261
ORIGINAL ARTICLE
Step Counting and Energy Expenditure Estimation in Patients
With Chronic Obstructive Pulmonary Disease and Healthy
Elderly: Accuracy of 2 Motion Sensors
Karina C. Furlanetto, PT, Gianna W. Bisca, PT, Nicoli Oldemberg, PT, Thaís J. Sant’Anna, PT,
Fernanda K. Morakami, PT, Carlos A. Camillo, PT, Vinicius Cavalheri, PT, Nidia A. Hernandes, PT,
Vanessa S. Probst, PT, Ercy M. Ramos, PhD, Antonio F. Brunetto, PhD, Fábio Pitta, PhD
ABSTRACT. Furlanetto KC, Bisca GW, Oldemberg N,
Sant’Anna TJ, Morakami FK, Camillo CA, Cavalheri V, Her-
nandes NA, Probst VS, Ramos EM, Brunetto AF, Pitta F. Step
counting and energy expenditure estimation in patients with
chronic obstructive pulmonary disease and healthy elderly: accu-
racy of 2 motion sensors. Arch Phys Med Rehabil 2010;91:261-7.
Objective: To compare the accuracy of 2 motion sensors (a
pedometer and a multisensor) in terms of step counting and
estimation of energy expenditure (EE) in patients with chronic
obstructive pulmonary disease (COPD) and in healthy elderly.
Design: In this descriptive study, all participants wore both
motion sensors while performing a treadmill walking protocol
at 3 different speeds corresponding to 30%, 60%, and 100% of
the average speed achieved during a six-minute walk test. As
criterion methods, EE was estimated by indirect calorimetry,
and steps were registered by videotape.
Setting: Research laboratory at a university hospital.
Participants: Patients with COPD (n⫽30; 17 men; mean
age ⫾ SD, 67⫾8y; mean forced expiratory volume in the first
second [FEV1] predicted ⫾ SD, 46%⫾17%; mean body mass
index [BMI] ⫾ SD, 24⫾4kg·m2) and matched healthy elderly
(n⫽30; 15 men; mean age ⫾ SD, 68⫾7y; mean FEV1 pre-
dicted ⫾ SD, 104%⫾21%; mean BMI ⫾ SD, 25⫾3kg·m2).
Interventions: Not applicable.
Main Outcome Measure: Step counting and EE estimation
during a treadmill walking protocol.
Results: The pedometer was accurate for step counting and
EE estimation in both patients with COPD and healthy elderly
at the higher speed. However, it showed significant underesti-
mation at the 2 slower speeds in both groups. The multisensor
did not detect steps accurately at any speed, although it accu-
rately estimated EE at all speeds in healthy elderly and at the
intermediate and higher speeds in patients with COPD.
Conclusions: In both patients with COPD and healthy el-
derly, the multisensor showed better EE estimates during most
From the Laboratory of Research in Respiratory Physiotherapy, Department of
Physiotherapy, Universidade Estadual de Londrina, Londrina, Paraná (Furlanetto,
Bisca, Oldemberg, Sant’Anna, Morakami, Camillo, Cavalheri, Hernandes, Probst,
Brunetto, Pitta); Postgraduate Program in Physiotherapy, Department of Physiother-
apy, Faculdade de Ciências e Tecnologia, Universidade Estadual Paulista Júlio de
Mesquita Filho, Presidente Prudente, São Paulo (Cavalheri, Hernandes, Ramos, Pitta);
Universidade Norte do Paraná, Londrina, Paraná (Probst), Brazil.
Supported by the National Council for Scientific and Technological Development,
Brazil.
No commercial party having a direct financial interest in the results of the research
supporting this article has or will confer a benefit on the authors or on any organi-
zation with which the authors are associated.
Reprint requests to Fábio Pitta, PhD, Departamento de Fisioterapia, CCS, Hospital
Universitário da UEL, Av Robert Koch, 60, Vila Operária, 86038-440, Londrina, PR,
Brazil, e-mail: fabiopitta@uol.com.br.
0003-9993/10/9102-00496$36.00/0
doi:10.1016/j.apmr.2009.10.024
walking speeds than the pedometer. Conversely, for step count-
ing, accuracy is observed only with the pedometer during the
higher walking speed in both groups.
Key Words: Pulmonary disease, chronic obstructive; Reha-
bilitation.
