Can the Wolf Motor Function Test Be Streamlined?

Abstract

Background

To assess upper extremity (UE) capabilities following stroke, the Wolf Motor Function Test (WMFT) measures time to complete 15 UE tasks and 2 strength tasks, but takes 30 to 45 minutes for the clinician to complete.

Objective

In an effort to streamline the WMFT, this study evaluated the association between the magnitude of improvement on any timed task of the WMFT and the change score on all other tasks among participants in the Extremity Constraint Induced Therapy Evaluation (EXCITE) trial.

Methods

This association was evaluated using regression methods according to chronicity and controlling for key covariates (functional level, gender, concordance) for log mean WMFT scores.

Results

After controlling for covariates, 6 tasks (hand to table (front), hand to box (front), reach and retrieve, lift can, lift pencil, and fold towel) influenced the overall WMFT score for survivors meeting EXCITE criteria and treated within 3 to 9 months poststroke. Six different tasks (extend elbow weight, hand to box (front), lift can, lift pencil, turn key in lock, and fold towel) influenced the overall WMFT score for those receiving constraint-induced movement therapy (CIMT) 1 year later. The importance of certain tasks relative to others may best represent overall UE function, but this streamlining enables the clinician to prioritize these tasks in the evaluation.

Conclusions

The delineation of those tasks depends on the time poststroke from enrollment to CIMT. This study demonstrates that the WMFT can be streamlined from 17 to 6 tasks.

Stroke is the leading cause of adult disability and also the third leading cause of death in the United States.¹ Over the past 30 years, mortality from stroke has declined, resulting in an increasing number of patients receiving rehabilitation.² Residual upper extremity (UE) hemiparesis persists long after the initial stroke in 50% of stroke survivors,³ and a considerable rehabilitative effort is required to maximize functional potential.

Among the many outcome measures to assess UE capabilities following stroke, the Wolf Motor Function Test (WMFT), a laboratory-based test, measures time to complete 15 UE tasks and 2 strength tasks. The WMFT is typically administered to patients with mild to moderate stroke⁴ to assess the impact of specific interventions, such as constraint-induced movement therapy (CIMT),⁵ functional electric stimulation-assisted exercise therapy (FES-ET),⁶ robotic therapy,⁷ and repetitive bilateral arm training with rhythmic auditory cueing (BATRAC).⁸ The reliability⁹ and validity^4,10,11 of the WMFT have been confirmed in previous studies. Unlike other assessments for stroke, the WMFT tests a wide variety of tasks (progressing from simple to complex) and is composed of 3 parts: (1) time, the speed of the completion of the tasks; (2) functional ability, the quality of movement during the task; and (3) strength.¹² The progressive tasks are relative to the number of joints used to perform the task, beginning with selective activation at the shoulder joint to complex but functionally relevant movements. The primary outcome measure is the mean or log mean WMFT score for the 15 timed tasks, keeping the 2 strength task scores separate.¹³ Because of the skewed distribution of normative data,^10,14 the transformed log mean of the WMFT has also been used to describe the data⁵ and in performing statistical analyses.

This test can differentiate patients who are considered higher or lower functioning among those individuals with mild to moderate stroke as defined by wrist and digit active range of motion.¹³ Moreover, the test is not influenced by whether the affected limb is dominant; however, women tend to exhibit slower performance times than men.¹⁰

The WMFT was used as a primary outcome measure in the Extremity Constraint Induced Therapy Evaluation (EXCITE) trial, whose primary outcomes⁵ and protocol have been detailed elsewhere.¹³ This randomized clinical trial demonstrated that CIMT produced greater functional changes in patients with subacute stroke (3–9 months) than did usual and customary care, and that both higher and lower functioning participants could improve significantly. Both the WMFT and the Motor Activity Log (MAL), a survey that measures everyday activities, and the Stroke Impact Scale, a measure of health related quality of life,⁵ were used for assessment in this trial.

