If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Correspondence to Fabio Pitta, PhD, Departamento de Fisioterapia – CCS, Hospital Universitário de Londrina, Rua Robert Koch, 60 – Vila Operária, 86038-350 – Londrina, Paraná, Brazil
Sant'Anna T, Escobar VC, Fontana AD, Camillo CA, Hernandes NA, Pitta F. Evaluation of a new motion sensor in patients with chronic obstructive pulmonary disease.
Objective
To assess the criterion validity and reproducibility of a new pedometer in patients with chronic obstructive pulmonary disease (COPD).
Design
Cross-sectional study.
Setting
Outpatient physiotherapy clinic from a university hospital.
Participants
Patients with COPD (N=30; 17 men; forced expiratory volume in the first second, 44±17% predicted) were videotaped while performing 2 protocols: one including 2 slow and 2 fast 5-minute walks, and another including a circuit of activities of daily living (ADLs). Concomitantly, patients wore 2 motion sensors: the new pedometer and a multisensor accelerometer.
Interventions
None.
Main Outcome Measures
Step counting (SC), energy expenditure (EE), walking distance (WD), activity time (AT), and walking intensity (WI) registered by the pedometer were compared with video and the multisensor as criterion methods.
Results
Correlations between the pedometer and the criterion method were high for SC during slow and fast walking (r=.79 and r=.95) and for EE during fast walking (r=.83). Correlation was more modest for EE during slow walking (r=.65) and for WD and WI during both speeds (.47<r<.68). The agreement between methods was also good, according to Bland-Altman plots. The device was reproducible for registering SC, WD, and EE during slow walking and for all variables during fast walking (intraclass correlation coefficient >.79 for all). During the ADLs circuit, the pedometer underestimated AT by an average of 55% but provided an acceptable EE estimation in a group basis (average difference of 6% with the multisensor).
Conclusions
In patients with COPD, the new pedometer analyzed in the present study is reproducible for most outcomes and highly valid for SC during slow and fast walking and EE during fast walking. The device's validity is more limited for EE during slow walking, and WD and WI at both speeds. Furthermore, during the performance of ADLs, it significantly underestimates activity time but provides an acceptable estimation of EE in a group basis.
Sedentary lifestyle in this population is commonly related to dyspnea (the most frequent symptom reported by patients) and impairment in exercise capacity.
In addition, inactivity in patients with COPD is also associated with poor functional status and higher risk of mortality, as evaluated by the Body Mass, Obstruction, Dyspnea and Exercise Index,
Potential consequences for stable chronic obstructive pulmonary disease patients who do not get the recommended minimum daily amount of physical activity.
These are the main reasons for the growing interest in physical activity monitoring, especially in a markedly sedentary population such as patients with COPD.
In order to accurately measure the patient's level of physical activities of daily living in clinical practice, valid and affordable motion sensors are needed. Pedometers are very simple motion sensors that provide information on step counts and are not expensive in comparison with more complex activity monitors. Despite its practicality, the use of classic pedometers (ie, those not involving accelerometry) has been not recommended in slow-walking populations,
Step counting and energy expenditure estimation in patients with chronic obstructive pulmonary disease and healthy elderly: accuracy of 2 motion sensors.
because of their low sensitivity. Possibly to counteract this limitation, a new pedometer with an embedded 3-dimensional accelerometer was developed. In contrast to the classic pendulum mechanism of regular pedometers, this 3-dimensional electronic sensor system could theoretically improve the pedometer's sensitivity to step counting and other activity outcomes. Its price, despite being higher than regular pedometers, is considerably lower than technologically advanced activity monitors. Therefore, the new pedometer could be a potential motion sensor for objective measurement of physical activities of daily living in patients with COPD in clinical practice and epidemiologic large studies developed with a limited budget, in which more advanced activity monitors might not be an option. However, to the present date, validity and reproducibility of this new pedometer have not yet been investigated in depth in patients with COPD or in any other population. Therefore, the objective of the present study was to evaluate the criterion validity and reproducibility of this new motion sensor in patients with COPD.
