European Journal of Applied Physiology

, Volume 95, Issue 4, pp 335–343

Comparison of two waist-mounted and two ankle-mounted electronic pedometers

  • Murat Karabulut
  • Scott E. Crouter
  • David R. BassettJr.
Original Article

DOI: 10.1007/s00421-005-0018-3

Cite this article as:
Karabulut, M., Crouter, S.E. & Bassett, D.R. Eur J Appl Physiol (2005) 95: 335. doi:10.1007/s00421-005-0018-3

Abstract

This study compared two ankle-mounted pedometers [StepWatch 3 (SW-3Ankle) and Activity Monitoring Pod 331 (AMPAnkle)] and two waist-mounted pedometers [New Lifestyles NL-2000 (NLWaist) and Digiwalker SW-701 (SW-701Waist)] under controlled and free-living conditions. In part I, 20 participants walked on a treadmill at speeds of 27–107 m min−1. Actual steps were counted with a hand counter. In part II, participants performed leg swinging, heel tapping, stationary cycling, and car driving. In part III, 15 participants wore all pedometers for a 24 h period. The SW-3Ankle displayed values that were within 1% of actual steps during treadmill walking at all speeds. The other devices underestimated steps at slow speeds but all gave mean values that were within ±3% of actual steps at 80 m min−1 and above. The SW-3Ankle registered some steps during heel tapping, leg swinging, and cycling, while the AMPAnkle was only responsive to leg swinging. During car driving no devices recorded more than eight steps, on average. Over 24 h, the AMPAnkle recorded 18% fewer steps than the SW-3Ankle (P<0.05), while the SW-701Waist and the NLWaist recorded 15 and 11% less than the SW-3Ankle, respectively (NSD). The SW-3Ankle has superior accuracy at slow treadmill walking speeds (although it was also more likely to detect “fidgeting” activities). Over 24 h, the SW-3Ankle tended to give higher estimates of steps per day than the other ankle- and waist-mounted pedometers.

Keywords

Walking Step counter Physical activity Treadmill Motion sensors 

Introduction

Investigators are interested in assessing physical activity (PA) in order to evaluate its relationship to health variables including obesity (Dunn et al. 1998), hypertension (Blair et al. 1984; Pescatello et al. 2004), and glucose tolerance (Moreau et al. 2001; Swartz et al. 2003). Traditionally, PA has been measured with questionnaires. These self-report instruments generally have good validity for recall of vigorous, structured PA but are less valid when it comes to capturing ubiquitous, light-to-moderate activity. Thus, researchers have become interested in devices that objectively monitor PA.

Electronic pedometers have the potential to provide inexpensive, accurate, and reliable measures of ambulatory activity. Pedometers and accelerometers do not capture all types of activity (e.g., they cannot assess swimming, cycling, and arm activities), but they excel in the measurement of walking-based activities. Walking is one of the most common types of leisure-time and transport-related PAs, and many domestic and occupational tasks also result in step accumulation. The main disadvantage of pedometers, compared to accelerometers, is that they cannot distinguish the pattern of activity (i.e., frequency, intensity, and duration). However, for the total accumulation of ambulatory activity, both types of devices give similar information. For instance, pedometers show high correlations with total counts from accelerometers (r=0.80–0.90) (Tudor-Locke et al. 2002b).

Pedometers have a number of useful applications. Because they are inexpensive devices that can help motivate people to increase their activity levels, they are often used in PA promotion campaigns (Lindbergh 2000; Pronk 2003). Pedometers are also useful for longitudinal studies employing walking interventions (Croteau 2004; Moreau et al. 2001; Swartz et al. 2003; Yamanouchi et al. 1995). In addition, pedometers have been used in descriptive, epidemiological studies (McCormack et al. 2003; Sequeira et al. 1995) and to compare different populations around the world (Vincent et al. 2003). Most researchers report pedometer data as “steps,” since that is the most direct expression of what the pedometer measures (Tudor-Locke and Myers 2001). Although some pedometers display data on distance and energy expenditure, the accuracy of these measurements is not as great as for steps (Bassett and Strath 2002; Crouter et al. 2003).

