This protocol describes a method to quantify upper limb performance in daily life using wrist-worn accelerometers.
A key reason for referral to rehabilitation services after stroke and other neurological conditions is to improve one’s ability to function in daily life. It has become important to measure a person’s activities in daily life, and not just measure their capacity for activity in the structured environment of a clinic or laboratory. A wearable sensor that is now enabling measurement of daily movement is the accelerometer. Accelerometers are commercially-available devices resembling large wrist watches that can be worn throughout the day. Data from accelerometers can quantify how the limbs are engaged to perform activities in peoples’ homes and communities. This report describes a methodology to collect accelerometry data and turn it into clinically-relevant information. First, data are collected by having the participant wear two accelerometers (one on each wrist) for 24 h or longer. The accelerometry data are then downloaded and processed to produce four different variables that describe key aspects of upper limb activity in daily life: hours of use, use ratio, magnitude ratio, and the bilateral magnitude. Density plots can be constructed that visually represent the data from the 24 h wearing period. The variables and their resultant density plots are highly consistent in neurologically-intact, community-dwelling adults. This striking consistency makes them a useful tool for determining if upper limb daily performance is different from normal. This methodology is appropriate for research studies investigating upper limb dysfunction and interventions designed to improve upper limb performance in daily life in people with stroke and other patient populations. Because of its relative simplicity, it may not be long before it is also incorporated in clinical neurorehabilitation practice.
Over the last two decades, there has been an explosion of interest in wearable sensors to measure movement. A wearable sensor that has generated a great deal of interest in the neurorehabilitation field is the accelerometer.1,2,3 Accelerometers, as the name implies, measure accelerations in gravitational units (1 g = 9.8 m/s2) or in arbitrary units called activity counts (1 activity count = a manufacturer-specified gravitational value). Accelerations, like human movement, are typically measured and recorded in three dimensions, corresponding to the different axes of the device. The devices are commercially available and resemble large wrist watches; they can be worn during daily activities with minimal disruption. Because of the reasonable cost and their ready availability, the use of accelerometers (termed accelerometry) is being integrated into neurorehabilitation research.
The value of accelerometry to the field of neurorehabilitation is that it offers a non-invasive, unbiased, quantitative measure of upper limb motor activity outside the clinic or laboratory.3 A key goal of rehabilitation services for people with stroke and other neurological conditions is to improve one's ability to function in daily life, and not just in the clinic or laboratory. The World Health Organization's International Classification of Function distinguishes between the capacity for activity, as measured in a structured environment with clinical tests, and performance of activity, as measured in an unstructured environment.4 Accelerometry enables measurement of upper limb performance in the unstructured environment, i.e. what someone actually does when they are not in the clinic or laboratory, not just what they could do. Incorporation of accelerometry into stroke rehabilitation research is now challenging the long-held assumption that functional improvements in a structured clinical environment translate to improvements in performance in unstructured, daily life.5,6,7,8
Our group9,10,11,12,13,14 and others7,15,16,17,18,19,20,21,22,23,24 have been spent a great deal of time and effort on developing accelerometry methodology for use in research and clinical practice. Accelerometry has become well established as a valid and reliable tool for measuring upper limb performance post stroke.1,2,15,16,17,25 The most recent challenge has been turning the raw accelerometer data into clinically meaningful information (see reference3 for a summary of this development process). The methodology described here can be used for distinguishing upper limb performance in daily life in healthy control participants10,12 from that in participants who have suffered from stroke6,9,11 or have other disorders. The variables derived from this methodology are responsive to change and quantify improvements over time.14 The accelerometer methodology is appropriate for research studies investigating upper limb dysfunction and interventions designed to improve upper limb performance in daily life in people with stroke and other neurologic populations. Because of its relative simplicity, it may not be long before it is also incorporated in clinical neurorehabilitation practice.
This protocol was approved by the Washington University Human Research Protection Office.
NOTE: Instructions were written specific to commercially available accelerometers and their related software for data collection (see Table of Materials).