© 2010 by the American Congress of Rehabilitation
Medicine
HYSICAL ACTIVITY IN daily life can be considered as
P the totality of voluntary movement produced by skeletal
1
muscles during everyday functioning. Its correct quantifica-
tion has become a challenge in order to obtain an adequate
assessment of the relationship between free-living physical
activity and health.2
COPD is characterized by air flow limitation; dyspnea; and
reduced exercise capacity, muscle strength, and quality of life.3
Patients with COPD spend less time walking in daily life than
age-matched subjects and walk at a lower intensity.4 Moreover,
previous studies5-8 have shown that physical inactivity is an
important predictor of hospital readmission and morbidity/
mortality risk in this population. Because of the close relation-
ships among physical inactivity, disability, and mortality, the
interest in objective measurement of daily physical activity in
patients with COPD has gained growing interest.9
Energy expenditure and step counting are common outcomes
when assessing physical activity in daily life. To obtain an accu-
rate assessment of these outcomes, the application of reference
methods is recommended. For energy expenditure assessment, the
literature usually recommends the doubly-labeled water method or
indirect calorimetry assessment.10-13 For step counting, direct ob-
servation and videotaping have been considered as reference
methods.14 However, these techniques are not easily used in
everyday life because of their methodologic complexity, limited
practicality, and/or high cost. Recently, the use of motion sensors
has gained widespread recognition. These are instruments for
detection of body movement. They are used to quantify physical
activity in daily life objectively during a period14 of time. Pedom-
eters (eg, Digiwalker SW701)a and multisensors (eg, SenseWear
Armband)b are among the most used motion sensors. Both of
them quantify steps and estimate total energy expenditure, pro-
viding information from free living conditions and not just infor-
mation derived from laboratory tests. However, the instruments
List of Abbreviations
BMI body mass index
COPD chronic obstructive pulmonary disease
FEV1 forced expiratory volume in the first second
MET metabolic equivalent
6MWT six-minute walk test
Arch Phys Med Rehabil Vol 91, February 2010
262 MOTION SENSORS IN CHRONIC OBSTRUCTIVE PULMONARY DISEASE, Furlanetto
present marked differences concerning technologic complexity
and cost. The SenseWear is more costly because it is composed of
a biaxial accelerometer and diverse physiologic sensors, in con-
trast with the lower cost and technologic simplicity of the pedom-
eter, which involves a simple system for detection of the hip
vertical movement as a step.13-16
Schneider et al15 showed that the Digiwalker SW701 pedo-
meter is more accurate for step counting in healthy adults than
a diversity of other devices. The SenseWear, on the other hand,
was compared with uniaxial, biaxial, and triaxial accelerome-
ters and showed the most accurate total energy expenditure
estimation at most treadmill speeds.17 However, there are no
previous studies comparing the accuracy of a simple instrument
such as the Digiwalker SW701 pedometer and a technologi-
cally advanced multisensor such as the SenseWear concerning
step counting and energy expenditure estimation during phys-
ical activity, especially in subjects characterized by inactivity
and slow walking such as patients with COPD. Such informa-
tion could help inform decision-making when choosing an
activity monitoring device for this patient group.
Because of the increasing need for objective techniques which
accurately detect physical activity, the objective of this study was
to investigate the accuracy of the Digiwalker SW701 pedometer
and the SenseWear multisensor in estimating energy expenditure
and counting steps compared with reference methods and with
each other in patients with COPD and healthy elderly subjects.
METHODS
Participants
The study involved 43 patients with COPD from the Outpatient
Respiratory Physiotherapy Clinic at the Hospital Universitário de
Londrina (Brazil) and 39 healthy elderly subjects who were rela-
tives or acquaintances of students from the aforementioned uni-
versity hospital. Groups were paired for age, BMI, and sex.
Inclusion criteria for healthy subjects were (1) absence of spirom-
etry abnormalities; (2) absence of bone, nervous, and/or muscle
dysfunction that could interfere on the assessment of physical
activity; and (3) BMI less than 30kg·m⫺2. Besides criteria 2 and 3,
inclusion criteria in the COPD group were (1) diagnosis of COPD
based on spirometry, clinical, and radiologic internationally ac-
cepted criteria,3 and (2) clinical stability (absence of exacerba-
tions) for at least 3 months before inclusion in the study. All
subjects were informed about study procedures and provided
written formal consent to their participation. The study was ap-
proved by the Ethics and Research Committee of Universidade
Estadual de Londrina/HU-UEL.
Study Design and Protocol
In this descriptive study, all subjects were submitted to an initial
assessment of lung function and functional exercise capacity
(6MWT) as screening measures. For step size determination,
subjects were instructed to walk (as in their daily walking) 10
steps in a straight line. Total distance was measured and divided
by 10. Afterward, each subject performed the following protocol:
walk on a treadmillc with no inclination at 3 different speeds
corresponding to 30%, 60%, and 100% of the average speed
achieved during the 6MWT (speeds 1, 2, and 3, respectively).