Various laboratories have studied the accuracy of electronic pedometers for step counting during treadmill and track walking (Crouter et al. 2003; Leenders et al. 2003; Schneider et al. 2003). In these studies, the criterion was direct observation. However, examination of pedometer accuracy under free-living conditions is problematic because it is not feasible to perform direct observation of step counts over 24 h. Thus, we decided to test other types of step counters to see if any of them was more accurate than waist-mounted pedometers, over a wide range of conditions.

Recently two ankle-mounted pedometers have been developed: AMP 331 (AMPAnkle; Dynastream Innovations Inc., Cochrane, AB, Canada) and the StepWatch 3 (SW-3Ankle; Cymatech Inc., Seattle, WA). Dynastream validated their device by having participants walk a prescribed number of steps (155) at different activity levels and found that the AMPAnkle gave mean values that were within ± 1% of actual steps under all conditions. To test the validity of the AMPAnkle for estimating distance they had participants walk five times around a 200 m track at three different speeds: slow (<45 m min−1), regular, and fast (>135 m min−1) and found that the AMPAnkle, on average, estimated the mean distance to within ±3% (Gildenhuys et al. 2003).

The SW-3Ankle is a new version of the Step Activity Monitor (SAM). It first became available in May 2004, and there are no published studies on the validity of SW-3Ankle. However, a previous version (SAM) was used in several studies and found to be more valid than waist-mounted pedometers, especially at slow walking speeds (<80 m min−1) and in those with gait impairments (Macko et al. 2002; Shepherd et al. 1999).

At slower walking speeds (<80 m min−1) the vertical acceleration of the hip is not always great enough to trigger the pedometer mechanism, resulting in an underestimation of the steps counted. Ankle-mounted pedometers, such as the SW-3Ankle and the AMPAnkle, should be able to overcome this problem because they respond not only to vertical acceleration, but also to horizontal acceleration.

The purpose of this study was to compare the validity of two ankle-mounted pedometers (SW-3Ankle and AMPAnkle) and two waist-mounted pedometers (SW-701Waist and NLWaist) under controlled and free-living conditions: part I-walking on a treadmill at six different speeds; part II-evaluation of potential sources of error including leg swinging, heel tapping, stationary cycling, and driving a car in city limits; and part III-wearing pedometers for 24 h.

Methods

Participants

Ten males (28±3.7 years) and ten females (28±3.9 years) were recruited from the University of Tennessee staff and student body and the surrounding community. Prior to participation in the study the participants were asked to read and sign an informed consent, which was approved by the University of Tennessee’s Institutional Review Board. They were also asked to complete a Physical Activity Readiness Questionnaire (PAR-Q) regarding their health status, and if they reported any contraindications to exercise they were not tested.

Anthropometric measurements

Participants had their height and mass measured (in light clothing, without shoes) using a stadiometer (Seca Corp., Columbia, MD) and a physician’s scale (Health-o-meter, Inc., Bridgeview, IL), respectively. Body mass index (BMI) was calculated by the formula: body weight (kg)×[height (m)]−2. Circumference measures of the waist and hip were taken according to the guidelines established by Lohman, Roche, and Martorell in the Anthropometric standardization reference manual (Callaway et al. 1988). Waist circumference was measured at the narrowest part of the torso (above the umbilicus and below the xiphoid process) and hip circumference at the maximal circumference of the hips or buttocks region (above the gluteal fold). Waist-to-hip ratio (WHR) was computed by dividing the waist circumference (cm) by the hip circumference (cm).

Pedometers

Participants wore two waist-mounted pedometers and two ankle-mounted pedometers during all parts of the study. All pedometers were positioned according to the manufacturer’s instructions.

The Yamax Digiwalker SW-701 (SW-701Waist) was placed on the right waistband in line with the patella. This pedometer uses a spring-suspended horizontal lever arm that moves up and down in response to the hip’s vertical accelerations. This movement opens and closes an electrical circuit; the lever arm makes an electrical contact and a step is registered. The SW-701Waist displays steps, energy expenditure, and distance traveled. The SW-701Waist does not have the ability to store data in specified time intervals, so the values displayed represent the cumulative value for the time it was worn. For the purpose of this study, steps and distance traveled were examined. For the estimate of distance traveled, the participants’ stride lengths were input into the pedometer. To determine the stride length, the participants were asked to take 20 steps at their normal walking speed down an indoor hallway. The length traveled was divided by 20 to obtain their stride length. This was performed twice and an average was taken.

