The positive relationship between physical activity and health is well documented (8,19); however, the study of this relationship is limited by the ability to accurately measure physical activity. Objective physical activity monitoring technologies (e.g., accelerometers) provide robust measures of physical activity compared with questionnaire-based instruments (14) and are beginning to be used in population-based surveillance (1,15,16). However, because accelerometer technology is evolving and because the calibration of these devices is somewhat contentious (16), those using accelerometers in surveillance have begun to supplement their data with step counts, which provide a simple, stable metric to monitor ambulatory physical activity (17). Evidence suggests that step-based ambulation accounts for the majority (3) of physical activity energy expenditure, providing an objective measure that captures a significant portion of daily physical activity. Therefore, capturing step counts in long-term population physical activity surveillance is advisable.
The primary tool used for the measurement of step counts is the pedometer (13). Although pedometers are reasonably accurate at counting steps, they cannot discriminate between steps accumulated in walking, running, or stair climbing. In an effort to provide information on activity intensity, some pedometer manufacturers have developed models that measure time-stamped step counts (e.g., Lifecorder EX, New Lifestyles). These new-generation pedometers use steps accumulated per unit of time, to estimate intensity level. A better solution for those that require the detailed outcome variables provided by an accelerometer (6) as well as the simple step count provided by a pedometer is to use a dual-mode accelerometer. For example, the Actigraph (Fort Walton Beach, FL) has an auxiliary function that provides simultaneous, time-stamped measurement of both accelerometer counts and step counts. Recently, another commonly used accelerometer, the Actical (Respironics Inc., Murrysville, PA) was modified to function as a dual-mode accelerometer (measuring accelerometer counts and step counts). However, unlike the step count function in the Actigraph, which has been validated previously (11), no external validation has been performed on the step count function of the Actical accelerometer. Therefore, the primary purpose of this study was to assess the criterion validity of the Actical step count function using two methodologies: 1) technical assessment using a mechanical shaker table, and 2) a practical assessment using visually counted steps during three bouts of treadmill ambulation. A secondary purpose was to assess the concurrent validity of the Actical step count function with a previously validated dual-mode accelerometer (i.e., Actigraph 7164).
Eight Actical accelerometers were initialized to collect data using 1-min epochs. The accelerometers were mounted to the surface of a mechanical shaker table. As described previously (7), the shaker table was driven by a hydraulic cylinder (Sheffer, 1-1/18HHSL6ADY) controlled by an electrohydraulic servo valve with cylinder position feedback. A position transducer (Lucas 5000, DC-E) was used to measure the position of the table, and a high-grade control accelerometer (calibrated at 98.1 mV·g−1 ± 3.6%) (B and K model 4371) was attached to the table to measure vertical acceleration. The acceleration signal was transmitted to a charge amplifier (B and K model 2635) and band-pass filtered at 3 kHz. The amplifier input was provided by a function generator, which was programmed to accurately and reliably oscillate the platform at the various testing conditions using a sinusoidal oscillation procedure. Care was taken to ensure that the monitors were secured firmly and positioned vertically along their sensitive axis to standardize the output of the piezosensors. Once the accelerometers were in place, the hydraulic shaker table was switched on and the first of the randomly ordered conditions was initiated, thereby accelerating all eight monitors simultaneously. Before data collection, the shaker table was warmed up to ensure optimal functioning of the hydraulics and the control electronics.
Six different conditions were selected to produce a range of physiologically relevant counts from low to high intensity (Table 1) within the limitations of the shaker table. All conditions began at the turn of a new minute on the initialization computer clock, which was recorded along with the condition end time for data-analysis purposes. This portion of the data collection lasted approximately 60 min (12-min warm-up + (six conditions × 7 min per condition) + (6 × 1 min between conditions)). The first and final minutes of each shaker condition were ignored to ensure that the data used for analysis (i.e., the middle 5 min) were free from errors attributable to variability in accelerometer synchronization. Intra- and interinstrument reliability was calculated for each of the six shaker conditions. The shaker table oscillations per minute acted as the criterion validity reference for the measurement of steps.