Methods
Study Design and Participants
Thirty patients with COPD (17 men) with no exacerbation episodes for at least 3 months were recruited to take part in this cross-sectional study. All participants had taken part recently or were currently enrolled in programs of respiratory physiotherapy at the University Hospital of the State University of Londrina, Brazil. The patient's latest assessment of lung function (<6mo) performed by hospital staff was used to confirm COPD diagnosis and classify degree of obstruction according to Global Initiative for Chronic Obstructive Lung Disease criteria.
None of the participants were long-term oxygen users or had other pathologic conditions that could impair physical activity performance. Initially, height and weight were registered and body mass index was calculated. Characteristics of study participants are described in table 1. In order to calculate the average step length, patients were asked to perform 10 steps at their own pace; then, the distance walked was registered and divided by 10. Additionally, patients were questioned about date of birth, smoking habits, and predominant hand. These data were necessary in order to configure the motion sensors used in this study. The study was approved by the institution's research ethics committee (HU-UEL n.04624), and all involved subjects gave formal and written consent prior to their inclusion.
Table 1Participants' Characteristics (N=30)
Variables
Mean ± SD
Sex (M/F)
17/13
Age (y)
67±7
Height (m)
1.60±0.10
Weight (kg)
70±16
BMI (kg/m−2)
27±6
FEV1 (% of predicted)
44±17
FEV1/FVC
53±12
Abbreviations: BMI, body mass index; F, female; FEV1, forced expiratory volume in the first second; FVC, forced vital capacity; M, male.
All participants were submitted to 2 physical activity protocols: a walking protocol and a circuit comprising different activities of daily living (ADLs). The first protocol consisted of 2 slow and 2 fast walks, in this order, each of a 5-minute duration, performed in a 30-m corridor. Speed of walking was established by asking patients to walk at a regular pace, as developed in daily life (slow walking), and to walk at a pace as if late for an appointment (fast walking). Between walks, patients were able to rest until they felt ready to start the next walk. The second protocol was performed in a 70m2 room and encompassed a 5-station ADLs circuit. The stations were comprised of the following: patients had to sit, climb up and down a step, lie down (1min each activity), and dress and remove a shirt, as if they were at home. Between stations, participants had to walk straight and in curves, while dodging objects (fig 1). This circuit was repeated 3 times successively. In each patient, both protocols were recorded by a video camera.a
Fig 1Activities circuit in a 70m2 room. Station 1 = sitting (1min); station 2 = up and down a step (1min); station 3 = lying (1min); station 4 = dressing in a shirt; and station 5 = removing the shirt.
During the protocols, patients wore 2 activity monitors simultaneously: SenseWear Armbandb (SAB) and Power Walker 610 (PW).c The SAB is a multisensor composed by a biaxial accelerometer and physiologic sensors. It is a small (8.8 × 5.6 × 2 cm) and lightweight (82g) monitor that was already validated for energy expenditure estimation in patients with COPD.
Step counting and energy expenditure estimation in patients with chronic obstructive pulmonary disease and healthy elderly: accuracy of 2 motion sensors.
In the present study, the SAB was configured to detect all physical activity movements that reached at least 2 metabolic equivalents because this intensity is equivalent to slow walking, typical in patients with COPD.
A final data report can be obtained by specific software (InnerView Professional 6.1b).
The PW is a pedometer combined with accelerometry, which comprises a 3-dimensional sensor instead of the classic pendulum system. It is relatively new, affordable, and its criterion validity and reproducibility were not yet studied. Outcomes provided by the PW are step counting (in steps), energy expenditure estimation (in calories), walking distance (in meters), activity time (in minutes), and walking intensity (in m/min). The manufacturer recommends its use inside a pocket of pants or a shirt; however, in a pilot study in our laboratory, we concluded that the most accurate position to wear the PW is attached to the waist, in the hemiclavicular line,
Fontana AD, Camillo CA, Oliveira NH, et al. Study of a motion sensor accuracy for quantification of physical activity in daily life in healthy individuals. In: Proceedings of Physiotherapy Congress: Educational, Research and Extension; 2009 Sep 26-26; Londrina, Brazil.
the same position as most pedometers are worn. Therefore, this position was used, always positioned on the right side. As criterion methods, the PW was compared with video recording concerning number of steps, walking distance, activity time, and walking intensity (speed) and was compared with the SAB concerning energy expenditure. Two PW devices were used for data collection in the present study. Both of them were tested for accuracy during a normal 5-minute walk prior to the protocols; less than 3% difference was presented between them.