The New Lifestyles NL-2000 (NLWaist) was placed on the left waistband in line with the patella. The NLWaist uses a piezo-electric accelerometer mechanism that has a horizontal cantilevered beam with a weight on the end, which compresses a piezo-electric crystal when subjected to acceleration. This generates voltage proportional to the acceleration and the voltage oscillations are used to record steps. Participant data (age, height, weight, and gender) can be input, so the net and gross energy expenditures can be displayed, although for the purpose of this study we only examined the step counts given. The NLWaist also has the ability to store 7 days of data in 1D epochs.

The StepWatch 3 (SW-3Ankle) was worn on the lateral side of the right ankle, using an elastic band with a hook-and-loop closure. The device measures 75×50×20 mm3 and weighs approximately 38 g. The SW-3Ankle uses an accelerometer, which measures directional (horizontal and vertical) acceleration to detect steps. For this study, the SW-3Ankle was programmed with the participants’ height, and the default response (normal) was used for questions pertaining to the participants’ “walking speed,” “range of speeds,” and “leg motion”. The SW-3Ankle can be set to record data in 3–255 s epochs. When a 3 s epoch is used it can store 60 days worth of data. For this study we had the SW-3Ankle record data in 1 min epochs (default setting), which was then downloaded to a computer using a docking station for determination of steps taken. The SW-3Ankle does not have a digital display, so the device must be downloaded to a computer to obtain the recorded information. The SW-3Ankle only records steps taken on one leg; therefore to determine the total steps taken for the SW-3Ankle the value displayed was multiplied by two.

The AMP 331 (AMPAnkle) was placed in a neoprene case which was placed securely around the left ankle, with the pod directly over the Achilles tendon. This device uses two accelerometers that measure acceleration of the shank in the horizontal and vertical directions throughout the gait cycle to count steps and measure stride length. The AMPAnkle is programmed with the participants’ gender, date of birth, height, and weight prior to testing. The user has the ability to choose various epochs for data storage; 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 12, 15, 20, 30, or 60 min. For the purpose of this study we had the device store the data in 1 min epochs, which were then downloaded to a computer using a docking station. The AMPAnkle has a digital display that can be used to obtain elapsed time, duration of data collection, total steps taken, total distance walked, average walking speed, and total energy expenditure. When downloaded to a computer, in addition to the information found on the digital display, the recorded values can be viewed based on the epoch chosen. For example, we used 1 min epochs which allowed for the viewing of minute-by-minute data. For this study, steps taken and distance traveled were examined.

Part I: treadmill walking

Participants walked on the treadmill (Quinton, Medtrack ST Control, Bothell, WA) at speeds of 27, 40, 54, 67, 80, and 107 m min−1 for 3 min at each speed, while wearing the pedometers. Before testing, treadmill speed and grade were calibrated according to the manufacturer’s instruction and the speed was verified using a hand-held digital tachometer (Nidec-Shimpo America Corp. Model DT-107, Itasca, IL) that had been calibrated to an accuracy of within ±0.1%. An investigator counted the number of steps taken using a hand-tally counter. The total walking distance was determined from the calibrated treadmill speed and bout duration.

Part II: heel tapping, leg swinging, car driving, and cycle ergometry

For the heel-tapping trial, the participants were in a seated position and tapped both heels lightly against the floor. During the leg swinging trial the participants sat on a table and swung both legs together. Both the heel tapping and leg swinging were performed for 3 min in beat to a metronome set at a rate of 120 times min−1. On the cycle ergometer the participants rode at a cadence of 60 rpm for 3 min.

Ten of the 20 participants volunteered to drive their car while wearing all four pedometers. The driving took place over a 6.4 km course on city streets and the principal investigator accompanied the participants during that time. Just before starting out around the driving loop, the waist-mounted pedometers were reset to 0 and the starting time was recorded. When the driving loop was completed, the values from the waist-mounted pedometers were noted, and the ending time was recorded. After the trials, data from the ankle-mounted pedometers were downloaded to a computer, and the number of steps recorded during each activity was noted.

Part III: 24 h study

A subset of 15 participants volunteered to wear all pedometers for a 24 h period. They were instructed to wear the devices for 24 h except when sleeping, bathing, and swimming and were encouraged to go about their normal daily activities. Participants were shown how to correctly position the pedometers. They were given a data sheet to record what time the devices were put on and taken off and the number of steps displayed by the waist-mounted pedometers at the end of the day.

