If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Validity of Accelerometry for Monitoring Real-World Arm Activity in Patients With Subacute Stroke: Evidence From the Extremity Constraint-Induced Therapy Evaluation Trial
Uswatte G, Giuliani C, Winstein C, Zeringue A, Hobbs L, Wolf SL. Validity of accelerometry for monitoring real-world arm activity in patients with subacute stroke: evidence from the Extremity Constraint-Induced Therapy Evaluation trial.
Objective
To examine the psychometric properties of an objective method for assessing real-world arm activity in a large sample with subacute stroke.
Design
Validation study.
Setting
Community.
Participants
Persons 3 to 9 months poststroke (N=169) with mild to moderate motor impairment of their hemiparetic arm enrolled in a multisite, randomized clinical trial of constraint-induced movement therapy.
Interventions
Not applicable.
Main Outcome Measures
Participants wore an accelerometer on each arm outside the laboratory for 3 days before and after treatment or an equivalent no-treatment period. They also completed the Actual Amount of Use Test (AAUT), which is an observational measure of spontaneous more-impaired arm use, and the Motor Activity Log (MAL), which is an interview assessing more-impaired arm use in daily life.
Results
Low-pass–filtered accelerometer recordings were reliable (r range, >.8) and stable (P range, >.48). Their validity was also supported. Correlations calculated across all participants at baseline between the ratio of more-impaired to less-impaired arm accelerometer recordings and AAUT and MAL scores were .60 and .52, respectively.
Conclusions
Accelerometry provides an objective, real-world index of more-impaired arm activity with good psychometric properties.
). Clinicians attempting to broaden assessment of the effects of upper-extremity rehabilitation interventions have used self-report or observational measures of functional independence in clinical (eg, FIM instrument
) environments. Although these approaches yield valuable information, they do not directly capture the effects of rehabilitation interventions on function of the impaired arm (ie, more-impaired arm) in daily life.
Measures of functional independence in the home are not adequate for this purpose because variations in scores on such instruments could be because of differences in impaired-arm use or compensatory strategies.
Directly measuring real-world impaired-arm activity is important because under certain conditions persons with stroke exhibit discrepancies between their motor capacity, as measured by laboratory performance tests, and use of that capacity in daily life.
it predicts that stroke survivors who have aversive experiences (eg, failure, excessive effort, or fatigue) when they attempt to use the impaired arm soon after stroke will learn to avoid using that extremity.
Therefore, several months later, even after some stroke survivors recover the capacity for coordinated movement of the impaired arm, they continue not to use it.
Accelerometry, by permitting unobtrusive monitoring of movement in a variety of environments, provides a promising tool for obtaining an objective index of arm activity in daily life. Uswatte et al
showed that a portable, wireless system of 4 accelerometers accurately measures the duration of spontaneous arm, torso, and ambulatory movements outside the laboratory in patients more than 1 year after stroke (n=9). Subsequently, Uswatte et al
showed that the duration of impaired-arm movement divided by the duration of unimpaired-arm movement is a reliable and valid real-world measure of upper-extremity treatment outcome. In Uswatte,
have developed an alternative method for monitoring arm movement. Their method uses a pressure transducer that measures vertical displacement of the wrist from the shoulder. It was validated in a study of 10 able-bodied persons and 10 persons with chronic stroke (ie, >1y postinjury).
Although their method has potential, problems may emerge with patient adherence. The device that measures vertical displacement extends the entire length of the arm, is taped to it at 3 points, and needs to be connected with a wire to a data logger worn by the patient.
This article examines the clinometric properties of an upper-extremity activity monitoring system that only requires patients to wear a single wireless accelerometer a little larger than a men’s watch above each wrist. The construct validity of this simple system is evaluated in a large sample (N=169) of research participants with subacute stroke (3–9mo after their first clinical stroke). We test whether the ratio of impaired-arm to unimpaired-arm accelerometer recordings converges with measures of real-world, impaired-arm use and diverges from a measure of overall physical activity.
Methods
Participants
Participants were subacute stroke patients with mild to moderate upper-extremity hemiparesis enrolled in the Extremity Constraint-Induced Therapy Evaluation (EXCITE) trial.
