Many self-report questionnaires are available to assess physical activity; however, the method of choice can depend on the design and purpose of the study as well as on the age of the individuals. Questionnaires differ on the types of activities assessed and whether intensity and/or duration of activity are probed. Also, the individual may be asked to recall recent activity (i.e., the past several days, such as in the previous day physical activity recall (PDPAR)) (18) or usual activity over the past year (such as in the self-administered physical activity checklist (SAPAC)) (14). Self-reported physical activity instruments are often preferable to more objective measures in large-scale studies due to lower costs and lessened staff and participant burden.
Accurately assessing physical activity is critical for examining associations between physical activity and health outcomes, such as obesity, cardiovascular disease, and diabetes. Although numerous physical activity questionnaires have been developed and validated in adults, fewer questionnaires have been developed to assess physical activity in children. One commonly used instrument, the PDPAR, was validated in adolescents in the 7th to the 12th grades, with correlations of 0.53–0.88 between relative energy expenditure from the PDPAR and either pedometer counts, Caltrac accelerometer counts, or daily heart rate (18), but has not been assessed in younger children. The reliability and validity of self-report questionnaires may be generally lower in children and adolescents compared with values observed in adults because of the imprecision in cognitive processing and in recall of physical activity, both of which are complex cognitive tasks (4).
Because the potential effect of inactivity on obesity and cardiovascular disease risk in children is a serious public health concern, longitudinal studies examining changes in physical activity/inactivity, adiposity, and cardiovascular disease risk within individuals are important to more fully understand how interactions between these important biological and lifestyle parameters affect health risk throughout childhood and adulthood. The Fels longitudinal study (6,7,10,15,19) is unique in that serial measurements are collected in individuals from birth throughout adulthood. Since 1988, annual measurements of physical activity have been collected from children 8 to 18 yr of age using a questionnaire (Fels physical activity questionnaire (PAQ) for children) that was extensively modified from the Baecke questionnaire for habitual physical activity (2,3), which was designed for use in adults. The validity of this questionnaire was not assessed at the time of its implementation into the Fels study; however, if valid, the long-term benefits of physical activity throughout childhood and into adulthood can be evaluated, the findings of which will have important public health value.
The purpose of this study was to evaluate the reliability and validity of the Fels PAQ for children in boys and girls 7–19 yr of age.
Throughout the school year, excluding winter, girls and boys, 7–19 yr of age and residing in a rural area (eastern shore of Maryland), participated in the study. None of the children had physical limitations, and all of them were able to participate in physical education in school. The study was approved by the Institutional Review Board for Human Subject Research for the Johns Hopkins Bloomberg School of Public Health (Baltimore, MD). In addition, approval was obtained by the county school board of education. Consent to participate was obtained from one parent, and assent was obtained from each boy or girl.
The final subject sample included 130 girls and 99 boys 7–19 yr of age. The sample was further divided into three school-age categories: elementary school (grades 3–5, N = 70); middle school (grades 6–8, N = 81), and high school (grades 9–12, N = 78).
Study design and protocol.
Students from one elementary school and one combined middle/high school participated in the study. All the study procedures were conducted within the school physical education classes, and all students of several classes were invited to participate. The study was conducted over a period of six study days and required two school visits for each participant (Table 1). On the first day, students were enrolled and oriented to study procedures, and measures of chronologic age, race/ethnicity, weight, height, and self-reported physical activity by the Fels PAQ for children were taken by field staff. The participants were then fitted with the accelerometer by the field staff and asked to wear the accelerometer during the entire day for six complete days, except while bathing or swimming. At the end of the sixth day, a staff member returned to the school to collect the accelerometer and to administer the Fels PAQ for children. A staff member was available at all times during the administration of the test to answer any questions. At the end of the monitoring period, there were approximately 1–2 d during which data from the accelerometers were downloaded and accelerometers were programmed for reuse.
Physical activity by accelerometry.
Activity was monitored by an omnidirectional accelerometer, the Actiwatch (Model AW16; MiniMitter, Sunriver, OR). The epoch length (sampling time) was set at 1 min for this study. The accelerometer was worn on the right hip on a belt fitted around the waist of the participants. The participants were instructed to wear the accelerometer for 6 d (four school days and two weekend days), except while bathing or swimming. The staff fitted the belt/accelerometer on the participant and answered any questions about them. On retrieving the accelerometers, the staff member downloaded the data through the Actiwatch Reader interface unit via a wireless link.
