that only the non-paretic limb and its reflection were visible; participants attended the reflection and performed synchronous bilateral movements. MT comprised intransitive movements (wrist flexion–extension, forearm pronation–supination, elbow flexion–extension) and transitive tasks (peg placement, card flipping). About 30 min of functional tasks followed MT and were individualized to daily needs.
• II.Robot-assisted training—proximal upper limb Training targeted shoulder–elbow movements using an end-effector system in active-assisted mode (with passive/ active/ resistive modes as appropriate). Sessions emphasized 45 to 50 min high-repetition, planar reaching with assistance titrated to performance, followed by 40 to 45 min functional tasks (e.g., reaching to shelves, lifting/transporting containers) to complete the 90-min dose.
• III.Robot-assisted training—distal upper limb Training focused on hand and wrist control: The hand training movements included passive grasp–release, pinch–release with thumb–index–middle. The wrist training movements included wrist flexion/extension or forearm pronation/supination. Robotic-assisted training was complemented by interactive game-based exercises for hand–arm integration, lasting 45 to 60 min, followed by 30 to 45 min of functional tasks (e.g., utensil use, buttoning, phone handling) completed the session.
• IV.MT combined with transcranial direct current stimulation (tDCS) Twenty to 40 min of Anodal tDCS (2 mA) was delivered concurrently or sequentially with 20 min of MT (anode over ipsilesional primary motor cortex at C3/C4 or ipsilesional premotor cortex at F3/F4; cathode over the contralesional supraorbital area), followed by 20 to 40 min of standalone MT, and then 30 min of functional task practice to complete the session.

Outcome measures

All pre- and post-intervention behavioral assessments were administered by evaluators who were independent of treatment delivery and blinded to group/intervention allocation. Evaluators completed standardized training using a harmonized operations manual and participated in cross-site calibration sessions before data collection. Assessments were scheduled within 7 days before the first session and within 7 days after the final session to minimize measurement error and ensure consistency. All assessments were conducted by research assistants who had completed assessment training and were authorized to perform the assessment after passing competency checks conducted by the principal investigator.

Assessment of UE motor impairment levels The FMA-UE subscale is one of the most widely used tools to assess UE sensorimotor impairment in patients after stroke [22]. The FMA-UE subscale examines 33 movements scored on a 3-point ordinal scale (score range: 0–66). A higher FMA-UE score suggests less impairment. The FMA-UE has good to excellent clinimetric properties.

Assessment of UE muscle strength The Medical Research Council (MRC) Scale is an ordinal metric used to evaluate muscle strength, where each muscle is scored on a scale from 0 to 5. A higher score reflects stronger muscle strength. The MRC’s reliability across all muscle groups has been classified from good to excellent among stroke patients [38].

Assessment of UE motor function The study used the WMFT for assessing the functional capabilities of the patient’s UE. Developed by Wolf and colleagues, the WMFT enables quantitative evaluation of UE motor skills through 15 function-based tasks and two strength-based tasks [39]. The WMFT measures the duration required for task completion (WMFT-time) and functional ability on a 6-point ordinal scale (WMFT-quality), with lower WMFT-time indicating quicker execution and higher WMFT-quality reflecting superior movement quality. The WMFT’s reliability has been verified as excellent [40].

Assessment of the independence in daily function The Nottingham Extended Activities of Daily Living Scale (NEADL) is a self-reported assessment that gauges IADL across four domains: mobility, kitchen tasks, domestic chores, and leisure activities. The scale ranges from 0 to 66, where a higher total score signifies enhanced daily functional capacity. The NEADL’s psychometric validity is well documented [41].

Statistical analysis

Statistical analyses for this study were conducted using SAS/STAT 9.4 software, with the CALIS procedure, on the SAS System for Windows 11 (SAS Institute Inc., Cary, NC, USA). A p value of ≤ 0.05 was deemed statistically significant. Descriptive statistics summarized the demographic characteristics, pre- and postintervention measures, and the changes in these measures for the participants.

We used covariance-based SEM to explore the structural relationships among the changes in these measures, a multivariate statistical analysis technique that models both directional and nondirectional relationships among variables and latent constructs [19]. This study concentrated on direct and indirect effects among observed measures without considering measurement models or latent constructs, using only path analysis within the SEM framework to focus on changes among these measures. Changes in UE motor impairment, muscle strength, functional arm use quality, and speed, as well as daily living function, were indicated by difference scores between pretest and posttest assessments using FMA-UE, MRC, WMFT-quality, WMFT-time, and NEADL, respectively.

