VA Research and Development LOGO

Logo for the Journal of Rehab R&D
Vol. 40 No. 5, September/Ocober 2003, Supplement 2
Pages 45 — 58


Bodies in motion: Monitoring daily activity and exercise with motion sensors in people with chronic pulmonary disease
Bonnie G. Steele, PhD, ARNP; Basia Belza, PhD, RN; Kevin Cain, PhD;
Catherine Warms, PhD, ARNP, CRRN; Jeff Coppersmith, MS, CSCS; JoEllen Howard, BS, GCS
Primary Care and Specialty Medicine Service, Health Services Research and Development Department, Department
of Veterans Affairs (VA) Puget Sound Health Care System, Seattle Division; Department of Biobehavioral Nursing and Health Systems, University of Washington, Seattle, WA; Department of Biostatistics and Office of Nursing Research, School of Nursing, University of Washington, Seattle, WA
Abstract — A primary goal of pulmonary rehabilitation is to improve health and life quality by encouraging participants to engage in exercise and to increase daily physical activity. The recent advent of motion sensors, including digital pedometers and accelerometers that measure motion as a continuous variable, have added precision to the measurement of free-living daily activity. Daily activity and exercise are variables of keen interest to proponents of the national health agenda, epidemiologists, clinical researchers, and rehabilitation interventionists. This paper summarizes issues related to conceptualizing and monitoring activity in the rehabilitation setting; reviews motion sensor methodology; compares motion-sensing devices; presents analysis issues and current and potential applications to the pulmonary rehabilitation setting; and gives practical applications and limitations.
Key words: accelerometer, daily activity, exercise, pedometer, pulmonary disease, pulmonary rehabilitation.

Abbreviations: COPD = chronic obstructive pulmonary disease, ICC = intraclass correlation coefficient, VMU = vector magnitude units.
This material was based on work supported by the Health Services Research and Development Merit Review NRI 98-194.
Address all correspondence and requests for reprints to Bonnie G. Steele, PhD, ARNP; VA Puget Sound Health Care System (111-B), 1660 Columbian Way South, Seattle, WA 98108; 206-764-2496; fax: 206-768-5398; email: bonnie.steele@med.va.gov.

DOI: 10.1682/JRRD.2003.10.0045
INTRODUCTION

In chronic pulmonary disease, dyspnea and deconditioning profoundly constrain physical activity and are known to produce, over time, spiraling losses in global functioning and life quality. Pulmonary rehabilitation, which includes graded exercise, strength and flexibility training, and collaborative self-management education, improves physical functioning and life quality and is now considered an integral component of optimal care for persons with severe lung disease [1,2]. It is likely that the most salient benefits of pulmonary rehabilitation come through program-related improvement in the ability to carry out daily physical activities, and in particular, to undertake the ubiquitous behavior of walking. The measurement of free-living physical activity and walking has recently been found to be particularly suited to devices that measure motion, such as accelerometers, which can objectively quantify even low levels of physical activity as a continuous variable and can detect subtle incremental changes as a result of intervention [3]. This article provides an overview of the potential utility of motion sensors to measure physical activity in persons with chronic pulmonary disease in the setting of pulmonary rehabilitation. We address the conceptualization of activity, exercise rehabilitation, motion sensing, comparison of motion sensors, methodological and analysis issues, applications to pulmonary rehabilitation, and practical considerations and limitations.

CONCEPTUAL DISTINCTIONS

For purposes of clarity, a number of conceptual distinctions should be made. First, exercise, such as those activities undertaken in a pulmonary rehabilitation program, is defined as the planned, structured, and repetitive bodily movement carried out to improve or maintain one or more aspects of physical fitness [4]. Daily physical activity, a variable only recently quantifiable, is the totality of voluntary movement, produced by skeletal muscles during everyday functioning [4]. Daily physical activity includes exercise. Because daily physical activity is both voluntary and community-based, the additional descriptor of "free-living" daily activity is often used. Tudor-Locke conceptualizes physical inactivity as a human behavior characterized by lack of participation in vigorous activities and minimal physical movement [5]. Persons who experience daily, incapacitating dyspnea due to chronic pulmonary illness fall readily into this group.

According to Webster's Dictionary [6], motion is defined as the act of moving the body or any of its parts; motion sensing is therefore the measurement of movement of the body, or in selected instances, depending on the location of the device, the movement of a body part, such as the arm or leg. While devices that measure motion of the body in toto or in one of its parts would seem to have strong face validity for characterizing daily physical activity measurement, this issue is less than clear, particularly in persons with very low levels of activity. For example, the issue of sensing extraneous motion that is not associated with voluntary movement and energy expenditure (arm movement without movement of the body, pendulous abdomen movement, and movement associated with car trips) may be responsible for considerable error variance.

MEASURING PHYSICAL ACTIVITY WITH MOTION SENSORS
Overview

Motion sensors in current use include pedometers and accelerometers. These devices may be used for purposes of surveillance, clinical, research, and program evaluation [7].

Surveillance

Motion sensors have been used to characterize population-based activity levels for the purposes of monitoring national physical activity levels and evaluating the attainment of physical activity recommendations, both at an individual and a population level [8,9].

Clinical Settings

Clinical uses of motion sensors include measurement of the processes and outcomes of programs in which exercise enhancement and increased daily activity are variables of interest [10]. Much work has been conducted with pedometers as a means to motivate clinical groups to exercise, including people with diabetes, obesity, and congestive heart failure [11-13].

