Journal Logo

APPLIED SCIENCES: BIODYNAMICS

Comparison of pedometer and accelerometer measures of free-living physical activity

TUDOR-LOCKE, CATRINE; AINSWORTH, BARBARA E.; THOMPSON, RAYMOND W.; MATTHEWS, CHARLES E.

Author Information
Medicine & Science in Sports & Exercise: December 2002 - Volume 34 - Issue 12 - p 2045-2051
  • Free

Abstract

Recent technological advances have spurred a tremendous interest in objective monitoring of physical activity. Evidence of this shift is readily apparent from recent special supplement publications of both the Research Quarterly for Exercise and Sport (9) and Medicine & Science in Sports & Exercise® (14) in which most articles addressed the topic. Objective monitoring of physical activity by accelerometers is now supported by technology that is capable of capturing free-living physical activity information expressed as activity counts on a minute-by-minute basis for weeks at a time. For example, the CSA accelerometer (model 7164, version 2.2, Computer Science and Applications, Inc., Shalimar, FL) possesses a timing mechanism and a memory capacity that is capable of recording movement parameters over researcher-determined units of time. Activity count cut points have been developed in laboratory studies to translate these activity data obtained in the field into estimates of activity duration in specific intensity categories (6). This technology is not inexpensive, however; accelerometers may cost as much as $450 per unit and require technical expertise and additional hardware and software to calibrate, input, distill, and analyze data (19). Although accelerometers have become an important activity assessment tool, they may be less feasible for use in certain clinical applications (e.g., screening, program evaluation, self-monitoring) or for physical activity surveillance purposes because of the cost and technical requirements for their use.

A low-tech option for objective monitoring is the simple and inexpensive ($10–$50 per unit) pedometer. Unlike accelerometers, pedometers are not designed to capture pattern, intensity, or type of physical activity. They do, however, detect steps taken with acceptable accuracy (2,7). Pedometers have shown evidence of reliability (17) and convergent and discriminative validity (20). Although a number of electronic pedometers are commercially available, the only brand comparison study ever conducted reported that a pedometer manufactured by the Yamax Corporation (Model SW-500, Tokyo, Japan) was the most accurate at detecting steps taken, recording within 1% of all steps taken under controlled conditions (2). This particular model is no longer produced by the Yamax company (1). Another model of the Yamax pedometer (SW-200) has shown a strong relationship (r = 0.80–0.90) under laboratory conditions with more expensive accelerometers, including the CSA (3). Leenders et al. (8) reported significant correlations (r = 0.84–0.93) between pedometer-determined steps per day and Tritrac and CSA accelerometer daily total activity counts under controlled field conditions and concluded that the Yamax SW-200 pedometer output provided representative information on total accumulated activity.

A practical application of objective monitoring is to ascertain which individuals attain recommended minimal levels of health-related physical activity (e.g., 30 min in moderate-intensity activity). An index of pedometer steps per day is required to approximate these recommendations, specifically regarding ambulatory activity. Because accelerometers provide more information about activity than pedometers yet the pedometer is the more practical instrument for use in large populations, we wanted to describe intensity patterns of activity across steps taken. Therefore, the purposes of this investigation were 1) to evaluate further the relative concordance of total accumulated CSA outputs and Yamax pedometer outputs measured during concurrent monitoring under free-living conditions; 2) to investigate the relationship between pedometer-determined steps and CSA-determined time spent in inactivity and in light-, moderate-, and vigorous-intensity activities; and 3) to identify a value of pedometer-determined steps per day that could be used as a proxy for public health recommendations for time in moderate-intensity ambulatory activities (21).

METHODS

Participants.

A convenience sample (recruited by word of mouth and fliers within and around a university community) of 60 adult volunteers were enrolled in a 2-wk study to evaluate validity of the International Physical Activity Questionnaire in South Carolina. The CSA and Yamax pedometer were worn concurrently during waking hours by participants for seven consecutive days as a strategy to validate the physical activity questionnaire responses with objective measures of physical activity. Participants were instructed to go about their normal lives unrestricted. Of this original sample, 52 were white, 7 were black, and 1 was Asian. Before participation, all subjects read and signed an informed consent form. The Institutional Review Board of the University of South Carolina granted approval for this study. This study conforms to the policy statement regarding the use of human subjects and written informed consent as published by Medicine & Science in Sports & Exercise®.

