INTRODUCTION

Public health authorities, including the Centers for Disease Control and Prevention, the National Institutes of Health, and the World Health Organization, agree that children and adolescents must become more active and maintain higher levels of physical activity throughout adulthood if they are to enjoy healthy and productive lives. To help children and adolescents become more active, researchers and practitioners need valid and reliable measures of youth physical activity.

During the past decade, there have been many significant advances in the area of youth physical activity assessment. Arguably, the most notable of these developments has been the proliferation of “second-generation” motion sensors and heart rate (HR) monitors that provide real-time estimates of the frequency, intensity, and duration of free-living physical activity for periods up to 1 mo (depending on sampling interval). The purpose of this article was to provide an overview of recent developments related to the use of activity monitors (accelerometers, pedometers, and HR monitors) to quantify free-living physical activity in children and adolescents. Important methodological issues are discussed, and future research priorities are identified. In keeping with the theme of Exercise and Sports Science Reviews, this article does not provide a comprehensive review of youth physical activity assessment. For detailed information regarding the use of self-report questionnaires, doubly labeled water, and direct observation, the reader is referred to excellent reviews by Harro and Riddoch (⁴) and Welk and colleagues (¹²).

OBJECTIVE MEASUREMENT OF PHYSICAL ACTIVITY

Because many children and adolescents have difficulty accurately recalling their past physical activity behavior, objective activity measures are being used with increasing regularity. However, in the use of these devices in field-based research, investigators are being confronted with a new set of methodological dilemmas. Some of the most pertinent include the following: Which objective measure is more valid among children: HR monitoring or accelerometry? Are triaxial accelerometers better than uniaxial accelerometers in quantifying free-living activity in children? How many days of monitoring are needed to reliably estimate habitual physical activity? Are low-cost objective measures of physical activity such as electronic pedometers valid among children and adolescents? These important questions are addressed here.

Heart Rate Monitors Versus Accelerometry

HR monitoring remains an attractive approach to assessing physical activity because of the linear relationship between HR and energy expenditure during steady-state exercise. There are, however, several problems associated with this method. First, it is widely recognized that factors such as age, body size, proportion of muscle mass utilized, emotional stress, and cardiorespiratory fitness influence the HR-o₂ relationship. Second, because HR response tends to lag momentarily behind changes in movement and tends to remain elevated after the cessation of movement, HR monitoring may mask the sporadic activity patterns of children.

Relative to HR monitors, accelerometers present fewer burdens to subjects (no electrodes or chest straps) and are capable of detecting the intermittent activity patterns characteristic of small children. On the other hand, accelerometers are insensitive to many forms of physical activity (i.e., stair climbing and bicycling), and it is difficult to convert their output (i.e., counts) to units of energy expenditure. Recently, several investigators derived algorithms or “count cutoffs” to convert accelerometer output to units of energy expenditure, but the predictive validity of these equations in field settings has not been determined. An additional problem related to the use of prediction equations or “count cutoffs” is that they assume steady-state exercise over a 1-min period. Consequently, if a child alternates between vigorous physical activity and rest within a given minute (a likely occurrence), the accumulation of counts for that minute will reflect only the average activity level during that period, and no credit will be given for engaging in vigorous physical activity. Therefore, given that both methods have significant limitations, which is better for assessing free-living physical activity in children? The results of recent methodological studies may help researchers and practitioners make a more informed decision as to which measure to use.

Eston and colleagues (²) examined the relative validity of HR monitoring and accelerometry in predicting the energy cost of children’s physical activity. Thirty children between the ages of 8 and 10 y performed two treadmill walking trials (4 and 6 km·h⁻¹), two treadmill running trials (8 and 10 km·h⁻¹), and three nonregulated “play” trials consisting of playing catch, hopscotch, and sitting and crayoning. Across all seven activities, both HR and accelerometer counts were strongly correlated with oxygen consumption (scaled to body mass raised to the power of 0.75). However, the correlation observed for the Tritrac accelerometer (r = 0.91) was notably higher than that observed for HR monitoring (r = 0.80), suggesting that accelerometers may be a more appropriate measure of physical activity in children. Welk et al. (¹³) conducted a study to determine whether HR monitoring or accelerometry provided better assessments of children’s overall level of physical activity. Validity was assessed via direct observation in two different conditions: (a) a low activity condition consisting of a regular 40-min classroom period and (b) a high activity condition consisting of a 30-min physical education class. During the classroom and physical education time periods, the mean within-subject correlation between the Tritrac accelerometer and observed physical activity was 0.70 and 0.77, respectively. Within-subject correlations between HR and observed physical activity were similar during physical education (mean r = 0.79) but were substantially lower during classroom activities (mean r = 0.49). The authors concluded that HR monitoring provided valid measures of physical activity during periods of increased activity but not during periods of relative inactivity. In contrast, accelerometry provided valid assessments of physical activity under conditions of both high and low activity.