Statistical treatment

The data were analyzed using SPSS 11.5.0 for Windows (SPSS Inc., Chicago, IL). For all analyses, an alpha of 0.05 was used to indicate statistical significance. For the treadmill walking portion of the study the actual steps, NLWaist steps, and SW-701Waist steps were divided by three to obtain steps per minute. The steps per minute values from the AMPAnkle and SW-3Ankle were taken from the time corresponding to the second minute of each treadmill speed. Due to the small sample size we could not examine the pedometer × speed interaction for steps and distance. Therefore, one-way repeated measures ANOVAs with five levels (four pedometers and actual steps/distance) were used to analyze steps per minute and distance at each speed. Where appropriate, post hoc analyses were performed using Bonferoni corrections.

Modified Bland–Altman plots were used to graphically show the variability in individual step counts per minute, at all speeds, around zero (Bland and Altman 1986). This allows for the mean error score and 95% prediction interval to be shown. Devices that are accurate will display a tight prediction interval around zero. Data points below zero signify an overestimation, while points above zero signify an underestimation.

One-way repeated measures ANOVAs with five levels (four pedometers and actual steps) were used to determine if the pedometers recorded a significant number of non-ambulatory steps during heel tapping, leg swinging, cycling, and car driving. For heel tapping, leg swinging, and cycling the steps from the NLWaist and SW-701Waist were divided by three to get steps per minute, and the downloaded steps per minute values from the SW-3Ankle and AMPAnkle were used. For car driving, total steps recorded during the 6.4 km course were used. A one-way repeated measures ANOVA with four levels (representing the four pedometers) was used to compare differences in 24 h step counts between the pedometers. Where appropriate, post hoc analyses were performed using Bonferoni corrections.

Results

The physical characteristics of the participants are shown in Table 1. There were significant differences between pedometers for steps taken at 27 m min−1 (F4,16=56.383, P<0.001), 40 m min−1 (F4,16=7.335, P=0.001), 54 m min−1 (F4,16=6.117, P=0.003), 67 m min−1 (F4,16=3.240, P=0.040), 80 m min−1 (F4,16=7.089, P=0.002), and 107 m min−1 (F4,16=8.498, P=0.001). For all analyses the observed power was greater than 0.94, except for 67 m min−1, which had an observed power of 0.7. Figure 1 shows the steps recorded as a percentage of the actual steps taken at each speed. On average the participants took 68±9.2 steps min−1 at 27 m min−1, 84±9.2 steps min−1 at 40 m min−1, 95±7.4 steps min−1 at 54 m min−1, 105±6.3 steps min−1 at 67 m min−1, 114±5.1 steps min−1 at 80 m min−1, and 128±5.5 steps min−1 at 107 m min−1. The SW-3Ankle was the only pedometer that gave mean step counts within 1% of actual steps at all speeds. The other pedometers tended to underestimate the actual steps at the slowest speeds, but accuracy improved as the speed increased. The mean step counts for the AMPAnkle and NLWaist were within 3% of actual steps at 67 m min−1 and faster, while the SW-701Waist was within 1% of actual steps at 80 m min−1 and faster.
Table 1

Physical characteristics of participants (mean ± SD)

 

Men (n=10)

Women (n=10)

All participants (n=20)

Age

28±3.7

28±3.9

28±3.7

Height (cm)

179±5.9

165±6.0

172±9.3

Weight (kg)

92±23.2

63±7.2

77±22.6

BMI (kg m−2)

28.9±7.5

23.0±2.0

26.0±6.1

Waist (cm)

95±15.9

74±4.9

84±15.6

Hip (cm)

110±13.4

98±9.1

104±12.6

WHR

0.86±0.1

0.74±0.04

0.80±0.08

Stride length (cm)

70±6.4

72±5.7

71±5.9

BMI body mass index; WHR waist-to-hip ratio

Fig. 1

Effect of treadmill walking speed on pedometer accuracy for counting steps (n=20). Error bars are standard deviation. *Significantly different from actual steps (P<0.05)

Figure 2 shows the individual error scores in step counts per minute across all treadmill walking speeds. Since most pedometers show greater mean errors at slower walking speeds a quadratic curve was used to represent the mean error and 95% prediction intervals. In terms of accuracy the SW-3Ankle was the most accurate across all walking speeds with a mean difference (actual minus SW-3Ankle) of 0.9 steps min−1 and a prediction interval of −2.3 to +4.1 steps min−1. The other three pedometers greatly underestimated steps at slow walking speeds, and there was far more variability in individual error scores across all speeds.
Fig. 2

