The Reliability of a Survey Question on Television Viewing and Associations With Health Risk Factors in US Adults

nature publishing group articles The Reliability of a Survey Question on Television Viewing and Associations With Health Risk Factors in US Adults Kelley K. Pettee 1, Sandra A. Ham 2, Caroline A. Macera 3 and Barbara E. Ainsworth 4 Research into the accuracy of self-reported measures used to quantify physical inactivity has been limited. The purposes of the current report were to examine the reliability of a survey question assessing time spent watching television and to describe associations between television watching and physical activity and health risk factors. Data from this cross-sectional investigation were obtained from a study designed to evaluate a physical activity module for potential use in the 21 Behavioral Risk Factor Surveillance System. Participants were 93 men and women (aged 45.9 (15.4) years) who answered the question pertaining to television watching during an initial visit and three follow-up visits to the study center. Intra-class correlation coefficients (ICCs) between administrations of the survey question were used to assess test retest reliability. Spearman rank order correlation coefficients were used to examine the associations of television viewing with physical activity and health risk factors. The test retest reliability of the television-watching question suggested moderate agreement (ICCs of.42 and.55 over a 3-week and 1-week period, respectively). After adjustment for age and sex, reported television-watching hours were positively associated with BMI (P =.2), percentage fat (P =.1), and light-intensity physical activity (P =.6) and negatively associated with cardiorespiratory fitness (P =.4) and moderate-intensity and hard-intensity physical activity (P =.3 and P =.3, respectively). Increased time spent in sedentary behaviors has been identified as a major modifiable risk factor in the development of chronic diseases and conditions. The single-item survey question evaluated in this study was shown to be a reliable measure of television watching and was associated with physical activity and health risk factor outcomes. Obesity (28) 17, 487 493. doi:1.138/oby.28.554 Introduction The majority of current public health campaigns focus on strategies aimed at increasing physical activity levels to prevent obesity and chronic disease (1), but interest is growing in interventions that specifically target reductions in sedentary behaviors (2,3). Physical inactivity is a state in which body movement is minimal; the term is typically used to describe time spent in sedentary behaviors (2,3). Although the US Surgeon General has determined that physical inactivity is a major health risk factor for various chronic diseases (4), the accurate measurement of inactivity has received far less attention than has the measurement of physical activity (2,4,5). When attempting to measure levels of physical inactivity, researchers tend to focus on time spent in modifiable sedentary behaviors. In both children and adults, the average amount of time spent watching television or videos, playing computer games and videogames, and sometimes even reading or napping (6 16) is used to quantify overall inactivity. Data from the Nielsen Reports suggest that television watching remains the major contributor to sedentary behavior in the United States (11,17). During the 24 25 television season, the television was on in the typical household for >8 h a day (17). Television watching is a particular problem because it has a lower associated metabolic cost than other types of sedentary behaviors such as reading, writing, and driving a car (18,19). Sedentary behaviors, particularly television watching, are significantly associated with obesity (6,11,2), risk factors for cardiovascular disease (9,14), and type 2 diabetes (11). However, there is limited evidence in the existing literature regarding the accuracy of questions used to measure sedentary behaviors (8). Therefore, the purposes of the current report were to determine the test retest reliability of a question regarding time spent watching television and to examine the associations between television viewing and physical activity and health risk factor outcomes. 1 Department of Health Promotion, Social & Behavioral Health, University of Nebraska Medical Center, Omaha, Nebraska, USA; 2 Centers for Disease Control and Prevention, Division of Nutrition, Physical Activity, and Obesity, Atlanta, Georgia, USA; 3 Division of Epidemiology and Biostatistics, San Diego State University, San Diego, California, USA; 4 Department of Exercise and Wellness, Arizona State University, Mesa, Arizona, USA. Correspondence: Kelley K. Pettee (kpettee@unmc.edu) Received 12 March 28; accepted 23 October 28; published online 18 December 28. doi:1.138/oby.28.554 obesity VOLUME 17 NUMBER 3 MARCH 29 487

