Validity of Movement Pattern Kinematics as Measures of Arm Motor Impairment Poststroke
Background and Purpose—Upper limb motor impairment poststroke is commonly evaluated using clinical outcome measures such as the Fugl-Meyer Assessment. However, most clinical measures provide little information about motor patterns and compensations (eg, trunk displacement) used for task performance. Such information is obtained using movement quality kinematic variables (joint ranges, trunk displacement). Evaluation of movement quality may also help distinguish between levels of motor impairment severity in individuals poststroke. Our objective was to estimate concurrent and discriminant validity of movement quality kinematic variables for pointing and reach-to-grasp tasks.
Methods—A retrospective study of kinematic data (sagittal trunk displacement, shoulder flexion, shoulder horizontal adduction, elbow extension) and Fugl-Meyer Assessment scores from 86 subjects (subacute to chronic stroke) performing pointing and reaching tasks was done. Multiple and logistic regression analyses were used to estimate concurrent and discriminant validity respectively. Cutoff points for distinguishing between levels of upper limb motor impairment severity (mild, moderate to severe) were estimated using sensitivity/specificity decision plots. The criterion measure used was the Fugl-Meyer Assessment (upper limb section).
Results—The majority of variance in Fugl-Meyer Assessment scores was explained by a combination of trunk displacement and shoulder flexion (51%) for the pointing task and by trunk displacement alone (52%) for the reach-to-grasp task. Trunk displacement was the only variable that distinguished between levels of motor impairment severity. Cutoff points were 4.8 cm for pointing and 10.2 cm for reach-to-grasp movements.
Conclusion—Movement quality kinematic variables are valid measures of arm motor impairment levels poststroke. Their use in regular clinical practice and research is justified.
Descriptions of motor patterns used for task performance can help better quantify impairment levels in individuals poststroke. Kinematic variables provide detailed measures at 2 levels: motor performance (end point error, velocity, etc) and movement quality (joint ranges, trunk movement). Both levels of movement description are increasingly being used as outcomes in studies involving upper limb (UL) movement re-education in individuals poststroke.1,2 Kinematic measurement has revealed differences in arm movement patterns during reaching tasks between (1) healthy control subjects (dominant arm) and individuals with stroke (more-affected arm)3; and (2) individuals with stroke (more-affected versus less-affected arms).4 For example, motor performance measures revealed that tapping with a stylus, requiring wrist flexion/extension, and 3-dimensional pointing movements were less precise and slower in the hemiparetic arm compared with healthy control subjects despite clinically assessed mild levels of arm motor impairment (strength ≥4/5 on the Medical Research Council Scale5 or ≥50/66 on UL Fugl-Meyer Assessment (FMA).6 This suggests that motor performance variables may better identify motor control deficits than clinical outcome measures.
In addition to motor performance measures, movement quality measures also identify and quantify motor compensations used for task accomplishment. During UL reaching tasks, children and adults with hemiparesis use excessive trunk displacement, compared with control subjects, to assist arm end point displacement when they have a restricted range of voluntary elbow extension and/or disrupted elbow/shoulder interjoint coordination.3,7 Thus, measurement of only motor performance variables may provide an incomplete assessment of motor abilities of individuals poststroke.
Test–retest reliability and minimal detectable change of 3-dimensional kinematic variables of a midline reaching task have recently been investigated in 13 individuals with chronic stroke.8 Only 2 trials were recorded for each reaching task. Motor performance (end point error, peak velocity, movement time, reach extent) and movement quality variables (ranges of shoulder flexion, shoulder abduction, elbow extension, and elbow–shoulder crosscorrelation) had moderate to excellent reliability (intraclass correlation coefficients ≥0.6). Minimal detectable change values differed with the lowest variability for reach-path ratio (7.9% to 20.9%) and elbow–shoulder crosscorrelation (10.2% to 18.2%) and highest (approximately 44% to 99%) for temporal measures.
