| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
ORIGINAL RESEARCH COMMUNICATION |
1 From Temple University, Philadelphia, PA (KEB, GDF, SVV); the University of Pennsylvania, Philadelphia, PA (HM, RBF); The Food Trust (AK, SS), Philadelphia, PA; Bryn Mawr College, Bryn Mawr, PA (KG); and the School District of Philadelphia, Philadelphia, PA (JN)
2 Supported by grants from the Centers for Disease Control and Prevention (R06/CCR321534-01) and the US Department of Agriculture/Food and Nutrition Service through the Pennsylvania Nutrition Education Program as part of Food Stamp Nutrition Education.
3 Address reprint requests and correspondence to KE Borradaile, Center for Obesity Research and Education, Temple University, 3223 North Broad Street, Suite 175, Philadelphia, PA 19140. E-mail: borradak{at}temple.edu.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Objective: The purpose of this study was to investigate the relation between body mass index (BMI) z score and self-report measures of diet and activity, the Youth/Adolescent Questionnaire (YAQ) and the Youth/Adolescent Activity Questionnaire (YAAQ), respectively.
Design: Participants were 1092 students in grades 4 through 6 from 10 schools in a US city in the middle Atlantic region with 
50% of students eligible for free or reduced-price meals. Students were assessed at baseline and again after 2 y. The relation between self-reported energy intake (YAQ) and activity (physical and sedentary) (YAAQ) and BMI z score was explored from both a cross-sectional and longitudinal perspective.
Results: The YAQ (energy intake) and YAAQ (physical and sedentary activity) did not relate to BMI z score in the expected directions from either a cross-sectional or longitudinal perspective.
Conclusion: In this large, racially diverse sample, the YAQ and the YAAQ were not significantly associated with BMI z score or changes in BMI z score.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The Youth/Adolescent Questionnaire (YAQ) and the Youth/Adolescent Activity Questionnaire (YAAQ) are widely used self-report measures of diet and activity, respectively, in children and adolescents aged 9–18 y. More than 50 studies reporting on results obtained with use of the YAQ and/or YAAQ have been published (contact the corresponding author for an exhaustive list). Fewer than half of these studies either directly or indirectly assessed the relation between these instruments and biomarkers or measures of adiposity. The YAQ and the YAAQ have been used as primary and secondary outcomes in areas such as childhood obesity, nutrition, physical activity, eating disorders, body satisfaction, bone density, pulmonary function, leukemia, and metabolic function (1-12).
Accurate measures of diet and activity are critical in understanding how these lifestyle and behavioral factors covary to affect health status. A validation study was conducted by the test developers who administered the YAQ, via mail, to the sons and daughters of nurses participating in an earlier study (13). The researchers compared mean-adjusted nutrients between the average of 3 24-h recalls and the average of 2 YAQs and observed a moderate average correlation (r = 0.54). This original validation study has its limitations, however. First, the method of administration (mail) may not represent the typical method of administration, and it does not ensure that these children completed the instrument independently. Second, the validation sample (children of nurses) may be too restricted and homogeneous to be generalizable to the larger intended population of 9–18-y-olds. Third, the criterion variable was obtained via self-report rather than with an objective measure.
Two additional studies, not conducted by the test developers, directly assessed the validity of the YAQ and concluded that the instrument has low validity compared with multiple food records in low-income African American and Hispanic youth (14) and does not provide an accurate estimation of mean energy intake for an individual compared with the criterion total energy expenditure by doubly labeled water (15). To our knowledge, no study has validated the YAAQ, and no study has jointly assessed both the YAQ and the YAAQ against a measure of adiposity such as body mass index (BMI) z score in a low-income inner-city school-based setting.