Methods Study population Data were obtained in 2 from a broader cross-sectional study designed to determine the validity and reliability of various physical activity questionnaires (21,22). Participants were recruited from the Columbia, South Carolina, metropolitan area, with most participants obtained from the University of South Carolina, through churches and community centers, or from newsletter advertisements. Participants had to be aged 18 years or older and to be able and willing to read and complete physical activity questionnaires, walk on a treadmill at 3. miles per hour, wear accelerometers and pedometers for 2 weeks, and have their body fat measured using bio-impedance analysis. The participants in the study had a broad range of physical activity habits. In all, 132 adult men and women were screened, and 13 (78.%) were enrolled in the study. Among those who were not enrolled (n = 29), reasons included lack of time (n = 3) and not being interested in completing the study activities required (n = 26). Of the 13 initial enrollees, 1 were excluded from the present analyses because 7 failed to complete the survey question on television watching at both the first and second visits and 3 did not complete this question at the second visit only. Research design Participants completed four clinic visits within the 3-week study period (each visit lasted between 3 min and 1 h). At visit 1, after providing informed consent, participants completed demographic, health history, and physical activity questionnaires, and they had their height and weight measured. In addition, at all four visits, participants were asked to respond to the television-watching question (how many hours they spent watching television per week). At the end of the first visit, participants were instructed in the use of a 1-week physical activity log (5) and given two activity monitors: a CSA accelerometer (model 7164; CSA, Shalimar, FL (now ActiGraph, Pensacola, FL)) and a Yamax Digiwalker pedometer (SW-2; New Lifestyles, Lee s Summit, MO). They were asked to wear the motion detectors for the next 7 days. Eight days later (visit 2), participants returned to the clinic with their completed physical activity logs and CSA accelerometer for data retrieval. They also responded to the physical activity questionnaire, including the television-watching question, a second time and were given a new physical activity log and asked to wear the CSA accelerometer for the next 7 days. One week later (visit 3), participants returned to the clinic to respond to the physical activity questionnaire a third time and were given another physical activity log and the CSA accelerometer to wear for the next 7 days. Eight days later (visit 4), participants returned to the clinic with their completed physical activity logs and their activity monitors. At this visit, participants also completed a submaximal treadmill graded exercise test, had their body composition, including percentage body fat, measured via bio-impedance analysis, and completed physical activity questionnaires for the fourth time. Study variables Data used in this report include demographic factors, the one-item question about watching television, and behavioral and health risk factor outcomes. Behavioral outcomes included objective measures of physical activity (pedometer steps and minutes of physical activity recorded on an accelerometer), and health risk factor outcomes consisted of measures of body composition (BMI and percentage body fat) and cardiorespiratory fitness (estimated VO 2 max and submaximal heart rate). Physical fitness, which includes cardiorespiratory fitness, can be defined as a set of attributes relating to the ability to be physically active that individuals either have or can achieve (23). Cardiorespiratory fitness measures have been widely used in research studies as a surrogate measure of higher-intensity physical activity. Estimated VO 2 max and submaximal heart rate measures were included in the current report to examine whether their relationship with television watching followed similar trends to physical activity. Demographic data. Demographic data were collected on all participants using a questionnaire at visit 1. The demographic factors used in the current report include age in years, sex, race/ethnicity (white, African American, or other), and education (less than a high school diploma, high school diploma, or any college education). Television viewing. The question on television watching was phrased thus: In a usual week, how many hours do you spend watching television while sitting or lying down? This item was completed at all four visits. Measures of physical activity. Physical activity was measured using the Yamax Digiwalker pedometer and the CSA accelerometer. The participants wore the pedometer and accelerometer clipped to their waistband for the entire 3 weeks of the study (pedometer on the right and accelerometer on the left) during all waking hours of the day, excluding periods of bathing or other activities involving water. Participants were asked to record in a 7-day physical activity log provided by study staff the time at which they put on the monitors in the morning, the time they took off the monitors at night, and the total number of accumulated steps (pedometer only). Participants were instructed to reset the pedometer to zero every morning. At