The accuracy of the TriTrac-R3D accelerometer to estimate... : Medicine & Science in Sports & Exercise (original) (raw)
The Surgeon General's Report on Physical Activity and Health concluded that there are increased health risks associated with an inactive lifestyle (13). Many studies that have examined the relationship between physical activity and health have relied on questionnaires to measure physical activity. However, questionnaires are typically designed to measure specific types of activity (e.g., occupational versus leisure time) and are dependent on the ability of the individual to accurately report their activity. Therefore, to better understand the dose-response relationship between physical activity and health, it is important that objective measures of physical activity be developed.
Doubly labeled water (DLW) is considered the "gold standard" for the measurement of physical activity. However, DLW is not without limitations. The stable water isotope 2H218O is expensive, and the analysis requires sophisticated and expensive equipment. Further, DLW provides information on total daily energy expenditure over a 1- to 2-wk period but cannot be used to examine acute patterns of physical activity, such as time spent in specific activities or intensity of exercise sessions.
The use of accelerometers to objectively quantify physical activity has been recommended. The first generation of these devices were single-plane accelerometers, and this may have limited the accuracy of these units under certain conditions (10,11). Therefore, triaxial accelerometers were developed to detect body acceleration in three planes (forward/backward, up/down, and left/right), and in theory this should improve the accuracy when compared with the single-plane accelerometers. In fact, Meijer et al. (7) found that use of a triaxial accelerometer improved the estimate of energy expenditure during graded treadmill walking when compared with the single-plane accelerometer. Further, the output from triaxial accelerometers is significantly correlated with other objective measures of energy expenditure, including DLW (3) and indirect calorimetry (2). However, despite a significant correlation between the triaxial accelerometers and the reference methods, few studies have assessed whether commercially available triaxial accelerometers can accurately estimate energy expenditure when compared to an acceptable reference method (i.e., indirect calorimetry).
The TriTrac-R3D accelerometer (Hemokinetics, Inc., Madison, WI) is a commercially available triaxial accelerometer that incorporates accelerometer output into a regression equation to compute energy expenditure at 1-min intervals. The purpose of this study is to examine the intra- and inter-instrument reliability of the TriTrac-R3D Accelerometer and to examine whether the TriTrac-R3D provides a valid estimate of energy expenditure during various forms of exercise when compared with indirect calorimetry.
METHODS
Subjects
Twenty participants (14 female, 6 male) volunteered to participate in this study. Eligibility criteria included the following: 18-35 yr of age, within 20% of ideal body weight (9), and no health problems or physical limitations that would prevent the subject from participating in the exercise associated with this study. The female participants were 20.6 ± 2.5 yr of age with a body mass index of 22.1 ± 2.7 kg·m−2. The men were 23.5 ± 4.6 yr of age with a body mass index of 25.9 ± 4.0 kg·m−2. All subjects completed a Physical Activity Readiness Questionnaire (1) and provided written informed consent before participating in this program. All procedures were approved by the Institutional Review Board at the University of Pittsburgh.
Design
Subjects participated in five different exercises, with each exercise performed on a separate day. The details of each of the exercises are shown in Table 1. The order in which the exercises were performed was random and included the following: walking on a treadmill (Trackmaster TM500, JAS Manufacturing, Carrollton, TX), running on a treadmill, riding a stationary cycle ergometer (Monark 818E, Varberg, Sweden), stepping up-and-down on a Reebok Step (Reebok International, Ltd., Ronks, PA), and lateral sliding (slideboard) on a Reebok Slide. Each exercise was performed for a total of 20-30 min. To assess whether the TriTrac-R3D accelerometer was capable of detecting changes in workload, the grade of the treadmill was increased and the speed at which the subject performed the cycling, stepping, and slideboard exercises was increased at 10-min intervals. A metronome was used to set the pace for the cycling, stepping, and slideboard exercises. During each exercise session, energy expenditure was measured using indirect calorimetry and two TriTrac-R3D Accelerometers.
Description of the exercise protocols.
Assessments
Height and weight. Height was measured using a calibrated wall-mounted stadiometer (Perspective Enterprises, Inc., Kalamazoo, MI) with the subject not wearing shoes. Height was measured at the first assessment and this measurement was then used for all subsequent assessment periods.