Each speed was sustained for 1 minute, with 1 minute of rest in
between. During the treadmill walking protocol, subjects wore
both motion sensors as study measures: the Digiwalker SW701
pedometer at the right side of the waist (hemiclavicle line) and the
SenseWear multisensor on the right arm. As a criterion method for
energy expenditure estimation, simultaneous indirect calorimetry
was performed by a portable gas analyzer. The device was cali-
Arch Phys Med Rehabil Vol 91, February 2010
brated before each test in accordance with manufacturer instruc-
tions. Energy expenditure (in kilocalories for standardization of units)
was derived from oxygen uptake assessment (mL·kg⫺1·min⫺1). The
exact beginning and ending of walking at each speed were
synchronized in all devices because of the presence of at least
3 investigators during each test (ie, 1 investigator initiating the
portable gas analyzer, 1 initiating the camera, 1 initiating the 2
motion sensors simultaneously). At the same time, the tread-
mill walking protocol was videotaped by a digital camera
(Sony Cyber Shot, DSC-W50d) as criterion method for step
counting. The results from both motion sensors (energy expen-
diture and number of steps) were compared with the criterion
methods and with each other.
METHODS
Lung Function Assessment
Simple spirometry was performed by the Pony spirometer.e The
technique was in accordance with American Thoracic Society.18
FEV1 and forced vital capacity were obtained postbronchodilator.
Reference values were those by Knudson et al.19
Six-minute walk test. The 6MWT was performed in ac-
cordance with international standards.20 Patients were encour-
aged to walk 6 minutes as fast as they could in a straight
leveled 30-m corridor. Two tests were performed with each
subject, and the longest distance was used to calculate speed
average and, consequently, the 3 protocol walking speeds.
Normative values were those by Troosters et al.21
Multisensor SenseWear armband. The Multisensor
SenseWear armband is a small (8.8⫻5.6⫻2.1cm) and light
(82g) monitor that is worn on the upper-posterior region of the
right arm. Information regarding various parameters including
accelerometry, multiple physiological sensors, and demo-
graphic characteristics such as sex, age, weight, height, and
dominant arm are used to estimate energy expenditure through
algorithms developed by the manufacturer.11,14 Among the
device’s main outcomes, the most commonly used are total
energy expenditure, average of MET, energy expenditure in
activities requiring above 3 MET, time spent in sedentary (⬍3
MET), moderate (3– 6 MET) and vigorous activities (6 –9
MET), as well as the number of detected steps. A final report
is obtained through analysis of the data by a specific software
(Inner View).b
Pedometer Digiwalker SW701. The Digiwalker SW701 is
a simple and relatively inexpensive device, worn attached to
the waist, providing the number of steps performed, distance,
and energy expenditure estimation in a determined period. For
this, the device requires a few characteristics of the wearer such
as weight and step length. Its mechanism consists of an internal
spring-levered system that is sensitive to vertical hip move-
ments. This spring lever is connected to an electric circuit that
computes each deflection as a step. Furthermore, based on the
device movement counting, it also provides an active energy
expenditure estimation.
Portable gas analyzer. The portable metabolic system
VO2000 AeroGraphf is a transducer for metabolic analysis of
pulmonary gas exchanges, projected to operate connected to a
computer, previously tested and validated.11 The system pro-
vides energy expenditure estimation by indirect calorimetry
executing continuous analysis of oxygen uptake, carbon diox-
ide production, and expired volume.
Statistical Analysis
The analysis was performed with Prism software.g The Kol-
mogorov-Smirnov test was used to analyze normality of the data.
 MOTION SENSORS IN CHRONIC OBSTRUCTIVE PULMONARY DISEASE, Furlanetto 263

Because data presented normal distribution, parametric tests Table 2: Number of Steps Registered by Each Method in Patients
were used, and results were expressed as mean and SD. Com- With COPD at Speeds 1, 2, and 3 (30%, 60%, and 100% of the
6MWT Average Speed, Respectively) and During the Entire
parison between the COPD and healthy elderly groups was Protocol (Summing All Speeds)
performed by the unpaired t test. One-way analysis of variance
was used in the comparison of devices (number of steps and Speed 1 Speed 2 Speed 3 Sum of All
Steps 1.4⫾0.3km/h 2.9⫾0.5km/h 4.8⫾0.8km/h Speeds
energy expenditure estimative at each speed and on the sum of
the 3 speeds), with Tukey as post hoc test. In addition, agree- Digiwalker 26⫾26* 73⫾35* 124⫾31 224⫾83*
ment between the measures was studied by the Bland and pedometer
Altman graphic method. For all analysis, statistical significance SenseWear 19⫾20* 59⫾31* 91⫾38*† 166⫾73*†
was set at P less than .05. multisensor
Video 79⫾17 105⫾20 139⫾25 324⫾57
RESULTS
NOTE. Values are mean ⫾ SD.
