Global Position Sensing and Step Activity as Outcome Measures of Community Mobility and Social Interaction for an Individual With a Transfemoral Amputation Due to Dysvascular Disease
- A. Jayaraman, PT, PhD, Max Nader Center for Rehabilitation Technologies and Outcomes Research and Center for Bionic Medicine, Rehabilitation Institute of Chicago, and Department of Physical Medicine and Rehabilitation, Northwestern University, Chicago, Illinois. Mailing address: Max Nader Center for Rehabilitation Technologies and Outcomes Research, Rehabilitation Institute of Chicago, 345 E Superior St, Room 1771, Chicago, IL 60611 (USA).
- S. Deeny, PhD, Max Nader Center for Rehabilitation Technologies and Outcomes Research, Rehabilitation Institute of Chicago, and Department of Physical Medicine and Rehabilitation, Northwestern University.
- Y. Eisenberg, MUPP, Department of Disability and Human Development, University of Illinois at Chicago, Chicago, Illinois.
- G. Mathur, PT, Max Nader Center for Rehabilitation Technologies and Outcomes Research, Rehabilitation Institute of Chicago, and Department of Physical Medicine and Rehabilitation, Northwestern University.
- T. Kuiken, MD, PhD, Center for Bionic Medicine, Rehabilitation Institute of Chicago, and Department of Physical Medicine and Rehabilitation, Northwestern University.
- Address all correspondence to Dr Jayaraman at: a-jayaraman{at}northwestern.edu.
Abstract
Background and Purpose Community mobility of individuals following lower limb amputation is highly variable and has a great impact on their quality of life. Currently, clinical assessments of ambulatory ability and motivation influence prosthetic prescription. However, these outcome measures do not effectively quantify community mobility (ie, mobility outside of the clinic) of individuals with an amputation. Advances in global positioning systems (GPSs) and other wearable step-monitoring devices allow for objective, quantifiable measurement of community mobility. This case report will examine the combined use of a GPS unit and a step activity monitor to quantify community mobility and social interaction of an individual with transfemoral amputation due to dysvascular disease.
Case Description A 76-year-old woman with a unilateral transfemoral amputation due to vascular disease carried a commercial GPS unit and step activity monitor to quantify her community mobility and social interaction every day over a period of 1 month. The step activity monitor was affixed to her prosthesis. The patient used a wheelchair as well as her prosthesis for everyday mobility.
Outcome Information from the GPS unit and step activity monitor provided quantitative details on the patient's steps taken in and out of the home, wheelchair use, prosthesis use, driving trips, and time spent on social and community trips.
Discussion This case report describes a potential clinical measurement procedure for quantifying community mobility and social interaction of an individual with lower limb amputation. Future efforts are needed to validate this measurement tool on large sample sizes and in individuals with different mobility levels. Additionally, automatization of data analysis and technological approaches to reduce compromised GPS signals may eventually lead to a practical, clinically useful tool.
Approximately 1.6 million individuals in the United States currently live with limb loss, and approximately 82% of these individuals have experienced amputation as the result of vascular disease.1 Based on existing trends and statistical projection, it is predicted that this figure will more than double by the year 2050 as the population ages and the prevalence of vascular disease increases.1
A primary goal of the physical therapy and prosthetics team is to improve community mobility of individuals with an amputation.2 This goal requires choosing the best prosthetic components and rehabilitation protocols for each patient based on his or her functional ability, societal requirements, and motivation.3–5 Without accurate measures of a patient's ambulatory capabilities outside of a controlled clinical setting, it is difficult for clinicians to know the extent to which patients benefit from their personalized prosthetic devices or adhere to their rehabilitation plans. There is a need, therefore, to develop quantitative outcome measures to accurately assess patients' community mobility and participation in everyday society. Accurate and objective measures of home activity and community participation may improve prosthesis prescription and reimbursement, assist in the evaluation of new prosthetic technologies, aid in the development of better rehabilitation strategies, and clarify the relationship between quality of life and community mobility.