Objective monitors.

A dual-mode CSA (model 7164, Version 2.2) that collects both activity count data and the number of cycles in the signal, or “cycle counts,” is now available. When the instrument is worn at the waist, cycle counts approximate the number of steps taken in the sampling interval (steps per minute) and can be totaled to represent accumulated steps taken over the monitoring frame. The instrument’s acceleration signal is filtered by an analog bandpass filter and digitized by an eight-bit A/D converter at a sampling rate of 10 samples per second storing data in researcher-defined intervals (e.g., 1 min) (18). The CSA requires a force ≥0.30 ×g to register and record a movement. The intensity of the movement is determined by the magnitude of the force in 0.05 ×g increments up to 2.0 ×g. The Yamax pedometer model SW-200 contains a horizontal, spring-suspended lever arm that deflects with the up-and-down motion of the hips during ambulation (e.g., walking). An electrical circuit opens and closes with each deflection detected, and an accumulated step count is displayed digitally on a feedback screen. The Yamax pedometer requires a force ≥0.35 ×g to register and record a movement and is unable to discriminate the magnitude of the force. Both the CSA accelerometer and the Yamax pedometer are small, lightweight, unobtrusive objective motion sensors that count movement and that can be worn concurrently with comfort. Instruments were worn at the waist. The pedometer was clipped to clothing, and the CSA was hung from a pouch attached to a snug elastic belt. Both instruments were checked for calibration (using a mechanical shaker) before use in the field. Participants were asked to write day-end values for pedometer steps on a simple calendar-type log.

Data treatment.

An automated data reduction program that we have previously used to run quality assurance checks and summarize accelerometer data was adapted for these analyses (10,11). Minute-by-minute data were summarized into daily averages for activity counts (counts per minute per day, total counts per day), cycle counts (CSA-steps per day), and activity durations (minutes per day) in specific intensity levels (inactive [0–499 counts·min−1], light [500–1951 counts·min−1], moderate [1952–5724 counts·min−1], and vigorous [5725+ counts·min−1]) (6,15). For these analyses, we further stratified the original light-intensity category of Freedson et al. (6) (i.e., 0–1951 counts) into the inactive and light categories (described above) in an effort to evaluate more carefully the lower end of the activity spectrum. The count cutpoints of Freedson and colleagues (6) were selected because the light- and moderate-intensity categories were specifically calibrated to ambulatory activity (i.e., walking). An overall average of the daily values from the accelerometer were obtained for each participant who met CSA compliance criteria defined as accumulating at least 3 d with a minimum of 12 h·d−1 of monitoring (i.e., 75% coverage for 16 waking hours). Pedometer data (day-end steps taken) were summed and divided by number of days worn to compute mean steps per day.

No equipment malfunction was noted. Participants’ data were excluded when they did not have concurrent CSA and pedometer data (four participants missing CSA data, two participants missing pedometer data). Only two participants did not meet the CSA minimal compliance criteria, and their data were therefore not included in these analyses. These data reduction strategies left data from 52 participants (27 men, 25 women) for analysis with complete 7-d data.

Data analysis.

Descriptive data are presented as means ± SD and 95% confidence intervals, having confirmed normal data distributions. A coefficient of variation (CV) for pedometer-derived physical activity was calculated as (SD/mean) × 100. Correlation analyses (Pearson product-moment correlation coefficients) were computed to quantify the linear relationship between CSA activity counts (expressed as both counts per minute per day and total counts per day), CSA-steps per day, and pedometer outputs (PED-steps per day). Dependent sample t tests were used to test differences in steps detected by the two instruments. Agreement between the CSA-steps per day and the PED-steps per day was evaluated using the Bland-Altman method (4), computed as the differences between CSA- and PED-steps per day plotted against the mean of both. This type of plot tests the relationship between measurement error and the best estimate of the true value, or the mean value of the monitors. Limits of agreement were calculated as the mean difference ± 2 SD (between CSA- and PED-steps per day) (4).

Independent t tests were used to test differences in steps per day between genders. Differences in age, BMI, and CSA-determined variables were tested between quartiles using a general linear model (PROC GLM) procedure following data stratification of PED-steps per day using the 25th, 50th, and 75th percentiles for distribution. In the case of a significant finding, post hoc analyses (Student-Newman-Keuls test) were used to compare specific differences between quartiles. Furthermore, effect size was calculated as the difference between the lowest and highest quartile means, divided by the group SD. Calculating effect size presents differences in relation to expected interindividual variation (16). An effect size of 0.2 is considered small, 0.5 is moderate, and 0.8 is large (5). Significance was set at an alpha level of P < 0.05. Statistical analyses were conducted using SAS Version 8.01.