the end of each week, the participant returned the physical activity log to study staff and was given another physical activity log to complete during the following week. The number of pedometer steps recorded in the log each week was averaged to obtain a 7-day daily average. The steps were weighted, however, to reflect differences in steps taken according to wear time (a weighted average was produced for a 24-h period) (24). The acceleration signal was filtered using an analog bandpass filter and digitized using an 8-bit analog-to-digital converter at a sampling rate of 1 samples per second, with data stored in user-defined intervals of 1 min. Calibration of the accelerometer was verified each week using CSA s calibration standard. To evaluate the average time (in minutes) spent at various activity levels, the Freedson (25) cut points were used. The Freedson cut points, which were developed in a controlled laboratorybased setting, are commonly used to summarize continuous ambulatory movement from the CSA. Light-intensity activity was defined as 1,951 counts per minute, moderate-intensity activity as 1,952 5,724 counts per minute, and hard-intensity activity as 5,725 counts per minute. Complete accelerometer data were available for 83 participants (89%). Body composition. BMI was calculated from the heights and weights measured at visit 1 by dividing the weight in kilograms by the square of the height in meters (kg/m 2 ). Percentage body fat was obtained at visit 4 using a bioelectrical impedance analyzer (RJL Systems, Clinton Township, MI; Model 11). For quality control purposes, participants were asked a series of questions that assessed their current state of hydration and exercise levels before test administration. All measures were taken on the right side of the body with the participant lying supine on a nonconductive surface. Four electrodes were placed on cleaned surfaces of the body: two proximal electrodes on the dorsal surface of the wrist and ankle and two distal electrodes on the base of the second or third metacarpophalangeal joints of the hand and foot. The distance between proximal and distal placements was at least 5 cm. Multiple assessments were performed for each participant, and measurements were repeated until two trials were within 3 Ω for both resistance and reactance. Percentage body fat was estimated using sex-specific and fatness-specific prediction equations established by Segal et al. (26). On the basis of the distribution, tertiles were established for BMI ( 3., 25. 29.9, and <25. kg/m 2 ) and sex-specific percentage body fat (men: 25.1, 21. 25.1, and <21.%; women: 35., 25.9 35., and <25.9%) to further explore the relationship between body composition and time spent watching television. Cardiorespiratory fitness. Before the treadmill test, electrodes were placed on the participant s chest, and electrocardiograph tracings (12- and 3-lead) and blood pressure were obtained with the 488 VOLUME 17 NUMBER 3 MARCH 29 www.obesityjournal.org

participant in the supine, sitting, and standing positions. After the resting measures were obtained, the test procedures were explained to the participant, and the participant was briefed on the ratingsof-perceived-exertion scale, mounting/dismounting the treadmill, test endpoints, signals for test termination, and expected discomforts. The participants were asked to begin walking on the treadmill at 3 miles per hour and % grade; testing began when gait normalized. The test protocol had five stages, each lasting 2 min and increasing in effort by ~1 MET (metabolic equivalent) per stage (27). During the final seconds of each stage, ECG tracings, blood pressure, and ratings of perceived exertion were obtained, and if the participant was able to continue, the grade of the treadmill was increased by 2.5%. The test was terminated after the completion of all five stages (final stage: 3 miles per hour at 1% grade) or if there were any indications of an abnormal response to the exercise. Indications for termination were based on the American College of Sports Medicine guidelines for exercise testing (28). After completion of the test, the speed and grade were reduced to 1.5 miles per hour and %, respectively, and participants were encouraged to walk for 3 min for recovery. Participants were then asked to sit quietly for the next 3 min. ECG tracings and BP measures were obtained at the end of each 3-min recovery period. VO 2 max (maximal oxygen uptake) was estimated using prediction equations developed by Ainsworth (B.E. Ainsworth, unpublished data) using the Study of Activity, Fitness, and Exercise (SAFE) population (13 men and women aged 2 59 years). The prediction equation was based on the participant s sex, age, BMI, and stage-5 heart rate: VO 2 max = 12.3 (.25 age) (.28 heart rate at end of stage 5) (.47 BMI) (3.67 sex: male = 1 and female = 2) (R 2 =.7314). To explore the relationship between physical fitness and time spent watching television further, sex-specific estimated VO 2 max (men: 43.9, 36. to <43.9, and <36. ml/kg/min; women: 34.3, 28.8 to <34.3, and <28.8 ml/kg/min) and submaximal heart rate categories (men: <115, 115 to <127, and 127 bpm; women: <137, 137 to <15, and 15 bpm) were classified as low, middle, or high fit on the basis of tertiles of the distribution. Statistical methods. Descriptive statistics included demographic data, television-viewing habits, and behavioral and health risk factor outcomes both in the entire study population and stratified by sex. All variables were assessed for normality. Normally distributed variables were reported as mean ± s.d., non normally distributed variables were reported as medians with twenty-fifth and seventy-fifth percentiles, and proportions were noted for categorical variables. Depending upon the characteristics of the variable, t-tests, Wilcoxon rank sum, or χ 2 -tests were used to compare men and women. Table 1 Characteristics of the study participants (n = 93) Demographic data a All study participants (n = 93) Men (n = 48) Women (n = 45) Age (years) 45.9 ± 15.4 45. ± 15.7 46.8 ± 15.3 Race/ethnicity White (%) 74.2 77.1 71.1 African American (%) 22.6 16.7 28.9 Any college education (%) 85. 93.8 75.6* Television viewing (hours per week) Survey 1 b 8. (4., 14.) 1. (3.5, 15.) 8. (4., 14.) Survey 2 c 7. (3., 14.) 7. (3., 14.) 7. (3., 14.) Survey 3 d (n = 9) 7. (3.5, 13.) 7. (3., 14.) 6. (3.5, 8.) Survey 4 e (n = 89) 7. (3., 14.) 7. (2., 14.) 6. (3., 14.) Physical activity e Digiwalker pedometer, steps per day (n = 87) 8,58.8 ± 3,183.6 7,877.5 ± 3,138.8 8,262.3 ± 3,259.8 CSA accelerometer (n = 83) Light activity, min/day ( 1,951 counts) 1,418. (1,41.3, 1,429.1) 1,412.7 (1,393.4, 1,429.6) 1,421.9 (1,47.4, 1,428.6) Moderate activity, min/day (1,952 5,724 counts) 2. (1., 38.6) 21.1 (1.4, 42.7) 17.7 (9.4, 29.6) Hard activity, min/day ( 5,725 counts).3 (, 2.1).3 (, 2.3).4 (, 2.) Body composition BMI, a kg/m 2 (n = 92) 26.2 ± 4.4 26.7 ± 3.9 25.7 ± 4.8 Percentage body fat d (n = 9) 27. ± 8.1 22.5 ± 5.2 31.7 ± 8. Cardiorespiratory fitness Heart rate at final stage of submaximal treadmill 133.4 ± 19.7 123.9 ± 18.9 143.9 ± 14.7 test d (bpm) (n = 89) Predicted VO 2 max from submaximal treadmill test d (ml/kg/min) (n = 88) 35.9 ± 8.1 4.2 ± 7.4 31.2 ± 6. Descriptive statistics are presented on a study sample of 93 participants unless otherwise noted. Normally distributed variables are reported as mean ± s.d., non normally distributed variables are presented as medians (twenty-fifth and seventy-fifth percentiles), and proportions are noted for categorical variables. a Collected at baseline. b Collected at visit 2, day 8. c Collected at visit 3, day 15. d Collected at visit 4, day 23. e Collected from visit 1 through visit 4 (i.e., 23 days). Study cohort characteristics by sex: *P <.5; **P <.1; ***P <.1; P <.1. obesity VOLUME 17 NUMBER 3 MARCH 29 489

The 1-week and 3-week test retest reliability of the televisionwatching question was assessed using the intra-class correlation coefficient (ICC) by employing a two-way fixed-effects model. The reported television-watching hours from surveys 1 and 2 were used to assess 1-week test retest reliability, and reported televisionwatching hours from surveys 1 and 4 were used to assess 3-week test retest reliability. The strength of agreement for the ICC ranges can be interpreted as follows: <., poor;..2, slight;.21.4, fair;.41.6, moderate;.61.8, substantial; and.81 1., almost perfect (29). Spearman rank order correlation coefficients were used to determine the association between television viewing and physical activity, body composition, and physical fitness using the first administration of the survey. For graphic representation, we evaluated the median hours spent watching television across tertiles of body composition (BMI and percentage fat) and cardiorespiratory fitness (submaximal heart rate and estimated VO 2 ). Jonckheere Terpstra tests were used to estimate the linear trend in hours spent watching television across body composition and cardiorespiratory fitness categories. Partial correlations were used to explore whether the results warranted further adjustment by age and sex. SAS 9.1 (SAS, Cary, NC) was used for the analysis and correlation plots, and ICCs were generated using SPSS 14. (SPSS, Chicago, IL). Approval to conduct the study was obtained from the University of South Carolina s institutional review board. All 93 participants read and signed an informed consent form before data collection. Results The characteristics of the study population are presented in Table 1. The mean age of the participants was 45.9 ± 15.4 years (22.6% were aged 6 years or older), 48.4% were female, and 22.6% were African American. The mean BMI was 26.2 ± 4.4 kg/m 2 (34.4% were overweight (BMI: 25 29.9 kg/m 2 ) and 2.4% were obese (BMI: >3 kg/m 2 )), and 85.