Based on this evidence, one could conclude that accelerometers provide better assessments of children’s free-living physical activity than HR monitors. However, it is important to note that several techniques have been devised to overcome some of the limitations of HR monitoring; these include the use of relative HR indices that, in theory, control for individual differences in cardiorespiratory fitness and age.

Many relative HR indices can be found in the research literature, but three of the most popular are the activity HR index (AHR), the PAHR-25 index, and the PAHR-50 index. The AHR is empirically defined as the mean of the recorded HR minus the resting HR, whereas the PAHR-25 and PAHR-50 are defined as the percentage of HRs that are 25 and 50% above resting HR. Because all three indices depend on accurate measures of resting HR, it should not be surprising that the operational definition of resting HR and the protocol used to measure it have profound effects on physical activity estimates. Logan et al. (⁷) examined the impact of different definitions of resting HR on the apparent activity level of children aged 3 and 4 y. Resting HR was measured five different ways: (a) mean of the lowest HR plus all HRs within 3 beats, (b) mean of the lowest 5 heartbeats, (c) mean of the lowest 10 heartbeats, (d) mean of the lowest 50 heartbeats, and (e) actual resting HR assessed using a standardized protocol. Depending on the protocol used, the PAHR-25 varied by 10–50%, the PAHR-50 varied by 16–65%, and the AHR varied by 9–44%. The authors concluded that a consensus must be obtained for deriving or measuring resting HR before relative HR indices can be used to effectively quantify physical activity in children.

A more burdensome approach to assessing physical activity via HR is to calibrate HR and o₂ on an individual basis. Of the various approaches to obtaining individualized HR-o₂ relationships, the HR FLEX method is one of those most studied. This method is based on the assumption that above a given intensity threshold, there is a linear relationship between HR and oxygen consumption. Below this threshold, the relationship is more variable. Therefore, to estimate o₂ or energy expenditure from the HR, the linear prediction is used above the HR FLEX point, and the average of a series of HR values obtained during rest are used below it. The HR FLEX point is empirically defined as the average of the lowest HR during exercise and the highest HR during rest.

Livingstone et al. (⁶) evaluated the accuracy of the HR FLEX method in 36 free-living children between the ages of 10 and 15. Compared with the doubly labeled water method, HR-based estimates of total daily energy expenditure exhibited large individual differences ranging from −16.7 to 18.8%, with mean group differences ranging from −9.2 ± 4.5% to 3.5 ± 6.6%. Similarly, Emons and co-workers (¹) evaluated the validity of the HR FLEX method in children using indirect calorimetry and doubly labeled water as criterion measures of energy expenditure. Energy expenditure predicted by 24-h HR monitoring was not significantly different from that estimated by doubly labeled water. However, relative to indirect calorimetry and doubly labeled water, the HR FLEX method overestimated 24-h energy expenditure by 10.4 and 12.3%, respectively.

In an effort to improve the precision of HR-derived estimates of free-living energy expenditure, several investigators have used a combination of HR monitoring and accelerometry. In this approach, two distinct HR-o₂ relationships are individually established: one for active periods and another for periods of inactivity. When a motion sensor or accelerometer worn on the trunk or a limb records movement above a given threshold, energy expenditure is predicted using the active HR-o₂ equation. Below the movement threshold, energy expenditure is predicted using the inactive HR-o₂ equation. Treuth et al. (⁹) tested the validity of this approach in children by comparing energy expenditure estimated through a combination of HR monitoring and accelerometry with the energy expenditure measured by whole room calorimetry. The mean level of error associated with the prediction of o₂, co₂, and energy expenditure was −2.6 ± 5.2, −4.1 ± 5.9, and 2.9 ± 5.1%, respectively. Given the small magnitude of these errors, the authors concluded that the combination of HR monitoring and accelerometry was an acceptable method for estimating energy expenditure not only for groups of children but also for individuals.

Based on this evidence, it appears that the decision to use HR monitoring or accelerometry depends on the outcome of interest and the scope of the study. If the outcome of interest is relative participation in physical activity and the measurement of resting HR is not a viable option, then accelerometers might be the method of choice. If, however, the goal is to estimate energy expenditure, then HR monitoring using an individualized HR-o₂ calibration curve would be the method of choice. The enhanced precision afforded by the combination of HR monitoring and accelerometry represents an exciting development in youth physical activity assessment, and this approach warrants further study.