In today's health care climate, the physical rehabilitation of individuals with amputation is increasingly directed toward functional outcomes, assessment of which typically falls under the categories of self-report, professional-report, and physical performance measures.6–8 Self-report and professional-report measures are based on subjective recall and evaluation by the patient (or family member) or the clinician, respectively.9 These measures, by their nature, may be subject to bias. Although performance measures such as the Six-Minute Walk Test or the Berg Balance Scale are objective and have been validated in numerous disease conditions, these measures assess the patient's ability at a specific point in time, leading to performance variation between “good days” and “bad days.” In addition, these measures may not reflect dynamic progress or encompass all aspects of real-life mobility in the community that may be influenced by additional environmental or logistic confounds such as weather, social constructs, or living environments.2,8,10
Readily available technology such as global positioning systems (GPSs) and commercial accelerometers can be used to provide objective and quantitative outcome measurements outside of the clinical setting.11–15 Global positioning system technology has been successfully introduced in health care research16–19 and shown to accurately detect when the patient is walking and to provide estimates of walking speed and distance traveled.15,17–19 It also has been used to collect data on location, speed of movement, and time taken to perform movements.20–22 More importantly, GPS technology presents a way to objectively quantify the unique interaction of the patient with his or her physical environment. Combined with an accelerometer or step activity monitor, GPS data place measurement of the patient's physical activity and prosthesis use into the context of the community in which the patient participates.18,23,24 Currently, the use of GPS monitoring devices combined with accelerometers has been reported in individuals with stroke and spinal cord injury and in controls who are healthy.18,24 However, there are no studies addressing use of this technology to monitor community mobility in individuals with amputations.
The purpose of this case report is to describe the objective quantification of community mobility and social interaction in an individual with an amputation using a commercial step activity monitor and GPS monitoring device over a period of 1 month.
Patient History and Review of Systems
The patient was a 76-year-old African American woman with a history of hypertension, congestive heart failure, and coronary artery disease. She was retired and had stopped smoking several years prior to the study. After retirement, the patient did not exercise regularly and gained more than 6.8 kg (15 lb). Three years prior to enrolling in the case study, she developed nonhealing dry gangrene in her right foot. After a failed revascularization surgery, she had an amputation at the local county hospital. The amputation was initially transtibial but was revised to transfemoral within 10 days. The patient progressed through the system of care in a traditional manner: she was initially treated in acute care and progressed through inpatient rehabilitation followed by home health and outpatient physical therapy. Her final prosthesis and socket were fitted 1 year after her surgery. She had occasional visits to the prosthetist for prosthesis adjustments and related physical therapy.
Before enrollment in the case study, the patient was classified by her physician and prosthetist, using the Medicare Functional Classification Level (MFCL), as level K2.25 This classification was based on her capacity and potential to accomplish expected postrehabilitation daily function. At the K2 level, a person is considered to have limited mobility and may walk for limited periods of time without significantly varying his or her speed, typical of a limited community ambulator. The individual also is expected to use the prosthesis to negotiate low-level environmental barriers such as curbs, stairs, or uneven surfaces. The functional K level thus contributes to the determination of what type of prosthesis components the patient receives. Based on her K2 functional level, the patient described in this case report had received a GeoFlex Knee (Ohio Willow Wood Co, Mt Sterling, Ohio), which is a friction-controlled polycentric knee that caters to stability needs, a liner suspension (a lanyard with buckle, Össur, Reykjavik, Iceland), a 2-ply fit sock, and a K2 Sensation foot (Össur), which is a full-length keel foot with a multiaxial ankle meant for slow speed and gentle modest walking with a mobility aid.
Clinical Impression
Individuals with an amputation due to vascular disease constitute the largest population of individuals with lower extremity amputations in the United States.26–29 The goal of this case study was to evaluate our outcome tool on a typical person with amputation due to vascular disease. The patient lived alone and used either her prosthesis or a wheelchair for community mobility. She was involved in social activities such as attending church and left her home for personal chores such as grocery shopping. The assessment plan prior to this case study was to use both standardized patient-report and performance-based tests to determine prosthesis use and ambulation within the home and community. In addition, the plan was to obtain the patient's self-reported perceptions of community mobility and social interaction.
