The increasing prevalence of obesity in U.S. children in the recent decades has led to a search for factors responsible for this secular trend (29). Although this increase in the prevalence of obesity does not appear to be due to an increase in caloric intake, no longitudinal information is available regarding physical activity levels in U.S. children in order to examine the role of activity in the increase prevalence of obesity in children (21). Cross-sectional studies suggest that a dramatic decrease in physical activity takes place during adolescence (6,30). The relative paucity of information on longitudinal changes in physical activity levels in the U.S. stems from the inherent difficulty of prospectively assessing levels of physical activity, particularly in children. Longitudinal data from the Amsterdam Growth and Health study indicate that weighted activity scores (METs·wk−1) declined by 26% from ages 12–13 through 17–18 in Dutch girls (31). However, there are virtually no longitudinal data available to estimate the extent of the decline and the specific time points during adolescence when the decline occurs in U.S. youths (30).
Physical activity can be assessed by subjective and objective methods that include questionnaires, direct observation, or mechanical devices (1,4,8,11,22,25–27). The physical activity questionnaire is probably the most commonly used method to assess physical activity in population studies because of its relatively low cost and ease of administration (17,18). Activity questionnaires are used to capture habitual activity which spans over a week, month, year, or even lifetime. Some of the questionnaires used to assess physical activity levels in children were adapted from existing questionnaires used in adult studies (1,2,4,5,7,27). However, there is still little information available on the validity and reliability of these instruments in children (1).
The diary method may be less subject to recall bias and also lends itself for validation studies because of the relatively shorter time involved. However, a short-term assessment with the use of an activity diary is less likely to reflect “usual” activity levels because of acute illness, time constraints, inclement weather or seasonal variations (15). Questionnaires with a longer time-frame, such as 1 yr, may be more likely to represent usual activity patterns and, therefore, have been used extensively in chronic disease epidemiology. Limitations of these questionnaires include selective recall of activities and errors made in quantifying duration and frequency of activity which occurred in the past. An optimal approach would be to use both of these self-report methods to capture the overall physical activity of an individual over both short and long time periods.
Objective measures of physical activity are often included in a study to validate the self-reported data collected through questionnaires. The doubly labeled water technique, which has become the gold standard for the objective measurement of physical activity levels in recent years (9,10), is not feasible for large population studies because of its high cost. A more widely used method to validate self-report assessments of physical activity is the use of an activity monitor such as Caltrac, Tritrac, or the CSA accelerometer (11,14,19,20).
The purpose of this report is to describe the development and use of two self-report methods and an activity monitor to assess longitudinal changes in physical activity from childhood through adolescence. This report also describes analytic methods used to examine whether the observed changes in physical activity levels represented real changes in activity level or numerical artifact due to modifications in data collection forms. Finally, the longitudinal trend in physical activity in a biracial cohort of adolescent girls is presented using these three different measures of activity assessment.
Overview of the NHLBI Growth and Health Study (NGHS)
In 1985, the National Heart, Lung, and Blood Institute initiated the NHLBI Growth and Health Study (NGHS), a multicenter study to examine the biologic, environmental, and psychosocial correlates of the racial difference in obesity development. A major emphasis was placed on the collection of the best information on energy intake and energy expenditure to address as fully as possible the question of energy balance in obesity development.
NGHS Study Population
A total of 2379 girls (1213 black and 1166 white), ages 9 and 10 yr at entry, were recruited from schools (the Richmond school district outside Berkeley, CA, and Cincinnati, OH) and a large health maintenance organization in the Washington, DC, area. The cohort was followed annually until ages 18–19 yr. The study design, protocol, and baseline characteristics of the NGHS have been previously described (24). The study population for this report consists of 2322 girls (51% black and 49% white), 97.6% of the NGHS cohort, who completed either or both physical activity self-reports at baseline examination. The NGHS protocol was approved by the Institutional Review Board at all participating centers and informed consent was obtained from all participants and their parents.
NGHS Physical Activity Measurement Instruments
At the inception of NGHS in 1985, there was virtually no published information on a physical activity assessment tool that had been used longitudinally from childhood through adolescence. Two self-report methods, a 3-d activity diary (AD) and a habitual activity patterns questionnaire (HAQ), were developed to capture both detailed short-term activity and also habitual activities during the past year. A Physical Activity Resource Unit at the University of Pittsburgh was established in the 7th study year with the goal to standardize and optimize the collection of data on physical activity in the multicenter, longitudinal NGHS. The tasks included central training of the nutritionists and interviewers in the collection of physical activity data, development of standardized coding and scoring systems for physical activity, and provision of epidemiologic expertise related to the management and analysis of physical activity data.