The YAQ and YAAQ are frequently used in evaluations of interventions and programs designed to improve BMI, but it is unclear whether these instruments correlate with BMI. The purpose of this study was to investigate the relation between BMI z score and the YAQ and the YAAQ as self-report measures of diet and activity.
|
SUBJECTS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Questionnaires
Description of YAQ
The YAQ, a widely used 152-item food frequency questionnaire, is a unidimensional assessment of the previous year's diet in 9–18-y-olds and requires 
20–30 min to complete (17). A typical item asks the respondents to report the frequency with which they consumed a particular food (eg, yogurt, potato chips, and noodles) over the previous year. Response categories differ by type of food; more popular items (eg, apple juice) have response categories represented by days, weeks, and months, whereas less popular foods (eg, raisins) have weekly or monthly options. Frequencies for each of the 152 items are synthesized and transformed by Channing Laboratory, Brigham & Women's Hospital, into a series of nutrient indices representing daily intake (eg, total energy, fat, fiber, and calcium) (17).
Description of YAAQ
The YAAQ is a 24-item inventory and a two-dimensional measure of the previous year's activity (physical and sedentary activity) in 9–18-y-olds and takes 5–10 min to complete (18). A typical question asks respondents to report the amount of time spent engaged in various physical (eg, basketball, biking, walking, and playing outdoors) and sedentary (eg, watching television, reading, and homework) activities over the previous year. Response categories differ by type of activity (physical or sedentary). Physical activity responses range from <1/2 h/wk to 6 h/wk, whereas sedentary activity response options, which are broken down by weekday and weekend, are represented by whole numbers of hours (0–5 h). Frequencies of individual items are summed by Channing Laboratory to reflect the average number of hours of physical and sedentary activities per 1 wk.
The activities included in the YAAQ are not mutually exclusive. Thus, if a child spent, on average, 1 h/wk watching television with friends, he or she would receive a score from the YAAQ for 2 h of sedentary activity (1 h for watching television and 1 h for hanging out with friends). Furthermore, because sedentary activity is recorded in integers, a child who spends 30 min engaged in sedentary activity is forced to choose either 0 h or 1 h.
Administration
The YAQ and YAAQ were not altered in any way and were obtained directly from the questionnaire developers. At least 2 research assistants (RAs) were present for each questionnaire administration. RAs administered the questionnaires and read the instructions and items aloud, keeping a constant pace between test administrations. RAs circled the room to answer questions, and they instructed the children not to talk while they completed the instruments. The questionnaires were usually administered in the student's classroom, but occasionally larger group administrations were conducted in cafeterias, theaters, or gymnasiums. Classroom teachers either aided in behavior management only or left the classroom completely.
Scoring
The completed YAQ and YAAQ questionnaires were collected by the RAs and were sent to Channing Laboratory where they were scanned and scored. Channing Laboratory provided our research team with the scores of the questionnaires, which included, for every individual, values for total energy (measured in kcal/d obtained from the YAQ) and for physical activity and sedentary activity (both measured in h/wk obtained from the YAAQ). Channing Laboratory also provided additional nutrient information obtained from the YAQ (eg, fat and fiber) as well as all individual-level data for both questionnaires. All analyses in this paper were performed on the aggregate scores (total energy, physical activity, and sedentary activity), which were received directly from Channing Laboratory.
BMI
Weights and heights were measured annually on a digital scale and wall-mounted stadiometer by a trained research team with use of a standardized protocol (19) that was adapted for a school-based setting. Student height was measured without shoes to the nearest 0.1 cm, and weight was measured in light clothing. BMI z scores and percentiles based on age and sex were calculated for each student with use of the 2000 Centers for Disease Control and Prevention growth charts (20). Age- and sex-normed BMI percentiles were used to assign each participant to a relative weight category based on Institute of Medicine guidelines (21). BMI z score was selected as the criterion for the YAQ and YAAQ for several reasons. First, because intake and activity drive BMI and because the YAQ and the YAAQ are used to measure intake and activity, BMI should be related to the YAQ and the YAAQ. Second, BMI for age correlates strongly with other measures of body fatness (22). Third, collection of height and weight data is a noninvasive procedure that is routinely performed in a school setting. Use of weight or raw BMI is an alternative criterion, but neither of these takes into account relevant information such as sex- and age-specific growth, unlike BMI z score.