Physical activity by self-report: Fels PAQ for children.
The original Fels PAQ for children was a self-administered questionnaire assessing habitual physical activity “since your [the child’s] last visit” without reference to a particular time frame (e.g., past day, month, or year). Because childhood physical activity is measured annually in the Fels longitudinal study, we modified the recall time frame for this study to ask about physical activity performed “in the last year” to mimic the time frame used in the Fels study. This eight-item questionnaire contains three “open” questions for which activities are listed by the participant and the frequency of participation for each activity is obtained. The remaining five questions use a Likert scale to evaluate physical activity. (See Appendix A for the components of the questionnaire.)
The Fels PAQ for children was extensively modified from the Baecke questionnaire for habitual physical activity, which assesses physical activity in adults. Questions were revised to more adequately reflect children’s activity patterns (e.g., more “open” response questions were added regarding sports activities and a “chore” section replaced the “work” section of the original questionnaire). In addition, some questions were deleted and others were added to create the new instrument. Although the Fels PAQ for children maintains little resemblance to the original Baecke questionnaire, the scoring mechanism for the Fels PAQ was modeled after the Baecke questionnaire in that it provides for a total physical activity score as well as for indices of sport, leisure, and work (or chore) activity. Activities listed in the “open” questions of the Fels PAQ for children were ranked as low, moderate, or high intensity based on MET values used for adults (1). Sport activities with MET values ≤4.5 were considered to be low intensity, those with values of 4.5–7.9 were considered to be of moderate intensity, and those with MET values of ≥8.0 were considered to be of high intensity. The MET cut points for sports were estimated from the activities listed as belonging to the low-, middle-, and high-intensity levels of sports as reported by Baecke et al. (2,3). Chore activities with MET values ≤3 were considered to be low intensity, those with values of 3–4.9 were considered to be of moderate intensity, and those with values of ≥5.0 were considered to be of high intensity. The thresholds for chores were selected to define three distinct intensity levels, and are very similar to those defined by others (1) of <3, 3–6, and >6 METs. The slight difference in thresholds for the highest intensity chores would likely not affect our results, as the chores in our study were typically in the low-intensity (<3 METs) range. As well, the chores in the high-level range that were noted in our study (e.g., mowing the lawn) would be classified as high intensity by others as well. In addition, the frequency of engaging in listed activities was also considered in the scoring of the questionnaire. Active games (e.g., tag) were not scored if listed in response to Question 1 that asked “In the last year, what sports did you play in school?” However, active games were included in response to Question 2, “In the last year, what sports or physically active games did you play outside of school?” The scoring algorithm was recomposed as follows: The statement “During leisure time I watch television or read: never, seldom, sometimes, often, or very often” was removed from the data analysis, as it was a measure of inactivity. Also, the statement (Question 4) “During leisure time I play sports: very often, often, sometimes, seldom, or never” was removed from the analyses for the sport score and was added to the leisure score. The exact scoring algorithm used for all analyses in the current investigation, complete with formulas, is shown in Appendix B.
The Actiwatch data were considered to be complete if the participant completed 70% of the day (in other words, at least 1000 min of activity must have been recorded per day). Each participant must have worn the accelerometer for at least 4 of the 6 d. These criteria excluded three boys. The raw data were downloaded directly from the accelerometers and were then summarized by using a specially designed macro in Microsoft Excel (Microsoft, Seattle, WA). The activity over the 6 d (counts per day) was calculated. In addition, the activity excluding sleep (counts per minute) was calculated by graphically determining the participants’ sleep time and then excluding counts during that time frame. Physical activity between boys and girls and across age groups was compared. Because the data were not normally distributed, nonparametric tests, such as the Wilcoxon test and Kruskal–Wallis test, were performed.
The reliability of the two administrations of the Fels PAQ for children (total activity score and each index for sport, leisure, and work) was determined using the intraclass correlation (ICC).
To test the validity of the Fels PAQ for children, Spearman correlations were performed between the Fels PAQ for children and the accelerometer measurements in counts per minute for all days, weekdays, and weekend days. These correlations were completed for each test administration of the Fels PAQ as well as for the mean total score and all indices (sport, leisure, and work) and the Actiwatch data. The Actiwatch data were also separated into weekdays and weekend days.