1. Preparing the Accelerometers to Collect Data
2. Placement and Wearing of the Accelerometers to Collect Data from Participants
3. Download the Data for Visual Inspection
4. Download the Data for Processing
5. Variables and Graphical Representations Created from the Accelerometry Data
NOTE: Upper limb movements associated with walking are included in the analyzed data. Previous work has established that walking does not influence the accelerometer ratio variables.15 Although inclusion of walking does not change the non-ratio variables for neurologically-intact adults,27 it is possible that the inclusion of walking could result in a small overestimation of the non-ratio variables for participants with stroke.
Data from a referent sample of community-dwelling, neurologically-intact adults can be used to interpret data from participants with stroke or other conditions affecting upper limb performance.10,11,12 Table 1 shows summary statistics for hours of use and the use ratio from a healthy referent sample. Overall, most people are active with their dominant and non-dominant hands for about the same amount of time throughout the day. The average is near 9 h, but there is a wide range, capturing more active and less active people. The average use ratio is just under 1.0 and has a small standard deviation. Thus, regardless of how active one is, the dominant and non-dominant limbs are used for similar durations throughout the day. Further, age does not influence upper limb performance measures in the presence of good health.12 Calculated values substantially outside these referent values (± 3-4 SDs) should be carefully checked to ensure that they are real, as suggested by Uswatte and colleagues.16
Average | Standard Deviation | Minimum | Maximum | |
Hours of dominant limb use | 9.1 | 1.9 | 4.4 | 14.2 |
Hours of non-dominant limb use | 8.6 | 2 | 4.1 | 15.5 |
Use ratio | 0.95 | 0.06 | 0.79 | 1.1 |
Table : Summary Accelerometry Statistics from Neurologically-intact, Community Dwelling Adults. Values are from referent sample of 74 community dwelling adults (average age 54 ± 11, 53% female, 84% right hand dominant), from reference12.
The density plots allow one to take a closer look at the data. Figure 1 is a representative density plot from a healthy adult, with data collected and processed as described above. Plots like this provide important information about upper limb performance in daily life. There are three key features of this plot that are highly consistent across adults of all ages.3,11 First, the picture is symmetrical. This indicates that the upper limbs are active together throughout the day, with the dominant and non-dominant limbs used similarly. The similarity of movement may not present be at a specific instance in time, with each limb taking its turn leading or lagging during various activities, but can be seen over the course of the day. Even the bars on either side at -7 and 7 (indicating solely dominant and solely non-dominant activity) are similar in color. The symmetry is contrary to common perceptions about hand dominance. Second, the plot is tree-shaped with a wide bottom portion and rounded edges. The 'rims' or rounded edges of the bottom portion represent activity where one limb is moving while the other is relatively still. An example of this would be placing objects in a container with one hand while holding the container with the other.10 The symmetry in the rounded edges indicates that both hands are active to perform and to stabilize similarly over the course of the day. The top peak represents the less frequent, higher intensity activities, such as placing large objects on a high shelf with both hands.10 And third, there is a warm glow in the center. This indicates that the most frequent upper limb movements are low intensity with approximately equal contributions from both limbs. Examples of this would be typing or cutting with a knife and fork.10
Figure 1: Representative Example from a Neurologically-intact Adult. The density plot shows 24 h of upper limb use in daily life, plotted on a second-by-second basis. The x-axis (magnitude ratio) indicates the contribution of each limb to activity. The y-axis (bilateral magnitude) indicates the intensity of movement. The color represents frequency, with the large color bar scale on the right side of the figure, where brighter colors indicate greater frequencies. The small bars at -7 and 7 represent unilateral dominant and non-dominant activity, respectively. Please click here to view a larger version of this figure.
Across this sample of adults, the density plots are remarkably similar in shape and color.11 People who are relatively inactive tend to have shorter, wider, pictures with cooler colors. People who are very active tend to have taller pictures with warmer colors. The striking consistency across adults makes it easy to identify participants with upper limb performance that is different than from these norms.
Figure 2 is an example of a density plot in a person with stroke. This person is a right-handed male who had an ischemic stroke affecting his brain on the right side 11 months prior to these data being collected. The right side of the brain controls the left side of the body, and his left upper limb had moderate paresis and dysfunction, as indicated by a Motricity Index28 score of 60/100 and an Action Research Arm Test29 score of 38/57. During the 24 h wearing period, the paretic, left limb was active for 1.5 h and the non-paretic, right limb was active for 5.8 h. His use ratio was 0.47, approximately half of the normal value. Compared to the density plot in Figure 1, this density plot is decidedly asymmetrical, indicating that the paretic upper limb was rarely active during daily life. The cool colors of the middle portion of the plot compared to the dark red colors of the single bar at -7 indicate a high frequency of movement with just the non-paretic limb. The overall peak is low, indicating only low intensity activities. Overall, the density plot indicates that the paretic limb participates only minimally in daily activity.