Research and Program Evaluation

Apart from the research substantiating the validity, reliability, and stability of these devices in specific groups and settings, motion sensors have been used to measure adherence to experimental exercise protocols and relationships between free-living physical activity and other key variables, such as functional capacity, self-efficacy for walking, and health status [14,15]. Accelerometers are particularly useful in providing objective feedback of ambulatory activity (dose quantification) to investigators and to study participants in exercise adherence research. This is especially important for pulmonary patients, for whom precise quantification of walking during daily living is essential because small improvements due to effective treatment can often produce large gains in overall functioning. Motion sensor technology may also be used to evaluate and improve the quality of rehabilitation programs and program changes.

Motion Sensor Methodology

Traditional methods for measuring daily, free-living physical activity are imprecise and suffer from a number of problems. For example, methods that rely on self-report of activity and exercise, such as diaries and questionnaires, are both time-consuming and unreliable, especially for the elderly because they depend on memory [16,17]. Direct observation is time-intensive and intrusive. Other more reliable methods, such as radioisotope techniques using doubly labeled water, are technologically complex and expensive [18].

A wide array of motion sensors exists that has the potential to more precisely measure free-living daily physical activity in rehabilitation and other settings. The Table contains an overview of the types of devices, ranging from the simplest (least complex) to the most complex, with a

Table. 
Comparison of activity monitors available in United States.
Type
Brand/
Manufacturers/
Price
Characteristics
and Features
Physical
Placement
of Device
Strengths
Limitations
Populations
Used in
Validation
Studies
Pedometer/
Step Counter
Yamax Digiwalker
(Yamax Inc., Tokyo,
Japan; New Lifestyles, Inc., Kansas City, MO) (most often used in research)
$20-$30
Many other brands,
including Freestyle
Pacer, Eddie Bauer,
and Accusplit
$19-$30
Measures vertical
accelerations at hip
to count steps taken.
Smaller than a pager,
extremely light.
LCD screen display.
4 models with variable
programmable functions:
steps, distance, calories, time.
Uses photo/electronic battery
with life up to 3 years.
Has safety strap to prevent loss.
Waist
Displays cumulative
data continuously.
Useful as
a motivational tool.
Easy to use
and unobtrusive.
Least cost of any
option.
Good measure
of walking activity.
Must remain
vertical.
Wearer must record
output if daily activity data required.
Healthy adults
 
 
 
 
 
 
 
 
StepWatch
(Prosthetics Research Study, Seattle, WA)
$3300 for monitor,
computer interface
dock, and
communication
software
Measures step counts via
a custom accelerometer
with programmable filtering
parameters adjusted for
cadence and motion.
Requires Mac computer,
reader interface unit, and
proprietary software.
Pager-sized.
Ankle
Displays walking
activity as time series. Allows long-term
continuous recording
of ambulatory function.
Expensive.
Adults with
amputations
Adults with chronic conditions affecting mobility
 
 
 
 
 
 
 
Uniaxial
Accelerometers
Caltrac
(Muscle Dynamics,
Torrance, CA,
$70-$90
Measures vertical
accelerations.
Pager-sized.
LCD screen display
with updates every 2 min.
Energy expenditure estimated
by entering age, height,
weight, and gender of wearer.
Programmable modes for
cycling and weight lifting.
Runs on two AAA batteries.
Waist
Displays cumulative
data continuously.
Useful as
a motivational tool.
Low cost.
No time-series data,
cannot show patterns
of activity.
Wearer must record
output if daily activity data required.
Healthy adults
Older adults
Children
 
 
 
 
 
 
 
 
Actigraph
(formerly CSA
Actigraph)
(MTI Health Services,
Fort Walton Beach, FL)
$1500 for monitor,
interface unit,
and software
Measures vertical
accelerations.
Analog filters reject
frequencies outside range
of normal human movement.
Slightly smaller than pager.
Programmable; requires PC,
reader interface unit, and
proprietary software.
Memory up to 256 k. Data
collection up to 22 days.
Uses coin cell battery.
Mainly used in research.
Waist
Wrist
Ankle
Collects time-series data; shows activity
patterns.
Output can be either
activity counts
or step counts.
Count ranges for
light, moderate, hard
and very hard have
been established.
Calibration device
available.
Water resistant.
Discriminates
change in speed
but not grade.
Higher cost.
No feedback
to wearer.
Healthy adults
Adults who use wheelchairs
Type
Brand/
Manufacturers/
Price
Characteristics
and Features
Physical
Placement
of Device
Strengths
Limitations
Populations
Used in
Validation
Studies
Multiaxial
Accelerometers
RT3 Triaxial
Research Tracker
(replaced Tritrac)
(StayHealthy, Inc.,
Monrovia, CA)
$500 for monitor and
docking station
CTI Personal
Calorie Tracker
(available for
personal/clinical use)
$150
Measures 3 planes (vertical,
horizontal, and sagittal);
records as vector magnitude
units.
Pager size.
Requires PC, docking station,
and proprietary software that is
downloadable through web site.
Data collection up to 21 days.
Reports activity units
and energy expenditure.
Has event marker.
Uses two AAA batteries.
Waist
Sensitive to low
levels of activity.
Reflects intensity
& frequency of
activity.
Collects time-series
data; shows activity
patterns.
Output available as
x, y, z axis plots, as
well as a triaxial
vector plot over time.
Moderate cost.
Possible vibration
artifact.
No feedback
to wearer.
No studies
using RT3.
Tritrac:
Young adults
Older adults
Adults with
multiple sclerosis
Adults with COPD
Children
 