P14 STATA for Windows (version 7.0; Stata Corporation, College Station, TX) was used. Data are presented as means ± SD or ± SE; a two-tailed P < 0.05 was taken as indicating significance.
The physical activity data from the Actiwatch (counts per minute and counts per day) and the Fels PAQ for children are shown in Table 2. There were significant differences between the girls and the boys for the measures by the Actiwatch. The boys have significantly higher accelerometry counts than the girls for both average counts for whole day and for counts per minute, excluding sleep. The accelerometry (counts per day and counts per minute) for the middle school participants were significantly lower than the elementary school children and higher than the high school participants.
For the Fels PAQ for children, the boys have a significantly higher sport score as well as a higher leisure score translating into a higher total score. No differences in the work score (P = 0.59) were found between the girls and boys. For the sport score from the Fels PAQ for children, the middle school participants had a higher average score than the elementary school children.
Reliability of the Fels PAQ for children.
The reliability (ICC) of the Fels PAQ for children for the girls, boys, and the elementary, middle, and high school age groups ranged from 0.48 to 0.76 (Table 3). Reliability was the highest for the sport score in the high school participants (r = 0.76).
Validity of the Fels PAQ for children.
Table 4 presents the validity of the self-report activity measure (Fels PAQ for children) and the Actiwatch for the different age groups. For the elementary school children, the correlations were significant for both total score and sport score for all days, weekdays, and weekend days, with r values ranging from 0.29 to 0.36. For the middle school participants, the corresponding correlations were low and not significant for any days (range, r = 0.008–0.13). For the high school participants, the correlation between the total score and the Actiwatch was significant for all days. The correlation between the sport score and the Actiwatch was also significant for all days, weekdays, and weekend days for the high school participants. The correlation between the leisure score and the Actiwatch was significant for all days and weekdays for the elementary and middle school participants.
In this study, the reliability of the Fels PAQ for children was determined by comparing two administrations of the measure spaced 6 d apart. We found an average reliability (0.62–0.71) in our three age groups for the total score. Our reliability of approximately 0.62–0.71 is similar or somewhat lower than that observed in children (14,16,18), and lower than that observed in adults (2,8). The SAPAC had a reliability of 0.65 in elementary school children (14). In young girls, the Girls Health Enrichment Multisite Studies (GEMS) physical activity questionnaire had a reliability of 0.80 (16). The PDPAR had a high reliability (r = 0.98) when administered on the same day (18). By comparison, in adults, the test-retest reliability of the original Baecke questionnaire for total activity was 0.93 for a 1-month period (r = 0.86 for leisure and r = 0.90 for sports) (8). For a 3-month period, the reliability was slightly lower (r = 0.88 for work score, r = 0.74 for leisure, and r = 0.81 for sports) (8). Lower correlations for reliability among children versus adults are expected due to the lower cognitive precision (4). In addition, the instruments are not directly comparable due to the extensive revision of the Fels PAQ for children from the original Baecke. Another possible explanation for the lower reliability is the less structured (open-ended) format of the Fels PAQ, since the questionnaire contains three items in which the participant must fill in the activity rather than choose from a list.
We also examined the reliability for the separate components of the total score (i.e., the sport, leisure, and work scores). The reliability of the sport score was lower than that of the total score in girls, boys, and the elementary school and middle school participants but not in the high school participants. These older children likely have better recall ability than the younger children. As well, this may be due to the higher prevalence of organized sports (e.g., basketball and soccer) in high school compared with middle school or elementary school. This would then be reflected in better reliability with the sport score, as was shown in this study. The reliability of the sport score was highest (r = 0.76) in the high school than in any other age group.