Figure 2: Representative Example from a Person with Stroke. The density plot shows 24 h of upper limb use in daily life, plotted on a second-by-second basis. The x-axis (magnitude ratio) indicates the contribution of each limb to activity. The y-axis (bilateral magnitude) indicates the intensity of movement. The color represents frequency, with the large color bar scale on the right side of the figure, where brighter colors indicate greater frequencies. The small bars at -7 and 7 represent unilateral dominant and non-dominant activity, respectively. Compare the symmetry, peak height, and color to Figure 1. Please click here to view a larger version of this figure.
While the accelerometry methodology has been developed for use in persons with stroke, the utility of this methodology extends to other populations. It can be beneficial for evaluating outcomes in a variety of patient populations. Figure 3 is an example of a density plot in a person with an upper limb amputation below the elbow. This individual was a 75 year old male, injured in an accident approximately 8 years ago. His right, previously dominant, hand was amputated at the time of the accident. He owns an upper limb prosthesis but wears it only 1-2 times per month to lift heavy objects. Most of the time, as in this figure, he does not wear it. During the 24 h wearing period, the intact, left limb was active for 6.9 h and the residual, right limb was active for 4.7 h (accelerometer was worn distally on the residual limb). His use ratio was 0.68, indicating a preference for engaging the intact limb over the residual limb. This density plot is less symmetrical and has cooler colors than that of a control subject (Figure 1), but is more symmetrical and shows more activity than the person with stroke shown in Figure 2. Thus, this person favors the intact limb, but still engages the residual limb in activities during daily life.
Figure 3: Representative Example from a Person with Upper Limb Amputation. The density plot shows 24 h of upper limb activity in daily life, plotted on a second-by-second basis. The x-axis (magnitude ratio) indicates the contribution of each limb to activity at the moment in time. The y-axis (bilateral magnitude) indicates the intensity of movement. The color represents frequency, with the large color bar scale on the right side of the figure, where brighter colors indicate greater frequencies. The small bars at -7 and 7 represent unilateral dominant and non-dominant activity, respectively. Compare the symmetry, peak height, and color to Figures 1 and 2. Please click here to view a larger version of this figure.
Another example of how this methodology might be used is in persons with limited mobility who need to increase activity. Figure 4 is an example of a density plot from an elderly, right-handed individual staying in a skilled nursing facility. This person was debilitated after an acute illness and was receiving nursing and rehabilitation services in order to regain independence and return home. The dominant limb was active for 2.4 h and the non-dominant limb was active for 2.0 h. The use ratio was 0.84, which is on the low end of the normative range (see Table 1). This density plot is nearly symmetrical, as would be expected from a general medical condition, but the peak is very low and the colors are mostly cool, indicating little activity during the wearing period.
Figure 4: Representative Example from a Person Recovering from Medical Illness in a Skilled Nursing Facility (SNF). The density plot shows 22 h of upper limb activity in daily life, plotted on a second-by-second basis. The x-axis (magnitude ratio) indicates the contribution of each limb to activity at the moment in time. The y-axis (bilateral magnitude) indicates the intensity of movement. The color represents frequency, with the large color bar scale on the right side of the figure, where brighter colors indicate greater frequencies. The small bars at -7 and 7 represent unilateral dominant and non-dominant activity, respectively. Compare the symmetry, peak height, and color to Figure 1. Please click here to view a larger version of this figure.