Activity Diary (AD)
A 3-d activity diary, in tandem with a 3-d food record, was self-administered on 2 consecutive weekdays and 1 weekend day in 8 of the 10 yr (years 1–5, 7, 8, 10) of the NGHS. Because of possible limited literacy among 9 and 10 yr old girls, the year 1 AD was formatted as a pictorial menu of 24 activities commonly performed by children (see year 1 AD in Appendix 1) (4). Twenty-two of these pictures were physical activities (e.g., bike riding, dancing, and swimming) and two were sedentary activities (e.g., board games, sitting watching TV or reading). Additionally, several blank spaces were offered to write in “other” activities not depicted on the menu. Times of waking up and going to bed were reported for each diary day. Thus, a complete 24-h cycle would be recorded including sleeping hours at night. In year 2, the pictorial menu of individual activities was rearranged to nine activity groupings aggregated by similar intensity level (See year 2 AD in Appendix 2).To accommodate children’s difficulty in accurately quantifying time, duration choices were given as three interval categories for each activity, 1–15 min, 16–30 min, and >30 min. Participants completed the AD by checking off the time duration of each activity they performed during specific parts of the day. Wristwatches were provided to all participants in year 1 as an incentive and to foster accuracy in their recording of time.
Derivation of the summary AD scores.
A comprehensive catalogue of MET values was developed from existing literature (based on adult studies) and modified to reflect more closely the energy expenditure associated with activities performed by NGHS participants given their age range and gender (3,12,13,32). A summary daily activity score for each participant was calculated for each of the three diary days. First, the assigned MET value of each activity grouping was multiplied by the duration of the specific activity menu item recorded. Because the duration choices were time intervals, the midpoint of the interval choices was assigned as the time expended (7.5 min for the 1–15 min interval; 22.5 min for 15–30 min and 45 min for greater than 30 min). These MET values were tallied across the day for each activity grouping and summed to obtain the summary MET value for that diary day.
When the total time reported for the listed activities in a given part of the day exceeded the maximum time available for that period (e.g., 7 h of activity in a 6-h period), that part of the diary day was deemed unreliable and excluded. If two or more parts of a day were excluded, then the entire diary day was deemed “unreliable.” When two/three diary days were deemed “unreliable” from the three days collected, this AD was coded as “unusable.” All diary days were deemed “unusable” if a girl reported being pregnant when she completed the AD. Thus, the final AD score, expressed as MET-min·d−1, was the average of two or three acceptable diary days. Further, a replacement AD was done if the corresponding days’ food record was replaced, if the day’s activities were markedly altered because of illness, or a completely blank AD was turned in. Replacement ADs always corresponded to the specific day of the week as the originally scheduled AD. For all study years, completed ADs were reviewed in a structured manner by centrally trained staff using a common protocol.
Form changes in the activity diary.
a) The first form change occurred in year 2. The change entailed the conversion of the pictorial menu of 24 individual activities to 9 grouped activities, aggregated around similar intensity.
b) New activity groupings were added in years 3, 8, and 10. Further discussion related to these additions follows in the Results section of this article.
c) Periodically, as age-appropriate, individual activities were added or deleted within a grouping, as described in the following section.
Establishment of uniform MET values for similar AD activity groupings for all years.
As the cohort matured, the pattern of daily activities changed. To accommodate this change, specific activities were periodically added or removed from activity groupings. As a result, the mean MET value of some activity groupings varied slightly across the years. To minimize this variation in the mean MET value of a given activity grouping, a “uniform” MET value for each activity grouping was derived by averaging the mean MET values of that activity grouping across all study years. An example is provided for the activities listed under “group 9” and the MET values for this activity grouping by study year (Table 1).
As can be seen from the example, the numerical values were very similar. The uniform group MET approach was adopted to minimize variability caused by these minor changes in the activity content of the groupings. Keeping a constant mean MET value for each activity grouping for all years would enhance longitudinal data analysis with an approach such as the generalized estimating equation (GEE) method (33).
Habitual Activity Questionnaire
A habitual activity questionnaire (HAQ), adapted from Ku et al. (16), was administered as a structured interview in years 1, 3, and 5 and self-administered for years 7–10. For all study years, the responses were reviewed in a structured “debriefing” interview by centrally trained interviewers. The HAQ was designed to assess the type and frequency of participation outside of school in sports, physical activities, and classes/lessons during the past year. Participants were asked to list all these activity categories for the school year and the summer months. They were also instructed to estimate the weekly frequency for each activity listed.