Methods for a multisite randomized trial to investigate the effect of constraint-induced movement therapy in improving upper extremity function among adults recovering from a cerebrovascular stroke.
Principal inclusion criteria were as follows: (1) 10° or more active wrist extension, (2) 10° or more active metacarpophalangeal (MP) and interphalangeal (IP) extension of 2 fingers of the impaired hand, (3) 10° or more active MP and IP abduction and extension of the thumb on the same hand, (4) ability to transfer to and from the toilet independently and safely, (5) ability to maintain standing for 2 minutes, and (6) no major cognitive deficit (Mini-Mental Status Examinations score ≥24
Methods for a multisite randomized trial to investigate the effect of constraint-induced movement therapy in improving upper extremity function among adults recovering from a cerebrovascular stroke.
Methods for a multisite randomized trial to investigate the effect of constraint-induced movement therapy in improving upper extremity function among adults recovering from a cerebrovascular stroke.
Accelerometer data were available at baseline from 76% of the 222 participants enrolled in EXCITE. Table 1 summarizes the demographic, stroke-related, mobility, and impaired arm characteristics of these subjects.
Table 1Demographic, Stroke-Related, Mobility, and Impaired-Arm Characteristics of Study Participants
Characteristic
Treatment Group (n=82)
Control Group (n=87)
All Participants (N=169)
Demographic
Mean age ± SD (y)
63.0±12.8
64.2±12.7
63.6±12.8
Mean education ± SD (y)
14.5±3.8
14.2±2.9
14.4±3.3
Women (n)
31
33
64
Race (n)
European American
56
68
124
African American
19
13
32
Other
7
6
13
Stroke related (n)
Paresis of right side
36
43
79
Concordance of paretic and dominant side
38
49
87
Type of stroke
Ischemic
74
74
148
Hemorrhagic
8
13
21
Real-world mobility
Mean Stroke Impact scale mobility scale points ± SD
Duration of impaired-arm movement was calculated by dividing the number of epochs in the accelerometer record with above-threshold values by the total number of epochs.
21.7±9.4
22.5±11.2
22.1±10.3
Mean ratio of impaired-to-unimpaired arm movement ± SD
Based on active range of motion of the hemiparetic hand, participants were divided into higher- (20° wrist extension, 10° MP and IP joint extension at all digits) and lower- (10° wrist extension, 10° MP and IP joint extension at 2 digits and thumb) functioning subgroups.9
Higher
64
68
132
Lower
18
19
37
NOTE. There were no significant differences between treatment and control participants.
Abbreviation: SD, standard deviation.
Duration of impaired-arm movement was calculated by dividing the number of epochs in the accelerometer record with above-threshold values by the total number of epochs.
† Based on active range of motion of the hemiparetic hand, participants were divided into higher- (20° wrist extension, 10° MP and IP joint extension at all digits) and lower- (10° wrist extension, 10° MP and IP joint extension at 2 digits and thumb) functioning subgroups.
Methods for a multisite randomized trial to investigate the effect of constraint-induced movement therapy in improving upper extremity function among adults recovering from a cerebrovascular stroke.
Manufacturing Technologies Inc, 709 Anchors St, Fort Walton Beach, FL 32548.
each about the size of a large wristwatch, were placed in snug pouches sewn onto cloth and elastic bands, and 1 unit was strapped on just above each wrist. In this study, pouches were color coded to help participants remember which unit belonged on which arm. Figure 1 shows an experimenter modeling the accelerometers. Accelerometers were worn on both arms because a single unit worn on the impaired side might act as a cue to use that extremity and thereby confound the measurement of the effects of rehabilitation on arm function.
Additionally, a previous experiment suggested that the ratio of impaired- to unimpaired-arm recordings controls adequately for variations in overall levels of physical activity (eg, ambulation).
Such variations affect recordings from the impaired arm alone because the impaired arm, of course, moves whenever there is movement of the torso. The insensitivity of the ratio of impaired- to unimpaired-arm recordings to changes in overall levels of physical activity is desirable because this ratio is proposed as an outcome measure for interventions designed to improve real-world use of the impaired arm.
Fig 1Accelerometer configuration. The accelerometersa were placed in snug pouches sewn onto cloth and elastic bands, and 1 unit was strapped on just above each wrist. With this placement, the accelerometers were sensitive to vertical displacement of the arm and arm movement parallel to the forearm.