Thirteen patients with COPD and 9 healthy elderly subjects *P⬍.05 vs video.
were excluded because of intolerance to any of the tests or †
P⬍.05 vs Digiwalker pedometer.
protocol interruption as a result of noncompliance to the met-
abolic system mouthpiece. The group of patients with COPD
who remained (n⫽30; 17 men) was characterized by reduced
pulmonary function and functional capacity and normative showed similar energy expenditure estimation to indirect cal-
BMI. Most subjects (73%) had moderate airflow obstruction orimetry at the second and third speeds, with significant un-
according to the Global Initiative for Chronic Obstructive Lung derestimation only at the first speed (P⬍.05). When comparing
Disease classification3 (table 1). The healthy elderly group energy expenditure estimation between the 2 motion sensors,
(n⫽30; 15 men) showed normal values of lung function, func- the pedometer provided significantly lower values at the first
tional exercise capacity, and BMI (see table 1). Walking speed speed (see table 3).
was significantly different between the groups at all speeds (see There was significant underestimation of the total number of
table 1). The groups did not show significant differences con- steps taken during the whole protocol (ie, summing the 3
cerning age, weight, height, and BMI. speeds) by both the pedometer and the multisensor (see table
2). Bland and Altman plots for the number of steps (fig 1A)
Patients With Chronic Obstructive Pulmonary Disease depict an average underestimation of 100 steps for the pedom-
Table 2 shows that, in the COPD group, the number of steps eter and 158 steps for the multisensor out of an average of 324
registered by the pedometer was significantly lower than the steps performed during the 3 speeds. For the pedometer (see fig
videotape at the first and second speeds (P⬍.001), with no 1A, left), the plot showed negative significant correlation
statistical difference at the third speed. With regard to the (r⫽⫺.50), indicating that the lower the total number of per-
multisensor step counting, there was significant underestima- formed steps, the higher the underestimation. On the other
tion compared with the videotape at the 3 speeds (P⬍.001). hand, total energy expenditure estimation provided by the
The comparison of step detection provided by the 2 motion multisensor during the whole protocol did not show statistical
sensors showed significant underestimation by the multisensor difference to indirect calorimetry, while the pedometer showed
compared with the pedometer at the third speed and in the sum significant underestimation (see table 3). Bland and Altman
of all speeds (see table 2). plots for energy expenditure (fig 1B) depict an average under-
Table 3 shows that the pedometer energy expenditure esti- estimation of 7.3kcal for the pedometer and 3.9kcal for the
mation was statistically similar to indirect calorimetry only at multisensor out of an average of 16kcal spent during the 3
the third speed, with significant underestimation at the first and speeds. Both for the pedometer and for the multisensor, the
second speeds (P⬍.001). On the other hand, the multisensor plots showed positive significant correlations (r⫽.60 and
r⫽.52, respectively), indicating that the higher the cumulative
energy expenditure, the higher the underestimation.
Table 1: Sample Characteristics of Patients With COPD Healthy Elderly
and Healthy Elderly
Table 4 shows that the number of steps registered by pedometer
Characteristics Patients With COPD (n⫽30) Healthy Elderly (n⫽30) in the healthy elderly group was statistically similar to the
Men/Women 17/13 15/15
Age (y) 67⫾8 68⫾7
Weight (kg) 63⫾12 64⫾11 Table 3: Energy Expenditure (kcal) Registered by Each Method in
Height (m) 1.62⫾0.07 1.59⫾0.08 Patients With COPD at Speeds 1, 2, and 3 (30%, 60%, and 100% of
BMI (kg/m2) 24⫾4 25⫾3 the 6MWT Average Speed, Respectively) and During the Entire
FEV1, % predicted 46⫾17* 104⫾21 Protocol (Summing All Speeds)
FVC, % predicted 67⫾19* 103⫾18 Speed 1 Speed 2 Speed 3 Sum of All
FEV1/FVC 52⫾12* 80⫾5 Kcal 1.4⫾0.3km/h 2.9⫾0.5km/h 4.8⫾0.8km/h Speeds
GOLD (I/II/III/IV) 1/12/10/7 NA
Digiwalker 0.8⫾0.9* 2.6⫾2* 5.4⫾2.5 8.7⫾4.9*
Speed 1 (km/h) 1.4⫾0.3* 1.6⫾0.2
pedometer
Speed 2 (km/h) 2.9⫾0.5* 3.3⫾0.5
SenseWear 3.2⫾1.7*† 3.9⫾2 5⫾2.5 12.1⫾5.6
Speed 3 (km/h) 4.8⫾0.8* 5.4⫾0.7
multisensor
NOTE. Values are mean ⫾ SD or as otherwise indicated. Indirect 4.6⫾3.2 5.3⫾3.3 6.2⫾3.6 16⫾9.8
Abbreviations: FVC, forced vital capacity; GOLD, Global Initiative for calorimetry
Chronic Obstructive Lung Disease; NA, not applicable; Speed 1, 2,
and 3, based respectively on 30%, 60%, and 100% of the average NOTE. Values are mean ⫾ SD.
speed during a 6MWT. *P⬍.05 vs indirect calorimetry.
†
*P⬍.05 vs healthy elderly. P⬍.05 vs Digiwalker pedometer.