% of the study cohort reported having at least some college education. Also presented in Table 1 are the median hours of television watching per week as reported at the four visits. Median weekly hours for the entire cohort ranged from 7. (reported at visits 2, 3, and 4) to 8. (visit 1), which suggests the relative stability of this measure over time. Compared with men, women had a significantly higher percentage body fat and submaximal heart rate (both P <.1) and a lower estimated VO 2 max (P <.1), and they were less likely to have any college education (P <.5). No additional significant differences were found by sex (Table 1). The test retest ICC with 95% confidence limits for the television-watching question is shown in Figures 1 and 2. The 1-week test retest reliability of the question was.55 (95% CI:.39,.68; P <.1), suggesting moderate reproducibility (Figure 1). When the time interval between clinic visits was increased from 1 to 3 weeks, the test retest reliability was slightly attenuated, but it still suggested moderate agreement between the two survey administrations (ICC =.42; 95% CI:.23,.58; P <.1) (Figure 2). Hours spent watching television were not significantly associated with pedometer steps, but television watching was related to time (minutes per day) spent in various levels of activity as measured by the CSA accelerometer (Table 2). Television hours were positively related to light-intensity activity (ρ =.3) and inversely related to time spent in moderate-intensity and hard-intensity activity (ρ =.25 and.33, respectively). Higher television hours were positively associated with higher BMI and percentage body fat (ρ =.39 and.31, respectively). The amount of time spent watching television was inversely associated with predicted VO 2 max from a submaximal graded exercise test (ρ =.24), but it was not significantly related to heart rate. Televisionwatching time significantly increased across increasing BMI and percentage body fat categories (Figure 3), and time spent watching television was highest in the least-fit categories and lowest among those with higher cardiorespiratory fitness levels (Figure 4). After adjustment for age and sex, the relationship between hours spent watching television and body composition, cardiorespiratory fitness, and objective measures of physical activity did not differ. On the other hand, the association between television hours and percentage body fat and predicted VO 2 max appeared to strengthen (ρ =.4 and.31, respectively) after adjusting for age and sex (Table 2). TV hours per week from survey 2 4 ICC =.55 (.39,.68) P <.1 3 2 1 1 2 3 4 TV hours per week from survey 1 Figure 1 One-week test retest reliability of television-watching question (n = 93). ICC, intra-class correlation coefficient. TV hours per week from survey 4 4 3 2 1 ICC =.42 (.23,.58) P <.1 1 2 3 4 TV hours per week from survey 1 Figure 2 Three-week test retest reliability of television-watching question (n = 89). ICC, intra-class correlation coefficient. 49 VOLUME 17 NUMBER 3 MARCH 29 www.obesityjournal.org

Table 2 Spearman rank order correlations between television watching (h/week) and body composition, cardiorespiratory fitness, and objective measures of physical activity Physical activity b n Unadjusted ρ P value Adjusted a ρ P value Digiwalker pedometer, steps per day 89.6.57.5.63 CSA accelerometer 83 Light activity, min/day ( 1,951 counts).3.5.3.6 Moderate activity, min/day (1,952 5,724 counts).25.2.25.3 Hard activity, min/day ( 5,725 counts).33.3.32.3 Body composition BMI, c kg/m 2 92.39.1.38.2 Percentage body fat d 92.31.3.4.1 Cardiorespiratory fitness Heart rate at final stage of submaximal treadmill test d (bpm) (n = 89) Predicted VO 2 max from submaximal treadmill test d (ml/kg/min) (n = 88) 89.13.24.2.6 88.24.2.31.4 a Adjusted for age and sex. b Collected from visit 1 through visit 4 (i.e., 23 days). c Collected at baseline. d Collected at visit 4, day 23. TV watching, hours/week 2 18 16 14 12 1 8 6 4 2 <25. (n = 42) P =.1 P =.2 25 29.9 (n = 32) BMI, kg/m 2 >3. (n = 19) Low % Fat (n = 29) Middle % Fat (n = 29) Percent (%) Fat High % Fat (n = 32) 75th Pctl 25th Pctl 5th Pctl Figure 3 Reported number of hours per week spent watching television (collected at baseline) by BMI (kg/m 2 ) (collected at baseline) and percentage fat (collected at visit 4, day 23). Pctl, percentile. TV watching, hours/week 2 18 16 14 12 1 8 6 4 2 Low fit (n = 32) P =.29 P =.3 Middle fit (n = 27) Heart rate, bpm High fit (n = 29) Low fit (n = 27) Middle fit (n = 28) High fit (n = 32) Estimated VO 2 ml/kg/min 75th Pctl 25th Pctl 5th Pctl Figure 4 Reported number of hours per week spent watching television (collected at baseline) by heart rate at final stage of submaximal treadmill test (bpm) (collected at baseline) and predicted VO 2 max from submaximal treadmill test (ml/kg/min) (collected at visit 4, day 23). Discussion In the current investigation, the test retest reliability of a one-item question designed to measure the amount of time spent watching television was evaluated, and associations of television viewing with physical activity, body composition, and physical fitness were examined. Findings from the report suggest that the television question is moderately reproducible and relatively stable over time. In addition, television watching was shown to be associated with physical activity and health risk factor outcomes. To our knowledge, no previous study has employed objective measures of physical activity, including an accelerometer, to evaluate these relationships with television viewing in adults. The