 
 
 
 
 
 
 
Mini-Motion Logger
Actigraphs
(Ambulatory Monitoring, Ardsley, NY)
$500-$2000/unit
+$1200-$2600 for
interface unit
and software
Measures 3 planes.
Analog filters reject
frequencies outside range
of normal human movement.
Multiple models available
from micro-mini (wristwatch
size, less than 1 oz) to
basic-mini (4¥3 cm, 1.7 oz).
Light sensor available.
Multiple programmable
parameters. Requires PC,
reader interface unit
and proprietary software.
Memory size 32 to 128 k.
Software programs for motor activity, sleep, and circadian rhythms.
Data collection 16-30 days,
depending on model.
Lithium battery.
Wrist
Sensitive to low
activity levels.
Collects time-series
data; shows activity
patterns.
Validated sleep
estimation algorithm.
Wrist placement is
convenient and familiar.
Possible vibration
artifact.
Expensive.
No feedback
to wearer.
Healthy adults
Women following coronary bypass
surgery
Older adults
 
 
 
 
 
 
 
 
Actiwatch
(MiniMitter Company, Inc., Bend, OR)
$1075 per unit + $1850
for reader and software
Measures 3 planes
("omnidirectional").
Watch size, 17 g wt.
Programmable epoch length.
Requires PC, reader interface unit, and proprietary software.
Memory size 16 to 64 k.
Software programs for motor
activity, sleep, and circadian
rhythms. Downloaded data can
be displayed both graphically
as actograms and numerically
as activity counts.
Data collection up to 44 days.
Lithium battery.
Wrist
Sensitive to low
activity levels.
Collects time-series
data, shows activity
patterns.
Validated sleep
estimation algorithm.
Very small and light.
Wrist placement is
convenient and familiar.
Waterproof.
Possible vibration
artifact.
Expensive.
No feedback
to wearer.
Adults with
Alzheimer's disease
Adults with cancer
Children and infants

similar continuum from the least to the most expensive. They also vary on continuums of sensitivity to motion and degree of information available to participants. Selection of a motion sensor requires consideration of the strengths and features of the motion sensing device and the amount and type of data required. Practical issues include cost of the device, comfort and ease of wearing the device, and the need for computers or other accessories.

Reliability and validity of physical activity monitors are specific to the device, the population, and the activity behavior being studied. Accuracy/precision depends on how the device is constructed, as well as how it is used. Concurrent validity is most often established by assessing the degree of correlation with other activity measures (calorimetry, self-report, observation) or with indicators of known outcomes of activity (fitness, functional capac-ity, heart rate, VO2max). Characteristics of the population under study may affect the accuracy of motion sensors. For example, older adults with limited mobility may move so slowly that the motion is not detected by the sensor. Finally, the specific activity behaviors of the individuals being monitored will affect the validity of activity measurement. Energy expenditure during static work (work done without movement) will not be measured by motion sensing technology.

As a measure of steps taken, electronic pedometers have demonstrated reasonable validity and high reliability. All pedometers tend to underestimate distance or steps for very slow walking [19,20]. This inaccuracy results from vertical movements at the hip being less pronounced at slow speeds and the sensor commonly failing to register some of them. A comparison of the accuracy of five electronic pedometers (Freestyle, Pacer, Eddie Bauer, Yamax, and Accusplit) for measuring distance walked found significant differences among models; the Yamax, Pacer, and Accusplit demonstrated the greatest accuracy [19]. The effects of walking speed were also examined, and the Yamax was found to be significantly more accurate than the Pacer and Eddie Bauer models at slow to moderate speeds. No significant differences were found at the fastest speed. We also assessed inter-unit reliability and found only the Yamax to be consistent between units. Other investigators have found similar variability among units due to differences in spring tension [20]. Step counts measured by the Yamax pedometer correlated only modestly with self-reported energy expenditure (r = 0.34-0.49) [20].

Studies exploring the validity of both uniaxial and multiaxial accelerometers as a measure of energy expenditure have substantiated significant correlations between the two (0.66-0.96) [20-22]. A major issue in the use of accelerometry for physical activity measurement is that the unit of measure (activity count, or vector magnitude units [VMU]) is not standardized, and no direct translation into energy expended exists. Several of the instruments include programs based on regression equations to calculate caloric expenditure; but differences in the accuracy of the calibration equations, rather than differences in the monitors themselves, have been shown to contribute to differences in recorded energy expenditure [23]. For research purposes, we recommend that motion sensor data be analyzed as counts [20].

Because uniaxial sensors track motion in the vertical plane only, they are not accurate for activities with static trunk movement, such as cycling and rowing [20]. The specific activities being performed also affect the accuracy of accelerometry for measuring energy expenditure. For example, the Tritrac accelerometer has been shown to overestimate the energy expenditure of walking and jogging and to underestimate the energy expenditure of stair climbing, stationary cycling, and arm ergometry [24]. Similarly, another study comparing three accelerometers and a pedometer for prediction of energy expenditure during moderate intensity activity suggested that all four motion sensing devices overpredicted energy expenditure during walking, but underpredicted energy expenditure in activities that included arm movement and static work [25].

A major advantage of a triaxial sensor over a uniaxial sensor is that the instrument is more sensitive to light activities, such as slow walking. A disadvantage, however, related to this greater sensitivity is that the device also becomes sensitive to vibrational artifact, recording background vibration (especially that related to being in a vehicle) as movement. Some manufacturers claim to set the device at a frequency response capable of capturing the range of human movement but to filter out rapid vibrations. These claims require researcher evaluation and can be tested by determining if measures obtained during vehicular transportation as a passenger differ significantly from measures obtained during quiet sitting [26].