A proxy value of pedometer steps per day for public health recommendations was determined using the mean steps per day associated with the quartile n which individuals accumulated an average of 30 min·d−1 of moderate activity.

RESULTS

Instrument concordance.

PED-steps per day averaged 9638 ± 4030 for the total sample (CV = 42%). Table 1 displays the descriptive variables, pedometer steps per day, CSA outputs, and CSA-derived time in increasing intensity levels for the total sample. CSA monitoring time averaged 14.4 ± 1.5 h·d−1; this translates to 90% coverage of a typical 16-h waking day (assuming 8 h·d−1 of sleep). There were no differences by gender for age, PED-steps per day, CSA activity counts (counts per minute per day, total counts per day) or CSA-steps per day. PED-steps per day were inversely correlated with age (r = −0.30, P = 0.02). The relationship between CSA output and age was also inverse but not significant (counts per minute per day vs age, r = −0.27, P = 0.06; total counts per day vs age, r = −0.23, P = 0.10). There was no relationship between CSA-steps per day and age (r = −0.15, P = 0.28). Adjusted for age (partial correlation), PED-steps per day were correlated with CSA-counts per minute per day (r = 0.74, P < 0.0001), total counts per day (r = 0.80, P < 0.0001), and CSA-steps per day (r = 0.86, P < 0.0001). Figure 1 displays the Bland-Altman plot depicting the difference between the CSA and pedometer step outputs (CSA-steps per day minus PED-steps per day). The mean difference in steps detected between devices was 1845 ± 2116 steps·d−1 (CSA > PED; t = 6.29, P < 0.0001). Differences were normally distributed. The limits of agreement (shown as bold lines) were −2387 to 6077 steps·d−1. In 7 of 52 cases, the pedometer detected more steps than the CSA. The mean difference in steps detected between devices was not statistically different between genders (t = 0.77, P = 0.44).

TABLE 1
TABLE 1:
Descriptive variables, pedometer steps per day, CSA outputs, and CSA-derived time in increasing intensity levels for total sample.
FIGURE 1
FIGURE 1:
Bland-Altman plot of error scores (CSA-steps per day minus PED-steps per day).

Relationship between PED-steps and CSA-time in intensity categories.

Tables 2 through 4 show a summary of the descriptive variables (age, BMI;Table 2), CSA outputs (counts per minute per day, CSA total counts per day;Table 3), and CSA-derived time in activities of increasing intensities (Table 4) for the total sample according to quartile of pedometer-assessed ambulatory activity. Results of the cross-quartile statistical comparisons and calculated effect size (between the highest and lowest quartiles) are also presented. Pedometer-determined physical activity was defined as quartile 1 (≤6529 PED-steps·d−1), quartile 2 (6530–9027 PED-steps·d−1), quartile 3 (9,028–12,571 PED-steps·d−1), and quartile 4 (≥12,572 PED-steps·d−1). There were no significant differences across pedometer quartiles in frequency of gender, mean age, or BMI. There were no significant differences between quartiles in instrument time monitored (according to CSA recording). CSA total counts per day and CSA-steps per day differed significantly between each pedometer-determined quartile, displaying an increasing pattern with each higher quartile. CSA counts per minute per day differed significantly only between the lower two and the upper two pedometer quartiles. The SD is inflated for each variable in the upper quartile because this quartile does not have an upper bound. The mean did not differ extraordinarily from the median in this quartile for any of the parameters discussed.

TABLE 2
TABLE 2:
Comparison of descriptive variables across quartiles of pedometer-determined ambulatory activity.
TABLE 3
TABLE 3:
Comparison of CSA outputs across quartiles of pedometer-determined ambulatory activity.
TABLE 4
TABLE 4:
Comparison of CSA-derived time in increasing intensity levels across quartiles of pedometer-determined ambulatory activity.