Examination
The patient was examined at the Rehabilitation Institute of Chicago (RIC) by a physical therapist and prosthetist prior to being provided with the GPS and step activity monitor. A summary of all tests conducted is shown in the Table. The patient's goals for using her prosthesis, stated during the examination, were to use her prosthesis inside and outside the home. The inside home activity goals included walking more (eg, from bedroom to bathroom, from kitchen to bedroom) with the prosthesis and an assistive device and standing with assistance to perform activities of daily living such as cooking and brushing teeth. Outdoor activities included walking from the car in the parking lot to a store, walking inside the store, walking to church, and walking inside the church. Because these indoor and outdoor goals were based in her home and community environment, it was necessary to find a reliable and valid way of assessing her progress toward these goals. Based on the tests listed in the Table, the patient was identified as a limited community ambulator (level K2), as established by:
A score of 32 on the Amputee Mobility Predictor Pro (level K2 score range=18–46)30
A gait speed of 0.55 m/s on the 10-Meter Walk Test (0.4–0.8 m/s is indicative of a limited community ambulator)31
A prosthetic socket fit comfort score of 8/10 (0=most uncomfortable, 10=most comfortable)
No phantom limb pain recorded
Baseline Outcome Measures for the Patient, Indicating Her Diagnosis as a Limited Community Ambulator
The patient was a good candidate to test use of GPS and a step monitor to assess community mobility and social interaction, as she was a limited community ambulator with specific goals of using her prosthesis to walk in the community and interact socially. She performed all of the clinical tests, showed great enthusiasm for participating in the case study, and did not feel uncomfortable about being monitored. Prior to her enrollment in the case study, institutional review board (IRB) approval had been obtained, and she provided written consent in accordance with Northwestern University's IRB regulations.
Measurement Procedure
A commercially available GPS data logger, the Travel Recorder XT (QStarz, Taiwan, Republic of China), was used to track the patient's location. This unit was selected because of its long battery life (42 hours), small size (fits in palm of hand), light weight (74 g), large storage capacity (400,000 waypoints), and precise accuracy (within 2.5 m). Global positioning system data logging has been validated and has been successfully used in numerous clinical studies.14–24 Coat pockets, pants pockets, and purses were found to be appropriate locations for storing the data logger. These placements were tested prior to this case study, and all placements resulted in accurate data logging.
A data quality check was completed to examine limitations of the GPS data loggers in calculating distances. The distance from origin to destination might be overestimated by interference from environmental variables (eg, tall buildings in downtown Chicago, cloud cover), causing a mapping route line that was longer than a straight line, or underestimated because the data were logged at 10-second intervals. To measure the accuracy level of the distances, we randomly selected 100 trips and calculated the distance traveled by overlaying satellite imagery and drawing a line that precisely followed the path of travel. We performed an independent-samples t test between the distance logged by the GPS and the distances that were drawn along the path of travel. There was no significant difference between the logged distance and the manually calculated distance (t=0.044, P>.5).
The patient was asked to wear the GPS data logger for 1 month and was instructed to have it on her at all times during the day. She was given a wristband with the words “Are you plugged in?” to act as a reminder to turn on the GPS device in the morning and charge it at night. She was given an instruction sheet for using the GPS data logger along with phone numbers to call if questions arose. After 1 month, the GPS data logger was returned, and the data were uploaded using the GPS data logger software, QTravel (QStarz). Step activity (accelerometer) data were collected by attaching the monitor (StepWatch 3.1 Activity Monitor, Orthocare Innovations, Oklahoma City, Oklahoma) to the prosthesis. The monitor was timed to start recording at the same time as the GPS unit. The StepWatch accuracy has been validated at gait speeds consistent with the gait speed of our patient in individuals with amputations and those with other disabilities.11–13,32–34
Data Analysis
The GPS unit was configured to record the patient's location every 10 seconds. The GPS data were configured in Microsoft Excel (Microsoft Corporation, Redmond, Washington) so that they could be read by the mapping and spatial analysis software (ArcGIS, Esri, Redlands, California). Once in ArcGIS, the entire dataset was split into specific days for detailed analysis.
In order to identify the patient's movements and community destinations, a data analyst determined each time the patient left a location and arrived at a destination for each day of the case study. Each day was segmented and coded into sections based on the patient's location (in her home or in the community), her mode of transportation (walking, driving, or using a wheelchair), and her transitions (to or from the car or community destination). This process allowed detailed quantitative data to be developed for every minute of the patient's day, which was recorded in a daily trip log in Microsoft Excel.