Weight was measured to the nearest 0.25 pounds using a calibrated balance-bean scale (Health-O-Meter, Inc., Bridgeview, IL), with subjects dressed in light-weight clothing (shorts and T-shirt). Body weight was measured before each assessment period. Body weight measured at the first assessment was used for the descriptive data provided earlier (see description of subjects).
Indirect calorimetry. Indirect calorimetry was used as the reference method to measure energy expenditure. Subjects breathed through a Hans Rudolph one-way nonrebreathing valve and expired gases were measured using a SensorMedics 2900 Metabolic Cart (SensorMedics, Yorba Linda, CA). The metabolic cart was calibrated before each assessment according to the directions outlined by the manufacturer. Oxygen consumption was measured at 20-s intervals, with the mean of the three intervals in each min (20 s, 40 s, and 60 s) used for data analysis. Energy expenditure at each minute was computed by multiplying the oxygen consumption (L·min−1) with the nonprotein caloric equivalents based on the respiratory exchange ratio.
Accelerometry. Accelerometry during each exercise session was measured using a TriTrac-R3D. According to the directions outlined in the manual provided by the manufacturer, the device should be worn on the torso, with no specific anatomical site recommended (4). Therefore, the TriTrac-R3D was worn on a belt that was firmly attached to the anterior torso of the subject at the level of the waist, with the device placed in a position that was perpendicular to the mid-line of the anterior thigh. To examine whether there were differences between TriTrac-R3D units, two units (TT-Unit1 and TT-Unit2) were worn during each exercise session, one on the left and one on the right side of the waist. The same two TriTrac-R3D units were worn by all subjects, and the side of the torso that the unit was worn on (right or left side) was randomly assigned for each exercise session. Data were collected at 1-min intervals and TriTrac-R3D System Software Version 2.04 was used to compute energy expenditure at each interval.
Data Analysis
Energy expenditure data measured by indirect calorimetry and estimated by the two TriTrac-R3D accelerometers were analyzed across 5-min intervals (min 1-5, 6-10, 11-15, 16-20, 21-25, and 26-30), and these mean values were used to examine inter-TriTrac reliability and validity of the TriTrac-R3D. When measuring energy expenditure using indirect calorimetry, it may take 3-5 min to achieve steady state conditions while exercising at a constant workload due to physiological adaptations that need to occur such as increases in heart rate and oxygen delivery, and as a result of limitations in the technology to assess oxygen utilization. To confirm this assumption, minute-by-minute data from indirect calorimetry were analyzed using intraclass correlation coefficients, and the intraclass correlation coefficients were higher during the second 5 min of each workload compared with the first 5 min of each workload. For example, during walking, the intraclass correlation was 0.69 for min 1-5 and 0.95 for min 6-10. The same pattern was observed during the other modes of exercise. Therefore, the mean energy expenditures measured using indirect calorimetry during min 6-10 of each workload were used for comparison with the mean energy expenditures estimated by the TriTrac-R3D acclerometers during the same time periods (e.g., min 6-10 of each workload).
The intra-instrument and inter-instrument reliability of the TriTrac-R3D was examined using intraclass correlation coefficients according to the procedure described by Strout and Fleiss (12). The mean differences in energy expenditure at each exercise workload between the two TriTrac-R3D accelerometers were examined using dependent _t_-tests. Two-factor (TriTrac × Workload) repeated measures analysis of variance (ANOVA) was used to assess changes in the movement detected across workloads between the two TriTrac-R3D units.
Correlations between energy expenditure estimated by each of the TriTrac-R3D units and energy expenditure measure by indirect calorimetry were also computed. A two-factor (Method × Workload) repeated measures analysis of variance (ANOVA) was used to examine the validity of each of the TriTrac-R3D accelerometers and to examine whether the validity of the TriTrac-R3D accelerometers was affected as the exercise workload increased. Each TriTrac-R3D unit was compared separately to indirect calorimetry. Further, the two-factor repeated measures ANOVA was performed separately for each type of exercise (walk, run, step, slideboard, cycle). Dependent _t_-tests were also used to assess differences in total energy expenditure for each exercise by comparing each TriTrac-R3D separately to indirect calorimetry. These analyses were performed using both SAS (version 6.1) and Statistical Packages for the Social Sciences software (SPSS version 7.5). Statistical significance was defined at P < 0.05.