Arch Phys Med Rehabil Vol 91, February 2010

264 MOTION SENSORS IN CHRONIC OBSTRUCTIVE PULMONARY DISEASE, Furlanetto

Fig 1. Bland & Altman plots comparing the results of (A) number of steps and (B) energy expenditure (kcal) registered by the pedometer and the
multisensor versus the criterion methods (video and indirect calorimetry) in patients with COPD during the entire protocol (summing all speeds).
Graphics on the left show the comparison between pedometer and the criterion methods, whereas graphics on the right show the comparison
between multisensor and the criterion methods. In each graphic, the central dotted line corresponds to the average difference between the
respective methods, whereas the upper and lower dotted lines correspond to the upper and lower limits of agreement, respectively.

videotape just at the third speed, whereas the multisensor showed motion sensors showed significant underestimation by the multi-
significant underestimation compared with the videotape at all sensor compared with the pedometer at the third speed.
speeds (P⬍.001). The comparison of step detection between the 2 Concerning energy expenditure, table 5 shows significant
underestimation by the pedometer compared with indirect cal-
orimetry at the first and second speeds (P⬍.001). On the other
Table 4: Number of Steps Registered by Each Method in Healthy hand, the multisensor estimation did not show differences to
Elderly at Speeds 1, 2, and 3 (30%, 60%, and 100% of the 6MWT
Average Speed, Respectively) and During the Entire Protocol indirect calorimetry at any speed. When comparing energy
(Summing All Speeds) expenditure provided by the 2 motion sensors, there was sig-
Speed 1 Speed 2 Speed 3 Sum of All
nificant underestimation by the pedometer at the first and
Steps 1.6⫾0.2km/h 3.3⫾0.5km/h 5.4⫾0.7km/h Speeds second speeds (see table 5).
On the healthy elderly group, as observed in the COPD
Digiwalker 36⫾27* 85⫾27* 139⫾24 260⫾67* group, there was significant underestimation on the total num-
pedometer ber of steps both by the pedometer and the multisensor (see
SenseWear 35⫾22* 82⫾29* 119⫾28*† 238⫾55* table 4). Bland and Altman plots for the number of steps (fig 2A)
multisensor depict an average underestimation of 88 steps for the pedom-
Video 85⫾14 114⫾14 148⫾17 348⫾40 eter and 110 steps for the multisensor out of an average of
NOTE. Values are mean ⫾ SD.
348 steps performed during the 3 speeds. For the pedometer
*P⬍.05 vs video. (see fig 2A, left), the plot showed negative significant correla-
†
P⬍.05 vs Digiwalker pedometer. tion (r⫽⫺.57), indicating that the lower the total number of

Arch Phys Med Rehabil Vol 91, February 2010

 MOTION SENSORS IN CHRONIC OBSTRUCTIVE PULMONARY DISEASE, Furlanetto
Table 5: Energy Expenditure (kcal) Registered by Each Method in
Healthy Elderly at Speeds 1, 2, and 3 (30%, 60%, and 100% of the
6MWT Average Speed, Respectively) and During the Entire
Protocol (Summing All Speeds)
Speed 1 Speed 2 Speed 3 Sum of All
Kcal 1.6⫾0.2km/h 3.3⫾0.5km/h 5.4⫾0.7km/h Speeds
Digiwalker 1⫾1* 3⫾1.5* 6.3⫾2.1 10.4⫾4.2*
pedometer
SenseWear 3.4⫾1.4† 4.9⫾1.8† 6⫾2.3 14.3⫾4.9†
multisensor
Indirect 4⫾3.1 5⫾2.8 6.3⫾2.8 15.3⫾8.4
calorimetry
NOTE. Values are mean ⫾ SD.
*P⬍.05 vs indirect calorimetry.
†
P⬍.05 vs Digiwalker pedometer.
performed steps, the higher the underestimation. Once again
similarly to the COPD group, total energy expenditure estima-
tion provided by the multisensor did not show a difference
from indirect calorimetry, whereas the pedometer showed sig-
nificant underestimation (see table 5). There was also signifi-
cant underestimation of energy expenditure by the pedometer
compared with the multisensor. Bland and Altman plots for
energy expenditure (fig 2B) showed positive significant corre-
lations both for the pedometer and for the multisensor (r⫽.67
and r⫽.61, respectively), indicating that the higher the cumu-
lative energy expenditure, the higher the underestimation.
DISCUSSION
Pedometer
The spring-levered pedometer used in this study was previ-
ously suggested as superior to other devices in different tread-
mill speeds22,23 and predetermined distances.22,24 However,
step counting by the pedometer in the present protocol was
inaccurate both in patients with COPD and healthy elderly
during low and moderate speeds compared with the criterion
method, whereas it was accurate for step counting at higher
speeds. Figures 1A and 2A (left) depict that pedometer’s un-
derestimation is higher at slower walking speeds. These results
corroborate some previous literature data showing that pedom-
eters adequately detected steps compared with a uniaxial ac-
celerometer (Computer Science and Applications Inch) at higher
speeds, but underestimated steps at slow walking, characteristic
usually observed in elderly people.16 Furthermore, Pitta et al4
showed that patients with COPD walk 25% less briskly compared
with healthy elderly. These facts raise important questions about
the use of pedometers to count steps during daily life in patients
with COPD, and likewise in healthy elderly who walk slowly. To
corroborate this concern, the sum of steps detected by the pedom-
eter during the whole protocol was significantly lower than the
reference method, confirming the device’s limitation for both
populations included in this study.