Because of the substantive improvements seen on the WMFT by EXCITE participants, efforts to determine if there is an association between the magnitude of improvement on any 1 task and the change score on all other tasks among participants receiving CIMT immediately or 1 year after enrollment seemed reasonable. The high interrater reliability and validity of the WMFT serve as the foundation for this evaluation. The purpose of the present study was to potentially streamline the WMFT by defining only those tasks that showed the greatest improvement relative to the overall WMFT change score. Such streamlining is important because it could reduce the time necessary to administer the test (time to administer the WMFT ranges from 30–45 minutes) and provide the most pertinent information about recovery potential to the clinician, since tasks are targeted to demonstrate movement contributions from specific joints. This study also examined the extent to which this association is modified by time poststroke after controlling for key variables including functional level, gender, and concordance.

Methods

The 222 participants in the EXCITE trial were recruited from 6 institutions including: Emory University, the Ohio State University, University of Alabama at Birmingham, University of Florida, University of Southern California, and University of North Carolina at Chapel Hill in collaboration with Wake Forest University. Eleven of the 222 participants had sustained hemorrhagic strokes while all others had experienced ischemic strokes. The present analysis includes data from 169 participants in the EXCITE trial who attended all 10 days of CI therapy training as well as preevaluations and postevaluations (42 did not attend all 10 days of training or did not have preevaluations and postevaluations).¹³ Of the 169 participants, 96 were in the immediate group (3–9 months poststroke) and 73 were in the delayed group (12–21 months poststroke). Of the 96 participants in the immediate group, 20 were classified as low functioning and 76 were classified as high functioning, 62 were men and 34 were female, and 46 were concordant and 50 were discordant (Table 1). The remaining 73 participants were randomized to the delayed group and received CIMT therapy 1 year after randomization. Of these 73 participants, 15 were classified as low functioning and 58 were classified as high functioning, 44 were men and 29 were women, 40 were concordant and 33 were discordant (Table 2). All sites had received approval from their respective institutional review board prior to recruiting participants whose data are examined in this article.

Table 1.

Demographic Representation of Immediate Participants in the EXCITE Trial

		Functional Level		Gender		Concordance
		HF	LF	M	W	C	D
	(n = 96)	76	20	62	34	46	50
Functional level	HF (n = 76)			49	27	36	40
	LF (n = 20)			13	7	10	10
Sex	M (n = 62)	49	13			28	34
	W (n = 34)	27	7			18	16
Concordance/discordance	C (n = 46)	36	10	28	18
	D (n = 50)	40	10	34	16

Abbreviation: HF, high functioning; LF, low functioning; M, men; W, women; C, concordant; D, discordant.

Table 2.

Demographic Representation of Delayed Participants in the EXCITE Trial

		Functional Level		Gender		Concordance
		HF	LF	M	W	C	D
	(n = 73)	58	15	44	29	40	33
Functional level	HF (n = 58)			33	25	33	25
	LF (n = 15)			11	4	7	8
Sex	M (n = 44)	33	11			25	19
	W (n = 29)	25	4			15	14
Concordance/discordance	C (n = 40)	33	7	25	15
	D (n = 33)	25	8	19	14

Abbreviation: HF, high functioning; LF, low functioning; M, men; W, women; C, concordant; D, discordant

CIMT

CIMT is a well-known intervention that combines behavioral and physical training by requiring a patient to use the impaired limb when the less affected limb is immobilized, usually by placing the less affected hand in a mitt that prevents opposition and grasp or, alternatively, by placing the entire UE in a sling. In the EXCITE trial, participants were randomly assigned to either the immediate (3–9 months poststroke) or delayed (over 1 year later) groups. Approximately half of those participants randomized to the delayed group received traditional physical therapy or alternative therapy treatments not associated with the EXCITE trial.^5,13 During the training period, all participants donned a safety mitt on their less impaired UE for 90% of their waking hours for 14 consecutive days. On weekdays (10 days), they received adaptive task practice (ATP) and repetitive task practice (RTP) progressing up to 6 hours. ATP is a form of behavioral training, also referred to as “shaping,” while RTP is less structured emphasizing repetition of common tasks such as eating or grooming for typical intervals of 15 to 20 minutes.⁵

WMFT

The 15 timed tasks of the WMFT for both CIMT groups (immediate and delayed) were standardized for performance of evaluators, instructions to participants, and positioning of items (eg, chairs and tables).¹³ This test was administered pre-treatment, immediately posttreatment, and every 4 months for 2 years. The outcome measure was videotaped so independent evaluators could objectively record scores.¹³ If the participant could not complete a task, a score of 121 was recorded. Scores of 121 were not included in the present analyses (Table 3).⁵

Table 3.