The analyst noted 2 types of signal problems. Total signal loss was noted when no GPS data were recorded; this problem could occur due to obstruction of the signal by buildings or loss of power in the GPS unit. Loss of signal accuracy was noted when GPS data were recorded but were not accurate to 2.5 m. If 1 of the 2 signal problems was recorded at a destination or during a trip, as long as the points of entry and exit were not compromised, the destination was kept as valid. If the signal problem occurred at a destination and the point of exit was compromised, the destination was marked as “poor,” and the time spent at that location was not used in the analysis. Of the 22,707 minutes logged by the GPS, 97% (21,955 minutes) of the patient's day were coded as “good” trips or destinations. The remaining trips were not logged because of total signal loss or poor signal accuracy. Although signal accuracy issues can make identifying destinations more difficult, the analyst was able to use geographic information system (GIS)–based methods to calculate the mean center of the points that were designated as the destination. This approach was helpful for urban areas where buildings are close together and it was not apparent which building the patient went to.
The data output from the step activity monitor was a table in Excel format with rows for each minute of the day and the number of steps recorded every minute. The data were integrated with the GPS data in ArcGIS so that each point on the map also included the number of steps taken in that minute of the day. The data analyst then could identify how many steps were taken at home, en route, and at each destination. Figure 1 illustrates the data flow and logic for combining the GPS and step data to determine the mode of transportation.
Flowchart illustration of analysis of combined step activity and global positioning system (GPS) data to determine when the patient was walking or using a wheelchair and where she was at the time of activity. a Based on the GPS data, if the patient was indoors and not traversing distances, we determined that she was likely moving her feet while stationary or making small ambulatory movements while in one location. b In some instances, there were not enough steps to indicate a walking trip given the distance traversed. c Some steps occurred during wheelchair trips from bumps, minor leg movements, and getting in and out of the wheelchair.
Determination of points of departure and arrival as well as mode of travel for each trip was done using 4 movement indicators: the patient's speed, the distance between points, the number of steps taken in that time period, and the patient's location according to digital aerials.
The analyst started by identifying days when the patient left her home. For these days, the first step was to identify the first car trip, which was done by looking for high speeds in the data. It was clear when the patient used a car because she would leave from the front of the house, did not follow bus lines, and stopped in parking lots. The analyst then identified the point of departure and arrival for the first car trip. Transitions to the point of departure (ie, traveling to and from the car) were marked by tight clusters of points and very low speeds. The analyst identified trips to the car using the 4 indicators to identify the start and end points.
Determining whether the mode of transport was walking or using a wheelchair was done by examining step activity and the distance traversed (Fig. 1). The method used to identify car trips (speed and distance between points) was not effective in distinguishing walking from wheelchair use because both have similar speeds. Step data were used to distinguish between these 2 modes; however, the step activity monitor was observed to record step measurements when the patient was clearly riding in the car (based on speed), likely as a result of bumps in the road or foot movements. Thus, the patient's mean standard stride length of 73.2 cm (SD=18.3) was used to determine whether the number of steps taken could account for the distance covered. If the number of steps was too low to account for the distance covered in the nonmotorized trip, the patient was considered to be using a wheelchair. Step data recorded during wheelchair usage were attributed to leg movements or bumpy terrain. The patient's mean stride length was determined using the gait speed from walking on an instrumented GAITRite (CIR Systems Inc, Sparta, New Jersey)—a device that determines temporal-spatial walking parameters—along with the GPS and step monitor data. Segments of clean GPS data during which the patient's speed was consistent with her walking speed determined from the GAITRite, and their GPS location (walking path or walkway) was consistent with a walking trip were used to determine if the patient was walking.
Determining Destinations
Destinations were identified when a cluster of points indicating a stop followed a continuous group of points indicating higher speeds. From there, the analyst also could see movement toward a building or destination. If there was too much static in the data, it was noted in the log, and no pedestrian trip was recorded. After determining that a valid trip took place, the next step was to determine the destination. The analyst used an aerial base map provided by ArcGIS along with Google Maps' rich interface of data: places of interest, local stores and services, aerials, and other visual information. The patient's destinations could be examined in detail using Google Maps Street View, which shows a street-level picture of the destination. As a result, it was clear what type of destination was visited (Fig. 2). Analysts took care to note the year of the aerial and street view image in order to make sure that the most up-to-date information was used. Destinations were categorized as commercial destinations (eg, grocery stores, shopping centers), religious (church), other residential (homes or apartments other than the patient's home), open space (parks or spaces not in or near buildings), mixed use (buildings with mixed commercial and residential uses), or medical (hospital or medical facilities). The analyst was able to identify all destinations for this patient.