associations between television watching and behavioral and health risk factor outcomes were further confirmed when the pattern of television hours was examined across categories of the specific health risk factor outcomes. More specifically, television-viewing levels followed a similar graded increase across BMI and percentage body fat categories. It should be noted that the inverse trend that was observed between television watching and estimated VO 2 max appeared to be stronger than the trend for submaximal heart rate, but this may be due to the contribution of BMI in the equation for predicting submaximal VO 2, which had a stronger association with television-watching time. These findings suggest that this question is appropriate for research studies examining television viewing. Only one previous report has examined the reproducibility of a question used to measure television viewing. A 25 report by Evenson et al. (16) examined the test retest reliability of a question used in the National Health and Nutrition Examination Survey that inquired about the time spent watching television and videos in a usual week. ICCs were generated to estimate test retest reliability, and findings suggested that the television watching question from the survey had fair reliability (ICC =.32) over a median of 16 days (range 9 3 days). The ICCs computed in the current study (.55 in the 1-week test retest and.42 in the 3-week test retest) were higher than obesity VOLUME 17 NUMBER 3 MARCH 29 491

those obtained in the report by Evenson et al. (16). One possible explanation for the discrepancy in observed correlations between studies is the wording of the question. Unlike the Behavioral Risk Factor Surveillance System, which utilizes a one-item question, the question from the National Health and Nutrition Examination Survey asked about time spent watching television and videos. The additional recall of a second activity (i.e., watching videos) could have led to a greater variability in responses given at each administration of the question and may partially explain the difference in observed test retest correlations between the current study and that of Evenson et al. (16). In addition, the time between administrations of the survey question was more structured in the current study (every 7 days vs. a range of 9 3 days), which may also have produced less variation in participant responses. In past research, physical inactivity was simply thought to be the inverse of physical activity (2). However, there is an alternative approach that says physical inactivity is distinct from physical activity and that one cannot be used as a proxy measure for the absence of the other. For example, it may be possible for an individual to meet current physical activity recommendations but still be relatively inactive for the majority of the day. In recent studies, correlations between physical activity and hours of reported television watching ranged from ρ =.3 to ρ =.11 (6,9,11). Although the findings from these studies suggest a null association, the observed correlations were not strong enough to suggest that these behaviors are simply the opposite of each other. In the current report, the comparison between television watching and objective estimates of physical activity yielded stronger associations than in previous reports (6,9,11). Yet the stronger associations observed in the current report are not surprising, given that past research in this area compared television watching with self-reported estimates of physical activity, which may be less accurate and subject to recall bias. In comparison with previous investigations, reported television viewing in the current report was low (7 8 h per week) and suggests social desirability bias. The desire of individuals to present themselves in a way that will be viewed favorably by others is evident when we compare subjective television viewing estimates with data from the Nielsen Reports, which utilize an objective measure of television viewing. According to the 24 25 report, television was on in the typical household for most of the day (17). However, although the television was turned on, it may not have been viewed by the same individual within the household for the entire time. The estimate obtained in the current report was also low compared with estimates in previous studies that have utilized subjective measures to ascertain television viewing. In a previous report by Fitzgerald et al. (6), participants were asked to recall the number of hours spent watching television per day, not over the previous week, as in the current report. It may be easier for individuals to recall an activity performed over the previous day, and this may partially explain differences in reported television watching. Furthermore, in a report by Hu et al. (11), the highest proportion of study participants reported watching 6 2 h of television per week, which represents a large response range. It is possible that the majority of participants in this category reported watching television at the lower end of the range (i.e., 6 h per week) but were grouped with those reporting