RESULTS
All 20 subjects had complete TriTrac-R3D and indirect calorimetry data for the walking exercise. However, due to failure of at least one of the TriTrac-R3D units, 19 subjects had complete data for the stepping exercise and 18 subjects had complete data for the slideboard exercise. Further, complete data were available for 13 subjects during the stationary cycling exercise due to difficulty keeping the TriTrac-R3D units securely attached while pedaling on four occasions, failure of a TriTrac-R3D unit on one occasion, and failure of the metabolic cart on two occasions. In addition, although the subjects were young and healthy, only 18 subjects were able to complete the first workload during the running exercise and only 9 subjects were able to complete the second running workload.
Intra-Instrument Reliability of TriTrac-R3D Accelerometer
Accelerometry should provide a consistent estimate of energy expenditure when exercise is performed at a constant workload. Therefore, the first set of analyses assessed the intra-instrument reliability of each TriTrac-R3D accelerometer, and this was assessed separately for each exercise. Each exercise stage was divided into 5-min intervals, and the minute-by-minute data within each interval were analyzed using intraclass correlation coefficients. As shown in Table 2, the intraclass correlations appear to be the highest under walking conditions (range = 0.86−0.96) when compared with the other exercise conditions.
Intraclass correlation coefficients to assess energy expenditure measured by two TriTrac-R3D accelerometers.
Inter-Instrument Reliability of TriTrac-R3D Accelerometers
The results of correlational analyses between the two TriTrac-R3D accelerometers are shown in Table 3. There was a significant correlation in energy expenditure estimated from the two TriTrac-R3D accelerometers for all activities and workloads, with the highest correlations found between the TriTrac-R3D accelerometers during walking and slideboard exercises (range = 0.92−0.98; P < 0.001). The correlations remained significant, but were weaker, when the running grade increased or the stepping speed increased. During cycling, the correlations between the two TriTrac-R3D accelerometers was 0.65 during the first 5 min of exercise but increased to 0.81 to 0.96 for the remaining 15 min of exercise.
Comparison of energy expenditure from two TriTrac-R3D accelerometers using correlation coefficients and dependent _t_-tests.
_t_-Tests were used to compare the mean energy expenditure per minute during each 5-min period between the TriTrac-R3D accelerometers. As shown in Table 3, there was a significant difference between the two accelerometers during walking, stepping, and slideboard exercises, with TT-Unit2 consistently producing 0.5-0.8 kcal·min−1 higher estimates of energy expenditure (P < 0.05). Differences between the two accelerometers were not significant when running or cycling.
The TriTrac-R3D uses the X (lateral), Y (horizontal), and Z (vertical) vectors to compute a vector magnitude (vector magnitude = square root of X2 + Y2 + Z2). To assess whether these components changed as workload changed during the various exercises, the data for each exercise were analyzed using a two-factor (TriTrac × Workload) repeated measures ANOVA, and the results of these analyses are presented in Table 4. Despite an increase in workload during walking and running, there was no significant change in the vector magnitude across workloads for either TriTrac unit, indicating that the TriTrac was unable to detect these changes in workload. However, there was an increase in the vector magnitude across workload for stepping, slideboard, and stationary cycling (P < 0.001). Analysis of each of the X, Y, and Z components (data not shown) also showed a significant increase as workload increased during stepping, slideboard, and stationary cycling (P < 0.006). In contrast, only the Z (vertical) component significantly increased across time during walking (P < 0.001), with neither the X, Y, or Z components changing across workload during running.
Comparison of vector magnitude accelerometer output between TriTrac-R3D units across workloads for various exercises.
TriTrac Accelerometers versus Indirect Calorimetry
Correlation between TriTrac-R3D and indirect calorimetry. The correlations between indirect calorimetry and each of the TriTrac-R3D accelerometers are presented in Table 5. Significant correlations (P < 0.05) ranging from 0.68 to 0.92 were found during the walking, running, and slideboard exercises, and the correlations were similar for both of the TriTrac-R3D accelerometers. Examination of the data collected at each workload during the stepping exercise revealed significant correlations of 0.54 (P < 0.05) to 0.65 (P < 0.01) for indirect calorimetry and TT-Unit1, whereas the correlations were 0.72 (P < 0.001) and 0.75 (P < 0.001) for indirect calorimetry and TT-Unit2. Energy expenditure estimated by accelerometry during cycling was not significantly correlated with indirect calorimetry.