Triaxial Versus Uniaxial Accelerometers

There is an ongoing debate as to whether triaxial accelerometers (Tritrac-R3D) provide better estimates of children’s physical activity than uniaxial accelerometers (CSA WAM 7164 and the Caltrac). Three-dimensional accelerometers were developed under the assumption that more is better. That is, by recording motion in more than one plane (vertical), triaxial accelerometers are better able to quantify the movements produced by children during normal play. While this argument has strong intuitive appeal, field- and laboratory-based studies testing the relative validity of uniaxial and triaxial accelerometers have produced conflicting results.

Welk and Corbin (¹¹) compared the validity of the Tritrac-R3D and Caltrac accelerometer in children using HR monitoring as an indicator of convergent validity. The correlation between HR and the Tritrac-R3D vector sum (r = 0.58) was marginally higher than that observed for Caltrac counts and HR (r = 0.52). Importantly, the correlation between the Tritrac-R3D and Caltrac was 0.88, suggesting that both approaches were providing similar information.

Freedson et al. (³) evaluated the validity and interinstrument reliability of the Tritrac-R3D and the CSA 7164 in 81 children ranging in age from 6 to 18 y. Each participant completed three treadmill trials consisting of walking/running at 4.4, 6.4, and 9.7 km·h⁻¹, respectively. To evaluate interinstrument reliability, participants wore two CSAs and two Tritrac-R3Ds during each trial. Consistent with previous laboratory-based validation studies, the CSA and Tritrac-R3D exhibited strong associations with energy expenditure measured by indirect calorimetry (r ≥ 0.90). The units did, however, differ considerably with respect to their interinstrument reliability. Across the three treadmill speeds, the CSA exhibited excellent interinstrument reliability (r = 0.89–0.94). In contrast, interinstrument reliability coefficients for the Tritrac-R3D were quite poor, ranging from 0.32 to 0.59.

Eston et al. (²) examined the relationships between oxygen consumption (relative to body mass raised to the power of 0.75) and output from the Tritrac-R3D and CSA 7164 accelerometers in children during laboratory-based exercise and unregulated play activities. Across all activities, the Tritrac-R3D vector sum exhibited stronger correlations with scaled oxygen consumption (r = 0.91) than did the CSA (r = 0.78). Notably, during primarily locomotor activities (walking, running, playing hopscotch), the largest accelerations were recorded in the vertical plane of the Tritrac-R3D (z-axis), whereas during crayoning and playing catch, the largest accelerations were recorded in the anteroposterior plane (y-axis). These findings were consistent with the view that three-dimensional accelerometers such as the Tritrac-R3D provide better evaluations of children’s free play activities than do uniaxial accelerometers.

Ott et al. (⁸) investigated the relative validity of the Tritrac-R3D and CSA 7164 with respect to their ability to measure children’s free play activities. Twenty-eight children between the ages of 9 and 11 y completed a circuit of eight free-play activities consisting of video game playing, throwing and catching, walking, bench stepping, hopscotch, basketball, aerobic dance, and running. During the activities, each participant wore an HR monitor, a CSA 7164 accelerometer, and a Tritrac-R3D accelerometer. Across all eight activities, both accelerometer types were significantly correlated with HR and an observation-based intensity score. However, the correlations observed for the Tritrac-R3D vector sum (r = 0.66–0.73) were greater than those observed for the CSA 7164 (r = 0.53–0.64). Similar to the study of Welk and Corbin, output from both accelerometer models were strongly correlated (r = 0.86), suggesting that over a range of free-living activities, both uniaxial and triaxial accelerometers provide useful information about children’s physical activity.

Based on this evidence, it appears that uniaxial and triaxial accelerometers provide comparable assessments of free-living physical activity in children. There are, however, concerns about the interinstrument reliability of the Tritrac-R3D, and this should be further investigated in other settings and study populations. Aside from issues of validity and reliability, the relative size of the two most popular accelerometers remains a practical consideration. The Tritrac-R3D measures a hefty 11.1 × 6.7 × 3.2 cm, which is considerably larger that the 5.1 × 3.0 × 1.5 cm measurement of the CSA 7164. The size of a measurement instrument is of the utmost importance in field-based studies, given that children are more likely to wear a device that is unobtrusive, can be worn underneath clothes when necessary, and does not preclude participation in common activities such as dance lessons and youth sports. The development of a compact and highly reliable three-dimensional accelerometer remains a research priority.