A convenience sample of 38 volunteers, ranging in age from 9 to 59 yr, participated in the study. The study was approved by the University of Saskatchewan and the Health Canada research ethics boards. Each participant signed a written informed consent form (parental consent for minors (<18 yr of age)) and completed a Physical Activity Readiness Questionnaire (PAR-Q) before participating in the study. A positive response to any of the PAR-Q questions excluded a participant from the study. Blood pressure was measured before and after activity to ensure appropriate hemodynamic recovery; no adverse incidents occurred.
Participants had their height, sitting height (while sitting in erect position on a flat wooden box), and weight measured (in light clothing, without shoes) before the testing session, using a calibrated digital stadiometer (ProScale M150, Accurate Technology, Inc., Fletcher, NC) and scale (2256 VLC, Mettler Toledo, Columbus, OH). Subjects' leg length was calculated by subtracting their sitting height (minus the box height) from their standing height. Body mass index (BMI) was calculated by dividing body mass (kg) by height squared (m2).
Before initiating the treadmill testing, each participant was fitted with eight Actical and eight Actigraph 7164 accelerometers (specifications described elsewhere (7)). Four Actical and four Actigraph accelerometers were positioned vertically in each of two nylon pouches and were attached with a belt to the waist of each participant. The midline of each pouch was aligned with the left and right anterior iliac spines. After a warm-up, participants performed three 6-min bouts of treadmill ambulation: slow walk (50 m·min−1), normal walk (83 m·min−1), and run (133 m·min−1) at 0% grade on a speed- and grade-calibrated treadmill (Reebok 5500C, Canton, MA). After each treadmill bout, participants were given a 1- to 5-min rest period (self-selected rest duration).
The Actical and Actigraph accelerometers were initialized to produce synchronized results at 1-min epochs. Similar to the technical assessment, the first and last minutes of each treadmill bout were ignored to ensure that the data used for analysis (i.e., the middle 4 min) were free from errors attributable to variability between participants during the mounting/dismounting of the treadmill. Steps were also counted by a trained observer using a handheld tally counter during the second and fourth minutes of each treadmill condition. Visually counted steps acted as the criterion reference for the measurement of steps. Finally, the concurrent validity of the Actical and Actigraph step count function was assessed.
For both the technical and practical assessments, the intrainstrument (i.e., within accelerometer for a given model) step count reliability was calculated using standard deviation, standard error of the measurement, and coefficient of variation (CV). More specifically, the intrainstrument reliability measures were calculated from the replicate minutes (i.e., minutes 1-5 for each of the six shaker conditions and minutes 1-4 for each of the three treadmill speeds). This minute-by-minute variability characterizes the accelerometers' ability to consistently measure the given condition rendered by the shaker table and/or treadmill.
For both the technical and practical assessments, the interinstrument (i.e., between accelerometers for a given model) step count reliability was calculated using standard deviation, standard error of the measurement, and CV. More specifically, these interinstrument reliability statistics act to characterize the differences between the eight accelerometers for a given model (i.e., Actical or Actigraph).
Data from the practical assessment were examined for step count performance differences by sex, age (9-15, 16-49, and 50-59 yr), leg length (30.0-76.5, 76.5-79.3, and 79.3-90.0 cm), and BMI (obese (BMI ≥ 30 kg·m−2) vs nonobese (BMI < 30 kg·m−2)) using analysis of variance. Age, sex, leg length, and BMI did not have a significant effect on validity coefficients; therefore, data were collapsed into one group. Paired-samples t-tests also showed no systematic positional effect (i.e., no left- vs right-pouch effect, and no within-pouch left vs right or front vs back position effects). Therefore, all analyses were performed on the pooled data. Differences between criterion step counts and accelerometer (Actical and Actigraph) step counts were assessed using a paired-samples t-test. Differences between Actical and Actigraph step counts were assessed in the same fashion, and Pearson correlation coefficients were computed to assess the concurrent validity of the Actical versus the Actigraph. Significance levels were set at P < 0.05. All statistical analyses were carried out using SPSS version 15.0 (SPSS Ins., Chicago, IL).