Accelerometers have been shown to be more sensitive in detecting activity differences in inactive populations and more sensitive at detecting short activity periods than recall measures [14,27-29]. Field evaluation studies comparing accelerometers to self-reported activity have usually suggested that accelerometers underestimate the amount of vigorous activity and energy expended in activity [30,31]. However, it is well known that self-report of activity is subject to recall bias. Multiple assessment devices (accelerometers and self-report measures) can be used together to improve the accuracy of activity profiles [25,32].

Like waist-mounted triaxial accelerometers, wrist-worn accelerometers have been shown to differentiate light, moderate, and heavy activity levels, as well as to record differences in sedentary activities. They can more accurately represent activities that are underrated by waist-mounted devices, such as cycling and rowing, but may overrate activities requiring rapid hand movements (especially typing) [33].

Several recent studies have compared the relative accuracy and validity of various motion sensing devices. Tudor-Locke and colleagues [34] compared CSA uniaxial accelerometer counts to Yamax pedometer steps and found the two to be highly correlated (r = 0.74-0.86), but because of the CSA's greater sensitivity, it tended to record more steps than the Yamax pedometer. Leenders et al. [31] compared a pedometer (Yamax) with a uniaxial (CSA) and a triaxial (RT3) accelerometer and found all three to be highly correlated (r = 0.84-0.93), with the highest correlations between the two accelerometers. The authors concluded that the accelerometers were comparable for assessing the amount and intensity of activity. They also suggested that the high correlations between the accelerometers and step count indicated that a large portion of physical activity was determined simply by measuring the number of steps taken each day.

Welk and colleagues [23] evaluated the absolute and relative validity of three accelerometers (CSA, Tritrac, and Biotrainer) during choreographed lifestyle activities and treadmill activity. Correlations among the three devices were high for both treadmill (r = 0.86) and lifestyle activities (r = 0.70). Kochersberger et al. [35] similarly compared the Tritrac and Mini-Motion Logger Actigraph. Again, correlations between the two were high and significant (r = 0.77, p = 0.001).

Individual physical activity is known to vary based on day of the week due to differences in work and leisure activity profiles. This variability may be less important in those who do not work a typical work week. Matthews et al. [36] measured day of the week effects in a large sample of healthy adults using an accelerometer (CSA) and found that, to guarantee 80 percent reliability for measuring activity counts and time spent in moderate to vigorous activity, at least 3 to 4 days of monitoring were required, with at least 1 weekend day included. To reliably measure inactivity, these same authors recommend at least 7 days of monitoring.

MOTION SENSORS IN PULMONARY REHABILITATION

For most adults, the health benefits of an exercise program are believed to be dependent on the dose of physical activity undertaken. Dose is characterized by the frequency, duration, intensity, and type of activity. Although less is known about specific dose-response relationships, a positive association between training intensity and maximal oxygen consumption, muscle strength, and other exercise outcomes has been reported in the literature [37,38]. Motion sensors have the potential to document more accurately the actual dose of a prescribed exercise regimen. They are especially useful because in most outpatient or home-based programs, only a small part of the program is supervised by program staff, who assess and record the quantity of exercise and the responses of participants. Since there are greater risks of injury and other health emergencies associated with higher intensity exercise, it is important to identify the optimal level of activity that produces the greatest health and immediate outcomes benefits while minimizing potential risks [39].

More recently, digital pedometers and accelerometers have shown promise as adjuncts to reinforce exercise adherence by allowing self-monitoring, and as process and outcome measures for physical conditioning programs [23,28,35,40]. A number of studies support the validity, reliability, feasibility and clinical utility of accelerometers in the care of people with chronic respiratory disease. Preusser and Winningham first used a Caltrac accelerometer to measure four days of home activity in a group of 17 patients with chronic obstructive pulmonary disease (COPD) and found significant correlations with inspiratory muscle strength (r = 0.66; p = 0.004) and inspiratory muscle endurance (r = 0.71; p = 0.001) [41]. In a study of 47 outpatients who had COPD as they entered a pulmonary rehabilitation program, we determined that a triaxial accelerometer, the Tritrac R3D, had excellent test-retest reliability during three standardized 6 min walk tests (intraclass correlation coefficient, rICC = 0.84). Pearson correlations between accelerometer-measured movement during walking compared to walking distance varied from 0.84 to 0.95. During measurement periods of free-living activity over three days at home, the device had an rICC of 0.69, implying good stability over longer periods of meas-urement. These data support the use of this device in meas-uring daily activity, even in very inactive populations [14].

To determine concurrent validity of accelerometer measurement of daily activity under free-living conditions, we measured daily activity over three full days at home in 63 outpatients with COPD who did not exercise regularly. We found significant correlations between accelerometer-measured daily activity and exercise capacity (maximal 6 min walk, r = 0.60; p < 0.001), level of obstructive pulmonary disease (percentage of predicted forced expiratory volume in 1 s, r = 0.37; p < 0.01), dyspnea (Functional Status and Dyspnea Questionnaire, r = -0.29; p < 0.05) and activity self-efficacy (Activity Self-Efficacy Questionnaire, r = 0.27; p < 0.05) and physical health status (SF-36, Physical Functioning Subscale, r = 0.40; p < 0.01). Multivariate analysis demonstrated that the only predictor of physical activity was the 6 min walk test. Like other studies, physical activity measured by accelerometer was not associated in this sample with self-report of functional status [15]. These findings indicate that the accelerometer has good concurrent validity with other indicators of function and above all, represents with precision walking behavior in chronic respiratory disease.