The accelerometers contain a single piezoelectric crystal mounted in such a way that the units are sensitive to movement in 2 axes (see fig 1). The charge produced by the piezoelectric crystal when it is subject to acceleration is sampled at 10Hz and integrated over a user-specified epoch. The integrated value is called a raw count; it represents a rough index of the amount of movement by the object to which the accelerometer is attached. For example, lifting a can from a table to a shelf produces roughly 20 raw counts/s from a wrist-mounted accelerometer. The recording epoch in this study was 2 seconds.
One might expect that an accelerometer sensitive to movement in all 3 axes would be required to characterize upper-extremity activity. However, Redmond and Hegge
have shown that acceleration recorded in any single axis closely approximates that recorded in any 2 other axes when monitoring upper-extremity movement during everyday activities. Redmond and Hegge report that after 1 second of recording arm movement during everyday activity the correlation between any 2 of 3 uniaxial accelerometers mounted in orthogonal orientations on an arm ranges between .60 and .70. After 4 minutes, these correlations range between .95 and .98. (For the accelerometer data reported here, recordings were collected for a minimum of 8 hours.) Recording in a single axis is adequate for everyday activities because daily activities typically contain movement components in all 3 axes.
were used to obtain convergent measures of impaired-arm activity outside the laboratory. The MAL is a structured interview during which respondents rate how well (quality of movement [QOM] scale)
they use their impaired arm for accomplishing upper-extremity activities of daily living (ADLs). Both the QOM and AOU scales are anchored at 6 points (0, never used; 5, same as prestroke). Several reports
Uswatte G, Taub E, Morris D, Light K, Thompson P. The Motor Activity Log–28: assessing daily use of the hemiparetic arm after stroke. Neurology. In press.
suggest that the MAL has high internal consistency, stability, test-retest reliability, and convergent and concurrent validity for measuring real-world arm use. In this study, MAL respondents rated their impaired-arm use on each of 30 ADLs
Uswatte G, Taub E, Morris D, Light K, Thompson P. The Motor Activity Log–28: assessing daily use of the hemiparetic arm after stroke. Neurology. In press.
for 3-day periods. We report QOM scores because analyses suggest that this scale is more reliable than the AOU scale and that it captures components of the amount as well as quality of arm use outside the laboratory.
Uswatte G, Taub E, Morris D, Light K, Thompson P. The Motor Activity Log–28: assessing daily use of the hemiparetic arm after stroke. Neurology. In press.
For example, the correlation between the 2 scales in this study was .92 (P<.001). The AAUT gauged real-world arm use on the basis of spontaneous behavior in the laboratory. Participants were videotaped without their awareness while they were unobtrusively led through a standardized scenario of 17 tasks; objects to be manipulated were presented in midline. (Participants gave consent to be videotaped before testing.) Trained, masked observers evaluated the quality (QOM) and amount (AOU) of impaired-arm use from videotape with a 6-point scale (0, attempts task but does not use more-impaired arm; 5, more-impaired arm is used for task and movement appears normal). Analyses suggest that this behavioral observation system has adequate test-retest reliability (r=.76, P<.001) and convergent validity with the MAL (r range, >.45; P range, <.05).
It contains 8 subscales, 4 of which assess different aspects of physical function. The other subscales assess cognitive problems, mood disorders, communication difficulties, and social participation, respectively. On the SIS mobility scale, participants rate their ability to move about in their daily environment with a 5-point scale (1, could not do at all; 5, not difficult at all).
Procedure
Participants were asked to wear accelerometers outside the laboratory during all waking hours, except when washing themselves, for two 3-day periods (periods 1 and 2). For CIMT patients (treatment group), period 1 and 2 were immediately before and after treatment, respectively. For participants receiving delayed treatment (control group), period 1 and 2 were before and after an equivalent no-treatment interval. The MAL, AAUT, and SIS were also completed at these times. This testing was conducted in the laboratory by trained, blinded project members. Their compliance with trial procedures was periodically checked by reviewing video of their performance.
Methods for a multisite randomized trial to investigate the effect of constraint-induced movement therapy in improving upper extremity function among adults recovering from a cerebrovascular stroke.