Derivation of the summary HAQ scores.
A summary weekly activity score was calculated by multiplying the MET value for each recorded activity by the frequency (times·wk−1) and by the fraction of the year each activity was performed. For scoring purposes, the following fractions were assigned to a given time frame: classes/lessons during the year (“most” of the year = 1, “half” = 0.5, and “small part” = 0.25); sports/physical activities during the school year (“most” of the school year = 0.75, “half” = 0.375, and “small part” = 0.1875); and sports/physical activities during the summer (“most” of the summer = 0.25, “half” = 0.125, and “small part” = 0.0625). The final HAQ score (MET-times·wk−1) was the sum of the weekly score for all activity categories (i.e., school sports, summer sports, classes/lessons) for the previous year.
Form change in the habitual activity questionnaire.
Only one HAQ form change was made in year 2. This change involved limiting the weekly frequency choices to three categories (less than once a week, 1 or 2 times·wk−1 or ≥ 3 a week) instead of an open-ended response in year 1. For the HAQ longitudinal database file, the activity frequencies for year 1 were made equivalent to the other years by collapsing all weekly frequencies greater than 3 times a week to the category, “≥3 times a week.” For example, on the year 1 HAQ form, if a participant had written 5 times a week and another participant wrote 6 times a week, both responses would be recoded as “≥3 times a week.”
Caltrac Activity Monitor
Because longitudinal assessment of physical activity was novel and the AD and HAQ were self-report methods developed for NGHS, an activity monitor was included to objectively measure the level of physical activity. A Caltrac activity monitor (Hemokinetics, Inc., Madison, WI) was used to measure daily activity for 3 d, concurrently, with the 3-d food record and activity diary for the whole cohort for three consecutive years (NGHS years 3–5). The Caltrac accelerometer detects not only the frequency of movement, but also the acceleration and deceleration of the body in a vertical plane (23). Instructions on the proper use of the Caltrac were provided by centrally trained personnel using a common protocol. The Caltrac was placed horizontally in a securely tightened pouch fastened to a belt over the right hip. Each night before going to bed, participants were instructed to remove the Caltrac and record the daily reading shown on the display window and the time the Caltrac was removed. At “wake-up,” participants were asked to record the reading on the display window and the time the Caltrac was put on. Because the Caltrac tends to underestimate activities in which the torso is immobile such as biking, and is not worn during swimming, all participants were instructed to log the amount of time they engaged in biking and swimming during the Caltrac assessment period on their Caltrac record.
Each accelerometer was programmed to measure only body movement without regard to the individual’s age, weight, height, or sex. A daily Caltrac score was computed by subtracting the “wake-up” reading from the “bedtime” reading. If two of three Caltrac daily scores were unavailable, this Caltrac score was coded as “unusable.” Thus, the final Caltrac score, expressed as counts·d−1, was the average of the daily score over the two or three day assessment period.
Caltrac was introduced in year 3 and was administered concurrently with the 3-d AD. Because the goal was to examine the “longitudinal validity” of our AD and HAQ tools, we limited our data analysis to a cohort on whom we had complete information from AD, HAQ and Caltrac for years 3, 4, and 5, which were the years when Caltrac was administered. For instance, in year 3, there were 2054 girls with valid ADs, 2186 with valid HAQs, and 1398 with valid Caltrac scores. When all 3 yr were included, there were 1369, 1450, and 1374 girls who had valid scores for all three measures (Caltrac, AD, and HAQ) for years 3, 4, and 5, respectively. When the data files were merged, our final sample size for the longitudinal examination of these three PA instruments for these 3 consecutive years became 683. A major reason for this reduced sample size for the validation analysis was the difficulty of administering Caltrac to such a large cohort of free-living active young girls. For example, on weekends, the participants often neglected to wear the Caltrac or their parents “reprogrammed” the instrument to give results in calories rather than raw readings. The Caltrac malfunctioning was another frequently cited reason for incomplete data collection. Further, if two of three Caltrac daily scores were not complete, the Caltrac score for that individual was coded as “unusable” for that year. To see whether there was bias with those who had valid Caltrac scores for those 3 consecutive years as compared with those who did not, a Wilcoxon rank-sum test was used to compare median AD and HAQ scores between those 683 with valid Caltrac readings and those without such data for years 3, 4, and 5. Results showed that there were no significant differences in the median activity scores between the two groups (P > 0.10 for each comparison).