Within Group Analyses of Immediate and Delayed Participants Showing the Number of Participants Who Could Not Complete the Task^a

Task	Immediate (n = 96)		Delayed (n = 73)
	Pretest	Posttest	Pretest	Posttest
Forearm to table	0 (0%)	0 (0%)	2 (3%)	1 (1%)
Forearm to box	5 (5%)	1 (1%)	4 (5%)	5 (7%)
Extend elbow side	3 (3%)	1 (1%)	5 (7%)	6 (8%)
Extend elbow weight	5 (5%)	1 (1%)	5 (7%)	9 (12%)
Hand to table (front)	0 (0%)	1 (1%)	1 (1%)	3 (4%)
Hand to box (front)	1 (1%)	1 (1%)	10 (14%)	8 (11%)
Reach and retrieve	1 (1%)	0 (0%)	5 (7%)	1 (1%)
Lift can	14 (15%)	3 (3%)	21 (29%)	13 (18%)
Lift pencil	15 (16%)	4 (4%)	19 (26%)	12 (16%)
Lift paper clip	20 (21%)	3 (3%)	17 (23%)	12 (16%)
Stack checkers	31 (32%)	4 (4%)	32 (44%)	14 (19%)
Flip cards	8 (8%)	2 (2%)	17 (23%)	7 (10%)
Turn key in lock	11 (11%)	4 (4%)	16 (22%)	10 (14%)
Fold towel	3 (3%)	1 (1%)	17 (23%)	6 (8%)
Lift basket	16 (17%)	9 (9%)	30 (41%)	14 (19%)

^aPercentage of the total number (n) of participants attempting the task in parentheses.

Definition of the Covariates

Three covariates were used for the analysis, including functional level, gender, and concordance. Gender requires no definition.

Functional level

The functional level of each participant was determined according to the wrist and digit active extension range of motion, a measure that had been determined to predict improvement in UE function use in past biofeedback studies^15,16 and had adopted as an inclusion criterion for many subsequent CIMT studies. Those in the high functioning group demonstrated active wrist extension of at least 20 degrees and active extension of the metacarpophalangeal joints and interphalangeal joints of at least 10 degrees. Those in the low functioning group had at least 10 degrees of active wrist extension (but >20 degrees), 10 degrees of thumb abduction/extension, and at least 2 additional digits with 10 degrees of extension. To assess the inclusion/exclusion criteria for functional groups, participants placed their prone forearms on a table with the wrist over the edge and lowered it to a resting, flexed position. They were then asked to raise the wrist into extension without moving the forearm from the table and to repeat these movements 3 times over 1 minute. Additionally participants had to be capable of at least 45 degrees of active shoulder flexion and abduction and 20 degrees elbow extension from a 90 degree flexed position.¹³

Concordance

The distinction between concordance and discordance was based on the participant’s premorbid handedness. Concordant participants had a more affected UE, which was their dominant hand based on assessment using the Edinburgh test for hand dominance or self-report, while discordant participants had a more affected UE, which was their nondominant hand.⁵

Statistical Analyses

To assess the importance of each item in the WMFT, the change score over the CIMT training intervention was examined. The change score is the difference between the time required to complete the task pre- and post-CIMT, and is an evaluation of the impact of CIMT on UE rehabilitation. The overall WMFT change score is the difference between the total time required to complete all tasks pre- and post-CIMT. For a given task item, mean scores were converted to log mean scores. The log mean change score in the immediate and delayed groups was plotted against the total log mean change score on the WMFT excluding the given task item (although both mean and log mean scores are used in various WMFT publications,^17–19 log mean values are preferred for analysis, because the transformation reduces problems related to the skewness of the time values in the WMFT).¹⁰ Tasks that could not be completed within 120 seconds for any given participant were excluded for this analysis. There were a small number of these cases, and their assigned value of 120 might skew the results.^4,13 The plots then revealed the line of best fit for the data set by each task.