As indicated in Table 1, the average steps per minute detected by the Actical matched identically with the shaker plate oscillations per minute. This relationship persisted across all six conditions; in fact, of the 240 condition minutes assessed (8 units × 5 min × 6 conditions), only 2 min were detected incorrectly (both minutes were off by a single step). Therefore, the correlation between steps detected by the Actical accelerometer and shaker table oscillations per minute was excellent (r = 1.00). Furthermore, the intra- and interinstrument variability of the Actical step count function was minimal (CVintra and CVinter < 0.1% (data not shown)). Comparing the Actical accelerometer count data with previous data from our lab (7) confirmed that the addition of the step count function to the Actical accelerometer did not negatively affect the accelerometer count function (data not shown).
The characteristics of the 16 male and 22 female participants are shown in Table 2. The treadmill step count reliability statistics are summarized by speed for both the Actical and Actigraph accelerometers in Table 3. In addition, Table 3 displays the mean difference in steps between the Actical and Actigraph devices and the criterion measure of visually counted steps. Both devices performed poorly at the slow walking speed, but no differences from the criterion were detected for either device at the normal walk and running speeds. In addition, modified Bland-Altman plots help refine the understanding of the error in step detection for both the Actical and Actigraph (Fig. 1). These plots reveal considerable variability at the individual level and highlight the fact that both accelerometers underestimate steps at the slow walking speeds while overestimating steps at running speeds greater than approximately 160 m·min−1.
Comparing the visually counted steps versus the steps detected by each of the accelerometers resulted in criterion-related validity correlations (r) of 0.73 and 0.52 at the slow walk condition and 0.99 and 0.99 at both the normal walk and run conditions for the Actical and Actigraph, respectively. Comparing the steps detected by each of the accelerometers between themselves resulted in concurrent validity correlations of 0.70, 1.00, and 1.00 for the slow walk, normal walk, and run conditions, respectively.
This is the first study to assess the performance of the step count function within the newly designed dual-mode Actical accelerometer. Another unique aspect of this study is that it employed both a technical assessment of validity using a mechanical shaker table and a more practical assessment using bouts of treadmill ambulation. Under the controlled conditions of the mechanical shaker table, the Actical step count function performed flawlessly in terms of its agreement with the criterion (i.e., the shaker table oscillations per minute). However, data from the treadmill assessment indicate that the accuracy of both the Actical and Actigraph step count functions differed from the criterion of visually counted steps at the slow walking speed, with both accelerometers underestimating the number of steps taken at this speed (errors of 7.4 and 5.3% for the Actical and Actigraph, respectively). Even the small errors detected here in the slow walking condition (i.e., five to seven steps per minute) can amount to large differences during a 24-h period.
The fact that step-counting accuracy is compromised at slow speeds may be one of the most consistent findings in the pedometer literature (2,5,10,11,18). Although this problem has been readily identified, it is difficult to correct because there is an inevitable sensitivity/specificity trade-off; the greater the sensitivity (i.e., ability to detect low step forces), the less the specificity (i.e., ability to discriminate between actual stepping movements and nonambulatory oscillations of one's center of gravity) (4,9).
To detect and register movement, the Actigraph dual-mode accelerometer must meet or exceed the manufacturer-set 0.30g acceleration threshold. Unfortunately, no such threshold-detection information is available on the Actical. If one presumes that the Actical movement-detection threshold is similar to that of the Actigraph, then it is not at all surprising that it preformed well during mechanical testing, because the lowest-intensity shaker condition was 0.50g. However, the greater step-counting error (i.e., the underreporting of steps) exhibited by both accelerometer models during the slow walk condition suggests that the gravitational forces imparted on the center of gravity of at least some of the participants may have been below the 0.30g threshold (although we have no data to prove this). The implication of these thresholds is that they will disproportionately affect populations who walk slowly or with a shuffling type of gait (e.g., the frail elderly) (12,20). Reflecting on the reliability data for the slow walk condition (Table 3), one can see that the interinstrument variability of the Actical is greater than that of the Actigraph. This discrepancy may be attributable to poor between-unit calibration at the low-intensity end of the Actical's dynamic range; however, this is speculative.