Modified Bland–Altman plots depicting error scores (actual steps per minute minus pedometer steps per minute) for each pedometer across all walking speeds: a Step Watch 3 (SW-3Ankle), b AMP-331 (AMPAnkle), c New Lifestyles NL-2000 (NLWaist), d Digiwalker SW-701 (SW-701Waist). Dashed lines represent mean difference; solid lines represent 95% prediction intervals

Two pedometers (SW-701Waist and AMPAnkle) displayed the distance traveled. There were significant differences between pedometers for distance traveled at 27 m min−1 (F2,18=22.905, P<0.001), 54 m min−1 (F2,18=11.004, P=0.001), 67 m min−1 (F2,18=18.082, P<0.001), 80 m min−1 (F2,18=18.794, P<0.001), and 107 m min−1 (F2,18=64.623, P<0.001), while there was no difference in distance at 40 m min−1 (F2,18=1.867, P=0.183). For all analyses the observed power was greater than 0.97, except for 40 m min−1 which had an observed power of 0.34. Figure 3 shows the steps recorded as a percentage of the actual distance traveled at each speed. The AMPAnkle underestimated the distance traveled at all speeds. At speeds of 40 m min−1 and above, the AMPAnkle provided mean estimates that were within 11% of the actual distance traveled. The SW-701Waist underestimated the distance traveled at all speeds except at 40 and 54 m min−1. For speeds between 40 and 80 m min−1 the SW-701Waist was within 7% of the actual distance.
Fig. 3

Effect of treadmill walking speed on pedometer estimates of the distance traveled (n=20). Error bars are standard deviation. *Significantly different from actual distance (P<0.05)

Table 2 shows the mean steps detected during heel tapping, leg swinging, cycle ergometry, and car driving. The ANOVA tests revealed significant differences between devices for heel tapping (F3,17=7.653, P=0.002), foot swinging (F4,16=5395.538, P<0.001), cycling (F4,16=19965.015, P<0.001), and car driving (F2,18=11.777, P=0.004). The observed power was greater than 0.94 for all tests. In general, the waist-mounted pedometers (NLWaist and SW-701Waist) recorded very little artifact with these activities. The SW-3Ankle was responsive to heel tapping, leg swinging, and cycling, while the AMPAnkle was responsive to leg swinging.
Table 2

Mean steps (± standard deviation) detected during heel tapping, leg swinging, car driving, and cycle ergometry

 

Heel tapping (steps min−1, n=20)

Leg swinging (steps min−1, n=20)

Cycle ergometry (steps min−1, n=20)

Car driving (total steps for 6.4 km route, n=10)

NLWaist

0.35±0.18

0.38±0.17*

5.9±3.8**

7.6±1.5***

SW-701Waist

0.27±0.11

0.17±0.08

3.3±1.3**

3.1±1.5

AMPAnkle

0.0±0.0

63.7±12.6*

5.0±2.6**

0.0±0.0

SW-3Ankle

28.7±6.4*

118.2±0.75*

120.2±0.41

0.0±0.0

*Significant detection of erroneous steps min−1 (P<0.05)

**Significant underestimation of cycles min−1 (P<0.05)

***Significant detection of erroneous steps in car driving (P<0.05)

During the free living condition (24 h trial) the participants wore the motion sensors for 13.7±1.1 h on average, similar to previous studies (Le Masurier et al. 2004; Leenders et al. 2000; Tudor-Locke et al. 2002a). The ANOVA showed significant differences between the devices over the 24 h period (F3,12=8.138, P=0.003), with an observed power of 0.95. For the 24 h data, the average steps recorded by each of the pedometers were: NLWaist, 11,087±970.2; SW-701Waist, 10,611±960.3; AMPAnkle, 10,269±854.7; and SW-3Ankle, 12,454±986.2 (Fig. 4). The only difference among the pedometers over the 24 h period was that the AMPAnkle recorded a significantly lower average step count versus the SW-3Ankle (P=0.013).
Fig. 4

Mean step counts (± standard deviation) recorded by pedometers over a 24 h period (n=15). Error bars are standard deviation. *Significantly different from SW-3Ankle (P<0.05)

Discussion

This study found that two new ankle-mounted pedometers provide superior validity at slower speeds (<67 m min−1). Specifically, the SW-3Ankle had mean step counts that were within 1% of the actual, at all speeds. Consequently, ankle-mounted pedometers such as the SW-3Ankle might be useful as a criterion for validating waist-mounted pedometers, over a wide range of speeds.