Statistical Analysis
Data distribution was checked by a Kolmogorov-Smirnov test, and, as variables presented normative distribution, data were reported as mean and SD and parametric statistics were used. Comparisons between the PW and SAB and between the PW and video were performed using paired Student t test. To evaluate the PW's validity, Pearson correlation coefficient was used. Bland-Altman plots and intraclass correlation coefficient (ICC)3,2
were used to analyze the agreement with criteria methods and pedometer reproducibility. Statistical significance was set at 5%. Statistical analysis was performed using Prism 5.0d and SPSS Statistics version 17.0.e
Power Analysis
A sample of 30 individuals had a power of 93% to detect a difference of at least 15 steps (∼3%) between the PW and video recording in the walking protocol of the present study, using the SD of the difference between methods (23 steps) and adopting the statistical significance of 5%. We assumed these values based on our own data regarding slow walking, and we considered a difference of 3% as not denoting clinical significance concerning step counting.
Results
All of the 30 recruited patients successfully completed the protocols. Because there was no difference between first and second walk in both speeds concerning almost all variables of interest (P>.170), we standardized to use only the first walk of each speed in the analysis. Regarding the number of steps, walking distance, and walking intensity, there was a statistical difference between the first and second slow walking (P<.002), although not of a clinically important difference (1.5%, 2%, and 3%, respectively). Average speed during the walking protocol for slow and fast walks was 3.8±0.5 and 4.7±0.5km/h, respectively. During the ADLs circuit, although we were not able to detect the real walking speed, the PW registered it as 3.7±0.8km/h, strongly resembling the walking speed obtained during the slow walking in our study (3.8±0.5km/h). The results concerning validity (correlations), agreement of the device with the criterion methods, and reproducibility (Bland-Altman analysis and ICCs) are described below.
Walking Protocol
During slow walking, there was strong correlation between the PW and video concerning the number of steps (r=.79, P<.001), and moderate correlation concerning walking distance (r=.63, P<.001) and walking intensity (r=.61, P<.001). Regarding energy expenditure, there was also moderate correlation between the PW and SAB (r=.65, P=.0001) (fig 2). The ICC between the 2 slow walks was excellent for the PW step counting, energy expenditure, walking distance, and walking intensity (.89, 95% confidence interval [CI], .77–.95, P<.0001; .98, 95% CI, .96–.99, P<.001; .91, 95% CI, .81–.96, P<.001; .79, 95% CI, .56–.90, P<.001). The ICC was low and nonstatistically significant regarding the PW activity time between the 2 slow walks (−.42, 95% CI, −1.99 to .32, P=.820).
Fig 2Correlations between the PW, video recording, and the SAB during a self-selected 5-minute slow walk. Correlation between the PW and video recording concerning the number of steps, walking distance, and intensity of walking (r=.79, P<.0001; r=.63, P<.001; r=.61, P<.001, respectively) and between the PW and SAB concerning energy expenditure estimation (r=.65, P=.0001).
During fast walking, there was excellent correlation between the PW and video regarding step counting (r=.95, P<.001), with moderate correlation regarding walking distance (r=.48, P=.006) and walking intensity (r=.47, P=.009). Still concerning the fast walking, there was strong correlation between the PW and SAB on energy expenditure estimation (r=.83, P<.001) (fig 3). High ICCs between the 2 fast walks were found for all outcomes provided by the PW (.96, 95% CI, .92–.98, P<.001 for step counting; .99, 95% CI, .98–.99, P<.001 for energy expenditure; .95, 95% CI, .90–.97, P<.001 for walking distance; .79, 95% CI, .57–.90, P<.001 for activity time; .95, 95% CI, .88–.97, P<.001 for walking intensity).