The greater accuracy of accelerometer technology has excellent potential in documenting pulmonary rehabilitation outcomes [2]. Walking is a key indicator of improvement in pulmonary rehabilitation programs and similar regimens aimed at reducing the consequences of physical inactivity. Although these devices poorly represent strength training activities that do not involve much bodily movement, they can capture improvements in endurance and exercise capacity. We have found that the Tritrac R3D is a sensitive process measure of physical activity undertaken during a pulmonary rehabilitation program. In a study of 41 men and women with COPD enrolled in an 8 week outpatient pulmonary rehabilitation program, subjects underwent 5 days of daily activity measurement using the Tritrac R3D accelerometer. Physical activity, measured in vector magnitude units (VMU) per minute, was 87.4 + 38.8 (mean, standard deviation) before the start of the program; VMU per minute increased to 115.2, standard deviation + 59.4, during the final weeks of the program (p < 0.01) [10]. Figure 1 demonstrates differences between a patient's daily physical activity measured in VMU before starting a pulmonary rehabilitation program and on the same days of the week while attending the program. Significant differences are evident between exercise and nonexercise days.

Figure 1. Average daily activity (VMU/min) for 2 days before beginning pulmonary rehabilitation program .

In addition to providing information about group performance, accelerometer measurement of daily physical activity provides a more precise indicator of individual performance during unsupervised exercise than previously possible. Figure 2 includes individual tracings of free-living daily activity of one pulmonary rehabilitation participant and clearly differentiates walking behavior during his two 6 min walk tests, sitting activity, and free-living walking to his vehicle, as well as the movement contributed by riding home in his car. Physical activity measured during an observed pulmonary rehabilitation exercise session is represented in Figure 3 and includes VMU measurement while patients used the treadmill, a seated stepper (NuStep®), an arm ergometer, and free weights. Static exercise (free weights, seated stepper, and arm ergometer) is generally less well represented than moderately paced walking on the treadmill. Figure 4 reflects two full days of activity accelerometer monitoring of one participant, comparing a day when he attended his pulmonary rehabilitation program with a day when he did not attend. The graphical display demonstrates more activity (in VMU) on the day he attended the exercise session.

Figure 2. Vector magnitude unit (VMU) measurements of minute-by-minute tracings for pulmonary rehabilitation program patient.
Figure 3. Vector magnitude unit (VMU) measurements of minute-by-minute tracings for patient in pulmonary rehabilitation program.
Figure 4. Full day vector magnitude unit (VMU) measurements.
CHOOSING A MOTION SENSOR FOR
PULMONARY REHABILITATION

Accelerometer technology, with its relatively complex data collection and management methodology to study daily activity, begs the question of why a digital pedometer would not suffice for this purpose. Certainly, in most circumstances with healthy individuals, step-counting will readily document walking activity and, as noted earlier, is useful as an adjunct to reinforce walking behavior and other exercise. A number of reasons support the use of accelerometer technology for persons with chronic pulmonary disease and others with diminished physical functioning. First, even the most reliable digital pedometers are unable to accurately measure slow walking speeds under 2 mph, because of less-pronounced accelerations at the hip [19]. For this reason, pedometers will underrepresent much of the walking activity of these groups. For the same reason, pedometers will be less sensitive to small improvements in walking activity in inactive individuals. Field-monitoring of exercise and daily activity over time is also better undertaken with an accelerometer because of its longer memory and its lack of dependence on having subjects record a daily value or otherwise keep a separate record of the pedometer data. This may be particularly true in persons with chronic pulmonary disease because of their high, preexisting burden of self-management and other considerations relating to their severe, chronic illness. Finally, the capability to document both the duration and relative intensity of movement, e.g., during walking, provides a unique quantitative index of the vigor with which exercise or activity is undertaken. Process and outcome measures that include these fundamental exercise elements can provide much more precise and valuable information to both program staff and participants.

DATA ANALYSIS AND PRACTICAL
CONSIDERATIONS OF USING MOTION
SENSORS IN PULMONARY PATIENTS
Data Analysis Concerns with Pedometers

Step counters or pedometers are simple, lightweight, and inexpensive devices, useful in the documentation of ambulatory activities. They rely on the patient to simply wear it and record the daily steps. Because of the necessity of gravity's action on the mechanical arm to record the step, the physical orientation is key to accurate counting. We have found in more obese patients that "abdominal overlap" prohibits this vertical orientation when attached at the waist, thus reducing the number of steps. Also, step counters, while correlating moderately well with accelerometer information, are unable to distinguish the intensity and the temporal distribution of activity. Finally, some devices, such as the Digiwalker®, will not accurately measure steps in individuals who walk under 2 mph, a major concern in severely functionally impaired patients.

Data Analysis Concerns with Accelerometers

Our present work involves the use of the StayHealthy RT3® triaxial accelerometer (Figure 5). We measure activity over 1 min epochs so as to capture brief bouts of activity over longer periods of time (up to 21 days). The intensity of the VMU correlates positively with the velocity of walking, allowing us to examine various levels of activity during daily living, as well as during "exercise activities." This accelerometer will provide a large amount of data in the form of VMU "counts," a unitless index of movement (Tritrac R3D® and StayHealthy RT3®). For most uses, it is necessary to analyze these data to come up with one or a few summary measures that characterize the subjects' activity level over the days.