Our results are generally consistent with other studies indicating that waist-mounted pedometers have compromised accuracy in those who walk at slower speeds. For example, Cyarto et al. (2004) investigated the impact of walking speed and gait disorders on the accuracy of waist-mounted pedometers (Yamax Digiwalker SW-200) in older adults. The SW-200 underestimated steps by 74, 55, and 46% at slow (25.2±10.2 m min−1), normal (38.4±16.8 m min−1), and fast (48±21 m min−1) paces in the nursing home residents and by 25, 13, and 7% at slow (57±12 m min−1), normal (76.2±12 m min−1), and fast (96.6±12.6 m min−1) paces in seniors’ recreation center members, respectively. Similarly, Macko et al. (2002) investigated the effects of gait disorders on the accuracy of an ankle-mounted SAM and a waist-mounted pedometer (Elexis Trainer, Model 3FM-180; International Microtech, Miami, FL) in stroke patients. The SAM recorded 98.5% of actual steps, while the Elexis Trainer pedometer recorded approximately 87% of actual steps during a 1 min floor walk at self-selected paces (average cadence 46 steps min−1). The current study shows that the newer SW-3Ankle has similar accuracy to the previous version (SAM), and both have superior accuracy compared to other pedometers at slow walking speeds.

Two of the pedometers (SW-701Waist and AMPAnkle) estimated the distance traveled. The AMPAnkle consistently underestimated distance by about 10% at 40 m min−1 and above. This error is greater than that reported by the manufacturer, but it could be due to differences in the test protocol. The present study examined treadmill walking with the pod secured in a neoprene sleeve, while the manufacturer’s testing was done on a track and used athletic tape to secure the pod over the Achilles tendon. In our study, there was a tendency for the neoprene sleeve that the AMPAnkle was placed to rotate around the ankle during the walking trials, which could have affected the accuracy in the present trial. The SW-701Waist underestimated the distance by more than 10% for the slowest and fastest speeds, but provided mean estimates that were within ±10% of the actual distance at other speeds.

It is important to distinguish between how the AMPAnkle and SW-701Waist calculate the distance traveled. The SW-701Waist simply takes the input stride length and multiplies that by the number of steps taken to get the distance traveled. On the other hand the AMPAnkle measures each individual stride using accelerometer data to get the distance traveled. To do this they have developed what is called the “smart” stride detection algorithm. Rather than just using acceleration peaks, which could cause erroneous stride detection due to vibrations from outside sources, their algorithm uses the changes in the shank angle and the shank angular velocity during the gait cycle to detect strides, to get stride length (Dynastream Innovations Inc. 2003).

Some habitual movements such as heel tapping and leg swinging represent potential sources of error. The placement and the mechanism of the device are important in determining whether they detect steps in these types of movements. As expected, the waist-mounted pedometers detected virtually no steps during heel tapping and leg swinging, while the ankle-mounted pedometers recorded some erroneous steps. The SW-3Ankle recorded 29 erroneous steps during heel tapping and 118 erroneous steps during leg swinging, while the AMPAnkle recorded no steps during heel tapping and 64 erroneous steps during leg swinging. Nevertheless, the number of steps recorded during these movements would be unlikely to have a large impact over a 24 h period.

The SW-3Ankle detected 100% of cycle pedal revolutions, while the AMPAnkle, NLWaist, and SW-701Waist recorded only a few steps. While this could be considered a source of error, the ability of the SW-3Ankle to detect bicycling activity could be advantageous in some instances. For instance, the SW-3Ankle could be used to capture locomotion (e.g., walking, jogging, and cycling) rather than ambulation, per se. This may have a practical application in European countries where bicycling is a common mode of transportation. For example, residents of the Netherlands and Denmark take 28 and 20% of all trips by bicycle, respectively, whereas Americans take only 1% of trips by bicycle (Pucher and Dijkstra 2003).