The institutional review board of each university participating in the EXCITE trial approved the study protocol. Each subject signed an informed consent.
Data Reduction
Data were downloaded from the accelerometers and sent to the EXCITE Training Core.
Methods for a multisite randomized trial to investigate the effect of constraint-induced movement therapy in improving upper extremity function among adults recovering from a cerebrovascular stroke.
Training Core staff removed segments from the accelerometer recordings when subjects were thought to have the accelerometers off. Such segments were marked by extended periods of inactivity (ie, any 3-h segment for which the raw accelerometer counts recorded in every epoch were <2).
Training Core staff also corrected recordings from participants who had clearly worn the impaired-arm unit on the unimpaired arm and vice versa for the entire data-collection period. Such recordings were marked by persistently high accelerometry values for the impaired arm relative to the unimpaired arm. They were corrected by simply relabeling the impaired-arm raw data file by using the suffix for an unimpaired-arm file and vice versa. Training Core staff deemed recordings invalid if they were of insufficient length to yield reliable data (3.9%; <8h of data on a single day or 16h of data in total) or had summary variable values that exceeded benchmarks (ie, >3 standard deviations above mean values) from a previous study.
Details regarding these procedures, which were partially automated, are available from the corresponding author. The Training Core was masked to whether the recordings were from treatment or control subjects and scores on other measures.
Methods for a multisite randomized trial to investigate the effect of constraint-induced movement therapy in improving upper extremity function among adults recovering from a cerebrovascular stroke.
transformed the accelerometer recordings by dichotomizing the raw value recorded for each epoch around a low threshold, which was 2. Summary variables for the transformed recordings from each arm (impaired-arm summary variable, unimpaired-arm summary variable) were calculated by dividing the number of epochs with raw values above threshold by the total number of epochs and expressing this value as a percentage. The need for a “threshold filter,” also known as a low-pass filter, and the rationale for selecting a threshold of 2 have been described in a previous study.
showed that “threshold-filtered” recordings provide a more accurate measure of the duration of extremity movement than raw recordings in persons with chronic stroke. The EXCITE Data Management Center also calculated the ratio of impaired-to-unimpaired arm threshold-filtered recordings (ratio summary variable).
Data Analysis
We evaluated the convergent validity of the ratio summary variable by calculating its correlation with 2 real-world measures of arm use: the MAL and AAUT. Discriminant validity of the ratio summary variable was examined by calculating its correlation with a real-world measure of overall physical activity, the SIS mobility scale. We selected the ratio summary variable, rather than the impaired-arm summary variable, as a measure of impaired-arm activity because, as noted, previous findings suggested that the ratio summary variable captures changes in upper-extremity activity specific to rehabilitation of the impaired arm.
Stability of the recordings was evaluated by using paired t tests to detect whether there were significant changes in summary variable values in the control group from period 1 to 2, which were separated by approximately 2 weeks. We evaluated the test-retest reliability of the summary variables by calculating Pearson correlations between period 1 and 2 accelerometer recordings from the control group. A square root transformation was applied to the AAUT arm use score because its distribution was positively skewed. Square root transformations attenuate positive skew by reducing the distance between large values (eg, scores of 100, 81, and 64 are transformed to 10, 9, and 8, respectively).
We described the strength of the correlations observed by using conventions from the meta-analysis literature, which hold that 0.1, 0.3, and 0.5 are weak, moderate, and strong associations, respectively.
Reliability and Validity of Accelerometry for Measuring Real-World Impaired-Arm Activity
The accelerometry summary variables were reliable and stable. Correlations between period 1 and 2 values from control participants for the unimpaired-arm, impaired-arm, and ratio summary variables were .81, .87, and .90, respectively. Changes in the summary variable values were not significant from period 1 to 2 (P range, >.48). Additionally, there were no significant differences in summary variable values between treatment and control participants at period 1 (P range, >.58) (see table 1).