Representative global positioning system plots from (A) a day spent making multiple trips outside the house and (B) a day spent entirely in the home.
Once all of this information was determined, it was recorded in a trip log for that day. Trip logs included all of the trip attributes: mode of transit, trip start and end times, total trip time, time spent, activity at origin and destination, total number of steps taken, quality of GPS signal, trip distance, purpose of trip, and any other notes that could add detail to one of the above. The trip logs were then reorganized and summarized to develop new variables for data analysis.
Outcomes
Based on the measurement procedure described above, we determined the number of days in the month when the patient left the house, the destinations the patient visited over the course of the month, the time spent at those destinations, the distances traveled, the modes of travel used, the number of steps taken on walking trips, and the number of steps taken in and out of the home.
GPS Use
The patient turned on the GPS unit every day during the 31-day study month, except for 2 days in which the unit was misplaced. The GPS unit collected data for an average of 15 hours a day. The number of hours recorded by the GPS unit each day remained fairly consistent over the study (X̅±SD=15.1±4.0 hours); the participant was typically asleep when the GPS unit was off.
Time Outside the Home
The patient left home on 4 of the 20 weekdays and on 6 of the 9 weekend days (Fig. 3A). Data for one of the weekday visits was not analyzed due to inconsistent GPS signals. The mean (±SD) amount of time spent at home (extrapolated from time out of the home) was 23.2±1.9 hours per day on weekdays and 20.1±2.6 hours per day on weekends.
Time spent out of the house and at community locations. (A) Days when the patient left the house coded by weekdays and weekend days. Although she left the house 10 out of the 31 days, one of the days had missing global positioning system data. Therfore, only 9 days are presented here. Asterisk (*) indicates a day when she left the house, but time could not be calculated due to missing data. (B) The different community places visited during the month and the average duration of time spent at each visit.
Community Trips
Most time (X̅±SD) was spent at religious destinations (5.9±2.6 hours), followed by other residential (2.4±0.5 hours), outdoor open space (2.18±0.4 hours), and finally commercial (1.12±0.2 hours) (Fig. 3B). However, the most frequently visited type of destination was commercial (8 trips), followed by religious (5 trips), medical (2 trips), open space (1 trip), mixed use (1 trip), and other residential (1 trip). The mean (±SD) time taken from home to each of these destinations and calculated distance traveled indicate that the commercial trip took the least amount of time (6.2±3.3 minutes/1.35±0.63 miles), followed by religious (9.9±1.6 minutes/3.02±0.2 miles), other residential (11±1.0 minutes/3.27±0.47miles), and finally medical (31.3±2.9 minutes/10.94±2.69 miles). The 2 medical trips were to a large medical facility with thick walls that obstructed the GPS signal. The GPS signal failed as soon as the participant entered the building and did not resume until she was back home or en route to her home. Therefore, the total time at medical destinations is not reported.
The patient used a car to get to all destinations. The mean (±SD) travel time from home to any of these destinations was 11.1±8.1 minutes, and the mean (±SD) distance was 3.4±2.5 miles.
Trip Modes
The patient predominantly traveled by car, and only walked or used a wheelchair to get to and from the car (Fig. 4A). Twelve of the 28 pedestrian trips to or from the car were accomplished using a wheelchair, and 16 were accomplished by walking.
Trip modes and steps taken for the days when the patient left the house (10 out of 31 days). (A) Number of car trips, wheelchair trips, and walking trips on these days. Undetermined trips indicate trips where the step and global positioning system data did not allow a clear distinction between walking and wheelchair trips. (B) Steps taken at home and outside the home on days when the patient left the house.