Analyses
Outliers
We defined outliers as scale scores (total energy, physical activity, and sedentary activity) >1.5 times the interquartile range above the third quartile (75th percentile) or <1.5 times the interquartile range below the first quartile (25th percentile) (23). About 8% (n = 85) of observations were outliers and were removed from all analyses. These were from individuals who typically selected the maximum frequency for every item. Outliers did not differ from nonoutliers with respect to sex (P = 0.17), race (P = 0.29), relative weight status (P = 0.75), age (P = 0.88), or BMI z score (P = 0.64). Models were run with and without outliers; the presence of outliers did not change the results.
Cross-sectional validity
Baseline BMI z score was regressed on all 3 scales (total energy, physical activity, and sedentary activity) at baseline. Demographic variables (eg, sex, race/ethnicity, and age) were included as additional predictors in the multiple regression model. Subsequent analyses were rerun by demographic subgroup to determine whether the relation between self-reported diet (YAQ) and self-reported activity (YAAQ) and BMI z score varied systematically between groups. Thus, models were run separately for boys and girls to account for developmental variability as well as for race/ethnicity, age, and relative weight categories. Bonferroni corrections were made to adjust for the type I error rate associated with making multiple comparisons. Scale scores (total energy, physical activity, and sedentary activity) were also compared across relative weight categories with use of Tukey's honestly significant difference to account for multiple comparisons.
Longitudinal validity
Because there are limitations to cross-sectional data, a longitudinal analysis was our primary method for establishing the validity of the YAQ and YAAQ. Analysis of covariance (ANCOVA) models, which held constant an individual's observed and unobserved attributes, were used to assess whether changes in an individual's diet and activity, over a 2-y period, were related to BMI z score at year 2 (after controlling for baseline BMI z score). ANCOVA is preferable to modeling differences (year 2 – baseline) when growth in the outcome (BMI z score) increases over time (24), which we observed in our sample. Nonetheless, an analysis of differences yielded similar results. In these analyses, each individual served as his or her own control, which accounted for intraindividual characteristics that may be associated with BMI z score and, therefore, isolated the relation between changes in BMI z score and total energy, physical activity, and sedentary activity.
In the ANCOVA model, BMI z score at year 2 was regressed on change (year 2 – baseline) in total calories, as measured by the YAQ and changes in physical and sedentary activity, as measured by the YAAQ, as well as on BMI z score at baseline, and the aforementioned demographic variables (race/ethnicity, sex, and age). Models were run for the full sample and then separately by demographic subgroup. Models were also reestimated to include height to account for adolescent height growth spurts, with changes in BMI and weight as dependent variables. The results from these models were similar with respect to direction and significance.
A second, more flexible, longitudinal model was analyzed, which allowed the effects of the change in scale scores (energy intake, physical activity, and sedentary activity) on the change in BMI z score to vary depending on the individual's initial BMI z score. This was accomplished through the inclusion of interaction terms between the changes in energy intake, physical activity, and sedentary activity and initial BMI z score.
General analytic details
All analyses were conducted with use of SAS 9.1.3 (25). School was included as a fixed effect to account for the clustering of students within each school, and model assumptions were assessed (eg, multicollinearity, normality of the dependent variable, and model fit). Analyses that were run on only the control sample did not differ substantively from analyses run on the full sample (treatment and control groups).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Of the 1349 participants assessed at baseline, 1092 completed both the YAQ and the YAAQ and received values for total energy, physical activity, and sedentary activity. Of these 1092 participants, girls comprised 54.1% of the sample and participants were, on average, 11.2 ± 1.0 y of age. About 46% of the sample was African American, followed by Asian (22.2%), Hispanic (15.5%), white (12.3%), and other (eg, more than one race) (4.4%). Most of the participants were categorized, according to Institute of Medicine guidelines (21), as normal weight (57.8%), and 40.2% of the population was either overweight or obese (Table 1
). Item responses in this sample were consistent with the profile of urban children in grades 4 through 6. Popular foods and activities included potato chips, fruit drink, and television watching, whereas unpopular foods and activities included kasha/couscous/bulgur, beets, and hockey/lacrosse.