Even though the NLWaist recorded a significant number of steps (eight steps) during car driving, from a practical standpoint this source of error would have a minimal impact. Americans drive an average of 63 km day−1 (Pucher and Renne 2003), which would cause approximately an extra 78 steps day−1 to be detected by the NL. Since healthy adults take between 7,000 and 13,000 steps day−1 (Tudor-Locke and Myers 2001), the number of erroneous steps is likely to be less than 1% of their daily total.

The SW-3Ankle detected no steps during car driving, where individuals can spend a substantial amount of time. This is in contrast to a CSA accelerometer which has been shown to record 168 erroneous steps during a 32.6 km automobile trip, which for the average American could result in a 4–7% error over the course of a day (Le Masurier and Tudor-Locke 2003). Previous studies have suggested that the CSA accelerometer is a good criterion for validation of other pedometers because it yields mean values within ±1.1% of the actual steps between 54 and 107 m min−1 (Tudor-Locke et al. 2002a). However, the SW-3Ankle is equally valid and is not susceptible to recording erroneous steps during car driving.

Over a 24 h period, the AMPAnkle gave a significantly lower step count than the SW-3Ankle by 18%. The SW-701Waist and NLWaist were not significantly different from the SW-3Ankle, but they still gave mean values that were 15 and 11% less, respectively, than that recorded by the SW-3Ankle. The main reason appears to be that the SW-3Ankle records a higher percentage of actual steps during slow walking and possibly lifestyle activities.

Ankle- and waist-mounted pedometers each have advantages and disadvantages. The ankle-mounted pedometers have better accuracy for step counting at slow speeds (especially the SW-3Ankle), which may make them more suitable for use in those who walk at slow speeds or have gait abnormalities. In addition, they could be used as a criterion method for validating less expensive waist-mounted pedometer. However, the SW-3Ankle and AMPAnkle are expensive. The SW-3Ankle costs $500, plus an additional $1,500 for the computer software and docking station. The AMPAnkle costs $450 per device, plus an additional $750 for the computer software and docking station. Both of these ankle-mounted devices can store step data using various epoch lengths, which allows researchers to examine the pattern of activity. The AMPAnkle also provides information on the amount of time spent in different intensity categories (i.e., light, moderate, and vigorous).

On the other hand, the waist-mounted pedometers (NLWaist and SW-701Waist) are good choices for measuring PA in healthy, free-living adults. Waist-mounted pedometers are low-cost (SW-701Waist, $25; NLWaist, $50) and some have the ability to store data in 1-day epochs. Although waist-mounted pedometers tend to underestimate steps at slow speeds, most healthy adults walk at around 80 m min−1 (Temes 1994). Using waist-mounted pedometers also enhances comparability with previous studies (Le Masurier et al. 2004; Le Masurier and Tudor-Locke 2003; Schneider et al. 2004; Tudor-Locke et al. 2001, 2002a, b). Finally, waist-mounted pedometers have greater ease-of-use, and are well suited for behavioral interventions since they provide immediate feedback to the user.

In conclusion, the SW-3Ankle could serve as a useful criterion measure of steps, because it has superior validity and accuracy, especially at slow walking speeds. However, the SW-3Ankle is more likely to record erroneous steps during non-ambulatory activities such as heel tapping and leg swinging. In our view, these non-ambulatory activities probably account for only a very small percentage of total daily steps. Over 24 h, the SW-3Ankle gave higher estimates of steps day−1 than the AMPAnkle, because it detects a greater percentage of actual steps taken at slow walking speeds.

Acknowledgments

The authors would like to thank Cary Springer (UTK Statistical Consulting Services) for assisting with the statistical analyses. No financial support was received from any of the pedometer companies. Dynastream Innovations, Inc. loaned the AMP-331 equipment for the duration of the study. The results of the present study do not constitute endorsement of the products by the authors or ACSM.

Copyright information

© Springer-Verlag 2005

Authors and Affiliations

  • Murat Karabulut
    • 1
  • Scott E. Crouter
    • 2
  • David R. BassettJr.
    • 2
  1. 1.Department of Health and Exercise ScienceUniversity of OklahomaNormanUSA
  2. 2.Department of Exercise, Sport, and Leisure StudiesThe University of TennesseeKnoxvilleUSA
  3. 3.Division of Nutritional SciencesCornell UniversityIthacaUSA

Personalised recommendations