Regression methods were used to evaluate the relationship between overall change (excluding the given task being evaluating) and change on a given task for the overall groups (immediate and delayed). Analyses were performed to determine if the item-total relationships were significant (within groups). In this analysis, the main interest lies in determining how much variance in the overall difference can be accounted for by a single task. After graphing the slopes, the possible contributions made from distinct outlier observations were examined for each task. Outlier data points are considered as any observation that, based on best judgment, is thought to be very discrepant or different from other values, but may have contributed substantially to the significant association between the change score in a task and the change score for the remainder of the WMFT. Usually these consisted of 1 or 2 participants. Outliers were excluded to clarify the relationship for most participants, in a manner not distorted by these unusual cases. The data were then reexamined in the absence of those points to determine if a significant relationship between change in task score and total WMFT still existed. Last, the outlier data for log mean scores were again examined to determine if the same participants demonstrated consistent aberrations.

The overall relationship between item and total was determined with categorical covariates included (functional level, concordance, gender). When these covariates are excluded, the relationship may be determined partly by group differences (male/female, concordant/discordant, high/low functional level).

When these covariates are included in the model the analysis is done on a within-group basis, and group differences are statistically removed or partialled out. Items are considered to be representative of the whole test if there is a significant partial correlation between the item (log-transformed) change score and the overall log mean change score (excluding the item). The partial squared correlation is a value that corresponds directly to the significance of the item in the full linear model. Both the overall R² (including covariate components) and the R² with covariates partialled out are included and examined. The overall R² is the variance accounted for in the full model, including covariate components. The partial R² is the amount of influence the individual dependent measures has on the outcome without the influence of the covariates.

Results

After controlling for covariates, the immediate group showed a significant (P < .05) relationship between log mean change score for 6 tasks (hand to table [front], hand to box [front], reach and retrieve, lift can, lift pencil, and fold towel) and the log mean change score for the remainder of the WMFT (Table 4). These items have significant partial correlations for the immediate group.

Table 4.

Within Group Analyses of the Immediate and Delayed Participants Showing Overall and Partial R² Values (and Associated P Values) for Log mean Scores

Task	Immediate				Delayed
	Overalla		Partialb		Overalla		Partialb
	R2	P Value	R2	P Value	R2	P Value	R2	P Value
Change forearm to table (log)	.0228	.7134	.0139	.2603	.0008	.9996	.0001	.9402
Change forearm to box (log)	.0129	.8791	.0002	.9056	.0155	.8982	.0001	.9332
Change extend elbow side (log)	.0359	.4993	.0231	.1725	.0461	.5162	.0421	.1097
Change extend elbow weight (log)	.0495	.3225	.0452	.0509	.1565	.0194	.1535	.0014
Change hand to table (front) (log)	.0732	.1362	.0659	.0141	.0368	.6292	.0302	.1566
Change hand to box (front) (log)	.0602	.2219	.0511	.0364	.1221	.0615	.1185	.0061
Change reach and retrieve (log)	.0834	.0913	.0674	.0130	.0425	.5589	.0310	.1638
Change lift can (log)	.2897	< .0001	.2791	< .0001	.2259	.0014	.2045	.0006
Change lift pencil (log)	.1035	.0396	.1022	.0070	.1005	.1205	.0956	.0192
Change lift paper clip (log)	.0225	.7180	.0045	.5882	.0361	.6387	.0223	.2768
Change stack checkers (log)	.0631	.1995	.0496	.0960	.0478	.4968	.0216	.3410
Change flip cards (log)	.1104	.0293	.0432	.0715	.0746	.2535	.0596	.0724
Change turn key in lock (log)	.1252	.0152	.0472	.0577	.2111	.0026	.1848	.0009
Change fold towel (log)	.1232	.0167	.0716	.0164	.2226	.0016	.2086	.0004
Change lift basket (log)	.0193	.7743	.0028	.6781	.1335	.0424	.0746	.0575

^aOverall, refers to the analysis including the 3 covariates (functional level, concordance, gender) and the individual item.