Previous studies have not investigated concurrent validity of kinematic measures and their ability to detect UL impairment severity levels, including compensatory movement patterns in individuals with stroke. Our goal was to estimate the concurrent and discriminant validity of movement quality kinematic measures for UL pointing and reach-to-grasp tasks. The UL part of the FMA, with well-known psychometric properties,9,10 was used as the criterion measure. Arguably the gold standard for UL impairment measurement,11 it is widely used as the criterion measure to estimate concurrent validity of other outcome measures.12,13
Kinematic data were available from subjects with subacute to chronic poststroke hemiparesis from 4 previous studies conducted by our group between 2001 and 2008 of a pointing task—Study 1 (Cirstea and Levin)14 or Study 2 (Subramanian et al)15—or a reach-to-grasp task—Study 3 (Michaelsen et al)16 or Study 4 (Magdalon et al).17 The 4 study samples ranged from 12 to 28 subjects recruited according to common inclusion criteria: (1) first ischemic/hemorrhagic stroke; (2) >3 months poststroke; (3) score of ≥2 of 7 on the Arm and Hand section of the Chedoke-McMaster Stroke Assessment12; and (4) ability to communicate in French and/or English. Subjects were excluded if they had an (1) occipital, brain stem, or cerebellar lesion; (2) other neurological or orthopedic conditions; (3) major cognitive deficits assessed by standard tests; (4) sitting balance or trunk stability deficits; and (5) problems with wearing a head-mounted display (for Studies 2 and 4). All participants signed informed consent forms approved by the Centre for Interdisciplinary Research in Rehabilitation of Greater Montreal.
Study designs were cross-sectional (Study 4) or randomized clinical trials (Study 1 to 3). Although the 3 randomized clinical trials included specific practice regimens, only movement quality measures and UL FMA scores from the initial prepractice assessment were considered for analysis.
In Studies 1 and 2, pointing tasks consisted of repetitive pointing motions with the more-affected arm to a shoulder-height target placed at the functional arm’s length (ie, within reach; Figure 1A). Because initial arm positions varied between studies, only angular values at the end of movement (end position angles) were used. Subjects were instructed to “point to the target quickly and precisely in 1 smooth movement.” The subjects performed 12 to 25 movements with 10-second breaks between trials. For the grasping tasks (Studies 3 and 4; Figure 1B), participants reached and grasped a cylinder (diameter 35 mm, height 95 mm) using a whole-hand grasp at a self-paced speed for 10 trials. The cylinder was placed at xiphoid process height at a distance of 80% of the arm length. For all studies, a sound signaled movement start.
Kinematic Data Acquisition
Marker placement (infrared emitting diodes [IREDs]) had common anatomic locations as follows: index fingertip, distal ulnar head (wrist), lateral epicondyle (elbow), ipsilateral and contralateral acromion processes (shoulders), and sternal angle. Kinematic data were recorded for 2 to 5 seconds at 100 to 120 Hz with a 3-dimensional optical tracking system (Optotrak; Northern Digital, Waterloo, Canada). Kinematic data collected included trunk displacement, shoulder flexion, shoulder horizontal adduction, and elbow extension. Arm and trunk displacement were measured as the sagittal distance (in millimeters) moved by the sternal or fingertip marker between movement beginning and end. These were defined as times at which the fingertip/sternal marker tangential velocity rose above or fell below 10% of the peak tangential velocity, respectively.
Shoulder flexion was defined as the angle between vectors formed by elbow and ipsilateral shoulder IREDs and the vertical through the ipsilateral shoulder (arm alongside body=0°). Shoulder horizontal adduction was measured as the angle between 2 vectors defined by ipsilateral shoulder–elbow IREDs and contralateral–ipsilateral shoulder IREDs projected horizontally, where the full abduction in line with the shoulders was 0°. Elbow extension was measured as the angle between vectors formed by wrist and elbow IREDs and elbow and ipsilateral shoulder IREDs (full elbow extension=180°). Angular data were low-pass filtered at 20 Hz. Definitions of beginning and end for angle measurements were based on end point peak velocity defined previously.
Preliminary analysis involved grouping data into 2 sets according to the task (pointing, reach to grasp) and ensuring that requirements of linearity, normality, and homogeneity were met. The strength of correlation among predictors (trunk displacement, shoulder flexion, shoulder horizontal adduction, elbow extension) and with the UL FMA scores was estimated with Pearson correlations. Wrist movements were not included in the analysis. Correlations were defined as strong (≥0.7), moderate (0.4 to 0.69), or mild (≤0.39).18 Strong correlation between any 2 predictors indicates measurement of the same construct, which does not allow the amount of variance predicted to be attributed uniquely to 1 variable.18 Thus, in cases of high correlations between predictors, 1 predictor is substituted for another.