Comparison of energy expenditure measured by indirect calorimetry and TriTrac-R3D accelerometers.
Validity of the TriTrac-R3D. The difference between energy expenditure estimated by each of the TriTrac-R3D accelerometers and energy expenditure measured by indirect calorimetry at each workload is shown in Table 5. For walking and running at the lowest workload, TT-Unit1 and indirect calorimetry yielded similar results, however TT-Unit2 significantly over-estimated energy expenditure by 1 kcal·min−1 (P < 0.05). For other walking and running workloads and for all other exercises, both TriTrac-R3D accelerometers significantly underestimated energy expenditure compared to indirect calorimetry (P < 0.05).
Each TriTrac-R3D was also compared separately to indirect calorimetry using a two-factor (Method × Workload) repeated measures ANOVA. Analysis of the walking data showed a significant Method × Workload interaction (P < 0.001) for each of the TriTrac-R3D accelerometers when compared to indirect calorimetry (see Table 5). During walking, the difference between each of the TriTrac-R3D accelerometers and indirect calorimetry increased as the workload increased. A similar pattern of results were shown when data were analyzed for the other exercises (see Table 5).
The minute-by-minute data were summed across each exercise, and the total energy expenditure estimated by each of the TriTrac-R3D accelerometers was compared with the total energy expenditure measured by indirect calorimetry. Total energy expenditure during walking was significantly underestimated by 50.0 ± 24.9 kcal and 29.8 ± 25.5 kcal in TT-Unit1 and TT-Unit2, respectively (P < 0.001). A similar pattern was shown for stepping, slideboard, and cycling, with both TriTrac-R3D units significantly underestimating total energy expenditure during these activities (P < 0.001). However, there was no significant difference in total energy expenditure between indirect calorimetry and either TriTrac-R3D during running. The data from these analyses are presented in Table 6.
Difference in total energy expenditure during various exercise between indirect calorimetry and TriTrac-R3D accelerometers (mean ± standard deviation).
Both Meijer et al. (8) and Bouten et al. (2) positioned the triaxial accelerometer on the lower back, which is different than the position used in the current study. To determine whether the position of the TriTrac-R3D affected the above results, an additional six subjects (age = 18.8 ± 1.2 yr; BMI = 22.5 ± 1.7 kg·m−2) participated in a treadmill walking protocol with the TriTrac-R3D worn on the lower back at the level of the waist and perpendicular to the spine. Again, the energy expenditure estimated by the TriTrac-R3D accelerometer was compared with the energy expenditure measured via indirect calorimetry. Walking speed was 3.5 mph, and the grade was 0%, 5%, and 10%. Results of a two-factor repeated measures ANOVA showed a significant Method × Workload interaction (P < 0.001), with the TriTrac-R3D underestimating energy expenditure by 1.0 ± 2.0, 2.2 ± 2.1, and 4.7 ± 2.0 kcal·min−1 at each of the three walking grades, respectively.
DISCUSSION
The first generation of accelerometers were capable of detecting motion in only one plane, and this may have limited their ability to accurately estimate energy expenditure for activities that had motion in more than one plane. For example, Montoye et al. (10) showed that a single-plane accelerometer was not able to detect changes in grade (incline) during treadmill walking. Montoye et al. (10) also showed that the single-plane accelerometer overestimated energy expenditure during level treadmill running and underestimated energy expenditure during graded treadmill running. Further, the single-plane accelerometer underestimated energy expenditure during stepping, and the magnitude of this error increased as the stepping speed increased (10). It was hypothesized that the use of a triaxial accelerometer, which is designed to detect horizontal, vertical, and lateral motion, would improve the estimate of energy expenditure for these types of activities. Because most movements typically involve motion in more than one plane, the triaxial accelerometer may be capable of detecting all motion that would contribute to energy expenditure. For example, when walking, the vertical force vector is a significant contributor to this movement. However, there is also motion occurring in the horizontal (moving forward) and lateral (side-to-side motion of the hips) planes, and the contribution of these force vectors may change as one moves from level to graded walking. Therefore, using a triaxial accelerometer to detect all motion contributing to the desired movement may improve the ability to accurately estimate energy expenditure. However, when the results of this study (see Table 4) are compared with the findings of Montoye et al. (10), the pattern of the results is similar. Therefore, based on this comparison, the triaxial accelerometer may not improve the estimate of energy expenditure compared to the single-plane accelerometer.