Descriptive statistics (median, mean, standard deviation, percentiles, frequency distributions) were generated using SAS® (28). The term, “empirical analysis,” is used in this manuscript to distinguish an analysis from traditional statistical analysis which is generally performed to test a scientific hypothesis. To examine whether activity scores significantly changed over the study period, a Wilcoxon matched pairs signed rank test or Friedman chi-square test was used to compare the change in median activity scores between study years. Spearman correlation coefficients were generated between the Caltrac scores and 1) activity diary, and 2) habitual activity questionnaire by study year. As a measure of internal validity of the Caltrac scores, the correlation between the Caltrac scores and the AD was analyzed by dichotomizing average daily minutes reported for biking and/or swimming reported on the Caltrac record into ≥ 30 and <30 min.
Physical Activity Scores by Study Year
Activity diary scores.
Distribution of AD scores. The frequency distribution of the AD scores by study year revealed a highly skewed pattern which was present in all years, but more so, in the latter years of NGHS. For illustrative purposes, the distributions for years 1 and 10 are presented (Fig. 1).
Because the log and square root transformations failed to normalize the data distribution, the AD scores are presented without data transformation. To accommodate the skewness of the data, however, median, rather than the mean, values are presented.
AD scores by study year.Table 2 demonstrates that the median AD scores by study year showed, in general, a consistent decrease in physical activity from year 1 through year 10. Activity levels declined sharply in year 2 from the previous year but rose by 11% in year 3 and then continued to decline steadily through year 8 by 175.4 MET-min·d−1 (a total of 48% from year 1). In year 10, AD scores increased by 26% from year 8.
Distribution of HAQ scores. The frequency distribution of the HAQ scores by study year also was highly skewed which was present in all years with a marked skewing from year 7 onward. For illustrative purpose, the distributions of HAQ scores for years 1 and 10 are presented (Fig. 2).
Log and square root transformations failed to normalize the data distribution. Again, median scores are presented without data transformation.
HAQ scores by study year. In general, the median HAQ scores declined steadily in the earlier years, and then declined even more sharply (70%) from year 5 to 7 (Table 2). From year 8 to 10, the HAQ scores increase marginally by 2.1 MET-times·wk−1. Figure 3 shows that HAQ scores were lowest for year 8 and increased in year 9 and then, plateaued thereafter. Though the difference in the scores between years 8 and 9 were statistically, significant, P < 0.001, the HAQ scores for years 9 and 10 did not significantly differ from year 7 (P > 0.10). Thus, year 8 appears to be an exceptionally low scoring year for HAQ rather than an actual increase in the level of activity in years 9 and 10 from all other previous years, save year 7. When HAQ score for year 8 (age 16–17 yr) was examined by its contributing components (school-year sports, summer sports, and classes), the median score for all three components was 0 MET-times·wk−1. In year 9, the school-year sports increased to 2.6 MET-times·wk−1 and summer sports increased to 1.3 MET-times·wk−1. However, this increase in the HAQ for year 9 was seen only among those from moderate to high income households.
Distribution of Caltrac scores. The frequency distribution of the Caltrac readings by study year also revealed a highly skewed pattern which was noted in all years.
As with the AD and HAQ scores, the skewness became greater with time. Log and square root transformations failed to normalize the data distribution. Hence, median scores are presented without data transformation.
Caltrac scores by study year. The results from the Caltrac monitor demonstrated a similar decrease in activity levels from year 3–5 similar to that found with the AD and HAQ (Table 2). The median Caltrac score declined by 10% (28.3 counts·d−1) and 13% (34.0 counts·d−1), respectively, between years 3–4 and years 4–5. For each study year, the Caltrac score showed a strongly significant positive correlation with the AD score for each of the 3 yr examined (range of rs = 0.20 to 0.25 for years 3–5, P < 0.0001). On the other hand, though statistically significant and positive, the association between Caltrac and HAQ scores was weak (rs = 0.09 for both years, P = 0.02).
Empirical Approaches to the Establishment of Longitudinal Data on Physical Activity
Assessment of changes in the AD score from year to year.