Fig 3Correlations between the PW, video recording, and the SAB during a self-selected 5-minute fast walk. Correlation between the PW and video recording concerning number of steps, walking distance, and intensity of walking (r=.95, P<.0001; r=.48, P=.006; r=.47, P=.009, respectively) and between the PW and SAB concerning energy expenditure estimation (r=.83, P<.0001).
The statistical software was not able to analyze correlations between the PW and video concerning activity time, because it represented a vertical or horizontal line. In other words, both methods registered the same activity time. Activity time registered by the PW during slow walking was very similar to real walking time (4.8±0.4 vs 5.0±0.0min) and identical to real walking time during fast walking (5.0±0.2 vs 5.0±0.0min).
Figure 4 shows the Bland-Altman plots comparing the PW with video and the SAB concerning step counting and energy expenditure, respectively, during both slow and fast walking. Regarding step counting, the ICC between the PW and video was .88 (95% CI, .75–.94, P<.001) during slow walking and .97 (95% CI, .93–.98, P<.001) during fast walking. The ICC between the PW and SAB concerning energy expenditure was .79 (95% CI, .56–.90, P<.001) for slow walking and .89 (95% CI, .78–.95, P<.001) for fast walking. The ICC concerning walking distance between the PW and video was .75 (95% CI, .48–.88, P<.001) for slow walking and .53 (95% CI, .02–.77, P=.003) for fast walking. No significant ICC was found regarding walking intensity between the PW and the real value for any speed.
Fig 4Bland-Altman plots comparing the PW with video and the SAB concerning step counting and energy expenditure, respectively, during both slow and fast walking. (A) and (B) Agreement between the PW and video recording concerning step counting in slow and fast walking, respectively. (C) and (D) Agreement between the PW and SAB concerning energy expenditure estimation in slow and fast walking, respectively. Abbreviations: LL, lower limit (−1.96 SDs); UL, upper limit (1.96 SDs).
For logistical reasons, during the ADLs circuit, we compared only energy expenditure and activity time between the PW and the criterion methods considered in this study. No significant difference between the PW and SAB energy expenditure estimation was observed (10±6 vs 9.4±3.9 calories, respectively, corresponding to an average difference of only 6% in a group basis; P=.647). However, there was no significant ICC between the methods (.22, 95% CI, −.63 to .62, P=.253). Furthermore, the PW significantly underestimated activity time by an average of 55% in comparison with video recording (2.9±1.2 vs 6.4±0.7min, respectively; P<.001) (ICC=−.97, 95% CI, −3.14 to .06, P=.96).
Discussion
The present study showed that, in patients with COPD, although the PW had a somewhat limited validity to estimate walking distance, walking intensity, and energy expenditure during the slow walk, it was highly valid to register the number of steps. Additionally, it was reproducible not only for step counting, but also in the estimation of walking distance and energy expenditure in the slow walk. During the fast walk, the PW was reproducible concerning all variables and highly valid to estimate the number of steps and energy expenditure, although it presented limited validity to estimate walking distance and walking intensity. Finally, during the performance of ADLs, the PW underestimated activity time and was shown to provide an adequate estimation of energy expenditure in a group basis, but not in an individual basis.
The present results clearly demonstrate that the PW has advantages regarding step counting in comparison with other available pedometers, especially considering patients with slow walking patterns. The poor sensitivity of the classic pedometers to detect steps during slow walking is because of the mechanism used to register movement counts: a horizontal, spring-suspended lever arm that deflects with vertical movement of the hips during walking. As this mechanism registers steps based on the up-and-down motion during ambulation, a slow walking pattern (ie, with less pronounced up-and-down hip movement) hinders the accurate detection of steps.
Step counting and energy expenditure estimation in patients with chronic obstructive pulmonary disease and healthy elderly: accuracy of 2 motion sensors.