To assess the fit of the structural model and the causal relationships within it, we applied SEM with ML estimation for path coefficients. A path diagram visualized the variable interconnections, where a single-headed arrow denotes a direct association and an indirect association indicates mediation by another variable.

No single model fitting index can account for all scenarios; therefore, various indicators were used to evaluate the fit of our hypothesized model based on prior research recommendations [42]. Goodness of fit was assessed using criteria such as the p-value of χ² statistic > 0.05 (p > .05), χ²/df < 2), a comparative fit index (CFI) of > 0.95, root mean square error of approximation (RMSEA) of < 0.05, and standardized root mean residual (SRMR) of < 0.08. To refine the model, we considered the multivariate Lagrange multiplier (LM) and Wald tests for empirical significance [43]. The multivariate LM test suggests parameters to improve model fit by identifying potential reductions in the model χ² value. By adding these parameters to the model, such as a critical path between two variables, we can enhance the fit of the model. Analogously, the Wald test resembles backward deletion of variables in stepwise regression, aiming for a nonsignificant change in R² when variables are left out.

Results

This study collected the evaluation data of 151 stroke patients (97 men), with mean age of 55.62 ± 11.50 years. Table 1 summarizes the demographic and clinical characteristics as well as evaluation data of the participants. Table 2 shows the association between observed variables.

Table 1.

Demographic and clinical characteristics, and results of outcome measures (N = 131)

Characteristics	n (%) or Mean ± SD
Male (%)	97 (74.04)
Age (years)	55.62 ± 11.50
Time after stroke (months)	28.62 ± 25.07
Education (years)	11.37 ± 4.28
MMSE	28.27 ± 1.93
Outcome measures	Mean ± SD
UE impairment levels (FMA-UE)
Pretest	34.63 ± 9.67
Posttest	39.57 ± 10.87
Change	4.94 ± 4.57
UE muscle strength (MRC)
Pretest	25.12 ± 6.43
Posttest	27.42 ± 6.52
Change	2.30 ± 3.07
UE motor function (WMFT)
Quality Pretest	2.67 ± 0.61
Posttest	2.93 ± 0.69
Change	0.26 ± 0.25
Time (seconds) pretest	11.01 ± 5.54
Posttest	8.29 ± 4.90
Change	−2.72 ± 3.20
Independence in daily function (NEADL)
Pretest	31.29 ± 13.43
Posttest	34.63 ± 13.24
Change	3.34 ± 8.15

UE: upper extremity; MMSE: Mini-Mental State Examination; FMA-UE: Fugl-Meyer Assessment for upper extremity; WMFT: Wolf Motor Function Test; MRC: Medical Research Council scale; NEADL: Nottingham Extended Activities of Daily Living Scale. The “pretest” and “posttest” values represent the measurements taken before and after the intervention, respectively. The “change” value is calculated by subtracting the pretest value from the posttest value. An increase in the change values for FMA-UE, MRC, WMFT-quality, and NEADL, alongside a decrease in the change value for WMFT-time, signifies improvements in these assessments

Table 2.

Pearson’s correlation coefficients (r) for changes between pretest and posttest in observed variables (N = 131)

FMA-UE	FMA-UE	MRC	WMFT-quality	WMFT-time	NEADL
		< 0.001	< 0.001	0.008	0.150
MRC	0.463***		0.003	0.017	0.506
WMFT-quality	0.500***	0.260**		0.017	0.548
WMFT-time	−0.230**	−0.208*	−0.208*		0.009
NEADL	−0.126	−0.059	0.053	−0.228**

The bottom-left triangular panel displays the correlation coefficients among the change values of observed variables, whereas the top-right triangular panel represents the corresponding p values. The change value is determined by subtracting the pretest value from the posttest value. *p < .05; **p < .01; ***p < .001. UE: upper extremity; FMA-UE: Fugl-Meyer Assessment for upper extremity; WMFT: Wolf Motor Function Test; MRC: Medical Research Council scale; NEADL: Nottingham Extended Activities of Daily Living Scale

Path analysis of the structural model by using SEM

Figure 1 describes the hypothesized model based on theoretical and empirical evidence. All fitting indexes indicated that the model showed a good fit (χ² = 1.304, df = 2, p = .521; χ²/df = 0.652; CFI = 1.000; RMSEA = 0.000; SRMR = 0.021). In the standardized model results, IADL was directly determined primarily by impairment and speed rather than strength or quality; thus, Hypothesis 1 is partially supported. Specifically, the UE motor impairment level and the speed of UE motor function both showed significant negative effects (β = −0.293, p = .004 and β = −0.246, p = .003, respectively); however, the UE muscle strength and the quality of UE motor function had no significant direct effect on the daily living function (β = 0.112, p = .232 and β = 0.120, p = .213, respectively).