There were distinct differences in average time in summed moderate- and vigorous-intensity activity between each pedometer-determined activity quartile. Individuals in the highest quartile averaged 57.8 min·d−1 more in moderate and vigorous activities combined than those in the lowest quartile. Differences in time (minutes per day) spent in inactivity or light activity were not statistically significant across quartiles. All significant differences detected corresponded with a large calculated effect size (>0.80). Three nonsignificant differences (time monitored, age, and time in light-intensity activity) corresponded with moderate effect sizes (0.50–0.60). This finding suggests that if sample sizes had been larger, then these nonsignificant findings might have been statistically significant.

Identify a pedometer-step proxy for public health recommendations.

Individuals in the second quartile of pedometer-assessed activity averaged 32.7 ± 14.4 min·d−1 of moderate activity as determined by the CSA. The mean value of PED-steps per day in this quartile was 8064 ± 766.

DISCUSSION

To interpret pedometer-assessed ambulatory activity effectively, researchers, clinicians, and practitioners need normative values or benchmarks for interpreting change and comparison purposes (12). The mean PED-steps per day obtained in this sample is comparable to that expected for similarly aged, ostensibly healthy adult samples (20). Accumulated total CSA activity count data are infrequently reported, making similar comparisons between studies more difficult. The total counts per minute obtained in the present sample seems slightly lower than that reported during 1 d of monitoring (approximately 400,000–600,000 counts·d−1) in a younger sample (mean age = 25 yr) (15). Consistency of reported values obtained from objective monitoring of similar samples lends confidence to our findings.

One purpose of this article was to compare CSA accelerometer outputs and Yamax pedometer outputs measured under free-living field conditions. The high correlation (r = 0.80) observed between CSA- and PED-steps per day suggests a strong linear relationship between the two instruments. However, closer inspection of the data revealed a notable lack of agreement in steps detected between the two instruments, evident from the Bland-Altman plot (Fig. 1). The mean difference in steps detected by the CSA and the pedometer was approximately 2000 steps·d−1, and the limits of agreement between the instruments were broad (−2387–6077 steps). Cumulatively, these data suggest that one instrument’s step output may not be easily substituted for the other if an absolute number of steps taken is desired. However, the high correlation between the two instruments does support the interchangeability of the two motion sensors with regard to steps taken if only relative values are required.

Although there were seven cases in which the pedometer detected more steps than the CSA, the reverse was more often true. The more common disagreement in step output between the two instruments likely is due to differences in sensitivity to detect vertical accelerations. As stated previously, the CSA accelerometer requires a force ≥0.30 ×g to register and record a movement; the corresponding value for the Yamax pedometer is 0.35 ×g. Previous research has shown that the Yamax pedometer underestimates the number of steps taken at slower walking speeds (<60 m·min−1) by approximately 25% (2,7). Hendleman et al. (7) reported that this walking speed is much slower than self-selected normal walking speeds and would not, therefore, be expected to be an important source of error in field studies. Similar research has not yet been undertaken to evaluate the cycle count feature of the dual-mode CSA. It is plausible that the disagreement observed in this sample is frequently due to greater sensitivity of the CSA (relative to the pedometer) to lower range vertical accelerations; the CSA may be better able to detect slow steps and/or detect more nonstep movements as steps taken. This theory remains to be proved empirically, and it is not supported in seven cases in the present sample. In these few cases, it may be that the instruments were worn in some different manner. Similarly, although the mean difference in steps detected between devices was not statistically different between genders, the difference for men was consistently larger than that for the women. Plausible explanations for the apparent gender difference include 1) that men make more low-force movements that are detected more readily by the CSA than by the Yamax pedometer and/or 2) that there is a gender difference in how the two instruments are worn. For example, both instruments need to be held in a vertical plane to detect vertical oscillations. The pedometer was clipped to the clothing, but the CSA was suspended in a pouch from a snug belt. Although we are unable to verify the truth, it is plausible that the CSA was occasionally rotated away from the vertical plane in these cases. The true number of steps taken in this field study is not known, so we are unable to claim the superiority of one instrument over the other to detect accurately the true number of steps taken. A criterion measure of steps taken (e.g., observation and tally) is not feasible for such field studies of free-living populations. It is probably a more valid conclusion that each instrument detected (with acceptable accuracy) those steps that fell within their specific (and different) ranges of sensitivity. Researchers and practitioners should select an objective monitor that meets their needs in terms of cost, purpose, and instrument sensitivity. Whichever instrument is selected, the researcher needs to ensure that the motion sensor is held in the vertical plane at all times.