|
4% of the variance in BMI z score at baseline for the total sample [F(18, 988) = 2.67, P < 0.01]. After we controlled for physical and sedentary activities, sex, race, and age, increases in total energy intake were associated with decreases in BMI z score (Table 2). Physical activity, on the other hand, was positively related to BMI z score such that increases in physical activity were associated with increases in BMI z score. There was no observed statistically significant association between sedentary activity and BMI z score in the total sample. Results were similar in the individual demographic analyses (eg, this relation was maintained when the analysis was performed separately for all levels of sex, race/ethnicity, age, and relative weight status categories).
|
Longitudinal
Baseline BMI z score, sex, race, age, and changes in energy intake, physical activity, and sedentary activity explained about 87% of the variance in BMI z score at year 2 [F(19, 578) = 197.19, P < 0.0001]. Most of this variance (82.5%) was attributed to baseline BMI z score values alone, whereas 4.5% was attributed to sex, race, age, and changes in total energy intake, physical activity, and sedentary activity. There was no statistically significant relation between change in BMI z score and changes in energy intake (P = 0.30), physical activity (P = 0.71), or sedentary activity (P = 0.35) (Table 3) from baseline to year 2. Thus, after controlling for change in physical and sedentary activity, as well as sex, race, and age, changes in energy intake were not associated with changes in BMI z score. Similarly, changes in physical or sedentary activities were not associated with changes in BMI z score. The findings were similar in the model that included interaction terms (Table 3). Again, these analyses held intraindividual characteristics constant and demonstrated that changes in BMI z score were not related to changes in energy intake, physical activity, or sedentary activity. The results from these analyses were similar across all demographic subgroups and were consistent among all choices of dependent variables (eg, age- and sex-adjusted weight and BMI).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
No statistically significant relations between changes in BMI z score and changes in any of the 3 scales (total energy, physical activity, and sedentary activity) were observed in the longitudinal analyses. Similar findings were observed in approaches that partialled out age- and sex-specific growth in height for children in this age range (eg, absolute BMI and weight) and in analyses that used data from year 1 instead of year 2. Modeling these relations separately by subgroup (sex, race, age, and relative weight category) did not change these findings.
It may be possible that the YAQ and the YAAQ are not appropriate instruments for in-school, group administration or for measuring dietary intake and activity in inner-city, low-income fourth through sixth grade children. Nonetheless, our results are consistent with those of previous studies that used different samples and methods and either directly or indirectly compared the instruments to weight-based criterion variables (eg, BMI, BMI z score, weight, and relative weight categories). Gordon et al (2) observed that BMI z score and weight, respectively, increased over time despite a decrease in self-reported energy intake. Counterintuitive findings were also observed in studies that compared BMI to subsets of items (eg, fruits and vegetables) (1-12). When expected findings were observed, they were not consistent across sex (6).
Studies that compared the instruments to non-weight-related biomarkers such as bone mass density, vitamin D deficiency, and stress fractures (2,7–11), also reported contradictory findings; however, these studies were largely limited by their cross-sectional design. Loud et al (9), for example, reported that self-reported low calcium intake, low vitamin D intake, and fewer servings of dairy products were not linked to stress fractures and concluded that these correlates of bone mineral density may not have a large enough effect to overcome the variability in bone mineral density attributable to genetics. A longitudinal design would have ruled out the variability in intraindividual characteristics, by using each participant as his or her own control. In the current study, which used a longitudinal design, we still did not observe the expected relation between changes in self-reported total energy, physical activity, sedentary activity, and BMI z score.
