Ever since the 1985 article published by Stephens et al. (19 23 World Health Report describing WHO's effort to identify the impact of selected behaviors on morbidity and mortality, the difficulty in studying physical activity was the result of "a number of direct and indirect data sources and a range of survey instruments and methodologies used to estimate activity levels in the domains of work, transportation, domestic duties, and leisure time" ((23
In 1997, a group of public health and physical activity researchers from 16 countries met in Geneva, Switzerland, to identify a common method to assess physical activity for the purpose of population surveillance. The outcome of the meeting was the development of the International Physical Activity Questionnaire (IPAQ). The instrument has since been examined for validity and reliability in an international study and deemed acceptable for use in physical activity research and surveillance activities (7
The IPAQ has been used in several countries (3,9,18 14-16 1,2,8,17 11,12,22 public health surveillance is to track health behaviors over time, the IPAQ has not been introduced for physical activity surveillance activities in the United States. To examine the feasibility of using the IPAQ for international physical activity surveillance, the IPAQ investigators initiated the IPAQ International Prevalence Study in 2002 to identify the population prevalence of physical activity in various countries using the IPAQ. As a participant in the international study, the United States conducted the National Physical Activity and Weight Loss Survey (NPAWLS). NPAWLS included the BRFSS physical activity module (referred to here as the BRFSS) and the IPAQ; the intention was to compare the physical activity prevalence estimates obtained from the two questionnaires. This paper presents the findings of the comparison of the BRFSS and the IPAQ from adults enrolled in NPAWLS.
METHODS
NPAWLS was a nationwide telephone survey conducted between September and December 2002. Survey questions included items pertaining to health status, quality of life, weight control measures, and participation in physical activity. The overall objective of the study was to obtain data on individual physical activity and nutrition risk behaviors. The BRFSS and the IPAQ were positioned in the larger survey as the second and fourth of 10 modules and separated by a weight-control module. To avoid an order effect in answering the physical activity questionnaires, two versions of NPAWLS were administered: version A and version B. In version A the IPAQ preceded the BRFSS. In version B the order was reversed. The participants were randomly assigned to each version such that an equal number of responses was obtained for each version (version A = 5654; version B = 5557). The sample of respondents for NPAWLS was drawn from the total noninstitutionalized U.S. adult population residing in telephone-equipped locations. Excluded from the referent population were institutionalized adults, those living in group quarters with 10 or more unrelated residents, adults without a telephone, and adults who did not speak English or Spanish well enough to be interviewed.
Sample design.
The design used a list-assisted, random-digit-dialed (RDD) sample of telephone-equipped households in the United States generated by preparing a list of all current operating telephone exchanges within each U.S. area code. These telephone exchanges, combined with all four-digit numbers from 0000 to 9999, were divided into blocks of 100 numbers; each block was assessed for containing at least one residential number. Qualifying blocks were combined to create the sampling frame, and numbers were systematically sampled from that frame. Nonresidential, nonworking, fax, and modem numbers were filtered to increase the connect rate of telephone interviews with viable respondents.
The study was designed to obtain adequate representation of Hispanic and non-Hispanic black respondents. A replicate design that mixed telephone numbers from three independent samples was used to achieve desired racial/ethnic percentages in the final sample. The first independent sample consisted of a national RDD draw. The second and third samples were drawn from frames constructed to yield higher percentages of Hispanic and non-Hispanic black respondents. The two samples, one targeting Hispanic households and the other targeting non-Hispanic black households, were constructed by creating subsets of telephone exchanges in the national frame. Telephone exchanges were included in each respective subset if ethnic household incidence was greater than 20%. Ethnic household incidence was determined by area code exchange and U.S. Census data. The University of South Carolina human subjects institutional review board approved the NPAWLS study.
Respondent selection within household.
A minimum of 15 attempts across at least 5 d was made to reach each sampled number. Interviewers first asked about the number of adults aged 18 and older in the contacted household, then asked all adult males and females in the household to be listed; the survey respondent was randomly selected from this roster. A total of 11,211 survey interviews were completed; the Council of American Survey Research Organizations response rate was 30.9%. The cooperation rate, defined as the proportion of all respondents interviewed from all eligible units that were actually contacted, was 51.4%. The cooperation rate was computed as completed interviews/(refusals + terminations + completed interviews).
Weighting.
The weights for the NPAWLS data were calculated as the product of three components: 1) a sample weight accounting for differential probabilities of selection, defined as the inverse of the product of the household selection probability and the respondent selection probability; 2) a poststratification factor adjusting weight totals to 2000 Census population figures by age, race/ethnicity, and gender, computed for each of the three independent samples, so that the sum of the adjusted weights for each sample would produce an estimate of the eligible U.S. population; and 3) a factor scaling of the weights for each sample to ensure that estimates were based on the combined data and would not overestimate population totals.
Participants.
All participants gave informed consent to participate in NPAWLS before answering any survey items. Exclusion criteria for the current analyses were incomplete data on the BRFSS (N = 496) and the IPAQ (N = 770). The final study sample for data analyses totaled 9945.
Measures
BRFSS.
The 2001 BRFSS physical activity module is an interviewer-administered questionnaire that classifies occupational physical activity (PA), quantifies the frequency and duration of moderate and vigorous leisure physical activity performed during nonworking hours in a usual week, and quantifies the frequency of strengthening physical activity in a usual week (http://www.cdc.gov/brfss/questionnaires/pdf-ques/2001brfss.pdf Table 1 , the moderate-intensity questions precede the vigorous-intensity questions. Prior to responding, study participants are read an introductory statement giving examples of moderate- and vigorous-intensity activities along with physiological cues for breathing and heart rate to help them recall activities at an appropriate intensity level. The reliability of the BRFSS is acceptable over a 1-wk period (κ = 0.84) (24 24 20 Table 2 .