We also examined the interinstrument reliability of the TriTrac-R3D accelerometer and found that there is a significant correlation in estimated energy expenditure between two accelerometers during walking, running, stepping, slideboard, and cycling exercises (see Table 2). These results are similar to the interinstrument reliability (r = 0.94) of single-plane accelerometers worn simultaneously during walking at three different speeds (11). However, despite finding that the interinstrument reliability was good, this study showed that the absolute difference in energy expenditure between the two TriTrac-R3D accelerometers was statistically significant for walking, stepping, and slideboard (see Table 3). This discrepancy between TriTrac-R3D units was less than 1 kcal·min−1, which may be considered to be clinically irrelevant. Meijer et al. (8) have shown that the difference in activity counts between two triaxial accelerometers during treadmill walking may be as much 22%, and the results of this study showed that the difference between TriTrac-R3D units was 10-15% for a similar mode of exercise. Thus, one should be sensitive to this potential difference between TriTrac-R3D units when using these devices in research and clinical settings. Because of this potential difference, it may be important that an individual always be assigned to wear the same TriTrac-R3D unit to minimize measurement error across time. In addition, when multiple units are available for use, random assignment of these TriTrac-R3D accelerometers may help to minimize potential biases in the data between individuals assigned to different treatment groups.
One possible explanation for the inaccuracy of the TriTrac-R3D accelerometer is the positioning of the device on the body. The instruction manual provided by the manufacturer of the TriTrac-R3D accelerometer indicates that the device needs to be attached firmly to the torso of the body, with the recommendation being to place it in a pocket or attach it to a belt (4). For this study, the TriTrac-R3D was attached to a belt and placed at the level of the waist, perpendicular to the mid-line of the anterior aspect of the thigh. However, with the TriTrac-R3D in this position, the device was unable to accurately estimate energy expenditure when the torso of the body had limited motion, such as during stationary cycling. Further, this positioning may have contributed to the inability of the TriTrac-R3D to accurately estimate energy expenditure when the walking and running grade increased, and when the stepping and slideboard speed increased. However, positioning the TriTrac-R3D on the lower back as reported by both Meijer et al. (8) and Bouten et al. (2) did not improve the estimate of energy expenditure in this study.
A second possible explanation for the inaccuracy of the TriTrac-R3D accelerometer lies in the computation of energy expenditure. The formula used to convert the raw motion data to energy expenditure from physical activity is a proprietary formula and will not be released by the manufacturer (4). However, this proprietary formula does incorporate all vectors of motion into the calculation of energy expenditure (vector magnitude = square root of X vector2 + Y vector2 + Z vector2). This is consistent with the published findings of Bouten et al. (2), which recommended that the raw data from the three vectors of motion be summed and this value entered into a regression equation to compute energy expenditure. However, it may be incorrect to assume that horizontal, lateral, and vertical should be weighted the same when estimating energy expenditure. For example, in this study, energy expenditure during level walking was underestimated by as much as 22% and this is similar to the 15% reported by Bouten et al. (2). However, as the walking grade increased, the error in the estimate of energy expenditure increased to as much as 45%. In addition, energy expenditure for both vertical motion (stepping) and lateral motion (slideboard) was underestimated by approximately 40%. Therefore, the existing formulas may need to be adjusted to account for this discrepancy in estimated energy expenditure. Further, it is important to examine whether the TriTrac-R3D can be used to accurately assess energy expenditure during forms of lifestyle activity such as gardening and house cleaning.
Despite finding that the TriTrac-R3D accelerometer cannot accurately estimate energy expenditure during a variety of activities, results of this study show that there are advantages to using this device. There is a significant correlation between energy expenditure estimated using a triaxial accelerometer and energy expenditure measured using indirect calorimetry (see Table 5), a finding that replicates the results of other studies that have examined the correlation between output from triaxial accelerometers and valid reference methods such as indirect calorimetry (2) and DLW (3). Therefore, in situations where one is interested only in correlational data, the TriTrac-R3D may be a reasonable option. For example, a triaxial accelerometer may be used to differentiate between active and less active individuals; however, the absolute energy expenditure accumulated in physical activity may not be accurate.