Number of Days Required to Characterize an Individual’s Usual Physical Activity Behavior

Studies that examined the day-to-day variability of children’s physical activity measured by HR and accelerometry have provided insight into this important methodological question. Most recently, Trost et al. (¹⁰) examined age-related trends in the reliability of objectively measured physical activity in a population-based sample of children and adolescents. Among children in grades 1–6, the single-day reliability of moderate-to-vigorous physical activity assessed by the CSA 7164 ranged from 0.46 to 0.49. In comparison, the single-day reliability among adolescents in grades 7–12 was notably lower, ranging from 0.31 to 0.32. At these levels of variability, it was estimated that between 4 and 5 d of monitoring would be necessary to achieve a reliability of 0.80 in children, and between 8 and 9 d of monitoring would be necessary to achieve a reliability of 0.80 in adolescents. Importantly, for both children and adolescents, 7 d of monitoring produced acceptable estimates of daily moderate-to-vigorous physical activity (R = 0.76–0.86) and accounted for significant differences in weekday and weekend physical activity.

Knowledge of the number of days to assess physical activity and getting children to wear the monitor for that time period are two entirely different matters. There are, however, some steps that one can take to minimize noncompliance. For studies that monitoring over a continuous 7-d period, one solution is to outfit participants with a “fresh” activity monitor on the beginning of each monitoring day. Few studies, however, have the resources to implement this strategy, and it is doubtful whether the average family or school would be willing to tolerate such an imposition on a daily basis. A related, yet less burdensome, approach is to sample single days of physical activity at weekly or monthly intervals. Over longer periods of time, this approach has the added advantage of controlling for seasonal variations in physical activity behavior.

Regardless of study design, effectively educating children, parents, and teachers about how and when to wear the activity monitor is absolutely critical. Age-appropriate written materials that can be easily displayed on bulletin boards and refrigerator doors are particularly useful for this purpose. Classroom demonstrations and practice monitoring periods are also useful educational strategies. Whenever possible, coaches, referees, and other sport officials should also be included in the education process, because children are frequently not permitted to wear activity monitors during sports practices and/or formal competitions. Other general strategies to minimize noncompliance include frequent contacts with parents and teachers (face-to-face or telephone), the use of activity monitoring log sheets to record daily use, and the provision of incentives such as gift certificates, coupons, or extra credit.

Validity of the Pedometer for Measuring Physical Activity in Children

While accelerometers have been shown to be useful tools for quantifying physical activity in children, their relatively high cost ($150–500 per unit) prohibits their use in small-budget feasibility studies and large-scale epidemiological and surveillance studies. A cost-effective alternative to accelerometers is to measure physical activity with an electronic pedometer. Pedometers have the same basic limitation as uniaxial accelerometers, in that they are insensitive to nonlocomotor forms of movement. In addition to this limitation, however, these devices are unable to record the magnitude of the movement detected (movement above a given threshold is counted as a step regardless of whether it occurred during walking, running, or jumping), nor do they possess real-time data storage capabilities. Consequently, pedometers can only provide an estimate of the relative volume of activity performed over a specified time period, assuming that most of the activity performed involves locomotor movement such as walking. Many commercially available pedometers provide users with estimates of energy expenditure; however, the algorithms used for these calculations are not appropriate for children.

Studies evaluating the concurrent validity of electronic pedometers have yielded positive results. Eston and colleagues (²) reported a correlation of 0.92 between steps recorded by the Yamax Digiwalker and scaled oxygen consumption during treadmill walking/running and unstructured play activities in 8- to 10-y-old children. Kilanowski et al. (⁵) observed correlations greater than 0.95 between pedometer steps per minute and directly observed physical activity in 12-y-old children.

Considering these findings, it appears reasonable to conclude that electronic pedometers provide valid assessments of the total volume of physical activity performed by children. These devices should be especially useful in studies in which the goal is to document relative changes in physical activity or to rank order groups of children on physical activity participation. Pedometers do not, however, provide information about the frequency, intensity, or duration of physical activity. In addition, because pedometer steps are influenced by factors such as body size and speed of locomotion, investigators should exercise caution when using pedometers in growing children or groups of children with different levels of maturation.

SUMMARY

Second-generation activity monitors such HR recorders and accelerometers have significantly enhanced our ability to objectively measure the frequency, duration, and intensity of children’s physical activity. These measures do, however, have strengths and weaknesses that should be carefully considered before embarking on a study or project evaluation (see Table 1). This article examined four practical issues related to the use of objective physical activity measures in field-based research and presented information that will help researchers and practitioners decide which objective measure to use and how long to use it.