Step Activity
On days when the patient stayed at home, she recorded only about 0 to 2 steps per day, with the exception of one day during which 199 steps were recorded. This finding indicates that for most of the days spent at home, the patient most likely did not wear the prosthesis with the attached step monitor. Figure 4B shows the total number of steps taken (in and out of the home) on days when the patient left the home. The average (±SD) steps taken during these outdoor activity days were 52±73 at home and 188±132 outside the home.
Discussion
In this case study, we sought to examine and describe the use of combined data from a GPS unit and a step activity monitor to quantify community mobility in an individual with transfemoral amputation due to dysvascular disease.
The ability to move around one's home and in the community is a basic requirement and is one of the core goals of physical rehabilitation for individuals postamputation. Identifying the constraints and goals of community mobility is a complex issue, given the fact that it drives prosthetic prescription, physical therapy, and associated reimbursement costs. Health-related quality-of-life measures regard independent mobility as a key indicator for quality of life.35,36
The extent of mobility of an individual with amputation may largely depend on his or her own ability, the prosthetic components available, and the environment in which he or she lives. The current approach to prescribing a prosthesis for individuals with amputations is based on the Medicare Functional Classification Level (MFCL), known as the K levels. K levels were developed by the Centers for Medicare & Medicaid Services (CMS) in response to the availability of increasingly advanced, expensive prosthetic components.25 The CMS intended to give insurance payers the information they need to determine which devices are medically necessary for each patient. Unfortunately, the K levels, as currently applied, have major limitations. The K-level classification is used by the physician and prosthetist to determine the patient's ability to reach a predetermined functional ambulatory state within a reasonable period of time, and they choose the most appropriate prosthetic device based on this predicted functional level and not based on current activity levels. The K level categorizes individuals with lower extremity amputations into broad levels, such as nonambulators (level K0), household ambulators (level K1), limited community ambulators (level K2), community ambulators (level K3), and high-level or athletic users (level K4). Within each functional K level, considerable variability in daily activity levels exists between individual patients; however, the K levels do not subcategorize individuals within a functional level.
Considerable variations in gait exist in prosthetic users, given the influence of the level of amputation, the type of amputation, and assistive devices used. These gait variations influence walking speed and overall community mobility. Thus, within a given K level, there could be significant variability in walking capabilities. The existing K-level classifications rely on subjective assessments2 and not on tested and validated quantitative outcome measures. The CMS, which oversees reimbursement, does not specify or suggest any objective measures that clinicians should use to assess community mobility and prosthetic prescription. Lack of objective assessment may lead to mobility limitations, given that patients might not be prescribed an appropriate prosthesis. Currently, many quantitative clinical outcome measures exist that have been validated and deemed reliable for populations of individuals with amputations.2,8,10 However, these measures may not reflect dynamic progress or encompass all aspects of actual mobility in the community that are affected by environmental or logistic confounds, such as weather, social constructs, or living environment.2,8,10 In addition, community mobility priorities for each individual are variable.
Our case study attempted to quantify one individual's preferences using the combination of the GPS and a step activity monitor. Step activity monitors have been used and validated extensively for steps taken in everyday life and for upper limb activity including wheelchair use.37–39 Halsne et al40 quantified step activity data in individuals with transfemoral amputations over a period of 1 year. Patients classified as levels K2, K3, and K4 averaged (±SD) 1,154±538, 1,446±641, and 2,560±100 steps/day, respectively.40 In the current case study, our patient took an average of 240 steps/day, which is much lower than published data, indicating the limitations and variability in the K-level classification. However, one major limitation of step monitors is that they purely count steps without taking into consideration the location and activity associated with the steps. They do not provide any demographic or quality-of-life information, such as time spent inside and outside the house or ability to go to commercial or other social locations, which suggests social involvement. For example, the ability to use public transport compared with a disability services van suggests the ability of an individual who is disabled to mingle with the able-bodied population, handle crowds, and move in locations that may lack ramps or other accessibility features.
Combining GPS with the step monitor gives us the ability to provide objective and quantitative tracking of patient mobility behavior over extended periods of time.11–24 Tapping these newer technologies may someday allow clinicians to continue to monitor how often, when, and where the prosthesis is used once the patient leaves the clinic. Validation of K-level assignments using this technology could help provide better reimbursement strategies. Furthermore, clinicians might be able to modify their physical rehabilitation strategies based on information obtained to improve the patient's use of the prosthesis. Combining these multiple technologies (GPS, accelerometers, and step counters) into one dedicated device could further increase the accuracy of data analyses by simultaneously considering location, movement, and speed. However, it is very important to acknowledge the limitations of this case study and technology in its current state.