The data presented in Table 3 shows that the estimate of energy expenditure by the TriTrac-R3D during each exercise is greater than what would be expected under resting conditions (approximately 1 kcal·min−1), suggesting that the TriTrac-R3D can detect motion even though the absolute estimates of energy expenditure may be inaccurate (see Tables 5 and 6). This, coupled with the ability of the TriTrac-R3D to collect data at 1-min intervals, may allow the accelerometer to be used to identify acute periods of physical activity. For example, in studies where duration of activity was of primary concern, we have used the TriTrac-R3D accelerometer to verify that exercise was being performed according to the prescribed duration (6). In addition, we have recently used the TriTrac-R3D to identify over-weight individuals who over-report the number of minutes of exercise that they perform (5) and found that these individuals lost less weight in a behavioral weight loss program than individuals who accurately reported their exercise. Therefore, use of the TriTrac-R3D to identify acute periods of activity can be advantageous.
In summary, it has been suggested that triaxial accelerometers can be used to estimate energy expenditure. The findings of this study show that estimates of energy expenditure by the TriTrac-R3D accelerometer during various forms of exercise are significantly correlated with energy expenditure measures using indirect calorimetry. However, these estimates of energy expenditure are less than the energy expenditure determined via indirect calorimetry, with the discrepancy increasing as workload increases. Therefore, the use of triaxial accelerometers to quantify energy expenditure has limitations and these data should be interpreted with caution.
REFERENCES
1. American College of Sports Medicine. ACSM's Guidelines for Exercise Testing and Prescription, 5th edition. Baltimore: Williams & Wilkins, 1995, pp. 14-16.
2. Bouten, C. V., K. R. Westerterp, M. Verduin, and J. D. Janssen. Assessment of energy expenditure for physical activity using a triaxial accelerometer. Med. Sci. Sports Exerc. 26:1516-1523, 1994.
3. Bouten, C. V. C., P. H. G. Wilhelmine, V-V. De Venne, K. R. Westerterp, M. Verduin, and J. D. Janssen. Daily physical activity assessment: comparison between movement registration and doubly labeled water. J. Appl. Physiol. 81:1019-1026, 1996.
4. Hemokinetics, Inc. TriTrac-R3D Research Ergometer Operations. Madison, WI: Hemokinetics, Inc. 1993.
5. Jakicic, J. M., B. A. Polley, and R. R. Wing,. Accuracy of self-reported exercise and the relationship with weight loss on overweight women. Med. Sci. Sports Exerc. 30:634-638, 1998.
6. Jakicic, J. M., R. R. Wing, B. A. Butler, and R. J. Robertson. Prescribing exercise in multiple short bouts versus one continuous bout: effect on adherence, cardiorespiratory fitness, and weight loss in overweight women. Int. J. Obes. 19:893-901, 1995.
7. Meijer, G. A., K. R. Westerterp, H. Koper, and F. ten Hoor. Assessment of energy expenditure by recording heart rate and body acceleration. Med. Sci. Sports Exerc. 21:343-347, 1989.
8. Meijer, G. A. L., K. R. Westerterp, F. M. H. Verhoeven, H. B. M. Koper, and F. ten Hoor. Methods to assess physical activity with special reference to motion sensors and accelerometers. IEEE Trans. Biomed. Eng. 38:221-229, 1991.
9. Metropolitan Life Insurance Company. Nutrition and Athletic Performance: Values for Ideal Weight. New York: Metropolitan Life Insurance, 1983.
10. Montoye, H. J., R. Washburn, S. Servais, A. Ertl, J. G. Webster, and F. J. Nagle. Estimation of energy expenditure by a portable accelerometer. Med. Sci. Sports Exerc. 15:403-407, 1983.
11. Pambianco, G., R. R. Wing, and R Robertson. Accuracy and reliability of the Caltrac accelerometer for estimating energy expenditure. Med. Sci. Sports Exerc. 22:858-862, 1990.
12. Strout, P. E., and J. L. Freiss. Intraclass correlations: uses in assessing rater reliability. Psychol. Bull. 86:420-428, 1979.
13. U.S. Department of Health and Human Services. Physical Activity and Health: A Report of the Surgeon General. Atlanta, GA: Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, 1996.
Keywords:
EXERCISE; CALORIC EXPENDITURE; RELIABILITY; VALIDITY; ACTIVITY MONITORING
© 1999 Lippincott Williams & Wilkins, Inc.