two to three times more television watching. Unfortunately, no information regarding the distribution of participants within this broad category was reported by the investigators, making it difficult to compare television-viewing levels with those in the current report. Several limitations need to be considered in interpreting our findings. First, the television-watching estimate was obtained through a method (self-reporting) that can be influenced by a variety of factors, including the inability to recall accurately the number of hours spent watching television. In general, it has been suggested that lower-intensity activities are more difficult to recall than higher-intensity activities (3,31). However, compared with other lower-intensity activities, time spent watching television may be more easily quantifiable, because programming is more structured and is commonly scheduled every 3 or 6 min. A second factor is that there may be differences in recall according to how the survey question is asked. In the current report, participants were asked to recall the total number of television hours watched in a usual or typical week. Past research has suggested that for many people, responding to a question that is framed in terms of a usual or typical week may require longer-term recall than does answering a question about the past week, especially if the pattern of the particular behavior is irregular (32). For example, if a person s time spent watching television varies from week to week, the first step would be to identify a week that he or she considers typical. The longer the interval between the week identified as most reflective of the usual number of hours spent watching television and the time of recollection, the less accurate the recall. However, the test retest reliability findings in our study suggest that the additional level of recall necessary to quantify time spent watching television in terms of a usual week did not affect the overall utility of this question in research settings. Another concern is that the study population for the current report consisted primarily of highly educated, higher socioeconomic status white men and women, which may limit the generalizability of our findings to more diverse populations. Of the individuals who were screened, 29 did not enroll in the study, which may have introduced a selection bias. Furthermore, 1 participants were excluded from the present analyses for not responding to the television-watching question. It is important to note that women were more likely than men not to complete the television-watching question (P <.5). No other baseline characteristics differed between responders and non- responders to the television-watching question (data not shown). One other important limitation is the concept of media multitasking, which can be defined as participating in more than one medium at a time (e.g., using the Internet while watching television). When asked to quantify the amount of time spent watching television in a usual week, individuals may have more difficulty recalling this information if they were engaging in another medium at the same time (33). 492 VOLUME 17 NUMBER 3 MARCH 29 www.obesityjournal.org

Unfortunately, we were unable to determine whether multitasking was present in this study sample. Furthermore, our data were collected before media multitasking was as prevalent as it is today. Finally, the validity of the television- watching question was not evaluated in the current report because of the lack of an available way to measure television viewing, such as a television Allowance Device (Miami, Florida). Accurate quantification is important to researchers interested in understanding the time spent in sedentary behaviors, particularly television watching, and developing successful interventions to reduce that time. A one-item question that quantifies time spent watching television in a usual week would be easy to add to epidemiological surveys as a proxy measure of time spent in sedentary pursuits. The use of a question that measures time spent watching television in research studies may be appropriate for estimating time spent in sedentary behaviors in relation to chronic disease and the prevention or management of disability. Acknowledgments This research was supported by the Centers for Disease Control and Prevention grant number U48/CCU49664, funded to the Prevention Research Center, Norman J. Arnold School of Public Health, University of South Carolina. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the Centers for Disease Control and Prevention. Disclosure The authors declared no conflict of interest. 28 The Obesity Society References 1. Haskell WL, Lee IM, Pate RR et al. Physical activity and public health. Updated recommendation for adults from the American College of Sports Medicine and the American Heart Association. Circulation 27;116: 181 193. 2. Dietz WH. The role of lifestyle in health: the epidemiology and consequences of inactivity. Proc Nutr Soc 1996;55:829 84. 3. Epstein LH, Roemmich JN. Reducing sedentary behavior: role in modifying physical activity. Exerc Sport Sci Rev 21;29:13 18. 