Figure 5. RT3® accelerometer from StayHealthy®, next to AAA battery to demonstrate scale.

Figure 5.
RT3® accelerometer from StayHealthy®, next to AAA battery to demonstrate scale. Buttons are event markers, useful during analysis for matching vector magnitude units with diary entries.

There are several issues relating to such analyses, which are discussed in this section.

Sources of Extraneous Variance
Physical Activity Versus Noise Signals

As in Figure 2, there is quite a bit of minute-to-minute variability in VMU, with frequent upward spikes punctuating a more low-level base that also varies. Some of this variability represents the type of actual activity that we wish to detect, such as a person getting up off the couch to walk around the house. This activity is easiest to see when it is for a sustained period, as demonstrated by the 6 min walk in Figure 2. In our experience with pulmonary rehabilitation patients, it is relatively easy to detect long periods of activity, as are seen when a person is intentionally engaging in exercise.

However, some of the variability in VMU may not actually be major activity, but rather something like a person shifting position on a couch, or just noise in the detection device. In our experience with the RT3, many of the devices record randomly varying, relatively low but nonzero values for VMU when the device is sitting on a table at night while the subject sleeps. For this very inactive population, this type of low-level random noise can be a problem when one is trying to summarize activity level during the day. This may reduce the reliability of activity monitors in very inactive populations.

Variability Between Devices

In our experience with RT3 devices, there is considerable variability from device to device in the VMU readings when the device is sitting on a table. What is not clear is whether these differences in sensitivity across devices represent an additive or a multiplicative effect. If additive (e.g., one device reads 10 VMU higher than another at any given level of activity), it would have some impact at very low levels of activity, but not at high levels. If multiplicative (e.g., one device reads 50% higher than another for any given level of actual movement), then this has serious implications for across-device comparability. This issue needs further investigation. Meanwhile, in our studies we assign the same device for the pre, post, and followup data collection periods for a given subject.

Car Rides

Another source of noise may be car rides and use of rocking chairs, wheelchairs, and other equipment contributing to movement without skeletal muscle activity. The motions and vibrations of a car cause fairly high VMU readings, when the person is not, in fact, being active. Bouten [18,40] and his investigators deleted data over a certain magnitude which they attributed to automotive transport, but this approach is time-consuming to develop and probably imprecise. We are currently investigating how to deal with this problem. It is probably not practical to identify each car episode and excise these segments from the data. However, if it is known for each individual how many minutes were spent in a car each day, and if we know the average VMU when riding in a car compared to sitting in a chair, then an adjustment can be made. In past studies, we found that there was no significant difference between time spent riding in a car between measurement periods, suggesting that people with chronic lung disease will have similar driving habits over the period of analysis, thus obviating the need to arduously remove the car noise from every driving episode.

Summary Measure Issues

Freedson and Miller [20] have recommended that the summary measure for accelerometers be movement "counts" rather than energy expenditure, which is computed from varied, often undefined formulae, such as the Harris-Benedict equation, which may not be readily applicable to every measurement setting. Nonetheless, temporal factors may add error variance. For example, the devices are usually not worn while sleeping. Therefore, it is important to identify the time period when it is being worn. Data from a 24 hour period can be summarized by either adding up the total VMU over the period in which the device is worn, or taking the average over this period (mean VMU per minute or per hour). The number of hours per day that the device is worn varies across people, so these two methods of summarizing will not be perfectly correlated. For example, on days during which the device was worn only a short period (e.g., 9 hours) because the person forgot to put it on when they first got out of bed, average VMU per minute may be an acceptable measure of daily activity. On the other hand, if someone has only 9 hours of data because they were in bed for 15 hours, then average VMU while it was being worn misses the fact that this person was, in reality, very inactive for most of the day. In this case, total VMU for the day may be a better measure.

Rather than treating VMU as a continuous variable, it may be desirable to define a threshold and compute how often VMU was above this threshold. For example, a threshold of 500 VMU might be good for indicating periods of fairly high activity, so the number of minutes (or percentage of time) during which activity is above 500 may be a good summary measure of activity.

Within and Between Subject Variability

When attached at the waist, the accelerometer measures the various motions of the pelvis during gait, including lateral displacement of the center of gravity, pelvic rotation, and lateral pelvic tilt or drop. These movements vary between individuals and within individuals at different walking speeds, and must be considered when analyzing accelerometer data. While this presents some issues in inter-individual comparisons, the recordings are usually internally consistent within individuals.

Another issue is that there may be differences across people that affect the sensitivity of the device. Such differences may be due to exactly where the device is worn, obesity, or gait differences. This possibility raises the issue of whether VMU values can be interpreted in an absolute sense or need to be somehow normalized to a person. For example, in our study we have VMU meas-ures during a 6 min walk, in which the subject is supposedly walking as fast as possible. We are exploring the use of person-specific thresholds to compute summary measures, e.g., the percentage of time with activity greater than 30 percent of that measured during the 6 min walk. Alternatively, VMU values could be normed to the mean value observed while the person was sitting in a chair filling out a questionnaire, e.g., percentage of time with VMU values more than 3 times as high as this baseline. Finally, handling, bumping, or dropping the unit is an additional source of error variance. To minimize this problem, subjects should be instructed not to handle the devices and to wear the devices in a protective holder that minimizes the possibility of dropping it.