^bPartial, refers to the item-specific explained variance after statistically removing the categorical covariates.

The delayed group showed a significant relationship between log mean time change score for 6 tasks (extend elbow weight, hand to box [front], lift can, lift pencil, turn key in lock, and fold towel) and the log mean change score for the remainder of the WMFT (Table 4). These items have significant partial correlations for the delayed group.

The items that are significant in both subsamples are hand to box, lift can, lift pencil, and fold towel. The turn key in lock task is significant in the delayed group, but is not significant in the immediate group (P = .0577). Similarly, extend elbow weight is significant in the delayed group, but is not significant in the immediate group (P = .0509). While these analyses are not significant for both groups, the significance is at the marginal level for the immediate group, suggesting that these items are worth further investigation. For 2 tasks, significant results differed between groups. The hand to table task is significant in the immediate group, but not in the delayed group (P = .1566). The reach and retrieve task is also significant in the immediate group, but not in the delayed group (P = .1638). The partial R² values, however, are not substantially different in the 2 groups (R² values are .0659 and .0674 in the immediate group and .0302 and .0310 in the delayed group). Thus, while the significance of the relationships differs between the samples, the R² values are, in these 4 cases, relatively small. Interestingly, the flip cards task, which requires considerable coordination across multiple joints and represents a fine motor skill with a precise pronation/supination component, is relatively close to significance in both immediate and delayed groups, (P = .0715 and .0724, respectively).

Unusual Observations

The importance of covariates that influence the relationship between individual tasks and total WMFT score should be tempered by the presence of outlier data. There were very few outlier data points. The extent to which individual patients may have influenced the data was examined by characterizing these outliers. Nine participants contributed to the 17 outlier points for the log mean scores. There was no apparent relationship between individuals who contributed aberrant data and the change score; that is, different participants were outliers for different tasks. Last, exclusion of outlier data is considered to be questionable in many evaluations, because the outlier observations are valid but somewhat discrepant. For the present data set, the issue is not the evaluation of an experimental result, but the comparative evaluation of tasks. For this reason, outlier exclusion is less problematic in this evaluation.

Discussion

To determine if the WMFT can be streamlined by assessing only those tasks that show the greatest improvement relative to the overall WMFT change score, this retrospective study represents the first attempt to evaluate the relative contribution made by each of the 15 WMFT timed tasks to the overall change in the WMFT among those EXCITE participants who underwent CIMT on each of the 10 days over which time the training occurred. We evaluated the association between the magnitude of improvement on any 1 task of the WMFT and the change score on all other tasks among participants receiving CIMT in the EXCITE trial. This effort was undertaken using linear model analyses controlling for functional level, gender, and concordance. Our results would suggest that administration of the WMFT could be reduced to specific tasks whose change score most closely represent the change in the total test. The delineation of those tasks depends on the time poststroke from enrollment to CIMT and the inclusion of selected covariates.

As is evident by this analysis, the relationship can be examined in several ways based on time poststroke and important covariates (functional level, gender, and concordance). When considering the time from stroke, the WMFT can be streamlined for patients in the immediate group to 6 tasks (hand to table [front], hand to box [front], reach and retrieve, lift can, lift pencil, and fold towel), which showed a significant relationship between change score and the log mean change score for the remainder of the WMFT (Table 4). For the delayed group, the WMFT can be reduced to 6 tasks (extend elbow weight, hand to box [front], lift can, lift pencil, turn key in lock, and fold towel), which showed a significant relationship between change score and the log mean change score for the remainder of the WMFT (Table 4). Taking both samples into account, the tasks that are most valuable are (in order of level of R²) lift can, lift pencil, fold towel, and hand to box. These 4 tasks are the best tasks for the WMFT core evaluation.