Construct validity of the ratio of impaired-to-unimpaired arm recordings (ratio summary variable) for measuring impaired-arm activity was supported. Table 2 lists correlations between the ratio summary variable and the MAL, AAUT, and SIS mobility scale. Corresponding statistics for the impaired-arm summary variable are also listed. Correlations between the ratio summary variable and real-world measures of impaired-arm use (MAL, AAUT) were strong, whereas the correlation between the ratio summary variable and real-world measure of mobility (SIS mobility scale) was weak. In contrast, the impaired-arm summary variable correlated moderately with the real-world measures of impaired-arm use and mobility, suggesting that this summary variable did not index arm activity associated specifically with use of the impaired arm.
Table 2Pearson Correlations Between the Accelerometry Summary Variables and Measures of Real-World Arm Use and Mobility (N=169)
Participants wore accelerometers, on average, for 37.9 hours during each data-collection period. If based on previous work participants slept for 7 hours a day,
these data suggest that subjects wore the units for approximately 76% of waking hours. The treatment and control groups did not show significant differences in compliance during period 1 nor did control participants show significant differences in compliance between period 1 and 2.
Reasons Accelerometry Data Were Not Available
Twenty-three percent of accelerometry data from period 1 were invalid or missing. Reasons were as follows: (1) errors in initializing accelerometers or downloading and storing recordings (8% of total possible observations), (2) subjects wearing accelerometers for an insufficient time (3%), (3) subjects putting on only 1 unit or putting units on after the programmed recording period (3%), (4) subjects switching limb placement of units for part of the recording period (4%), (5) recordings that did not appear veridical (5%; eg, very large values for extended periods), and (6) recordings that clearly indicated accelerometer malfunction (<1%; eg, many negative values).
Examination of factors that might be associated with missing accelerometry data showed that there were a larger proportion of African Americans (37%) than European American (22%) and other (7%) participants with missing data (χ2 test=6.0, P<.05). The breakdown of reasons for missing data was similar between African-American and European-American participants, except that 10% of African Americans versus 1% of European Americans (χ2 test=6.6, P<.05) wore only 1 unit or made a similar mistake. There were no other notable differences in the demographic, stroke-related, and motor characteristics of EXCITE participants with and without accelerometry data.
Discussion
The results suggest that the ratio of impaired- to unimpaired-arm accelerometer recordings is a valid measure of real-world impaired-arm activity. The ratio summary variable correlated strongly with 2 measures of impaired-arm use and correlated weakly with a measure of overall physical activity. Moreover, the strength of the correlations between the ratio summary variable and measures of impaired-arm use did not depend on the method of assessment (ie, physical vs self-report or observational), increasing confidence that the relations observed were because of the fidelity of the accelerometry measure and not a common method of assessment across measures. When validating 1 self-report measure against another, for example, the validity coefficients may be inflated by reporting biases that similarly affect both self-report measures.
In addition, the results support using the ratio of impaired- to unimpaired-arm recordings over using impaired-arm recordings alone to index real-world impaired-arm use. As noted, the ratio summary variable correlated more strongly with measures of impaired-arm use than overall physical activity. The impaired-arm summary variable, in contrast to the ratio summary variable, correlated moderately with both the impaired-arm and overall physical activity measures. This pattern suggests that the ratio summary variable captures impaired-arm activity associated mainly with its use, whereas the impaired-arm summary variable quantifies impaired-arm activity associated with its use for accomplishing upper-extremity tasks and its movement during overall physical activity. The reason for these differential relations is that changes in overall physical activity affect impaired- and unimpaired-arm recordings roughly equally. Thus, a large increase in ambulatory activity results in a correspondingly large increase in the impaired-arm summary variable but a much smaller change in the ratio summary variable.
The validity of the ratio summary variable for assessing changes in real-world arm activity after upper-extremity rehabilitation was not examined directly in this study because the authors were kept blind to EXCITE trial outcomes until all the controls were crossed over to CIMT. Previous studies,
however, show that the ratio summary variable is responsive to change. The change in the ratio summary variable from pre- to post-treatment in a group of CIMT patients more than 1 year after stroke was 4 times larger than that in a no-treatment control group over an approximately equivalent interval.
These studies also suggest that the ratio summary variable is a valid measure of treatment outcome. Correlations between pre- to post-treatment changes in ratio summary variable values and MAL scores ranged from .74 to .91.