Concurrent validity of kinematic measures (predictors) was estimated against FMA scores (dependent variable). To determine which predictor or combination thereof (trunk displacement, shoulder flexion, shoulder horizontal adduction, or elbow extension) explained the greatest amount of FMA score variance, the best fit model was obtained using multiple linear regression analysis.19 Adjusted r2 values were used because they consider multicollinearity. The most parsimonious model sought using various combinations of predictors was verified using robust regression procedure.
Discriminant validity of kinematic measures (predictors) was estimated against FMA scores (dependent variable) using logistic regression analysis. The cutoff score of 50/66 on the FMA20 was used to discriminate between subjects with mild (Group 0) or moderate-to-severe (Group 1) hemiparesis. Logistic regression estimated which predictor(s) contributed most to the probability of a subject moving between groups. Using receiver operating characteristic curves21 we calculated area under the curve and sensitivity and specificity values for predictor(s). Sensitivity/specificity decision plots were constructed to estimate cutoff values for transition between levels of severity for each predictor. Data were analyzed using SPSS (Version 17) and SAS (Version 9.3.1) with significance levels of α<0.05.
Data from 44 and 42 participants, respectively, were available for analysis of pointing and reach-to-grasp tasks (Table 1). Both data sets met assumptions of linearity, normality, and homogeneity.
Pointing Data Set
Correlations Between Predictors
All correlations were significant (P<0.05) except for shoulder horizontal adduction and trunk displacement. Correlations between FMA and individual predictors ranged from low to moderate (Table 2A) and were highly significant (P<0.005), except for those between FMA and shoulder horizontal adduction. Between predictors, correlations were low to moderate (Table 2A) except for shoulder flexion and elbow extension (r=0.71).
Multiple Regression Analyses
The best 4 models explaining the maximum amount of variance in FMA scores are shown in Table 2B along with amounts of variance explained by individual predictors. Trunk movement alone explained 46% of the variance. Two best fit models were identified (Table 2B, bold). One consisted of a combination of trunk displacement, shoulder flexion, and shoulder horizontal adduction, explaining 55% of the variance in FMA (analysis of variance F[3,40]=16.06, P<0.001; adjusted r2=0.51) and the other included trunk displacement and shoulder flexion, explaining 51% of the variance (analysis of variance F[2, 41]=21.32, P<0.001; adjusted r2=0.49). Robust regression revealed that shoulder horizontal adduction did not contribute significantly to the variance. Thus, the final best fit model consisted of trunk displacement and shoulder flexion.
Logistic regression revealed that trunk displacement was the only variable discriminating between mild and moderate-to-severe motor impairment levels (Table 2C). The final logistic regression equation was: Logit P (mild impairment for pointing)=−3.972−0.040 (trunk displacement)+0.024 (shoulder flexion)+0.005 (shoulder horizontal adduction)+0.043 (elbow extension). The odds of a subject moving from Group 0 (mild hemiparesis) to Group 1 (moderate-to-severe hemiparesis) increased by 96% with a 1-unit increase in trunk displacement. The area under the curve was 0.86 (P<0.001; 95% CI, 0.76 to 0.97). The cutoff value for transition between groups estimated by the sensitivity/specificity decision plot was 4.8 cm (Figure 2). Thus, subjects with mild and moderate-to-severe hemiparesis used ≤4.8 cm or >4.8 cm of trunk displacement, respectively, for the pointing task.
Reach-to-Grasp Data Set
Correlations Between Predictors
For the reach-to-grasp task, there were low-to-moderate correlations between FMA and shoulder flexion and shoulder horizontal adduction and elbow extension, whereas FMA and trunk displacement were highly inversely correlated (r=−0.72; Table 3A). Between predictors, shoulder horizontal adduction, trunk displacement, and elbow extension were moderately correlated, whereas the other predictors were highly correlated. All correlations were highly significant (P<0.005).
Multiple Regression Analyses
Trunk movement alone explained 52% of the variance in FMA scores (analysis of variance F[1,40]=42.91, P<0.001; adjusted r2=0.51) and accounted for 1 best fit model (Table 3B, bold). The other best fit model consisted of a combination of trunk displacement and shoulder horizontal adduction, which also explained 52% of the variance (analysis of variance F[2,39]=21.02; P<0.001; adjusted r2=0.49). However, robust regression revealed that shoulder horizontal adduction did not significantly contribute to the model. Thus, the best model was one involving only the trunk.