This study protocol was performed on a treadmill, and walking speeds were determined based on the average speed developed during the six-minute walk test, which is considerably different in comparison with the present study. Because of this methodologic difference, the walking speed was lower than the observed in the present study. However, the slow walking speed observed in the present investigation is similar to the walking speed of healthy older adults, which yielded inaccurate step detection by the Digiwalker in the previous study by Furlanetto et al.
Step counting and energy expenditure estimation in patients with chronic obstructive pulmonary disease and healthy elderly: accuracy of 2 motion sensors.
also showed that classic pedometers underestimate the number of steps as walking speed decreases, and the inaccuracy is more evident below 3mph (equivalent of 4.8km/h, which is similar to the average self-selected fast walking speed of our patients). Recent literature in patients with COPD shows a study evaluating the accuracy of a pedometer with a similar mechanism as the PW for step counting, the Omron HJ-720ITC.
However, this pedometer presented a difference of 14%±26% and 1.7%±6.9% from the actual value (manual count) regarding the number of steps during natural walking in a corridor in patients that walked slower and faster, respectively. In the present study, the PW presented a difference of only .09%±5.8% and 1.2%±6.9% from the actual value of step counting during slow and fast walking, respectively, being evaluated in similar walking speeds as the aforementioned study. This happens presumably due to some differences in the mechanisms used to register movements. Another accelerometer (StepWatch) also demonstrated high accuracy when registering steps during very slow walking in older adults
; however, it is considerably more expensive than the PW and has a more complex process when handling the data. Certainly prices may be highly variable, and this has also to be taken into consideration when one makes the decision of which pedometer to adopt.
The correlation between real value and PW registration of walking distance and walking intensity was poorer in fast walking than in slow walking. This may initially be seen as a surprising result, but we have a hypothesis for its explanation. This probably happened because these outcomes from the PW (walking distance and walking speed) not only depend on the number of steps registered, but also on the step length entered in the device's configuration before initiation of its use by the patient, which in our study was calculated based on a 10-step walk at a slow speed. However, during fast walking, people tend to naturally increase step length.
Gait patterns during different walking conditions in older adults with and without knee osteoarthritis−results from the Baltimore Longitudinal Study of Aging.
Age-associated differences in the gait pattern changes of older adults during fast-speed and fatigue conditions: results from the Baltimore longitudinal study of ageing.
Therefore, the fixed step length entered in the device's configuration to that specific patient naturally decreases the possibility to maintain a correct estimate of the walking distance and intensity as the step length increases along with the increasing speed.
The SAB, used in the present study as the criterion method for the energy expenditure estimation, was already validated for the evaluation of energy expenditure in patients with COPD.
Step counting and energy expenditure estimation in patients with chronic obstructive pulmonary disease and healthy elderly: accuracy of 2 motion sensors.
The energy expenditure estimation by the PW during slow walking was moderately correlated with the SAB estimation, although highly reproducible. During fast walking, the PW was not only highly reproducible but also strongly correlated with the SAB in estimating energy expenditure. In the ADLs circuit protocol, the PW and SAB estimation of energy expenditure presented no agreement when considering subjects individually, but they were similar when considering the whole group. Although the PW presented acceptable performance concerning energy expenditure estimation when compared with the SAB (error estimation of 4%, 6%, and 6% concerning slow walk, fast walk, and ADLs circuit, respectively), its performance is yet considerably better for assessing step counting in patients with COPD than for assessing energy expenditure. Our results are in accordance with the conclusion of Pitta et al,
in a comprehensive review about physical activity quantification in COPD, which states that motion sensors are more accurate for measurement of volume of physical activity in daily life than for the estimation of energy expenditure, especially in populations that typically present slow walking speeds. Besides the fact that PW estimates energy expenditure indirectly based on other variables programmed into the pedometer settings, it does not account for the additional energy spent in activities such as stair climbing, uphill walking, frequent changes in walking direction, carrying loads, and arm activities.
The same considerations could be used for the explanation of poorer estimation of walking distance and walking intensity in comparison with the better estimation of the number of steps.