Because of the periodic modifications in the AD format, it was deemed necessary to investigate those years when sizable changes in the AD scores were seen to determine whether the observed changes in the scores for that particular yr represented real changes in the level of activity or artifacts from changes in the data collection form. In year 2, a sharp decrease in the AD score was seen (Table 2). This may have been due to the aggregation of several individual activities into one activity grouping in year 2, which may lead to an under-ascertainment of the actual time duration recorded. To explore this possibility, the year 1 activity data file was reconfigured by aggregating the 24 activities into the 9 groupings analogous to the activity menu of year 2 AD, and the median score for year 1 was recalculated. The recalculated median AD score was 446.8 MET-min·d−1 as compared with the original value of 450.8 MET-min·d−1. Though very similar in value, to be more consistent across the years, the score from the reconfigured year 1 data was selected. However, greater ascertainment of activities by the visual “prompt” for individual activities on the pictorial menu from year 1 cannot be ruled out. The drop in the activity score in year 2 may be in part due to the switch from individual activities to activity groupings.
A higher median AD score was noted in year 3 which might have been influenced by the addition of “walking” (debriefed as leisure-time walking) as a separate activity group. Thus, the overall daily scores for years 1 and 2 could be lower as a result of under-ascertainment of leisure-time walking.
The third change in the diary format was made in year 8 when six new activity categories were added to accommodate the increasing number of participants who took on more household chores or after school jobs (see Appendix 3, activity categories nos. 12–17 on year 10 AD, which were the same in year 8). However, the median score in year 8 was still lower than the previous years’ score. Additionally, the contribution from the miscellaneous activities listed in the “Other” category for the daily METs decreased dramatically in year 8, from 26.8 to 4.7 MET-min·d−1, 82%. Therefore, the addition of these new groupings did not mask a decrease in the level of activity in year 8. In year 10, the median score significantly (P < 0.001) increased by 26% from year 8 (Table 1). The year 10 diary contained an additional new activity category which was related to child and infant care (see Appendix 3, activity category 18 on year 10 AD).
To assess whether or not the addition of the new activity grouping related to child-care for year 10, we rescored the AD by excluding the MET contribution from the child/infant care grouping. Although the revised AD score was 15.6 MET-min·d−1 higher, it was not statistically significant, P > 0.10. We also examined the percent contribution of this new grouping to the total AD score. The “child/infant care” accounted for 12% (60 MET-min·d) of the total AD score in year 10. However, in year 10, the contribution of “Other” (i.e., miscellaneous not included in the given activity groupings) activities also increased by 17.1 MET-min·d−1 (7%) from year 8. These findings suggest that there was a greater diversification of activities in year 10 because many of the participants engaged in the new activity grouping and also reported many activities not listed on the AD menu (e.g., occupational activities). Hence, the increase in activity in year 10 is most likely an actual increase in the overall level of activity and the inclusion of the new activity grouping was appropriate to capture the low/moderate physical activities that are being performed by young women age 18–19 yr.
Assessment of changes in the HAQ score from year to year.
The most notable longitudinal trend in HAQ scores was the sharp drop from year 5 to 7. No form changes had taken place between these 2 years. However, in year 7, it was noted that a very large number of participants scored “zero” on the HAQ. The percent of girls reporting “zero” activity increased from 1.7% in year 5 to 35% in year 7 and continued to increase in year 8 (43.7%). After year 8, “zero” activity stabilized at around 40%. Because the “zeros” entered for the HAQ might be due to the participant not completing the HAQ, hard copies of the questionnaires which reported “zero” activities were reviewed. The review of the hard copies confirmed that these were true “zero” activity because they had been verified by trained interviewers at the time of debriefing.
Examination of internal validity of the Caltrac assessment.
Because Caltrac tends to underestimate activities in which the torso is immobile and is not worn during swimming, the correlation between Caltrac score and the AD should be weaker for those participants reporting a greater amount of these types of activities. Correlation with Caltrac measures was assessed by stratifying the AD score by those who reported ≥ 30 min and those reporting <30 min·d−1 of biking and/or swimming during their Caltrac assessment period. For those who reported ≥ 30 min·d−1, there was no significant relationship between their AD score and the Caltrac score (rs = 0.04 to 0.21 for years 3–5, P = 0.14 to 0.78). On the other hand, the correlation between the AD score and the Caltrac score for girls reporting <30 min·d−1 of biking/swimming was significant (rs = 0.22 to 0.26 for years 3–5, P < 0.0001).