It would have been desirable to have less missing data than we did (ie, 23%). Primary reasons were experiment error when programming accelerometers to record or download data (8% of total possible data points) and subject error when wearing the units (10%). The most common type of experiment error was tagging accelerometer data files with the wrong identifying information (ie, start date and time of recordings) by running the software to download accelerometer data from a different directory than that used to program the accelerometers to collect data. Other examples of experimenter mistakes were programming accelerometers to start collecting data at the incorrect time (eg, after participants had stopped wearing the units) and selecting the wrong recording mode when programming accelerometers to collect data. Rewriting the software used to program accelerometers so that it prompts users to verify critical recording parameters could reduce experimenter error. Other helpful steps would be eliminating the need to run the accelerometry software from the same directory when setting recording parameters and downloading data, giving laboratory staff additional hands-on training, and checking the fidelity of the accelerometry procedures whenever there is staff turnover. Participant error could be reduced by providing a take-home instructional video, creating a telephone hotline, and systematically engaging caregivers to help participants with following the accelerometry procedures. Another helpful procedure would be to make daily telephone calls to remind participants to wear the units and troubleshoot any problems encountered. Of note, none of the participants refused to wear accelerometers, most wore the units for a large proportion of the requested time (mean, 76%), and no one reported an injury or other adverse event in connection with wearing them. Accelerometer malfunction appeared to affect less than 1% of recordings.
The main source of the differential missing data rates between African-American (37%) and European-American (22%) participants was that a larger proportion of African Americans put on only 1 unit or put on the units after the programmed recording period. The literature on minority participation in clinical trials
suggests several reasons for diminished enthusiasm about wearing accelerometers on the part of African-American participants. Possible reasons include mistrust of clinical research
Further research is needed to help identify the specific reasons and increase compliance among African-American participants in future accelerometry studies. With respect to generalizability of the current findings to African Americans, reliability and validity of the accelerometer recordings did not differ by race. African-American participants with missing data, however, might still have had a different pattern of accelerometer recordings than those with intact data.
Outside of the CIMT literature, the relation between hemiparetic arm deficits at the impairment level and activity in daily life has been largely unexamined with the result that relatively little is known about how recovery in these 2 domains is related.
It is hoped that the development of objective methods for assessing real-world activity, such as accelerometry, will encourage researchers to examine this relation more closely. Other largely unexplored territory that might be illuminated by the application of methods such as accelerometry is the relation between specific neurologic disorders such as spatial neglect
Exclusive attention to these domains does not permit evaluation of the effects of neurorehabilitation on actual function of the more-impaired extremity in daily life.
This application of accelerometry is particularly salient given the increasing emphasis on restoring function of the impaired extremity, as opposed to teaching compensatory strategies.
Conclusions
Ambulatory monitoring of arm movement performed with accelerometers provides an objective method for assessing real-world upper-extremity motor status and rehabilitation outcome in patients with mild-to-moderate hemiparesis subsequent to subacute stroke. Only 2 wrist-worn accelerometers are needed because the ratio of impaired- to unimpaired-arm recordings indexes upper-extremity activity associated with use of the impaired arm and is largely independent of overall physical activity.
Supplier
aManufacturing Technologies Inc, 709 Anchors St, Fort Walton Beach, FL 32548.
References
Uswatte G.
Taub E.
Implications of the learned nonuse formulation for measuring rehabilitation outcomes: lessons from constraint-induced movement therapy.
Methods for a multisite randomized trial to investigate the effect of constraint-induced movement therapy in improving upper extremity function among adults recovering from a cerebrovascular stroke.
Uswatte G, Taub E, Morris D, Light K, Thompson P. The Motor Activity Log–28: assessing daily use of the hemiparetic arm after stroke. Neurology. In press.
Supported by the National Center for Medical Rehabilitation Research of the National Institute of Child Health and Human Development and National Institute for Neurological Diseases and Stroke, National Institutes of Health (grant no. HD 37606), and the American Heart Association Southeast Affiliate (grant no. 0365163B).
No commercial party having a direct financial interest in the results of the research supporting this article has or will confer a benefit upon the author(s) or upon any organization with which the author(s) is/are associated.
Duration of impaired-arm movement was calculated by dividing the number of epochs in the accelerometer record with above-threshold values by the total number of epochs.
00529-6&id=gr1.jpg){kind=link}