Limitations
Using GPS devices to log community participation and mobility is a relatively new field, especially for people with disabilities. Although our case study showed promising results, there were several limitations, including technology limitations, analysis time, and human error, that are discussed below.
Some trips and destinations were not coded due to interference with or static in the GPS signal. Interference typically occurred when the patient entered large buildings with thick walls, encountered severe weather, or entered a significantly dense urban area with tall buildings. Previous studies have gone into more depth on GPS signal interference.23,41 Signal interference can result in missing data in the GPS file (reflecting total loss of GPS signal) or compromised accuracy of data points. In the latter case, data can show a person's location shifting in a way that is inconsistent with typical movement. When this behavior occurred, it was not possible to determine how the patient moved; we could only determine that she went from one location to another.
Another issue in this case study was missing data due to device malfunction, a drained battery, or lack of patient adherence. The analysis also was limited by the availability of up-to-date geographic data. Google's Street View data are from 2007. The Bing Maps aerial, ArcGIS aerial basemap, and Google Maps point-of-interest data are more recent (2010–2012). In addition, information on certain roads, structures, and businesses may not be up-to-date.
There also are inherent challenges in the type of data analysis used in this case study, including challenges in quantifying specific behaviors (eg, walking versus wheelchair motion), and false-positive recorded events. Of the total steps taken during the recorded GPS time (3,271), a daily average of 15% false positives was identified as occurring during car or wheelchair trips. Behaviors with overlapping speeds, such as walking and wheelchair motion, are more challenging to differentiate with GPS data. Other studies have combined GPS data with accelerometer data to distinguish between modes of transportation used by a patient, such as vehicle use versus walking18,42; however, the authors did not distinguish between wheelchair motion and walking behavior. For individuals with amputation, our interest is in determining use of the prosthesis for community mobility; therefore, distinguishing walking behavior from wheelchair use is crucial. In our dataset, car trips were easy to distinguish from wheelchair or walking trips using GPS velocity data; however, distinguishing between the 2 behaviors required a combination of step monitor data and the patient's known average stride length (calculation of stride length is described in the “Data Analysis” section). Baseline measures of gait, gait speed, and, if possible, GPS and step monitor data for both walking and wheelchair use can assist in future analyses. Further variability in analysis can occur if the patient spends a lot more outdoor time in areas where GPS signal clarity is poor or if the patient has a significant variable cadence during walking, affecting the stride length, which would cause further difficulty in differentiating walking steps and wheelchair activity. Future analysis strategies need to take into consideration all of these current limitations.
Finally, the datasets that result from long-term monitoring of these devices are very large, and analysis is time-consuming. The dataset for this study required approximately 2 weeks to analyze. New algorithms to automate data analysis will be necessary to make these tools practical and efficient for use by clinicians and researchers.
In conclusion, this case report demonstrated that it is possible to sensitively quantify community mobility and related social activities of an individual with amputation. The technique allowed us to quantify steps as well as the reasons the steps were taken during the everyday life of the person. These findings provided insight into the patient's prosthesis use in everyday life. However, clinical application of this technique will require future work in several areas, including resolving technological limitations and achieving efficient analysis of large data files. Future studies should include larger sample sizes and participants with a range of clinical diagnoses to further validate the measures for different levels of community mobility.
Footnotes
Dr Jayaraman, Mr Eisenberg, Ms Mathur, and Dr Kuiken provided concept/idea/project design. Dr Jayaraman, Dr Deeny, Mr Eisenberg, and Ms Mathur provided writing and data analysis. Dr Jayaraman and Ms Mathur provided data collection and project management. Ms Mathur and Dr Kuiken provided fund procurement and patient recruitment. Dr Jayaraman, Ms Mathur, and Dr Kuiken provided facilities/equipment. Ms Mathur provided institutional liaisons. Dr Deeny and Mr Eisenberg provided consultation (including review of manuscript before submission).
- Received December 27, 2012.
- Accepted October 1, 2013.
- © 2014 American Physical Therapy Association