We tested the validity of the self-report measures by comparing it with the accelerometer. The validity of the Fels PAQ for children with the Actiwatch was the highest with the elementary school children. Our validity of 0.34 in the elementary school children and 0.21 in the high school participants is somewhat lower than other studies in children (9,13,15). The validity of self-report assessments compared with accelerometers has been reported to vary from 0.26 to 0.60 in children and adolescents (11,13,16,17). Findings similar to those of the current study have been reported in African American girls completing the GEMS physical activity questionnaire (16). In that study, the R values were 0.27–0.29 with the self-report and the CSA/MTI accelerometer. Our validity results are slightly higher than the previous studies in adults using the original Baecke questionnaire (8,12). Jacobs et al. (8) found r = 0.19 for the total score, r = 0.01 for the leisure score, and r = 0.32 for the sport score. Our first test administration on day 1 (score 1) of the Fels PAQ for children to the high school participants yielded a low correlation with the activity counts, whereas our second administration on day 6 (score 2) did much better. The reason for this discrepancy is unknown, but perhaps the high school participants had a clearer understanding of the instrument on reading it a second time or remembered their activities better.
The validity of the instrument was also evaluated for its components. Of interest, we found that the Fels PAQ for children sport score worked well with both elementary school and high school participants. Although the correlation for the sport score was higher than that for total activity in high school students, the sport index performed equally well to the total activity score for elementary school children and was comparable with the sport score for high school students as well. It is unclear why the middle school participants had the poorest validity. Perhaps this is due to the lack of reported sports in elementary school children because their time is spent in more active games or play. The high school participants reported more sports. It is possible that, when answering the questionnaire, the elementary school and high school participants were able to more reliably answer the sport questions. The questionnaire contained two sport questions and thus the total score could be influenced to a larger extent with these questions. As well, better correlations for the total activity and sport index were found with activity counts during school days (weekdays) than during weekend days for the elementary school and high school students. The reason for this might be that the children have more regular sports activity at school than at home, making it easier to recall these types of structured activities. The middle school participants may not be able to answer the sport question as accurately as the elementary school and high school students because they may be less involved in both active play and organized sports. These middle school participants may be spending their time in leisure activities, which they may be better able to recall. Leisure activity may take up more of the total score. This is evidenced by the significant correlations observed between the leisure score and total activity and the activity on the weekdays and weekend days.
The Fels PAQ for children was designed to reflect “usual” activity over a 6- to 12-month period in the original study of children in the Fels longitudinal study. By using an objective measure of physical activity such as the accelerometer, we were able to provide precise and accurate data on children’s physical activity. The validation standard, the Actiwatch, was worn for 6 d. Although this is only a fraction of the time frame that the Fels PAQ for children is evaluating, the accelerometer data should reflect the child’s usual activities during a random week of the school year. However, we had to assume that the 6 d of activity monitoring reflected “usual” activity over the course of the year for these children and adolescents.
Given the moderate psychometrics of the Fels PAQ in children, the questionnaire is useful because it can assess various components of activity, specifically leisure, sport, work, and total activities. Each of these components may be useful to those developing and conducting intervention studies. Increased reliability and validity for activity measures in children may best be accomplished by using more than one instrument in certain research studies, particularly if physical activity is a key outcome variable.
Two previous studies have evaluated the reliability and validity of the Actiwatch (5,11). The Actiwatch was compared with direct observation (Children’s Activity Rating Scale (CARS)) in 41 preschool-age children for a 3- to 6-h period (5). The 10-min sensor readings correlated with the 10-min CARS (r = 0.60, P < 0.001). This was dependent on the child’s activity for those hours, with higher correlations in the most active children. In 26 children (7–14 yr of age) who spent 6 h in a room respiration calorimeter, the mean minute-to-minute correlation between energy expenditure and the Actiwatch counts was higher than energy expenditure and Actigraph counts (r = 0.78 ± 0.06 vs r = 0.66 ± 0.06) (11).
Although this study enrolled a large number of children and adolescents, it is limited in its generalizability to other ethnic populations (e.g., Hispanics, Asians). The sample was recruited from a rural area in our attempt to mimic the sample of children enrolled in the Fels longitudinal study. Another limitation is that accelerometers do not detect some activities well (e.g., biking), and we may have missed certain activities such as swimming. Furthermore, the MET values used to classify sport and chore activities were based on adult data rather than those among children. A comprehensive list of MET values of physical activities among children was not available at the time of development of the questionnaire. Because data were only used to classify sports and chores in terms of relative intensity, use of the adult data should have only a minor effect on results.