When energy intake was validated against total energy expenditure by doubly labeled water, the YAQ did not provide an accurate estimation of energy intake at the individual level, but it may estimate energy intake at the group level (15). Even if the YAQ estimates energy intake at the group level, it precludes researchers from using common methodologic procedures such as estimating inferential change analyses (eg, pretest/posttest), using individual-level predictors (eg, sex, race, or others), and estimating mediator, moderator, or subgroup analyses. In 2 studies, authors observed a plausible relation between self-reported activity, obtained from the YAAQ, and adiposity; decreases in television watching were associated with decreases in the prevalence of obesity in girls (12), and increases in physical activity were associated with decreases in BMI in girls (27). In the first of these studies, the authors expressed concern about the validity of the YAAQ in interpreting the effect of their intervention on the prevalence of obesity because of all the YAQ and YAAQ outcomes explored, only television was associated with obesity, and this relation existed only in girls (12). The second study had important methodologic differences from our study. First, the authors used a predominantly white, middle-class sample whereas our sample was racially diverse and low-income. Second, they mailed the questionnaires to participants, whereas we administered the questionnaires in person at schools. Third, they collected self-reported height and weight data, whereas we collected these data in schools with a standardized protocol.
Several theories suggest that the validity of self-report measures, not limited to the YAQ or YAAQ (eg, 24-h recalls, food records, food frequency questionnaires, activity questionnaires, and others), are compromised by under- and overreporting, which may be deliberate for reasons of social presentation/desirability (15, 28). This theory is consistent with our observation that obese participants self-reported lower energy intake than normal-weight participants. The effects of any over- and underreporting on the validity of the instruments are mitigated by holding constant each child in a longitudinal model. In other words, if overweight children consistently underreported and underweight children consistently overreported, we would still expect that over time, positive changes in energy intake, for example, would be associated with a positive change in BMI z score.
Other factors affecting the validity of self-report measures include cognitive barriers we observed in this sample of low-income, inner-city children who were administered the YAQ and YAAQ such as respondent confusion (29), level of mental arithmetic, portion size estimation (30), and fatigue. RAs reported that respondents were unclear as to whether they were to report ideal or actual frequencies of dietary intake and behavior. The children in this sample also expressed difficulty estimating actual frequencies. The variation in response frequencies across items may have exacerbated these issues of mental arithmetic. For example, missing data were most prevalent in the weekend section of the item addressing sedentary activity, which may be attributed to the fact that up until that point, participants were not used to breaking down activities by type of day.
Finally, the length of these instruments, particularly the YAQ, may induce fatigue on the part of the respondents. The administration of the YAQ took considerably longer (40–45 min) to complete than expected by the test developers (20–30 min) (17). The YAAQ also took longer (>10 min) than specified by the developers (18). As a result, RAs observed audible sighing, and students complained about the number of items on these scales.
Conclusion
In this large, racially diverse sample, self-reported energy intake, physical activity, and sedentary activity did not relate to BMI z score in the expected directions from either a cross-sectional or longitudinal perspective. Numerous studies, including this one, spanned many content areas, used multiple criterion measures on different samples of children, and ultimately reported results contradictory and in opposition to what the YAQ and YAAQ purport to measure at the individual level. The validity issues observed (eg, cognitive barriers and social presentation/desirability) are not specific to the YAQ and YAAQ but are characteristic of self-report instruments in general. Finally, these data further underscore the fundamental problems with self-reports of diet and physical activity. Given these well-documented issues, a dilemma is posed for the field. On the one hand, such assessments are appealing because they are inexpensive and provide quantifiable estimates of key variables such as energy, macronutrients, physical activity, and sedentary behavior. On the other hand, multiple studies, including this one, show that these measures are not valid when they are compared with objectively measured criterion variables such as BMI or doubly labeled water estimates of energy expenditure. Given such compelling data, the utility of collecting self-reported diet and physical activity data on low-income inner-city schoolchildren is unclear.
|
ACKNOWLEDGMENTS |
|---|
The authors' responsibilities were as follows—KEB and HM: analysis of data; GDF, SS, JN, and SVV: design of the experiment; GDF, AK, SS, KG, JN, and SVV: collection of data; KEB, GDF, HM, AK, and RFB: writing of the manuscript; and GDF, HM, AK, SS, KG, SVV, and RFB: provision of significant advice or consultation. The authors had no undisclosed financial or personal interest in any company or organization connected in any way with the research presented in the article.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