Pedometers offer the unique promise of a practical (inexpensive and feasible) method for surveillance, screening, program evaluation, and intervention through personal feedback. Despite the lack of agreement between the CSA accelerometer and the Yamax pedometer for steps taken, there was a strong relationship between their primary outputs (r = 0.74–0.86), consistent with the findings of others (3,8). The consistent strength of this relationship does permit us to explore the CSA-determined pattern of physical activity across pedometer-determined activity quartiles, a task related to the second purpose of this article. There were no differences in inactivity or light activity across pedometer quartiles despite increasing time in moderate- and vigorous-intensity activities with consecutively higher pedometer-derived quartiles of activity. A difference in time monitored could not explain this apparent paradox. Inspection of 95% confidence intervals showed considerable overlap of time between quartiles spent in light-intensity activities and in inactivity. There was a tremendous difference in absolute amount of time spent in these various categories compared with other intensity categories; a relatively greater portion of the day was spent at the lower end of the activity continuum (e.g., 815 min in inactivity and light activities combined) versus the higher end (e.g., 48 min in moderate and vigorous activities combined). This was true even in the most active individuals. To illustrate this point further, an absolute difference of 15 min spent in light activity between the first two quartiles was not statistically significant, whereas a similar absolute time difference in total moderate and vigorous activity was. Differences in time spent in moderate and vigorous activities across quartiles were evident and statistically significant. Individuals in the highest quartile (>12,572 PED-steps·d−1) engaged in more moderate activity (66 min) and more moderate and vigorous combined activity (79 min) than any other quartile group. The relevance of this finding is that individuals who accumulated more steps per day seemed to engage in moderate- to vigorous-intensity activities.

The final purpose of this article was to identify a value of pedometer-determined steps per day corresponding to the public health recommendations for time in moderate-intensity activity (13), specifically ambulatory activity. Evidence-based threshold values of pedometer-determined ambulatory activity would be useful for classification and intervention purposes (20). The U. S. Surgeon General (21) endorsed recommendations (13) for all individuals to accumulate 30 min or more of moderate activity on most, if not all, days of the week. It would be useful for researchers and practitioners using pedometers to have a steps per day value that is quantitatively linked with this recommendation. Individuals in the second quartile of pedometer-assessed activity averaged approximately 33 min·d−1 of moderate ambulatory activity as determined by the CSA corresponding with a mean value of approximately 8000 PED-steps·d−1. It is premature to conclude that this level of steps per day can be used as a proxy for public health recommendations for physical activity, specifically ambulatory activity, on the basis of this single study. The pedometer-derived quartiles emerged from data obtained from a small convenience sample with a narrow age range and BMI (evident from the lack of respective differences across quartiles). The pedometer data were similar to other published findings in similar populations, and the relatively large CV indicated that the sample displayed a range of activity levels. The findings are limited in generalizability, however, to populations of similarly aged, ostensibly healthy adults. Furthermore, although these arbitrary divisions may be indicative of higher levels of physical (ambulatory) activity, the pedometer alone does not discriminate intensity of movement or reflect the amount of time spent in specific intensity categories of activity. Evidence-based recommendations for threshold values must come from additional studies in diverse samples linking pedometer-determined ambulatory activity to important health outcomes.

In summary, there is a strong linear relationship between total physical activity outputs derived from the CSA accelerometer and the Yamax pedometer that is consistent with other reports. Step data derived from each instrument differs on average by 2000 steps·d−1, likely because of differences in instrument sensitivity thresholds. This finding suggests that one instrument’s step data may not be easily substituted for the other. Finally, in this study, approximately 8000 pedometer steps·d−1 was associated with a mean of 33 accumulated minutes of moderate-intensity ambulatory activity as determined by CSA accelerometer. This finding needs to be confirmed in different populations before it can be broadly used as a proxy value for classification (i.e., those meeting public health physical activity recommendations) and intervention purposes (i.e., for goal-setting and evaluation purposes).

This study was completed in South Carolina as part of a multicenter study to evaluate the International Physical Activity Questionnaire. We acknowledge the University of South Carolina Prevention Research Center for financial support of this project.

We appreciate the efforts of Katrina DuBose and Sharon Krumweide for assistance with data collection.

The authors do not have a professional relationship with companies or manufacturers who may benefit from the results of the present study. The results of the present study do not constitute endorsement of the products by the authors or the ACSM.