Data from this study suggest that the new step count function of the Actical is accurate at ambulatory velocities employed by most healthy people during locomotor behaviors (i.e., more than approximately 83 m·min−1). However, there was a tendency for both accelerometers to overestimate steps at higher speeds (i.e., more than approximately 160m·min−1). A possible reason for this overestimation could be a lack of specificity in terms of distinguishing between actual steps and spurious accelerometer movement caused by the bouncing of the accelerometers on the waist belt. Nonetheless, the Actical step count function performs similarly to that of the previously validated Actigraph across the speeds tested, and it does not differ from visually counted steps at faster speeds. However, care should be taken when generalizing these results beyond the present sample (e.g., young children and/or older adults (≥ 60 yr), or the severely obese). Also, it is unclear whether these results would be replicable in more free-living settings. Future work should assess the intra- and interinstrument variability of the threshold movement-detection levels of these and other activity monitors.
The authors would like to acknowledge Sean Schofield-Hurwitz for his assistance with the data collection and would like to thank the volunteer participants.
This study was not funded in any way by any of the accelerometer manufacturers. The results of this study do not constitute endorsement by the authors of the products described in this paper.
1. Andersen, L. B., M. Harro, L. B. Sardinha, et al. Physical activity and clustered cardiovascular risk in children: a cross-sectional study (The European Youth Heart Study). Lancet
2. Bassett, D. R., Jr., B. E. Ainsworth, S. R. Leggett, et al. Accuracy of five electronic pedometers for measuring distance walked. Med. Sci. Sports Exerc.
3. Bassett, D. R. Jr., A. L. Cureton, and B. E. Ainsworth. Measurement of daily walking distance-questionnaire versus pedometer. Med. Sci. Sports Exerc.
4. Crouter, S. E. Response to letter to the editor-in-chief. Med. Sci. Sports Exerc.
5. Crouter, S. E., P. L. Schneider, D. R. M. Karabulut, and J. R. Bassett. Validity of 10 electronic pedometers for measuring steps, distance, and energy cost. Med. Sci. Sports Exerc.
6. Esliger, D. W., J. L. Copeland, J. D. Barnes, and M. S. Tremblay. Standardizing and optimizing accelerometer data for physical activity monitoring. J. Phys. Act. Health
7. Esliger, D. W., and M. S. Tremblay. Technical reliability assessment of three accelerometer models in a mechanical set-up. Med. Sci. Sports Exerc.
8. Kesaniemi, Y. K., E. Danforth, M. D. Jansen, P. G. Kopelman, P. Lefebvre, and B. A. Reeder. Dose-response issures concerning physical activity and health: an evidence-based symposium. Med. Sci. Sports Exerc.
9. Le Masurier, G. C. Pedometer sensitivity and specificity. Med. Sci. Sports Exerc.
10. Le Masurier, G. C., and C. Tudor-Locke. Motion sensor accuracy under controlled and free-living conditions. Med. Sci. Sports Exerc.
11. Le Masurier, G. C., and C. Tudor-Locke. Comparison of pedometer and accelerometer accuracy under controlled conditions. Med. Sci. Sports Exerc.
12. Melanson, E. L., J. R. Knoll, M. L. Bell, et al. Commercially available pedometers: considerations for accurate step counting. Prev. Med.
13. Schneider, P. L., S. E. Crouter, and D. R. Bassett. Pedometer measures of free-living physical activity: comparison of 13 models. Med. Sci. Sports Exerc.
14. Shephard, R. J. >Limits to the measurement of habitual physical activity by questionnaires. Br. J. Sports Med.
15. Tremblay, M. S., and M. Wolfson. Canadian Health Measures Survey: background, rationale and overview. Health Rep.
16. Troiano, R. P. A timely meeting: objective measurement of physical activity. Med. Sci. Sports Exerc.
37(11 Suppl.):S487-S489, 2005.
17. Tudor-Locke, C. E., and A. M. Myers. Methodological considerations for researchers and practitioners using pedometers to measure physical (ambulatory) activity. Res. Q. Exerc. Sport
18. Tudor-Locke, C., J. E. Williams, J. P. Reis, and D. Pluto. Utility of pedometers for assessing physical activity: convergent validity. Sports Med.
19. U. S. Department of Health and Human Services. Physical Activity and Health: A Report of the Surgeon General
, 1st ed. McLean, VA: International Medical Publishing Inc., 1996.
20. Wilcox, S., C. E. Tudor-Locke, and B. E. Ainsworth. Physical activity patterns, assessment, and motivation in older adults. In: Gender, Physical Activity, and Aging
, R. J. Shepard (Ed.). Boca Raton, FL: CRC Press, pp.13-39, 2002.