4. U.S. Department of Health and Human Services. Physical Activity and Health: A Report of the Surgeon General. US Government Printing Office: Atlanta, GA, 1996. 5. LaMonte MJ, Ainsworth BE. Field Assessment of Physical Activity and Energy Expenditure among Athletes. CRC: Boca Raton, FL, 22. 6. Fitzgerald SJ, Kriska AM, Pereira MA, de Courten MP. Associations among physical activity, television watching, and obesity in adult Pima Indians. Med Sci Sports Exerc 1997;29:91 915. 7. Gordon-Larsen P, Adair LS, Popkin BM. Ethnic differences in physical activity and inactivity patterns and overweight status. Obes Res 22;1:141 149. 8. Gordon-Larsen P, McMurray RG, Popkin BM. Determinants of adolescent physical activity and inactivity patterns. Pediatrics 2;15:E83. 9. Fung TT, Hu FB, Yu J et al. Leisure-time physical activity, television watching, and plasma biomarkers of obesity and cardiovascular disease risk. Am J Epidemiol 2;152:1171 1178. 1. Andersen RE, Crespo CJ, Bartlett SJ, Cheskin LJ, Pratt M. Relationship of physical activity and television watching with body weight and level of fatness among children: results from the Third National Health and Nutrition Examination Survey. JAMA 1998;279:938 942. 11. Hu FB, Li TY, Colditz GA, Willett WC, Manson JE. Television watching and other sedentary behaviors in relation to risk of obesity and type 2 diabetes mellitus in women. JAMA 23;289:1785 1791. 12. Manson JE, Greenland P, LaCroix AZ et al. Walking compared with vigorous exercise for the prevention of cardiovascular events in women. N Engl J Med 22;347:716 725. 13. Feldman DE, Barnett T, Shrier I, Rossignol M, Abenhaim L. Is physical activity differentially associated with different types of sedentary pursuits? Arch Pediatr Adolesc Med 23;157:797 82. 14. Jakes RW, Day NE, Khaw KT et al. Television viewing and low participation in vigorous recreation are independently associated with obesity and markers of cardiovascular disease risk: EPIC-Norfolk population-based study. Eur J Clin Nutr 23;57:189 196. 15. Utter J, Neumark-Sztainer D, Jeffery R, Story M. Couch potatoes or french fries: are sedentary behaviors associated with body mass index, physical activity, and dietary behaviors among adolescents? J Am Diet Assoc 23;13:1298 135. 16. Evenson KR, McGinn AP. Test-retest reliability of adult surveillance measures for physical activity and inactivity. Am J Prev Med 25;28:47 478. 17. Nielsen Reports Americans Watch TV at Record Levels. <http://www.neilsenmedia.com>. Accessed 14 September 26. 18. Ainsworth BE, Haskell WL, Leon AS et al. Compendium of physical activities: classification of energy costs of human physical activities. Med Sci Sports Exerc 1993;25:71 8. 19. Ainsworth BE, Haskell WL, Whitt MC et al. Compendium of physical activities: an update of activity codes and MET intensities. Med Sci Sports Exerc 2;32(9 Suppl):S498 S54. 2. Ching PL, Willett WC, Rimm EB et al. Activity level and risk of overweight in male health professionals. Am J Public Health 1996;86:25 3. 21. Ainsworth BE, Bassett DR Jr, Strath SJ et al. Comparison of three methods for measuring the time spent in physical activity. Med Sci Sports Exerc 2;32(9 Suppl):S457 S464. 22. Yore MM, Ham SA, Ainsworth BE et al. Reliability and validity of the instrument used in the BRFSS to assess physical activity. Med Sci Sports Exerc 27;39:1267 1274. 23. Caspersen CJ, Powell KE, Christenson GM. Physical activity, exercise, and physical fitness: definitions and distinctions for health-related research. Public Health Rep 1985;1:126 131. 24. Schmidt MD, Blizzard CL, Venn AJ, Cochrane JA, Dwyer T. Practical considerations when using pedometers to assess physical activity in population studies: lessons from the Burnie Take Heart Study. Res Q Exerc Sport 27;78:162 17. 25. Freedson PS, Melanson E, Sirard J. Calibration of the Computer Science and Applications, Inc. accelerometer. Med Sci Sports Exerc 1998;3: 777 781. 26. Segal KR, Van Loan M, Fitzgerald PI, Hodgdon JA, Van Itallie TB. Lean body mass estimation by bioelectrical impedance analysis: a four-site crossvalidation study. Am J Clin Nutr 1988;47:7 14. 27. Jacobs DR Jr, Ainsworth BE, Hartman TJ, Leon AS. A simultaneous evaluation of 1 commonly used physical activity questionnaires. Med Sci Sports Exerc 1993;25:81 91. 28. Armstrong L, Balady GJ, Berry MJ et al. ACSM s Guidelines for Exercise Testing and Prescription, 7th edn. Lippincott Williams & Wilkins: Philadelphia, PA, 26. 29. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33:159 174. 3. Kriska AM, Caspersen CJ. Introduction to a collection of physical activity questionnaires. Med Sci Sports Exerc 1997;29:S5 S9. 31. Taylor CB, Coffey T, Berra K et al. Seven-day activity and self-report compared to a direct measure of physical activity. Am J Epidemiol 1984;12:818 824. 32. Chang L. Measuring the frequency of regular behaviors: comparing the typical week to the past week. Sociol Methodol 23;33:55 77. 33. Foehr UG. Media multitasking among American youth: prevalence, predictors and pairings <http://www.kff.org/entmedia/entmedia12126pkg. cfm>. Accessed 23 August 27. obesity VOLUME 17 NUMBER 3 MARCH 29 493