REFERENCES

1. Bassett, D. R. Validity and reliability issues in objective monitoring of physical activity. Res. Q. Exerc. Sport. 71: 30–36, 2000.
2. Bassett, D. R., B. E. Ainsworth, S. R. Leggett, et al. Accuracy of five electronic pedometers for measuring distance walked. Med. Sci. Sports Exerc. 28: 1071–1077, 1996.
3. Bassett, D. R., B. E. Ainsworth, A. M. Swartz, S. J. Stratch, W. L. O’Brien, and G. A. King. Validity of four motion sensors in measuring moderate intensity physical activity. Med. Sci. Sports Exerc. 32: S471–S480, 2000.
4. Bland, J. M., and D. G. Altman. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1: 307–310, 1986.
5. Cohen, J. Statistical Power Analysis for the Behavioral Sciences, 3rd ed. New York: Academic Press, 75–83, 1988.
6. Freedson, P. S., E. Melanson, and J. Sirard. Calibration of the Computer Science Applications, Inc. accelerometer. Med. Sci. Sports Exerc. 30: 777–781, 1998.
7. Hendelman, D., K. Miller, C. Baggett, E. Debold, and P. Freedson. Validity of accelerometry for the assessment of moderate intensity physical activity in the field. Med. Sci. Sports Exerc. 32: S442–S450, 2000.
8. Leenders, N. Y. J. M., W. M. Sherman, and H. N. Nagaraja. Comparisons of four methods of estimating physical activity in adult women. Med. Sci. Sports Exerc. 32: 1320–1326, 2000.
9. Mahar, M. T., and B. E. Ainsworth. Measurement of physical activity: The Cooper Institute Conference Series. Res. Q. Exerc. Sport. 71: S1–S158, 2000.
10. Matthews, C. E., P. S. Freedson, I. E. J. Stanek, et al. Seasonal variation of household, occupational, and leisure-time physical activity: longitudinal analyses from the seasonal variation of blood cholesterol study. Am. J. Epidemiol. 153: 172–183, 2001.
11. Matthews, C. E., J. R. Hebert, P. S. Freedson, E. J. Stanek, I. S. Ockene, and P. A. Merriam. Comparing physical activity assessment methods in the Seasonal Variation of Blood Cholesterol Study. Med. Sci. Sports Exerc. 32: 976–984, 2000.
12. Myers, A. M. Program Evaluation for Exercise Leaders. Champaign IL: Human Kinetics, 1999, pp. 1–149.
13. Pate, R. R., M. Pratt, S. N. Blair, et al. Physical activity and public health: a recommendation from the Centers for Disease Control and Prevention and the American College of Sports Medicine. JAMA. 273: 402–407, 1995.
14. Raven, P. B., A. W. Perrel, D. M. Robertson, H. W. Kohl, D. L. Kibbe, and S. Trost. Measurement of moderate physical activity: advances in assessment techniques. Med. Sci. Sports Exerc. 33: S439–S516, 2000.
15. Sirard, J. R., E. L. Melanson, L. Li, and P. S. Freedson. Field evaluation of the Computer Science and Applications, Inc. physical activity monitor. Med. Sci. Sports Exerc. 32: 695–700, 2000.
16. Thomas, J. R., W. Salazar, and D. M. Landers. What is missing in p< 0.05? Effect size. Res. Q. Exerc. Sport. 62: 344–348, 1991.
17. Tryon, W. W., L. P. Pinto, and D. F. Morrison. Reliability assessment of pedometer activity measurements. J. Psychopathol. Behav. Assess. 13: 27–44, 1991.
18. Tryon, W. W., and R. Williams. Fully proportional actigraphy: a new instrument. Behav. Res. Methods Instruments Comput. 28: 392–403, 1996.
19. Tudor-Locke, C. E., and A. M. Myers. Challenges and opportunities for measuring physical activity in sedentary adults. Sports Med. 31: 91–100, 2001.
20. 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. 72: 1–12, 2001.
21. Surgeon General’s report on physical activity and health. From the Centers for Disease Control and Prevention. JAMA. 276: 522, 1996.
Keywords:

MOTION SENSORS; MOVEMENT; BEHAVIOR; EXERCISE; AMBULATORY MONITORING

© 2002 Lippincott Williams & Wilkins, Inc.