Fig. 1.

Path analysis of the hypothesized structural model. The standardized regression coefficients of the hypothesized model are presented. Dark solid arrows represent statistically significant results (*p < .05, **p < .01, ***p < .001), gray solid arrows indicate marginally significant results (†p < .10), and dark dashed arrows denote insignificant results (p ≥ .10)

UE motor impairment level positively predicted the UE muscle strength (β = 0.463, p < .001), fully consistent with Hypothesis 2. UE motor impairment level also positively predicted quality of UE motor function (β = 0.500, p < .001), but the path from quality to speed reached only marginal significance (β = −0.165, p = .054), thus the indirect effect from impairment to speed via quality was not established at α = 0.05; accordingly, Hypothesis 3 is partially supported (direct to quality supported; indirect to speed not confirmed).

Finally, the direct paths strength-to-speed and quality-to-speed were both marginal and in the hypothesized direction (each β = −0.165, p = .054), but they did not meet the conventional significance threshold; therefore, Hypothesis 4 is not supported at α = 0.05 (trend-level evidence only). Overall, the pattern highlights impairment and speed as the key direct predictors of IADL, with strength and quality more likely to contribute through speed or other mediating pathways.

To improve the precision of our hypothesized model by removing the least critical path(s), the Wald test was used to identify paths that could be eliminated without adversely affecting the model’s fit. The Wald test outcomes indicated that three paths did not significantly alter the χ² variance (p = 0.091). Given the marginally significant effect of the path from UE motor function speed to daily living function (χ² = 3.583, df = 1, p = .058) and the documented significance of motor function in daily living activities in prior research [13], we decided to remove only the paths from UE muscle strength and UE motor function quality to daily living function (χ² = 2.892, df = 2, p = .236) because these also exhibited no significant effects.

Figure 2 presents the revised, more parsimonious model, which continued to fit the data well (χ² = 4.326, df = 4, p = .368; χ²/df = 1.081; CFI = 0.996; RMSEA = 0.025; SRMR = 0.039). Relative to the hypothesized model, overall fit remained comparable (CFI: 1.000 → 0.996; RMSEA: 0.000 → 0.025; SRMR: 0.021 → 0.039), supporting the deletion of the two non-significant direct paths while preserving the substantive structure of the model. In the standardized results for the revised model, UE motor impairment level positively predicted UE muscle strength (β = 0.463, p < .001) and the quality of UE motor function (β = 0.500, p < 0.001). Both UE muscle strength and the quality of UE motor function exhibited marginal negative associations with the speed of UE motor function (each β = −0.165, p = 0.055). Consistent with the retained direct effects on IADL, UE motor impairment level and the speed of UE motor function showed significant negative influences on daily function (β = −0.188, p = 0.024; β = −0.270, p = 0.001, respectively). Taken together, the revised model achieves greater parsimony with minimal change in fit relative to the hypothesized model and preserves the key pattern in which impairment and speed serve as the direct predictors of IADL.

Fig. 2.

Path analysis of the revised structural model. The standardized regression coefficients of the revised model are displayed. Dark arrows represent statistically significant results (*p < 0.05, **p < 0.01, ***p < 0.001), whereas a gray arrow indicates a marginally significant result (†p < 0.10). This revised model, which excludes the path from muscle strength and motor functional quality to daily living function, considers the results of the Wald test and aims for model simplicity

Discussion

To the best of our knowledge, this study is the first to use a path model to elucidate the impact of relationships among changes in UE motor impairment level, UE muscle strength, and UE motor function (both quality and speed) on the change in independence in daily activities after task-oriented training among individuals with chronic stroke. The results (refer to Fig. 2) suggest that the change in independence in daily activities for stroke survivors is influenced by the changes in their UE motor impairment level through both direct and two indirect pathways. The first indirect pathway is mediated by the transition from UE muscle strength to motor functional speed (i.e., muscle strength-efficiency pathway, as illustrated in the middle part of Fig. 2), whereas the second is mediated by the transition from motor functional quality to speed (i.e., functional quality-speed pathway, as illustrated in the top part of Fig. 2). The mediation effect refers to how the influence of an independent variable (i.e., UE motor impairment level) on a dependent variable (i.e., independence in daily function) is transmitted through other variables (i.e., UE motor function and UE muscle strength) [44].