Other practical considerations for the clinical usage of accelerometers include consideration of the participant's cognitive level, certainly necessary for remembering to put the unit on and recording possible interfering activities in a diary. In addition, care in application (more lateral placement often results in breakage of the holder, especially against arms of chairs), and using fresh, fully charged batteries at the beginning of the recording period are also important. The devices require only slight care, although dropping can dislodge the battery, resulting in loss of data.

Underrepresentation of Exercise Behavior

The inability of motion sensing devices to measure static exercise has been discussed previously and remains a major limitation to the use of these devices in pulmonary rehabilitation and other settings, where strength training and other exercise not employing body movement are used. Although walking will usually be accurately represented, accelerometers will not reflect the added work of walking up stairs and inclines. Additionally, motion sensors may not be able to accurately monitor frail older adults with slow gaits [42] or provide similar estimates of the time spent in resting/light, moderate, or hard/very hard physical activity [43].

CURRENT AND FUTURE RESEARCH

Our present research uses accelerometer technology to measure the efficacy of an exercise adherence intervention following a pulmonary rehabilitation program to promote daily physical activity and exercise following program completion. Clearly, as noted earlier, a number of methodological studies need to be undertaken to determine the best ways to limit the signal-to-noise ratio of the devices, particularly when they are used for measuring free-living activity. Motion sensors, particularly accelerometers, have considerable potential for answering questions related to pulmonary rehabilitation. For example, are participants carrying out exercise strategies at home with the same vigor as during supervised exercise sessions? Likewise, is it possible that frail pulmonary rehabilitation participants are reducing daily activity to accommodate their exercise program and are therefore doing less than before? To what degree are weight loss and other negative outcomes associated with a more strenuous program of walking in some individuals?

CONCLUSIONS

The purpose of this article is to provide an overview of the potential utility of motion sensors to measure physical activity in persons with chronic pulmonary disease in the setting of pulmonary rehabilitation. In particular, motion sensors are useful to measure walking behavior, a key outcome of these programs. Walking is also a core activity, fundamental to maintenance of an active and vigorous life in the face of normal aging, as well as the debilitating effects of chronic pulmonary disease. Clearly, the potential use of these devices to measure free-living daily activity, a variable heretofore immeasurable, transcends this limited area to include most domains of rehabilitation care wherein improving bodily movement is a central outcome of interest. It is likely that the future will include broad application of this methodology to clinical, research, and program evaluation venues.