Estimation of activity time by the PW was similar to the real activity time at both speeds during continuous walking. However, during the ADLs circuit, which better represents the patient's real daily activity pattern, the PW significantly underestimated activity time. One explanation for this might be that the PW, similar to other waist-mounted activity monitors, was not able to detect activities done only with the upper limbs (ie, without whole-body dislocation). Additionally, our ADLs circuit lasted an average of 6.4±0.7 minutes. We do not know if during a whole day of real-life measurement the underestimation of activity time by the PW would differ from the present results. Future studies are necessary to evaluate the utility of the PW to estimate the time spent in physical activity by patients in nonlaboratory based assessments (ie, daily life).
The lack of reproducibility of activity time estimation of the PW during slow walking probably occurred because in some patients there was a difference of 1 minute in the registration of the PW between the first and second walk. This happened because the PW's clock registers only full minutes. If there was even 1 second of difference between the times of starting the timer and starting the protocol walk, the PW would register 1 minute less or more than the actual time. Proportionally, 1 minute of under or overestimation in a 5-minute walk represents 20% inaccuracy but, during a longer continuous walk or during a whole day assessment in daily life, this difference might be less important. The same argument reinforces the lack of reproducibility of the PW concerning walking intensity during slow speed.
Study Limitations
Our study has some limitations. First, patients included in the study were participating or had already recently participated in programs of respiratory physiotherapy, composing a convenience sample. This might indicate that our results are not necessarily extendable for patients not under physiotherapy treatment, and this deserves further investigation. Another limitation is that we did not compare estimation of energy expenditure by the PW with a criterion standard method, as is recommended by the literature. Nevertheless, as the SAB was previously validated for estimating energy expenditure in patients with COPD,
Step counting and energy expenditure estimation in patients with chronic obstructive pulmonary disease and healthy elderly: accuracy of 2 motion sensors.
we believe it might be used as a criterion method for comparison with new tools aimed at estimating the same outcome. In addition, we were not able to manually count the number of steps performed by the patients during the ADLs circuit, because a lot of movements were undefined and would be difficult to standardize what should be considered a step and what should not. This did not allow us to investigate the validity of the PW's step counting during ADLs performance, although its capacity to correctly detect steps during walking in patients with COPD has been clearly shown. This is, in itself, valuable novel information to be added to the current literature, because classic pedometers are currently regarded as inaccurate (and therefore not indicated for use) in this population.
Conclusions
In patients with COPD, the PW is reproducible for most outcomes and highly valid for step counting during slow and fast walking and energy expenditure estimation during fast walking. The device's validity is more limited for energy expenditure estimation during slow walking and walking distance and intensity at both speeds. Furthermore, during the performance of a circuit of ADLs, the device significantly underestimates activity time but provides an acceptable estimation of energy expenditure in a group basis.
Potential consequences for stable chronic obstructive pulmonary disease patients who do not get the recommended minimum daily amount of physical activity.
Step counting and energy expenditure estimation in patients with chronic obstructive pulmonary disease and healthy elderly: accuracy of 2 motion sensors.
Fontana AD, Camillo CA, Oliveira NH, et al. Study of a motion sensor accuracy for quantification of physical activity in daily life in healthy individuals. In: Proceedings of Physiotherapy Congress: Educational, Research and Extension; 2009 Sep 26-26; Londrina, Brazil.
Gait patterns during different walking conditions in older adults with and without knee osteoarthritis−results from the Baltimore Longitudinal Study of Aging.
Age-associated differences in the gait pattern changes of older adults during fast-speed and fatigue conditions: results from the Baltimore longitudinal study of ageing.
Supported by Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq)/Brazil (grant no. 302061/2010-0 ); Coordenação de Aperfeiçoamento de Pessoal de Nivel Superior (CAPES)/Brazil (grant nos. 05768419942 and 00847546993 ); and Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)/Brazil (grant no. 2010/03223-2 ).
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 organization with which the authors are associated.
Reprints are not available from the author.
In-press corrected proof published online on Jul 31, 2012, at www.archives-pmr.org.
00407-8&id=gr1.jpg){kind=link}
00407-8&id=gr2.jpg){kind=link}
00407-8&id=gr3.jpg){kind=link}
00407-8&id=gr4.jpg){kind=link}