Our structural model elucidated the complex relationship between improvements in UE motor impairment levels and daily functioning in stroke survivors, uncovering a direct negative impact, contrary to the positive correlations highlighted in prior studies [8, 9]. Notably, Veerbeek et al. [45] identified that robot-assisted therapy for the paretic UE yielded modest yet significant motor control enhancements without corresponding improvements in basic ADLs. Further compounding this complexity, Ezema et al. [46] reported that poststroke depression adversely affects functional recovery, hinting at the potential role of psychological factors, particularly depression, in our model’s observed negative direct effect. This connection between UE motor impairment improvement and decreased daily functionality might reflect the adverse psychological states, although our study did not directly assess psychological conditions. This gap underscores the necessity for future research to integrate psychological assessments into intervention designs to more comprehensively understand the recovery processes of stroke survivors.

Building on these overall findings, we considered whether differences in intervention protocols and outcome definitions might account for the discrepant direct path observed in our model. Our pooled dataset integrated several task‑oriented protocols (MT, robot‑assisted proximal/distal training, MT+tDCS) and modeled change in instrumental activities of daily living (NEADL) as the primary outcome. By contrast, some of the prior reports we cited drew conclusions from different designs and targets. For example, Chen et al. [8] examined an ADL/IADL composite at a single inpatient timepoint (2–3 weeks after admission) under daily physiotherapy/occupational therapy focusing on facilitation, balance/transfers, strengthening, hand function, and ADL training, rather than longitudinal pre–post change. Rand [9] analyzed proprioception, FMA-UE, and IADL in a cross-sectional framework with no intervention, precluding causal interpretation of positive associations. Finally, Harris & Eng [14] found that strengthening improved grip strength and upper-limb function but not ADL. Considering the findings on Harris & Eng [14] and the present study, the impairment-level gains may not directly and positively generalize to complex daily function. The impairment gains may enhance complex daily function through motor-function improvements. Taken together, differences in protocol focus, study design, and outcome definitions across studies likely underline the discrepancy in certain aspects; residual heterogeneity (e.g., recovery phase, dose) merits stratified or sensitivity analyses.

Moreover, our findings spotlight two indirect pathways—muscle strength-efficiency (MS-E) and functional quality-speed (FQ-S)—that mediated positive influences on daily functioning through reduced UE motor impairment. This suggests that the beneficial effects on daily functions are attributed to these indirect impacts rather than the direct effect of UE motor impairment reduction. Specifically, the adverse direct impact underscores that simply reducing UE motor impairment, without concurrently enhancing muscle strength and motor functionality, fails to guarantee functional improvement. It indicates that functional enhancement is only assured when reductions in UE motor impairment trigger improvements via the MS-E and FQ-S pathways. Clinically, this implies that targeting abnormal synergy patterns, as evaluated by the FMA [22], alone may not suffice to boost daily functionality. Instead, addressing these patterns to improve muscle strength and motor function appears critical for enhancing daily functional performance.

In our framework, FMA-UE represents impairment/synergy release, MRC-type strength indexes force-generation capacity, and task completion time indexes speed/performance, consistent with a mechanistic chain of FMA-UE → strength → faster task execution. Across pooled trials, protocol emphases may slightly differ with respect to synergy reduction and strength/speed/mobility. Mirror therapy (with or without adjunct anodal tDCS) encourages individuated movement control that often involve desynergy movement patterns and functional control of joints and muscles, possibly reducing the severity of mass synergies [37]. Robot-assisted proximal/distal training not only delivered active-assisted, goal-directed practice that can be tuned toward independent joint control and reducing synergies [36] but also provides resistant mode that train the force generation of muscle when strengthening is appropriate for participants. This heterogeneity implies that some participants received more desynergy- and function-focused dosing whereas others received desynergy- and strength-focused dosing, which could attenuate an overall direct impairment → IADL path yet preserve indirect effects via motor quality/muscle strength and speed. To further elucidate the possible pathway regarding the elements of the impairments such as tone, stiffness, and passive range of motion and the severity of motor impairments, future studies should include dedicated measures of tone (e.g., MAS), passive/active range of motion, and quantitative kinematics for analysis and stratify patients with different levels of motor severity.