In conclusion, the Fels PAQ for children is moderately reliable for all age groups of children. The validity of the Fels PAQ for children is acceptable for elementary school and high school students when either the total activity score or the sport index is used. The sport index performed equally well with the total activity score for elementary school students but was a better measure of physical activity than the total score among high school students. The leisure score performed the best for the middle school participants. Although correlation coefficients are low for validity, these values are comparable (or in some cases higher) than validity coefficients between self-reported activity questionnaires and various criterion measures of activity among adults.
1. Ainsworth, B. E., W. L. Haskill, M. C. Whitt, et al. Compendium of physical activities: an update of activity codes and MET intensities. Med. Sci. Sports Exerc.
2. Baecke, J. A., J. Burema, and J. E. Frijters. A short questionnaire for the measurement of habitual physical activity in epidemiological studies. Am. J. Clin. Nutr.
3. Baecke, J. A. Questionnaire of habitual physical activity. Med. Sci. Sports Exerc.
4. Baranowski, T. R. Validity and reliability of self-report of physical activity: An information processing perspective. Res. Q.
5. Finn, K. and B. Specker. Comparison of Actiwatch sensors and CARS on assessment of physical activity in children. Med. Sci. Sports Exerc.
6. Guo, S. S., C. Zeller, W. C. Chumlea, and R. M. Siervogel. Aging, body composition, and lifestyle: the Fels Longitudinal Study. Am. J. Clin. Nutr.
7. Guo, S. S., C. Huang, L. M. Maynard, et al. Body mass index during childhood, adolescence and young adulthood in relation to adult overweight and adiposity: The Fels Longitudinal Study. Int. J. Obes. Relat. Metab. Disord.
8. Jacobs, D. R. Jr., B. E. Ainsworth, T. J. Hartman, and A. S. Leon. A simultaneous evaluation of 10 commonly used physical activity questionnaires. Med. Sci. Sports Exerc.
9. Kimm, S. Y. S., N. W. Glynn, A. A. Kriska, et al. Longitudinal changes in physical activity in a biracial cohort during adolescence. Med. Sci. Sports Exerc.
10. Maynard, L. M., S. S. Guo, W. C. Chumlea, et al. Total-body and regional bone mineral content (BMC) and areal bone mineral density (BMD) in children aged 8–18 y: The Fels Longitudinal Study. Am. J. Clin. Nutr.
11. Puyau, M. R., A. L. Adolph, F. A. Vohra, and N. F. Butte. Validation and calibration of physical activity monitors in children. Obes. Res.
12. Richardson, M. T., B. E. Ainsworth, H. C. Wu, D. R. Jacobs Jr., and A. S. Leon. Ability of the Atherosclerosis Risk in Communities (ARIC)/Baecke Questionnaire to assess leisure-time physical activity. Int. J. Epidemiol.
13. Roche, A. F. Growth, Maturation, and Body Composition: The Fels Longitudinal Study 1929–1991.
London: Cambridge University Press, 1992, pp. 1–282.
14. Sallis, J. F., P. K. Strikmiller, D. W. Harsha, et al. Validation of interviewer- and self-administered physical activity checklists for fifth grade students. Med. Sci. Sports Exerc.
15. Siervogel, R. M., W. Wisemandle, L. M. Maynard, et al. Serial changes in body composition throughout adulthood and their relationships to changes in lipid and lipoprotein levels. The Fels Longitudinal Study. Arterioscler. Thromb. Vasc. Biol.
16. Treuth, M. S., N. E. Sherwood, N. F. Butte, et al. Validity and reliability of activity measures in African-American girls for GEMS. Med. Sci. Sports Exerc.
17. Trost, S. G., D. S. Ward, S. M. Moorehead, P. D. Watson, W. Riner, and J. R. Burke. Validity of the computer science and applications (CSA) activity monitor in children. Med. Sci. Sports Exerc.
18. Weston, A. T., R. Petosa, and R. R. Pate. Validation of an instrument for measurement of physical activity in youth. Med. Sci. Sports Exerc.
19. Wisemandle, W., L. M. Maynard, S. S. Guo, and R. M. Siervogel. Childhood weight, stature, and body mass index among never overweight, early-onset overweight, and late-onset overweight groups. Pediatrics
APPENDIX A. FELS PAQ FOR CHILDREN
APPENDIX B. SCORING OF THE FELS PAQ FOR CHILDREN