A probable explanation for the inaccuracy of this kind of
device is the fact that vertical movements of the hip are less
marked at lower speeds (ie, with insufficient magnitude to
generate the contact of the spring to the electric circuit), and the
sensor usually fails to register some of these movements.1,16
Additionally, Tudor-Locke et al25 showed that the Yamax
SW-200 pedometer, a device with a similar mechanism as the
present study’s pedometer, needs a force of .35g to register a
step, underestimating steps compared with an accelerometer
whose sensitivity is .30g. This indicates that the pedometer
might not have adequately detected steps as a result of being a
265
device with relatively lower sensitivity. The sensitivity of each
motion sensor is determined not only by its type (pedometer,
accelerometer) but also by each device’s technical specifica-
tions, which implies that not necessarily all types of pedome-
ters are less sensitive than all types of accelerometers.
In the present study, the pedometer estimation of energy
expenditure was similar to the estimation provided by indirect
calorimetry only at the third speed for both groups. This
strengthens the concept that pedometers show better accuracy
at higher speeds. However, at the first and second speeds
(corresponding to 30% and 60% of the 6MWT speed, respec-
tively) and in the sum of all speeds, the pedometer did not
repeat this performance. According to Crouter et al,23 pedom-
eters usually have limited accuracy to estimate distance walked
and even lower accuracy to estimate energy expenditure. This
might occur because of pedometers’ mechanism for calculating
energy expenditure: it is based on the subjects’ step count. If step
counting is not accurate at lower speeds, energy expenditure will
not be correctly estimated either. However, this might not be the
only factor, because the Bland and Altman plots (see figs 1B and
2B, left) show that the underestimation is higher at higher meta-
bolic rates, regardless of the number of steps and speed. Reasons
for this might include increased energy expenditure at high met-
abolic rates linked to a high work of breathing (in the case of
patients with COPD) and the use of generic prediction equations
(in the case of both patients with COPD and the elderly), as
discussed in more detail below for the multisensor.
Multisensor
The multisensor did not count steps accurately in any group
at any speed in the present study. A probable explanation for
these findings is that this device is worn at the arm, although
the mechanism of steps detection is not clearly described by the
device’s manufacturer.
In contrast to the pedometer, the SenseWear mechanism for
calculating energy expenditure does not depend on step count.
The device combines signals of 4 sensor types: skin tempera-
ture, body heat flux, galvanic skin resistance, and a biaxial
accelerometer. Hence, regardless of the number of steps de-
tected or the isolated contribution of the biaxial accelerometer,
energy expenditure estimation is improved because of an im-
portant contribution of the additional physiologic sensors.9
This is a peculiar and useful characteristic of the SenseWear
compared with typical devices based on accelerometry alone,
because it was suggested that typical accelerometers are more
accurate for quantification and differentiation of body move-
ments than for estimation of energy expenditure26 (in contrast
with the multisensor SenseWear).
The multisensor was able to estimate energy expenditure
adequately at the 3 walking speeds in the healthy elderly group,
as well as at speeds corresponding to 60% and 100% of the
6MWT speed in patients with COPD, in addition to the total
protocol duration in both populations. On the other hand,
results from the Bland and Altman plots (see figs 1B and 2B)
showed correlation between higher metabolic rates and overall
underestimation of energy expenditure not only with the pe-
dometer but also with the multisensor, corroborating previous
findings by Patel et al.9 Underestimation of energy expenditure
at high metabolic rates in the COPD group might be at least
partly attributed to the fact that increased energy expenditure in
these patients is linked to a high work of breathing, which is
not reflected by accelerometry or step counting. In addition,
both for patients with COPD and for the elderly, energy ex-
penditure underestimation may be related to the use of generic
equations for energy expenditure prediction, because previous
Arch Phys Med Rehabil Vol 91, February 2010
266 MOTION SENSORS IN CHRONIC OBSTRUCTIVE PULMONARY DISEASE, Furlanetto

Fig 2. Bland and Altman plots comparing the results of (A) number of steps and (B) energy expenditure (kcal) registered by the pedometer
and the multisensor versus the criterion methods (video and indirect calorimetry) in healthy elderly during the entire protocol (summing all
speeds). Graphics on the left show the comparison between pedometer and the criterion methods, whereas graphics on the right show the
comparison between multisensor and the criterion methods. In each graphic, the central dotted line corresponds to the average difference
between the respective methods, whereas the upper and lower dotted lines correspond to the upper and lower limits of agreement,
respectively.