Logistic Regression Analysis
The results for the reach-to-grasp data set were similar to the pointing data set. Trunk displacement was again the only variable able to discriminate between levels of impairment (Table 3C). The final logistic regression equation was: Logit P (mild level of severity for reach to grasp)=6.093−0.049 (trunk displacement)+0.081 (shoulder flexion)−0.053 (shoulder horizontal adduction)−0.018 (elbow extension). Thus, the odds of a subject moving from Group 0 to Group 1 were increased by 95% with a 1-unit increase in trunk displacement. The area under the curve was 0.95 (P<0.001; 95% CI, 0.00 to 1.00) and the cutoff point was estimated at 10.2 cm (sensitivity/specificity decision plot; Figure 3). Thus, for the reach-to-grasp task, subjects with mild and moderate-to-severe hemiparesis used ≤10.2 cm or >10.2 cm of trunk displacement, respectively.
Concurrent and discriminant validity of movement quality kinematic variables as measures of UL impairment for pointing and reach-to-grasp tasks was investigated. We found that variability in FMA scores across subjects and data sets was explained by different combinations of kinematic variables for each task. A combination of trunk displacement and shoulder flexion best explained the majority of the variance (51%) in FMA scores for pointing with 46% attributed to trunk movement alone. For the reach-to-grasp task, trunk displacement was the most significant contributor to the variance (52%). Addition of other predictors in the model did not significantly change the amount of variance explained. Indeed, there were high autocorrelations between many of the predictors (Table 3B), which may be due to coupling between adjacent joints.22
Trunk displacement was the only variable that discriminated between mild and moderate-to-severe UL impairment for both tasks. A 1-unit increase in trunk displacement was associated with a 96% (pointing) or 95% (reach-to-grasp) increase in the odds of having a lower score on the FMA (greater impairment level). The sensitivity/specificity decision curve analysis indicated that cutoff points of 4.8 cm (pointing) and 10.2 cm (reach to grasp) for trunk displacement discriminated between different UL impairment severity levels. The cutoff point for pointing is consistent with previously reported values of trunk displacement in individuals poststroke with mild (5.1 cm) and moderate-to-severe (10.3 to 13.9 cm) UL impairment performing comparable reaching tasks.4 Additionally, the cutoff point for the reach-to-grasp task is similar to that characterizing this task in individuals with mild UL impairment (10.2 cm) compared with 1.7 to 2.5 cm in healthy age-matched subjects.3,23 This indicates that subjects with clinically estimated mild arm motor impairment could still use as much as 33% (pointing) to 300% (reach to grasp) more trunk displacement compared with age-matched control subjects. Thus, improvement in FMA scores may not necessarily indicate that test items are accomplished without using compensatory movement patterns. Our results are also consistent with findings that trunk displacement is used to assist arm extension during reaching and hand orientation during grasping.3,24
Our results suggest that movement quality variables are more sensitive in identifying UL deficits, even in well-recovered patients, as compared with clinical scales. Such information may be used to complement clinical assessment. Our results also demonstrate that movement quality kinematic variables are valid measures of arm motor impairment levels and can be used to distinguish between arm motor recovery and compensation.
Measuring movement kinematics in the clinical setting is difficult given the cost and specialization of recording and analysis technology. Alternatively, objective quantification of movement quality during task performance may be better estimated using multiple clinical outcomes25 including measures at 2 levels of the International Classification of Functioning.26 Thus, outcome measures can include clinical scales that focus on direct observation of movement patterns (eg, Reaching Performance Scale for Stroke27; International Classification of Functioning level: impairment) to be combined with those of motor function (eg, Wolf Motor Function Scale13; International Classification of Functioning level: activity).
Our kinematic assessments included only 2 UL joint movements, namely shoulder and elbow motion. Actions scored with FMA also include wrist movements, grasp types, and coordination. Inclusion of kinematic representations of these additional variables in the model may potentially alter the percentage of variance explained by the predictors. In addition, we only considered end position static angles. It remains to be determined if relationships between FMA and predictors would be altered when considering dynamic movement patterns.
We gratefully acknowledge the participants in the 4 studies and the Canadian Institutes of Health Research (CIHR), Canadian Foundation for Innovation, Heart and Stroke Foundation of Canada (HSFC), Physiotherapy Foundation of Canada, and CAPES-Brazil. M.F.L. holds a Tier 1 Canada Research Chair in Motor Control and Rehabilitation. S.K.S. holds a Focus on Stroke Doctoral Research Fellowship awarded by CIHR, HSFC, and the Canadian Stroke Network.