In the MS-E pathway, as depicted in the middle part of Fig. 2, the enhancement of muscle strength plays a pivotal role in mediating the impact of UE motor impairment level on the execution speed of motor function, leading to improvements in the daily functioning of stroke survivors. This mediation aligns with exercise physiology and motor control mechanisms that posit a critical influence of muscle strength on the speed of movement execution [28, 29]. Increased muscle strength enables rapid force generation, crucial for quick initiation and completion of movements, especially in tasks requiring speedy responses or adaptation to environmental changes [47]. Notably, this pathway reflects findings from previous studies indicating that muscle strength improvements, such as grip strength, can be predicted [23] and underscore the importance of strength training in rehabilitation protocols to increase the speed of motor function in individuals with muscle or motor control weaknesses [30]. However, the direct effect of muscle strength on daily functioning improvement was not found in our model, diverging from some previous research [15, 17], and indicating that the role of muscle strength in daily functioning is mediated by the efficiency of motor execution. Increased muscle strength is posited to accelerate motor execution speed, which, in turn, improves daily functioning.

In addition, our model did not reveal a direct significant effect of FMA-UE on WMFT execution time, as indicated by the LM test (p = .288), suggesting that the relationship between FMA-UE and WMFT execution time is mediated through muscle strength. This mediation reflects that those improvements in motor impairment or abnormal synergy patterns, assessed by FMA-UE, improve the UE muscle strength and subsequently lead to faster task completion. This finding guides future researchers and practitioners to focus on enhancing both muscle strength and task execution efficiency in stroke rehabilitation protocols, emphasizing their integrated impact on improving daily functioning outcomes for stroke survivors.

In the FQ-S pathway illustrated on top part of Fig. 2, improvements in the quality of motor function serve as a significant mediator between UE motor impairment levels and the execution speed of motor functions, ultimately contributing to the enhanced daily functioning of stroke survivors. The movement quality measured by the WMFT [40] evaluates whether the movement is performed by the affected UE and how well the affected UE performs the task. This finding suggested that improving abnormal synergy patterns enhances the use of the affected UE or the movement patterns of the affected UE executing the task. The improved quality of using the affected UE may further enhance the speed and efficiency of task execution.

Research demonstrates that motor recovery, as gauged by the FMA-UE, significantly affects WMFT quality scores. Improved FMA-UE scores are associated with better WMFT task performance quality, indicating that motor recovery enhances movement patterns for task execution [48, 49]. Despite existing studies exploring the link between FMA-UE scores and WMFT execution speed, suggesting potential effects of motor recovery on speeding up task completion, our model found no direct impact of FMA-UE on WMFT-time (refer to the LM test, p = .288). Instead, our results suggest the relationship between FMA-UE and WMFT-time is mediated by WMFT performance quality. This indirect effect implies that improvements in motor impairments, assessed by FMA-UE, lead to higher quality of task execution, which in turn, accelerates task completion.

This mediation concurs with literature suggesting that the connection between FMA-UE and WMFT-time stems from enhanced motor recovery promoting more efficient task performance, thus shortening completion times [9, 12, 49]. Contrary to previous findings [12], our model indicates that the direct effect of motor function quality on daily function was not present, suggesting that the previously observed effects of functional quality might actually stem from its influence on execution speed, which then affects daily functional outcomes in stroke survivors. The FQ-S pathway highlights the importance for future research and clinical practice to focus not only on enhancing functional execution quality but also on improving task execution speed within stroke rehabilitation protocols to optimize daily functioning outcomes for stroke survivors.

In summary, this model elucidates the pathways through which UE motor impairment affects the recovery of daily function in stroke survivors after intervention. It reveals that the level of UE motor impairment impacts daily function through one direct and two indirect pathways. Although the direct impact is negative, the indirect effects mediated by muscle strength and the quality and speed of motor function are positive. This complexity underscores the nuanced interplay between motor abilities in influencing daily living activities after stroke rehabilitation.

This study presents several limitations that need to be addressed in future research. First, to validate the robustness and generalizability of our model, subsequent studies should include larger sample sizes and use a greater number of indicators to represent each measure, rather than relying solely on observed measures, to mitigate the impact of measurement errors [50]. Such approaches will improve the model’s reliability and validity.

Second, our findings suggest only a marginal impact of muscle strength and motor functional quality on functional execution speed. Considering the varied metrics used in the assessment tools—for example, muscle strength assessment involves various parts of the UE, and motor function quality and execution speed is gauged through multiple functional movements—future research is needed to explore how these different measurements affect the recovery of daily functional independence in stroke survivors.