literature confirms that population-specific prediction equations
increase the estimation accuracy.27,28
Comparison Between Pedometer and Multisensor
For step counting at slow speeds, the motion sensors were
similar in the sense that neither of them provided adequate
estimations, and therefore they are equally inaccurate in sub-
jects who walk slowly. At the highest speed, however, the
pedometer showed to be more sensitive than the multisensor.
This indicates that the usefulness of the step counting feature of
the SenseWear is very limited even at higher waking speeds,
possibly because the device is worn at the arm and not at the
waist or ankle, as previously discussed. On the other hand, for
the estimation of energy expenditure, the multisensor was
clearly superior to the pedometer at slow walking speeds,
possibly because of the contribution of the physiologic sensors
in addition to the biaxial accelerometer.
Limitations and Future Perspectives
In the present protocol, each treadmill speed was sustained
for 1 minute, and one can correctly argue that it takes more
than 1 minute when changing exercise intensity to achieve
steady-state energy expenditure. However, the present protocol
did not aim at achieving the steady state for each speed. It
aimed simply to allow comparison of energy expenditure esti-
mations among the methods at the same points in time in
subjects walking at different speeds without necessarily achiev-
ing steady-state oxygen consumption at these speeds.
Conclusions of the present study can only be generalized to
subjects with BMI lower than 30 kg·m⫺2. The SenseWear
Armband has shown limited accuracy for the measurement of
energy expenditure in obese subjects.12 Furthermore, the pe-
dometer might not have ideal placement in subjects with a large
abdominal volume because it might stay in a different position

Arch Phys Med Rehabil Vol 91, February 2010

 MOTION SENSORS IN CHRONIC OBSTRUCTIVE PULMONARY DISEASE, Furlanetto
than the vertical alignment suggested for its use.24 For these
reasons, subjects with high BMI were not included in this
study. Future studies should investigate to what extent the
conclusions of this study also apply to obese subjects, an
important target population for physical activity interventions.
In the present study, the accuracy of movement sensors was
analyzed in a laboratory protocol, and therefore no assessment
in free-living conditions was involved. This suggests some
caution when generalizing the present results because previous
literature has shown that floor walking may induce higher
energy expenditure than treadmill walking.29,30 For the ad-
vance of this research field, future investigations should con-
sider nonlaboratory conditions, or at least simulate daily activ-
ities in a laboratory setting. These activities can include daily
tasks involving arms and legs, as well as walking with frequent
interruption and direction changes.
CONCLUSIONS
It can be inferred that, both in patients with COPD and in
healthy elderly, the SenseWear Armband multisensor had bet-
ter energy expenditure estimates during most walking speeds
than the Digiwalker SW701 pedometer. Conversely, for step
counting, accuracy was observed only with the pedometer
during high walking speeds in both groups. Technological
efforts in this research field should focus on making the devices
more sensitive during slow walking speeds.
References
1. Steele BG, Belza B, Cain K, Warms C, Coppersmith J, Howard J.
Bodies in motion: monitoring daily activity and exercise with
motion sensors in people with chronic pulmonary disease. J Re-
habil Res Dev 2003;40:45-58.
2. Le Masurier GC, Lee SM, Tudor-Locke C. Motion sensor accu-
racy under controlled and free-living conditions. Med Sci Sports
Exerc 2004;36:905-10.
3. Rabe KF, Hurd S, Anzueto A, et al. Global strategy for the
diagnosis, management, and prevention of chronic obstructive
pulmonary disease: GOLD executive summary. Am J Respir Crit
Care Med 2007;176:532-55.
4. Pitta F, Troosters T, Spruit MA, Probst VS, Decramer M, Gos-
selink R. Characteristics of physical activities in daily life in
chronic obstructive pulmonary disease. Am J Respir Crit Care
Med 2005;171:972-7.
5. Pitta F, Troosters T, Probst VS, Spruit MA, Decramer M, Gos-
selink R. Physical activity and hospitalization for exacerbation of
COPD. Chest 2006;129:536-44.
6. Yohannes AM, Baldwin RC, Connolly M. Mortality predictors in
disabling chronic obstructive pulmonary disease in old age. Age
Ageing 2002;31:137-40.
7. Garcia-Aymerich J, Lange P, Benet M, Schnohr P, Antó JM.
Regular physical activity reduces hospital admission and mortality
in chronic obstructive pulmonary disease: a population based
cohort study. Thorax 2006;6:772-8.
8. Pitta F, Troosters T, Probst VS, Lucas S, Decramer M, Gosselink
R. Potential consequences for stable chronic obstructive pulmo-
nary disease patients who do not get the recommended minimum
daily amount of physical activity. J Bras Pneumol 2006;32:301-8.
9. Patel SA, Benzo RP, Slivka WA, Sciurba FC. Activity monitoring
and energy expenditure in COPD patients: a validation study.
COPD 2007;4:107-12.