Our results now allow grouping of WMFT tasks by movement and functional characteristics, but this scheme has yet to be validated. The determination for why these 4 tasks among all participants were most sensitive to the total WMFT score is unclear. One possible explanation is that movement components related to other tasks are not performed or practiced as commonly. For example, turning a key, flipping cards, stacking checkers, and folding a towel are all complex object manipulation tasks; however, only turning a key in a lock (for the delayed group) and folding a towel (for both delayed and immediate groups) show a significant relationship with the overall WMFT change score. Grasping a smaller object, such as when turning a key, or a larger object, such as when folding a towel, may occur far more often and have been practiced with greater frequency than the movement components associated with flipping cards or stacking checkers. This is because the former are commonplace objects used with daily living. Additionally, the size of the object may affect the task feasibility. Larger objects, such as a can or a pencil, may require a less dexterous grasp pattern post-stroke. Thus, they may improve earlier in the rehabilitation period and have a significant relationship with the overall WMFT change score. On the other hand, grasping a paper clip, which does not significantly influence the overall WMFT change score, requires fine motor control since the object is smaller and lies flatter on the surface of the table. Similarly, stacking checkers, which also does not significantly influence the overall WMFT change score, requires goal-oriented finger motion and grip of a small object.²⁰

The 2 items significant for the immediate but not delayed groups (hand to table, reach and retrieve) are both gross motor tasks involving predominantly simple shoulder and elbow flexion, respectively. The 2 items significant for the delayed but not immediate groups (turn key in lock, extend elbow weight) involve either fine motor coordination or recruitment of elbow extensor musculature against a load, respectively. While the residual functional capability of participants in both groups is comparable, those individuals in the immediate group undergoing specific task training are well positioned to improve on simpler tasks involving isolated motions in synergy, while delayed group participants who have had more time to accommodate to their impairments may be better prepared to show additional improvements in resistive elbow extension movement and a complex task requiring interjoint coordination (turn key in lock).

The ability to manipulate objects in a safe and timely manner also depends on the kinetics of the movement. Hermsdorfer’s group^21,22 has previously demonstrated that an excessively high safety margin, or the difference between the actual grip force and the minimum force necessary to prevent the slippage of the object, is associated with poststroke object manipulation and indicates possible early fatigue. It also may indicate that there is a central sensorimotor lesion component and associated impairment of temporal coupling between grip and load force, with both hampering the rehabilitation of impaired fine motor skills. While the kinetics of movement was not specifically examined in this study, one would hope that improved kinetics, and thus improved object manipulation, would have resulted from CIMT. Alberts et al²³ demonstrated an improved grip pattern, with force and torque generations, when turning a key, which resembled that of able-bodied individuals after CIMT, providing us with a reason to believe that the kinetics underlying manipulation were, in fact, improved after CIMT.

Also, when considering the kinetics of 3 pinch methods (lateral, downward, and key, which is between the thumb and the side of the index finger) for gripping and lifting a small object, the key pinch necessitates the greatest force with the highest safety margin in those with acute stroke while the downward pinch necessitates the least force and lowest safety margin.²⁴ Thus, adults with impaired motor control because of hemiplegia often use the downward pinch. The higher safety margins for the lateral and key pinches validates 2 of the 6 tasks, lift can (lateral pinch) and turn key in lock (key pinch), which showed a significant change score when compared to the overall WMFT in this study, since their improvement involves a greater rehabilitation in grip force and thus a greater impact on total WMFT score.

The extent to which the hand to box (front), lift can, lift pencil, and fold towel tasks are representative for capturing the overall change in the WMFT seems to be pervasive, showing significant relationships for both immediate and delayed participants. Placing the hand of the affected extremity on a box through timing shoulder flexion in a sagittal plane is the most fundamental of all WMFT tasks, but involves motion over a greater arc than hand to table. Thus, the ability of participants to demonstrate substantial changes in this task is not surprising. Also, according to the ongoing study of classification of WMFT tasks,²⁵ lifting a can and folding a towel are classified as reach and grasp movements involving visual guidance, shoulder flexion and abduction, elbow extension, controlled wrist pronation and supination, and nonspecific grasping of an object with the hand. Lifting a pencil is classified as object manipulation involving visual guidance, maintained grasp, and shoulder, elbow, forearm, wrist, and distal hand movement. Given the multiple muscles and joints involved, necessity of aim (visual guidance) and the frequency with which these tasks are commonly practiced in daily life, their sensitivity to change is not unusual. As previously discussed, excessively high safety margins with grip forces associated with poststroke object manipulation indicate possible early fatigue and a central component that may hamper rehabilitation of fine motor skills.²¹ When considering 3 pinch methods (lateral, downward, key), the lateral pinch necessitates an intermediate force and safety margin and is used for lifting a can or pencil.²⁴ Thus, improving in the lift can and lift pencil task would be feasible but also challenging enough to represent greater improvement with the overall WMFT change score. On the other hand the fold towel task involves the most challenging pinch methods and use of the unimpaired UE, so its sensitivity to change would also represent greater improvement because of its difficulty and its coordinated use of the other limb to fulfill a goal-oriented task. Furthermore, a recent study demonstrated that rehabilitation of the affected UE is mainly determined by improvement of the affected hand, followed by synergistic independent movement of the affected arm. Lifting a can may best represent this motor sequence. Thus, UE improvement represented by the overall WMFT may best be indicated by the lift can taks.²⁶