REFERENCES
1. American Thoracic Society. Pulmonary rehabilitation. Am J Respir Crit Care Med 1999;159:1666-82.
2. Members of the British Thoracic Society Standards of Care Subcommittee on Pulmonary Rehabilitation; MDL Morgan, Chairperson. Pulmonary rehabilitation. Thorax 2001; 56:827-34.
3. Westerterp KR. Physical activity assessment with accelerometers. Int J Obes Relat Metab Disord 1999;23 Suppl 3: S45-49.
4. Casperson CJ, Powell KE, Christianson GM. Physical activity, exercise, and physical fitness: definitions and distinctions for health-related research. Public Health Rep 1985;100:126-31.
5. Tudor-Locke CE, Myers NM. Challenges and opportunities for measuring physical activity in sedentary adults. Sports Med 2001;31:91-100.
6. Webster's New World Dictionary and Thesaurus. New York: MacMillan; 1996.
7. Tudor-Locke CE, Myers AM. Methodological considerations for researchers and practitioners using pedometers to measure physical (ambulatory) activity. Res Q Exerc Sport 2001;72:1-12.
8. Schulz Y, Weinsier S, Terrier P, Durrer D. A new accelerometric method to assess the daily walking practice. Int J Obes 2002;26:111-18.
9. Trost SG, Pate RR, Sallis JF, Freedson PS, Taylor WC, Dowda M, et al. Age and gender differences in objectively measured physical activity in youth. Med Sci Sports Exerc 2002;34:350-55.
10. Steele BG, Belza B, Hunziker J, Holt L, Legro M, Coppersmith J, et al. Monitoring daily activity during pulmonary rehabilitation using a triaxial accelerometer. J Cardiopulm Rehabil. In press 2003.
11. Yamanouchi K, Shinozake K, Chikada K. Daily walking combined with diet therapy is a useful means for obese NIDDM patients not only to reduce body weight but also to improve insulin sensitivity. Diabetes Care 1995;18:775-78.
12. Fogelholm M, Kukkonen-Harjula K, Oja P. Eating control and physical activity as determinants of short-term weight maintenance after a very low calorie diet among obese women. Int J Obes Relat Metab Disord 1999;12:203-10.
13. Hoodless DJ, Stainer K, Savic N, Batin P, Hawkins M, Cowley AJ. Reduced customary activity in chronic heart failure: assessment with a new shoe-mounted pedometer. Int J Cardiol 1994;43:39-42.
14. Steele BG, Holt L, Belza B, Ferris S, Lakshminaryan S, Buchner DM. Quantitating physical activity in COPD using a triaxial accelerometer. Chest 2000;117:1359-67.
15. Belza B, Steele BG, Hunziker J, Lakshminaryan S, Holt L, Buchner DM. Correlates of physical activity in chronic obstructive pulmonary disease. Nurs Res 2001; 50:195-202.
16. Pols MA, Peeters PH, Kemper HC, Collette HJ. Repeatability and relative validity of two physical activity questionnaires in elderly women. Med Sci Sports Exerc 1996;28:1020-25.
17. Epstein LH, Paluch RA, Coleman KJ, Vito D, Anderson K. Determinants of physical activity in obese children assessed by accelerometer and self-report. Med Sci Sports Exerc 1996;28:1157-64.
18. Bouten CV, Verboeket-van de Venne WP, Westerterp KR, Verduin M, Janssen JD. Daily activity assessment: com-parison between movement registration and doubly labeled water. J Appl Physiol 1996;81:1019-26.
19. Bassett DR, Ainsworth BE, Leggett SR, Mathien CA, Main JA, Hunter DC, Duncan GE. Accuracy of five electronic pedometers for measuring distance walked. Med Sci Sports Exerc 1996;28:1071-1107.
20. Freedson PS, Miller K. Objective monitoring of physical activity using motion sensors and heart rate. Res Q Exerc Sport 2000;71 Suppl 2:S21-29.
21. Montoye HJ. Introduction: evaluation of some measurements of physical activity and energy expenditure. Med Sci Sports Exerc 2000;39:S439-41.
22. Sherman WM, Morris DM, Kirby TE, Petosa RA, Smith BA, Frid DJ, et al. Evaluation of a commercial accelerometer (Tritrac R3D) to measure energy expenditure during ambulation. Int J Sports Med 1998;19(1):43-47.
23. Welk GJ, Differding JA, Thompson RW, Blair SN, Dziura J, Hart P. The utility of the Digi-Walker step counter to assess daily physical activity patterns. Med Sci Sports Exerc 2000;32S:S481-88.
24. Campbell KL, Crocker PR, McKenzie DC. Field evaluation of energy expenditure in women using Tritrac accelerometers. Med Sci Sports Exerc 2002;34(10):1667-74.
25. Bassett DR Jr, Ainsworth BE, Swartz AM, Strath SJ, O'Brien WL, King GA. Validity of four motion sensors in measuring moderate intensity physical activity Med Sci Sports Exerc 2000;32 Suppl 9:S471-80.
26. Patterson SM, Krantz DS, Montgomery LC, Deuster PA, Hedges SM, Nebel LE. Automated physical activity monitoring: validation and comparison with physiological and self-report measures. Psychophysiology 1993;30:296-305.
27. Ng AV, Kent-Braun JA. Quantitation of lower physical activity in persons with multiple sclerosis. Med Sci Sports Exerc 2000;29(4):517-23.
28. Matthews CE, Freedson PS. Field trial of a three-dimensional activity monitor comparison with self-report. Med Sci Sports Exerc 1995;27:1071-78.
29. McDermott MM, Liu K, O'Brien E, Guralnik JM, Criqui MH, Martin GJ, et al. Measuring physical activity in peripheral arterial disease: a comparison of two physical activity questionnaires with an accelerometer. Angiol 2000:51(2):91-100.
30. Sirard JR, Melanson EL, Li L, Freedson PS. Field evaluation of the Computer Science and Applications, Inc., physical activity monitor. Med Sci Sports Exerc 2000;32(3): 695-700.
31. Leenders NY, Sherman WM, Nagaraja HN, Kien CL. Evaluation of methods to assess physical activity in free-living conditions. Med Sci Sports Exerc 2001;33(7):1233-40.
32. Allor K, Pivarnik J. Stability and convergent validity of three physical activity assessments. Med Sci Sports Exerc 2001;33:671-76.
33. Sugimoto A, Hara Y, Findley TW, Yoncmoto K. A useful method for measuring daily physical activity by a three-direction monitor. Scand J Rehabil Med 1997;29(1):37-42.
34. Tudor-Locke C, Ainsworth BE, Thompson RW, Matthews CE. Comparison of pedometer and accelerometer measures of free-living physical activity. Med Sci Sports Exerc 2002;34(12):2045-51.
35. Kochersberger GE, McConnell E, Kuchibhatla MN, Pieper C. The reliability, validity, and stability of a measure of physical activity in the elderly. Arch Phys Med Rehabil 1996;77:793-95.
36. Matthews CE, Ainsworth BE, Thompson RW, Bassett DR Jr. Sources of variance in daily physical activity levels as measured by an accelerometer. Med Sci Sports Exerc 2002; 34(8):1376-81.
37. Serres I, Varray A, Vallet G, Micallef JP, Prefaut C. Improved skeletal muscle performance after individualized exercise training in patients with chronic obstructive pulmonary disease. J Cardiopulm Rehabil 1997;17(4):232-38.
38. Neder JA, Jones PW, Nery LE, Wipp BJ. Determinants of the exercise endurance capacity in patients with chronic obstructive pulmonary disease. Am J Respir Crit Care Med 2000;162:497-504.
39. Haskell WL. What to look for in assessing responsiveness to exercise in a health context. Med Sci Sports Exerc 2000; 33 Suppl:S454-58.
40. Bouten CV, Westerterp KR, Verduin M, Janssen JD. Assessment of energy expenditure for physical activity using a triaxial accelerometer. Med Sci Sports Exerc 1994; 26:1516-23.
41. Preusser B, Winningham M. Validating a direct measure of energy expenditure with daily activity in patients with COPD. Am J Respir Crit Care Med 1995;151:A682.
42. Masurier GC, Tudor-Locke C. Comparison of pedometer and accelerometer accuracy under controlled conditions, Med Sci Sports Exerc 2003;35(5):867-71.
43. Ainsworth BE, Bassett DR Jr, Strath SJ, Swartz AM, O'Brien WL, Thompson RW, et al. Comparison of three methods for measuring the time spent in physical activity. Med Sci Sports Exerc 2000;32 Suppl 9:S457-64.
Submitted for publication January 16, 2003. Accepted in revised form July 7, 2003.

Go to TOP

Go to the Contents of Vol. 40 No. 5