10. Fruin ML, Rankin JW. Validity of a multi-sensor armband in
estimating rest and exercise energy expenditure. Med Sci Sports
Exerc 2004;36:1063-9.
11. Wahrlich V, Anjos LA, Going SB, Lohman TG. Validation of the
VO2000 calorimeter for measuring resting metabolic rate. Clin
Nutr 2006;25:687-92.
267
12. Papazoglou D, Augello G, Tagliaferri M, et al. Evaluation of a
multisensor armband in estimating energy expenditure in obese
individuals. Obesity 2006;14:2217-23.
13. Jakicic JM, Marcus M, Gallagher KI, et al. Evaluation of the
SenseWear Pro Armband™ to assess energy expenditure during
exercise. Med Sci Sports Exerc 2004;36:897-904.
14. Pitta F, Troosters T, Probst VS, Spruit MA, Decramer M, Gos-
selink R. Quantifying physical activity in daily life with question-
naires and motion sensors in COPD. Eur Respir J 2006;27:1040-55.
15. Schneider PL, Crouter SE, Basset DR. Pedometers measures of
free-living physical activity: comparison of 13 models. Med Sci
Sports Exerc 2004;36:331-5.
16. Melanson EL, Knoll JR, Bell ML, et al. Commercially available
pedometers: considerations for accurate step counting. Prev Med
2004;39:361-8.
17. King GA, Torres N, Potter TJ, Coleman KJ. Comparison of
activity monitors to estimate energy cost of treadmill exercise.
Med Sci Sports Exerc 2004;36:1244-51.
18. Miller MR, Hankinson J, Brusasco V, et al. Standardisation of
spirometry. Eur Respir J 2005;26:319-38.
19. Knudson RJ, Slatin RC, Lebowitz MD, Burrows B. The maximal
expiratory flow-volume curve: normal standards, variability, and
effects of age. Am Rev Respir Dis 1976;113:587-600.
20. ATS statement: guidelines for the six-minute walk test. Am J
Respir Crit Care Med 2002;166:111-7.
21. Troosters T, Gosselink R, Decramer M. Six minute walking
distance in healthy elderly subjects. Eur Respir J 1999;14:270-4.
22. Basset DR, Ainsworth BE, Leggett SR, et al. Accuracy of five
electronic pedometers for measuring distance walked. Med Sci
Sports Exerc 1996;28:1071-7.
23. Crouter SE, Schneider PL, Karabulut M, Basset DR. Validity of
ten electronic pedometers for measuring steps, distance and en-
ergy cost. Med Sci Sports Exerc 2003;35:1455-60.
24. Schneider PL, Crouter SE, Lukajic O, Basset DR. Accuracy and
reliability of ten pedometers for measuring steps over a 400-m
walk. Med Sci Sports Exerc 2003;35:1779-84.
25. Tudor-Locke C, Ainsworth BE, Thompson RW, Matthews CE.
Comparison of pedometers and accelerometers measures of free-
living physical activity. Med Sci Sports Exerc 2002;34:2045-51.
26. Leenders NY, Nelson TE, Sherman WM. Ability of different
physical activity monitors to detect movement during treadmill
walking. Int J Sports Med 2003;24:43-50.
27. Cole PJ, LeMura LM, Klinger TA, Strohecker K, McConnell TR.
Measuring energy expenditure in cardiac patients using the Body
Media Armband versus indirect calorimetry: a validation study.
J Sports Med Phys Fitness 2004;44:262-71.
28. Levine J, Melanson EL, Westerterp KR, Hill JO. Tracmor system
for measuring walking energy expenditure. Eur J Clin Nutr 2003;
57:1176-80.
29. Pearce ME, Cunningham DA, Donner AP, Rechnitzer PA, Fullerton
GM, Howard JH. Energy cost of treadmill and floor walking at self-
selected paces. Eur J Appl Physiol Occup Physiol 1983;52:115-9.
30. Heus R, Wertheim AH, Havenith G. Human energy expenditure
when walking on a moving platform. Eur J Appl Physiol Occup
Physiol 1998;77:388-94.
Suppliers
a. Yamax, 1-5-7, Chuo-cho, Meguro-ku, Tokyo 152-8691 Japan.
b. BodyMedia, 4 Smithfield St, 11th Fl, Pittsburgh, PA 15222.
c. Imbramed Millennium, Rua 7 de abril 266, Porto Alegre, RS
90220130, Brazil.
d. Sony, United States of America (www.sony.com).
e. Cosmed, Via dei Piani di Mt Savello, 37, Pavona di Albano, Rome,
I-00047 Italy.
f. Medical Graphics Corp, 350 Oak Grove Pkwy, St Paul, MN 55127.
g. GraphPad Software Inc, 2236 Avenida de la Playa, La Jolla, CA 92037.
h. MTI Health Services, 709 Anchors St, Fort Walton Beach, FL 32548.
Arch Phys Med Rehabil Vol 91, February 2010