Finally, this methodology may not be just for adults. The protocol is suitable for children, with minor adaptations to encourage wearing (e.g. colorful straps, suggestions that the devices 'make you look like a superhero'). Density plots from typically developing children show the same general shapes as adults, the tree-shape being narrower and the peak substantially higher. The children's shapes are consistent with their greater activity levels; an example of density plots from a typically developing child and a child with hemiparetic cerebral palsy can be seen on p. 25, Figure 5B and 5C in reference3. Further investigations are needed for the application to pediatric clinical practice. It is noted that the use ratio has a consistent moderate relationship to self-reporting of upper limb activity in adults with stroke,1 but, in children with cerebral palsy, the use ratio is not related to the parent report of upper limb activity.30 Whether the altered relationship between sensor-measured and reported values lies in the perceptions of the reporters or in some quantitative or qualitative difference of how children move is unknown. Future studies are sorely needed to determine normative values for typically developing children and investigate the interpretation of values in children with disabilities.
This report details a methodology for measuring upper limb performance in daily life using accelerometers worn on the wrists. Use of this methodology in rehabilitation research and clinical practice affords a significant advancement upon existing methods, i.e. the opportunity to learn how an experimental or typical treatment impacts functional performance in daily life, not just capability in the clinic or laboratory. Accelerometry can be used in conjunction with, or in place of, self-reported measures of daily performance,31,32,33 which may be more susceptible to cognitive deficits or unconscious bias.34,35,36,37 Early adoption of this methodology has yielded data contrary to expectations,5 which could force the field to rethink the content and delivery of rehabilitation services.
Critical steps in the protocol ensure accurate and real data were collected during the wearing period (protocol steps 2.2, 2.3, and 3.3). Failure to follow these steps could result in calculated values that have no meaning. It is relatively easy to make sure that the accelerometers are on the assigned wrists as the person leaves the clinic or laboratory. Visual inspection of the data after the accelerometers are returned is necessary, as participants often behave differently than instructed or anticipated. While relatively rare, participants have been known to remove the accelerometers shortly after leaving the investigative team, putting them back on again on the wrong sides, or trying to encourage others in their family to wear them. Much of this can be avoided if the accelerometers are clearly marked for each side, the wearing log is completed, and the data are inspected shortly after returning, i.e. in case a follow-up telephone call is needed to clarify wearing side and times.
While the accelerometry methodology quantifies general upper limb performance, it does not provide information about movement quality or about specific activities that were performed during the wearing period, such as knowing that a participant was eating; see reference3 for a discussion of this issue. As a tool then, accelerometry will be most useful as an outcome measure when the scientific question or rehabilitation intervention is focused on changing general upper limb performance in daily life, such as the amount of activity and the involvement of bilateral limbs in daily activity. Accelerometry will be less useful as an outcome measure when the scientific question or rehabilitation intervention is focused on changing the quality of movement or changing only a few specific movements in daily life. We anticipate that computational methods will improve over time and future generations of this methodology may be able to overcome these limitations.
In conclusion, accelerometry presents an opportunity for quantitative assessment of upper limb performance in daily life. The methodology described here can be considered the upper limb version of the more common mobility methodologies, where steps per day or minutes of moderate physical activity are recorded on wearable devices.38,39,40,41,42,43 While developed for persons with stroke, the versatility of the methodology will allow future application in a variety of other populations. Additional methodological development is needed in adult and pediatric neurorehabiliation populations other than stroke to help answer clinical and research questions related to bilateral activity of the upper limbs.
The authors have nothing to disclose.
We thank Brittany Hill, Ryan Bailey, and Mike Urbin for their contributions to the accelerometry methodology and data. Funding for this project comes from NIH R01 HD068290.
Accelerometers (2) | Actigraph LLC | wGT3X-BT | This is the most common device on the market. Similar products are available from other vendors. http://actigraphcorp.com/products-showcase/activity-monitors/actigraph-wgt3x-bt/ |
Hub | Actigraph LLC | 7 Port USB Hub | This device connects the accelerometers to the computer allowing for charging and communication. Includes hub, usb cables, power connector. http://actigraphcorp.com/products/7-port-usb-hub-2016/ |
Straps | Actigraph LLC | Woven Nylon Wrist Band | Other straps that are velcro or disposable are also available. http://actigraphcorp.com/product-category/accessories/ |
Actilife Software | Actigraph LLC | It is best to purchase the software from the same vendor as the accelerometers. Similar products are available from other vendors. http://actigraphcorp.com/products-showcase/software/actilife/ | |
Computational software | The most common software is MATLAB, but computation could also be done in Excel or other similar products. |