Article Text
Abstract
Objective To determine the relative accuracy during treadmill walking and motor vehicle travel of the ActivPAL, PALlite and Digi-Walker accelerometers.
Methods Forty healthy volunteers wearing all accelerometers undertook either five treadmill walks (n=20) at speeds ranging between 0.6 and 1.4 m/s or a 15 min motor vehicle journey (n=20). Step counts recorded by each accelerometer were compared with the actual step count determined by video analysis (treadmill walking) or with an actual step count of zero (motor vehicle). Mean percentage measurement error was calculated and compared between devices by one-way ANOVA and Student t test.
Results For treadmill walking, the measurement error was lowest for the ActivPAL, with no significant differences between the ActivPAL and the PALlite monitors. The measurement error was significantly higher for the Digi-Walker at speeds of ≤1 m/s. During vehicle travel erroneous steps were recorded by the PALlite (254 steps) and Digi-Walker (25 steps), but not the ActivPAL monitor (0 steps).
Conclusions The ActivPAL accelerometer accurately measures step count over a range of walking speeds and, unlike the other accelerometers tested, is not falsely triggered by motor vehicle travel.
Statistics from Altmetric.com
Introduction
There is increasing use of physical activity level as an outcome measure.1 2 Various types of accelerometer have been used to try to provide an objective measure of one or more components of physical activity, such as step count.1 3 An accurate step count is important as an outcome per se, and because it is used in the calculation of other outcomes, for example, distance walked, time spent walking.
Pedometers are the least expensive means of measuring a step count. They are easy to use and provide immediate feedback, but tend to become less accurate at slower walking speeds, that is, ≤1 m/s.4 5 The more technically complex accelerometers appear more accurate, can log activity against time and provide additional data on physical activity such as time upright.1 However, their cost is usually severalfold higher than pedometers and some models are ‘oversensitive’, recording erroneous steps during non-ambulatory movements or vibrations, for example, during motor vehicle travel.5
In this study, we have compared the relative accuracy of three accelerometers during treadmill walking and motor vehicle travel, in order to help inform our choice of accelerometer for use in our research programme.
Methods
Subjects
Forty healthy volunteers (15 male, mean (SD): age 28 (8) years, height 169 (8) cm, body mass index (BMI) 22.9 (2.4) kg/m2) were recruited.
Instruments
ActivPAL (PAL Technologies, UK)
This 5×3.5×0.7 cm, 20 g, £600 accelerometer is attached to the anterior thigh and uses piezoelectric technology to detect acceleration and change in posture. In addition to step count, it can record time spent sitting, standing and walking, sit to stand transitions and cadence rates for up to 10 days.6 7
PALlite (PAL Technologies, UK)
This 5×3.5×0.7 cm, 20 g, £200 accelerometer is attached to the ankle and uses piezoelectric technology to detect acceleration. Changes in electrical charge are logged against an internal clock to record a step count for up to 10 days.
SW-401 Yamax Digi-Walker (New Lifestyles Inc., Missouri, USA)
This 5×3.5×1.5 cm, 20 g, £20 pedometer is attached to the waistband in line with the anterior superior iliac spine and uses mechanical technology (a lever spring) to detect acceleration. The Yamax series are considered the most accurate and reliable pedometers.4 8
Treadmill walking
A mechanical belt treadmill (Enraf Nonius, Holland) located in a human performance laboratory was calibrated for speed on the flat by a medical physicist using a tachometer. After a period of walking at the lower and higher speeds for familiarisation purposes, the accelerometers were attached. Subjects undertook five treadmill walks, each of 5 min duration, starting at 0.6 m/s and increasing to a maximum of 1.4 m/s, reflecting the range of comfortable walking speeds in healthy older people.9Subjects were asked to avoid shuffling when the treadmill was accelerating or coming to a stop. The total duration of these periods did not exceed 10 s and steps taken during these periods were included. Step counts recorded by each accelerometer were noted after each walk. Actual step counts were determined from video recordings, viewed independently by two researchers at half speed.
Motor vehicle travelling
Subjects occupied the seat of a car, the accelerometers were attached and the engine started. A standardised route incorporating town, housing estate and single carriageway driving was followed for as close to 15 min as possible dependent upon finding a safe place to stop. Step counts for this period from each accelerometer were obtained.
Statistical analysis
Normally distributed data were expressed as mean±SD. Step counts recorded by each accelerometer were compared with the actual step count. Mean percentage error for treadmill walking was calculated using individual error values regardless of directionality, that is, underestimates or overestimates of step count. For motor vehicle travel, 15 min erroneous step counts were multiplied by four to calculate 1 h estimates. Measurement errors and their 95% CI were compared between devices by one-way ANOVA and Student t test. A p value of <0.05 was regarded as statistically significant. A sample size of 20 was based on a similar study5 and is sufficient to detect a 10% difference between actual and measured step counts (80% power, p=0.05).
Results
Treadmill walking
The mean number of steps for 20 subjects ranged from 375 (0.6 m/s) to 578 (1.4 m/s) with estimated stride lengths of 0.51 cm and 0.73 cm, respectively. Mean measurement error was low (<4%) and did not differ between the ActivPAL and the PALlite accelerometers. Compared with the ActivPAL and PALlite, the Digi-Walker tended to underestimate the number of steps taken and was significantly less accurate at speeds of ≤1 m/s (table 1).
Motor vehicle travelling
Ten journeys, each with two subjects, were undertaken lasting a mean duration of 15 min 3 s (range 14 min 59 s to 15 min 11 s) over distances of 9.3–9.8 km. The ActivPAL did not record any erroneous steps. Both the PALlite and Digi-Walker recorded a significant number of erroneous steps, the former 10-fold more than the latter (table 2).
Discussion
In this comparison, the ActivPAL accelerometer performed best, with the accuracy of the other two accelerometers compromised during slow walking (Digi-Walker) and/or motor vehicle travel (PALlite and Digi-Walker). Differences in accuracy may relate to differences in the sensitivity threshold, site of attachment, and the algorithm used in the processing of movement data.2 10
This study adds to existing work by examining the accuracy of accelerometers at walking speeds that are representative of older patient groups, and providing a direct comparison of accelerometers of varying cost.6 9 In addition, we have highlighted the importance of accurate measurement during motor vehicle travel and recommend, like others,5 that this aspect of free-living physical activity is routinely evaluated in validation studies.
We are further exploring the utility of the ActivPAL accelerometer in patients with cancer and a wide range of physical abilities. This includes examining its acceptability and identifying the optimal period of monitoring.
What is already known on this topic
An accurate step count is important for outcome measures concerning physical activity level.
There is little data on how accelerometers perform at slow walking speeds and during motor vehicle travel.
What this study adds
The ActivPAL monitor accurately measures step count over a range of walking speeds and is not falsely triggered by motor vehicle travel.
Accurate measurement during motor vehicle travel is important to consider in validation studies.
Acknowledgments
We thank Mr Nayan Karsan for assistance with data collection and Dr Victoria Owen for statistical support.
Footnotes
-
Competing interests None.
-
Ethics approval Ethics approval was obtained from the University of Nottingham Medical School Ethics Committee (PT/I/8/2007, A/6/2007).
-
Patient consent Obtained.