Longitudinal Changes in the Level of Physical Activity in the NGHS Cohort
From year 1 to 10, there was a significant (P < 0.001) decline of 35% in the median AD score from 446.8 to 292.1 MET-min·d (Table 2). This NGHS study period spanned 9–10 and 18–19 yr of age. There was even a greater (P < 0.001, by 83%) decline in the HAQ score during the same time period, from 29.3 to 4.9 MET-times·wk−1. For those 3 years when the measures from the three methods of assessment were available, there was a parallel trend in the pattern of the decline in activity among the two self-report methods and the objective method of physical activity assessment (Fig. 3). From year 3 to 5 (ages 11–12 to 13–14), the AD score decreased by 22%, whereas both the HAQ and Caltrac declined by 21%.
This is the first U.S. report that provides longitudinal documentation of the actual extent of decline in physical activity levels during adolescence utilizing three separate measures of physical activity. The data are from a multicenter, prospective study which tracked 2379 black and white girls for 10 yr. The level of girls’ daily activity declined by 35% from ages 9–10 to 18–19 yr, whereas the level of habitual activity declined by 83%. These dramatic findings are supported by data from three independent assessment methods, a diary, a questionnaire, and an accelerometer.
The NGHS data on physical activity have several unique features. First, they were collected in a standardized manner in a very large cohort of black and white free-living girls from three different regions of the United States. The information on physical activity is multidimensional in that both the level of daily activity as well as habitual activity were ascertained. An objective measure of activity with the use of Caltrac was also collected for 3 consecutive years of the study for the whole cohort. Because the diary and Caltrac days included two weekdays and one weekend day, these data were more likely to be representative of average daily activity. Because the habitual activity information also encompassed both the school year and summer months, any potential difference in seasonal variation was most likely accommodated. The relationship between the two self-report methods and the objective measure was examined cross-sectionally as well as longitudinally, further strengthening the effort to validate the diary and the questionnaire.
Another unique experience from the NGHS is the effort directed at ensuring the integrity of the longitudinal data that spanned 10 yr during which the study participants underwent rapid changes in the biologic and psychosocial developments of adolescence. Several empirical analytic approaches were used to address the impact of data collection form changes that occurred periodically during the study. From the outset, emphasis had been placed on selecting instruments that would minimize bias in data gathering in this biethnic cohort recruited from a wide range of socioeconomic background. The use of a pictorial format for the diary in year 1 was intended to minimize potential under-ascertainment of activities because of the limited literacy inherent in this young age group. However, as the pattern of activities diversified with age, the initial pictorial menu of individual activities was limiting. Hence, from year 2 onward, the menu for the diary was based on activity groupings. Another form change in the activity diary entailed the addition of new activity groupings in years 3, 8, and 10, particularly from year 8 onward when the participants were >16 yr of age and in grades 10+ in high school. The expansion of the activity menu was necessitated by the changing patterns of daily life in these young women. For example, 203 pregnancies (10.4% of girls) were reported in year 7, which led to more housekeeping and child-care chores. The nature of leisure activities had also changed. Despite the addition of six new activity groupings in year 8, the overall AD score actually declined, suggesting that this expansion of the activity menu did not appear to have led to higher ascertainment of activity levels. This was further verified by the marked decrease in the activities listed in the “Other” category for year 8 because the activities subsumed in the new categories would have been most likely listed under “Other” had these additional choices not been provided.
The final longitudinal data (Fig. 3) for the NGHS cohort confirm the previously reported cross-sectional observation of a dramatic decrease in the overall level of physical activity during the transition from childhood to adulthood. The parallel trends in the activity levels between years 3 and 5 among the three methods of assessment strengthened the finding of this decline in activity and further augmented the validity of the self-reported measures used in NGHS. The higher correlation between the Caltrac and AD scores, as compared with that between the Caltrac and habitual activity scores, further supported the internal consistency of the validation of AD and HAQ, because the AD and Caltrac assessed activity during the same 3-d period, whereas the HAQ score reflected activities during a whole year span. On the other hand, there was consistency in the longitudinal trend as assessed by the AD and HAQ scores, which suggests that each of these two instruments can be a useful method for assessing longitudinal changes in physical activity during adolescence. Further, these instruments accommodated the evolving pattern of activities which change inevitably as the participants enter young adulthood. To accommodate the female gender of the NGHS cohort, these instruments had been specifically developed to include occupational and household chores particularly relevant to young women, thus rendering greater sensitivity as well as specificity in assessing levels of physical activity in women.
In summary, the NHLBI Growth and Health study has provided a unique opportunity to develop useful approaches to collecting data to track changes in the level of physical activity in a multicenter, prospective study. Although some data collection form changes may be inevitable during the course of a long-term study, it is essential to ensure that the final statistics are free of the numerical impact of those changes to render greater scientific validity to the results.