Third, this study did not consider the psychological states of stroke survivors, such as the impact of poststroke depression on daily functions [46]. Future studies should incorporate an assessment of psychological states to evaluate whether the recovery of motor abilities influences daily functionality, potentially moderated by psychological conditions.

Fourth, although the MRC scale is widely used in clinics, it is ordinal and rater-dependent and therefore less sensitive to small changes, making it a suboptimal standalone index of UE strength [51–53]. We recommend supplementing MRC with objective dynamometry—standardized handgrip (e.g., Jamar^® or equivalent hand-held dynamometer) and joint-specific isokinetic testing—to improve measurement precision and reproducibility; recent studies report good-to-excellent reliability for isokinetic protocols in neurological/upper-limb populations [54].

Fifth, our pooled cohort spanned mild–moderate impairment, yet we did not stratify or fit multi-group SEM by baseline severity. Given evidence that training-related gains can vary by severity, reductions in abnormal synergies and downstream improvements in strength/speed may likewise differ between milder and moderate subgroups (e.g., Harris & Eng [14]). Future research should use a priori stratification or multi‑group SEM (e.g., mild vs. moderate FMA‑UE) with adequate power to test whether the MS–E and FQ–S pathways differ across severity levels.

Lastly, our study’s cross-sectional data focused on younger individuals with chronic stroke than are typically seen in stroke demographics (average age, 74.3 ± 13.6 years), and men comprised a larger proportion of our sample [55]. Therefore, the applicability of our structural model may not extend to the entire stroke population, such as those in the subacute and acute phases or with moderate to severe cognitive impairments. Future studies should seek to include stroke individuals with a broader range of characteristics to increase the model’s generalizability.

Conclusions

This study explored how various motor abilities interact to influence the recovery of daily living functions in stroke patients after intervention. The findings demonstrate that UE motor impairment level influences recovery of daily function through a direct negative effect and two indirect positive pathways. This highlights the importance for future researchers and clinical practitioners not only to consider the improvement in impairment levels (which may only have a negative effect) but also to ensure that such improvements lead to enhancements in muscle strength and the quality and speed of motor function. This approach is crucial for designing effective training programs and monitoring the recovery of daily living functions. Ultimately, this study emphasizes the significance of a comprehensive and nuanced understanding of motor recovery and its impact on daily living functions for clinical practice and rehabilitation protocols, aiming for a holistic improvement in stroke survivors’ quality of life.

Abbreviations

ADL

Activities of daily living

BCI

Brain–computer interface

CFI

Comparative fit index

FMA

Fugl-Meyer Assessment

FMA-UE

Fugl-Meyer Assessment for upper extremity

FQ-S

Functional quality-speed

IADL

Instrumental activities of daily living

LM

Lagrange multiplier

MAS

Modified Ashworth Scale

ML

Maximum likelihood

MMSE

Mini-Mental Status Examination

MRC

Medical Research Council

MS-E

Muscle strength-efficiency

MT

Mirror therapy

NEADL

Nottingham Extended Activities of Daily Living Scale

RMSEA

Root mean square error of approximation

SEM

Structural equation modeling

SIS

Stroke Impact Scale

SRMR

Standardized root mean residual

tDCS

Transcranial direct current stimulation

UE

Upper extremity

WMFT

Wolf Motor Function Test

WMFT-quality

Functional ability on a 6-point ordinal scale of the Wolf Motor Function Test

WMFT-time

Duration required for task completion of the Wolf Motor Function Test

Author contributions

SHL and CYW contributed to the conception of the study; SHL and TRY contributed to methodological design; SHL organized the database and performed the statistical analysis; SHL wrote the first draft of the manuscript; and SHL and CYW wrote sections of the manuscript. All authors contributed to manuscript revision and read and approved the submitted version.

Funding

This study was supported by Chang Gung Memorial Hospital (BMRP553, CMRPD1P0022), National Science and Technology Council in Taiwan (NSTC 113-2314-B-182-049-MY3), the National Health Research Institutes (NHRI-EX114-11105PI), and Healthy Aging Research Center (URRPD1Q0181) in Taiwan.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request. The data sharing adopted by the authors complies with the requirements of the funding institution and with institutional ethics approval.

Declarations

Ethics approval and consent to participate

This study was supported by Chang Gung Memorial Hospital (BMRP553, CMRPD1P0022), National Science and Technology Council in Taiwan (NSTC 113-2314-B-182-049-MY3), the National Health Research Institutes (NHRI-EX114-11105PI), and Healthy Aging Research Center (URRPD1Q0181) in Taiwan.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.