Unusual Observations

Among 5 070 task efforts recorded in 169 participants, 490 tasks could not be completed within 120 seconds. Only 54 participants were able to complete all 15 tasks. This observation illustrates the fact that many participants with mild to moderate stroke are not able to complete every task in less than 120 seconds, but represents a low percentage of total tasks that were not completed (Table 3). Furthermore, we could not consider excluding every participant who could not complete all tasks, as this restriction would have shrunk our sample size to a number too small for analyses.

The outliers were excluded if their inclusion would have favorably contributed to the significant association between the change score in a task and the change score for the remainder of the WMFT. As is evident by the persistence of significance for most of the tasks once outliers were considered, the outlier data did not contribute greatly to overall skewness in the data. There were very few outlier data points, and there was no relationship between individuals who contributed aberrant data and the change score.

Given the small number of outliers and low percentage of tasks that could not be completed, these data did not influence the analyses profoundly. On the other hand, they are indicative of the fact that while participants were considered to have sustained mild to moderate strokes, clearly there are WMFT tasks that could not be completed and hence indices of impairment.

Limitations

As suggested by Dobkin,²⁷ the infrequent acquisition of outcome measures throughout the intervention was a limitation in the EXCITE trial, thereby compromising the prospects for identifying a true dose-response relationship. However, increasing the frequency of recording outcome measurements could result in repeated task practice altering the WMFT results, rather than the WMFT truly reflecting improvements predominantly attributable to CIMT training.²⁸ This retrospective analysis evaluated only those participants (169/222) who completed the full 10 days of training and from whom posttreatment log mean WMFT scores were acquired. As a result, when analyses are undertaken by examining specific covariates, sample sizes become somewhat smaller. Furthermore, an examination of tasks that could be completed following CIMT therapy, but could not be completed prior to therapy, indicates that substantial improvement in the most distal tasks (most notably in stacking checkers, lifting a paper clip, and lifting a basket) does impact our ability to determine the extent to which these tasks might have influenced the overall change score. Another limitation results from the very specific inclusion rules used in the EXCITE trial that may represent 25% to 33% of potentially eligible stroke survivors. Thus, the restricted range of participants may limit the generalization of our conclusions.

Future Research

If indeed the size of the target object affects task performance, future studies could compare the extent to which novel tasks, or those that include grasp and release of smaller, less easily manipulated objects, influence improvement in function. Furthermore, because analyses using select covariates resulted in smaller sample sizes, replicating this study with a larger sample than studied in the EXCITE trial could enhance the interpretation and value of these data. Another possibility for future studies is to reassess the reliability of this method and to validate this streamlined WMFT against another standard, such as the Fugl Meyer UE Motor Assessment. Another factor illustrated by this study is the need to evaluate those tasks that cannot be completed and the impact of recovery on the ability to complete tasks that could not be previously completed.

In conclusion, our results suggest that administration of the WMFT can be reduced to the specific timed tasks whose change score most closely represents the change in the total test. This core WMFT subset includes the hand to box (front), lift can, lift pencil, and fold towel tasks, which are the tasks significantly related to overall WMFT changes in both sub-samples. If additional tasks are considered, then the turn key in lock and extend elbow weight tasks can be added. Finally, the flip cards task may be considered since the influence of this task on total WMFT change score is close to significant in both subsamples.