Finally, the NGHS tracked the extent of the actual decline in activity levels in 2322 girls from ages 9 through 19 yr. Closer examination of the pattern of this dramatic decline and the differential rate of decline in the daily versus habitual activity should aid in gaining important insights into the nature of this undesirable trend in physical activity during adolescence for the formulation of effective public health policy.
This research was performed under contract NO-HC-55023–26 and grant no. UO1-HL48941–44 of the National Heart, Lung, and Blood Institutes, National Institutes of Health, Bethesda, MD.
We wish to thank George Schreiber, D.Sc., and Barbara Campaigne, Ph.D., for their valuable contribution to the development of the physical activity protocol, and Barbara Schumann, M.A., for the compilation of the extensive NGHS activity data file. We gratefully acknowledge the investigators and staff of the NGHS centers for their dedicated effort and outstanding achievement in the cohort follow-up.
Participating NGHS Centers—Clinical Centers: Children’s Hospital Medical Center, Cincinnati, OH (S. Daniels, M.D., Ph.D., principal investigator); University of California, Berkeley, CA (Z. I. Sabry, Ph.D., principal investigator); Westat, Inc., Rockville, MD (G. B. Schreiber, D.Sc., principal investigator). Coordinating Center: Maryland Medical Research Institute, Baltimore, MD (B. A. Barton, Ph.D., principal investigator). N.I.H. Program Office: Division of Epidemiology and Clinical Applications, National Heart, Lung, and Blood Institute, Bethesda, MD (E. Obarzanek, Ph.D., project officer).
The funding for this study is solely from the National Heart, Lung and Blood Institute, and hence, there is no financial relationship that may lead to a conflict of interest.
1. Aaron, D. J., A. M. Kriska, S. R. Dearwater, et al. The epidemiology of leisure physical activity in an adolescent population. Med. Sci. Sports Exerc. 25:847–853, 1981.
2. Aaron, D., A. Kriska, S. Dearwater, J. Cauley, K. Metz, and R. Laporte. Reproducibility and validity of an epidemiologic questionnaire to assess past year physical activity in adolescents. Am. J. Epidemiol. 142:191–201, 1995.
3. Ainsworth, B. E., W. L. Haskell, A. S. Leon, et al. Compendium of physician activities: classification by energy costs of human physical activities. Med. Sci. Sports Exerc. 25:71–80, 1993.
4. Baranowski, T., R. J. Dworkin, C. J. Cieslik, et al. Reliability and validity of self-report of aerobic activity: Family Health Project. Res. Q. Exerc. Sport 55:309–17, 1984.
5. Bouchard, C., A. Tremblay, C. Leblanc, G. Lortie, R. Savard, and G. Theriault. A method to assess energy expenditure in children and adults. Am. J. Clin. Nutr. 37:461–467, 1983.
6. Centers for Disease Control, and Prevention. CDC Surveillance Summaries, August 14, 1998. MMWR
47(No. SS-3), 1998.
7. Crocker, P. R. E., D. A. Bailey, R. A. Faulkner, K. C. Kowalski, and R. McGrath. Measuring general levels of physical activity: preliminary evidence for the Physical Activity Questionnaire for Older Children. Med. Sci. Sport Exerc. 29:1344–1349, 1997.
8. Durant, R. H., T. Baranowski, H. Davis, T. Rhodes, W. O. Thompson, K. A. Greaves, and J. Puhl. Reliability and variability in heart rate monitoring in 3-, 4-, or 5-yr old children. Med. Sci. Sports Exerc. 24:265–271, 1992.
9. Fontvielle, A. M., I. T. Harper, R. T. Ferraro, M. Spraul, and E. Ravussin. Daily energy expenditure by five-year-old children, measured by doubly labeled water. J. Pediatr. 123:200–207, 1993.
10. Goran, M. I., W. H. Carpenter, and E. T. Poehlman. Total energy expenditure in 4- to 6-year-old children. Am. J. Physiol. 64:E706–E711, 1993.
11. Janz, K. F. Validation of the CSA accelerometer for assessing children’s physical activity. Med. Sci. Sports Exerc. 26:369–375, 1994.
12. Katch, F. I., and W. D. McArdle. Nutrition, weight control, and exercise.
Atlanta, GA: Houghton Mifflin, 1977, Appendix D.
13. Kemper, H. C. G., and W. van Mechelen. Methods of measurement used in longitudinal studies. In: The Amsterdam Growth Study: A Longitudinal Analysis of Health, Fitness and Lifestyle, H. C. G. Kemper (Ed.). Champaign, IL: Human Kinetics, 1995, pp. 28–51.