- Received June 11, 2010.
- Accepted July 30, 2010.
Cirstea MC, Ptito A, Levin MF. Feedback and cognition in arm motor skill reacquisition after stroke. Stroke. 2006; 37: 1237–1242.
Levin MF. Interjoint coordination during pointing movements is disrupted in spastic hemiparesis. Brain. 1996; 119: 281–293.
Platz T, Prass K, Denzler P, Bock S, Mauritz KH. Testing a motor performance series and a kinematic motion analysis as measures of performance in high-functioning stroke patients: reliability, validity, and responsiveness to therapeutic intervention. Arch Phys Med Rehabil. 1999; 80: 270–277.
Wagner JM, Rhodes JA, Patten C. Reproducibility and minimal detectable change of three-dimensional kinematic analysis of reaching tasks in people with hemiparesis after stroke. Phys Ther. 2008; 88: 652–663.
Duncan PW, Propst M, Nelson SG. Reliability of the Fugl-Meyer assessment of sensorimotor recovery following cerebrovascular accident. Phys Ther. 1983; 63: 1606–1610.
Gladstone DJ, Danells CJ, Black SE. The Fugl-Meyer assessment of motor recovery after stroke: a critical review of its measurement properties. Neurorehabil Neural Repair. 2002; 16: 232–240.
Gowland C, Stratford P, Ward M, Moreland J, Torresin W, Van HS, Sanford J, Barreca S, Vanspall B, Plews N. Measuring physical impairment and disability with the Chedoke-McMaster Stroke Assessment. Stroke. 1993; 24: 58–63.
Wolf SL, Catlin PA, Ellis M, Archer AL, Morgan B, Piacentino A. Assessing Wolf motor function test as outcome measure for research in patients after stroke. Stroke. 2001; 32: 1635–1639.
Cirstea MC, Levin MF. Improvement of arm movement patterns and endpoint control depends on type of feedback during practice in stroke survivors. Neurorehabil Neural Repair. 2007; 21: 398–411.
Subramanian S, Ptito A, Levin MF. Arm Motor Recovery in Chronic Stroke Survivors—Effects of Enhanced Feedback in Two Training Environments [Abstract]. Program number 553.3. Washington, DC: Society for Neuroscience; 2008.
Michaelsen SM, Dannenbaum R, Levin MF. Task-specific training with trunk restraint on arm recovery in stroke: randomized control trial. Stroke. 2006; 37: 186–192.
Magdalon EC, Levin MF, Quevedo AAF, Michaelsen SM. Kinematics of reaching and grasping in a 3D immersive virtual reality environment in patients with hemiparesis [Abstract]. Neurorehabil Neural Repair. 2008; 22: 601.
Streiner DL, Norman GR. Health Measurement Scales: A Practical Guide to Their Development and Use, IV ed. New York: Oxford University Press; 2008.
Cohen J, Cohen P. Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences, II ed. Hillsdale, NJ: Lawrence Erlbaum Associates; 1983.
Duncan PW, Goldstein LB, Horner RD, Landsman PB, Samsa GP, Matchar DB. Similar motor recovery of upper and lower extremities after stroke. Stroke. 1994; 25: 1181–1188.
Dewald JP, Pope PS, Given JD, Buchanan TS, Rymer WZ. Abnormal muscle coactivation patterns during isometric torque generation at the elbow and shoulder in hemiparetic subjects. Brain. 1995; 118: 495–510.
Michaelsen SM, Luta A, Roby-Brami A, Levin MF. Effect of trunk restraint on the recovery of reaching movements in hemiparetic patients. Stroke. 2001; 32: 1875–1883.
Michaelsen SM, Jacobs S, Roby-Brami A, Levin MF. Compensation for distal impairments of grasping in adults with hemiparesis. Exp Brain Res. 2004; 15: 162–173.
World Health Organization. ICF: International Classification of Functioning, Disability and Health. Geneva: World Health Organization; 2001.
Levin MF, Desrosiers J, Beauchemin D, Bergeron N, Rochette A. Development and validation of a scale for rating motor compensations used for reaching in patients with hemiparesis: the reaching performance scale. Phys Ther. 2004; 84: 8–22.