14. Klesges R. C., L. M. Klesges, A. M. Swenson, and A. M. Pheley. A validation of two motion sensors in the prediction of child and adult physical activity levels. Am. J. Epidemiol. 122:400–410, 1985.
15. Kriska A. M., and C. J. Casperson. Introduction to a collection of physical activity questionnaires. Med. Sci. Sports Exerc. 29:S3–S9, 1997.
16. Ku, L. C., L. R. Shapiro, P. B. Crawford, and R. L. Hunenemann. Body composition and physical activity in 8-year-old children. Am. J. Clin. Nutr. 34:2770–2775, 1981.
17. Laporte, R. E., L. L. Adams, D. D. Savage, G. Brenes, S. Dearwater, and T. Cook. The spectrum of physical activity, cardiovascular disease and health: an epidemiologic perspective. Am. J. Epidemiol. 120:507–517, 1984.
18. Laporte, R. E., H. J. Montoye, and C. J. Caspersen. Assessment of physical activity in epidemiologic research: Problems and prospects. Public Health Rep. 100:131–146, 1985.
19. Maliszewski A. F., P. S. Freedson, C. J. Ebbeling, J. Crussemeyer, and K. B. Kastango. Validity of the Caltrac accelerometer in estimating energy expenditure and activity in children and adults. Pediatr. Exerc. Sci. 3:141–151, 1991.
20. Matthews, C. E., and P. S. Freedson. Field trial of a three-dimensional activity monitor: comparison with self report. Med. Sci. Sports Exerc. 27:1071–1078, 1995.
21. McDowell, M. A., R. R. Briefel, and K. Alaimo. Energy and macronutrient intakes of persons ages 2 months and over in the United States:Third National Health and Nutrition Examination Survey, Phase 1, 1988–91.
Advance data from vital and health statistics; No. 265. Hyattsville, MD: National Center for Health Statistics, 1994, pp. 1–24.
22. McKenzie, T. L. Observational measures of children’s physical activity. J. Sch. Health 61:224–227, 1991.
23. 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.
24. The National Heart, Lung, and Blood Institute Growth, and Health Study Research Group. Obesity and cardiovascular disease risk factors in black and white girls: the NHLBI Growth and Health Study. Am. J. Public Health. 82:1613–1620, 1992.
25. Puhl, J., K. Greaves, M. Hoyt, and T. Baranowski. Children’s Activity Rating Scale (CARS): description and calibration. Res. Q. Exerc. Sport 61:26–35, 1990.
26. Sallis, J. R., M. J. Buono, J. J. Roby, D. Carlson, and J. A. Nelson. The Caltrac accelerometer as a physical activity monitor for school-age children. Med. Sci. Sports Exerc. 22:698–703, 1990.
27. Sallis, J. R., M. J. Buono, J. J. Roby, F. G. Micale, and J. A. Nelson. Seven-day recall and other physical activity self-reports in children and adolescents. Med. Sci. Sports Exerc. 25:99–108, 1993.
28. SAS Institute. SAS® Procedures Guide, Version 6. Cary, NC: SAS Institute Inc., 1990, pp. 617–634.
29. Troiano, R. P., K. M. Flegal, R. J. Kuczmarski, S. M. Campbell, and C. L. Johnson. Overweight prevalence and trends for children and adolescents. The National Health and Nutrition Examination Surveys 1963 to 1991. Arch. Pediatr. Adolesc. Med.
149: 1085–1091, 1995.
30. U. S. Department of Health and Human Services. Physical Activity and Health: A Report of the Surgeon General. Atlanta, GA: U.S. Department of Health and Human Services, Centers for Disease Controls and Prevention, National Center for Chronic Disease Prevention and Health Promotion, 1996, pp. 175–207.
31. Verschuur, R., and H. C. G. Kemper. The pattern of daily physical activity. In:Growth, Health and Fitness of Teenagers: Longitudinal Research in International Perspective
H. C. G. Kemper (Ed.). Basel-Karger, 1985, pp. 169–185.
32. Wilson, P. W., R. S. Paffenbarger, J. N. Morris, and R. J. Havlik. Assessment methods from physical activity and physical fitness in population studies: report of NHLBI workshop. Am. Heart J. 111:1177–1192, 1986.
33. Zeger, S. L., and K. Y. Liang. Longitudinal data analysis for discrete and continuous outcomes. Biometrics 42:121–130, 1986.