TABLE 1: Moderate- and vigorous-intensity questions from the 2001 Behavioral Risk Factor Surveillance System (BRFSS) physical activity module and moderate- and vigorous-intensity and walking questions from the short, interviewer-administered International Physical Activity Questionnaire (IPAQ).
IPAQ.
The short, interviewer-administered IPAQ identifies the frequency and duration of moderate and vigorous leisure, transportation, and occupational physical activity, walking physical activity, and inactivity during the past week (http://www.ipaq.ki.se Table 1 , the vigorous-intensity questions precede the moderate-intensity and walking questions. For each question, participants are given examples of moderate, vigorous, and walking activities and physiological cues for breathing and the heart rate to help them recall activities at an appropriate intensity level. One-week test-retest reliability of the short, interviewer-administered IPAQ is good (Spearman r = 0.70-0.97). Criterion validity for the IPAQ total minutes per week as measured against accelerometer total counts is acceptable (Spearman r = 0.23), as is the average correct classification of participants accumulating > 150 min·wk−1 of total physical activity (Spearman r = 0.74) (7 Table 2 . One MET-minute is defined as the MET intensity multiplied by the minutes per week of activity. A MET is roughly equivalent to the activity metabolic rate divided by the resting metabolic rate; MET represents the energy expended while sitting quietly at rest (21
TABLE 2: Scoring criteria for the 2001 Behavioral Risk Factor Surveillance System (BRFSS) physical activity module and the short, interviewer-administered International Physical Activity Questionnaire (IPAQ).
In this study, the IPAQ was scored using two methods. The first method used the BRFSS protocol, which includes the IPAQ moderate-intensity and vigorous-intensity questions. The IPAQ walking questions were excluded in these analyses. The second method used the IPAQ scoring protocol, which includes the IPAQ moderate-intensity, vigorous-intensity, and walking questions only.
Demographic variables.
Demographic variables were obtained from the 2001 BRFSS questionnaire (http://www.cdc.gov/brfss/questionnaires/pdf-ques/2001brfss.pdf 2 )), and educational level (≤ high school, some college, and college graduate). BMI was categorized as underweight/normal (≤ 24.9 kg·m−2 ), overweight (25.0-29.9 kg·m−2 ), and obese (≥ 30.0 kg·m−2 ). Age was categorized as young adult (18-34 yr), middle age (35-54 yr), and older adult (≥ 55 yr). Income was categorized as ≤ $25,000, $25,000-$49,999, and ≥ $50,000 total annual household income.
Data analyses.
SAS (version 8.2, Cary, NC) and SUDAAN software (version 8.0, Research Triangle Institute, Research Triangle Park, NC) were used for data analyses. Data were weighted to account for differential probabilities of sample selection and poststratified to the U.S. 2000 Census population. The kappa statistic and the 95% confidence interval (CI) were computed as a measure of agreement between the BRFSS and the IPAQ physical activity categories.
RESULTS
Participants included 4258 men and 5687 women ranging in age from 18 to 99 yr (Table 3 ). Every age and education level was well represented. Most were non-Hispanic white (72.9%), with a smaller percentage classified as non-Hispanic black (11.3%) and Hispanic (10.1%). Fifty-seven percent were overweight or obese. About 44% had a household income of more than $50,000 per year.
TABLE 3: Demographic characteristics of the study sample, National Physical Activity and Weight Loss Study (NPAWLS), 2002 (N = 9945).
Table 4 presents the comparison of the physical activity prevalence estimates for the BRFSS and the IPAQ; the IPAQ is scored using the BRFSS protocol. This comparison was made to eliminate differences in prevalence estimates that might result from each questionnaire's scoring protocol and activity cut points. No order effect was identified in the physical activity prevalence estimates for participants completing versions A or B of NPAWLS (data not shown, P > 0.05). Kappas showed fair to moderate agreement (κ = 0.34-0.49) between the BRFSS and the IPAQ for all participants combined and for groups stratified by sex, age, race/ethnicity, BMI, education, and income (6
TABLE 4: Age-adjusted, weighted prevalence, and kappa coefficients of agreement for physical activity categories using the Behavioral Risk Factor Surveillance System (BRFSS) scoring protocol for the 2001 BRFSS physical activity module and the short, interviewer-administered International Physical Activity Questionnaire (IPAQ).
Table 5 shows a comparison of the physical activity prevalence estimates for the BRFSS and the IPAQ using the scoring protocols designed for each questionnaire. The prevalence estimates for the least and most active physical activity categories were higher on the IPAQ than the BRFSS. These findings were observed for all participants combined as well as when the data were stratified by sex, age, race/ethnicity, BMI, education, and household income, except for two groups. For Hispanics, inactive estimates on the BRFSS and the IPAQ were similar, and for those earning more than $50,000 per year, the highest physical activity levels were similar on the BRFSS and the IPAQ. The prevalence estimates for the middle physical activity category (BRFSS = insufficient; IPAQ = moderate) were generally higher on the BRFSS than on the IPAQ. Kappa statistics for the comparisons ranged from 0.26 to 0.39 and were considered to demonstrate fair agreement (6
TABLE 5: Physical activity prevalence estimates for the 2001 Behavioral Risk Factor Surveillance System (BRFSS) physical activity module and the short, interviewer-administered International Physical Activity Questionnaire (IPAQ). Data are presented as the age-adjusted, weighted prevalence estimates and kappa coefficients of agreement for their respective physical activity categories.
DISCUSSION
The absolute difference in the prevalence estimates between the BRFSS and the IPAQ was notable. When the IPAQ was scored using the BRFSS protocol, the inactive category was 0.1-3.9% higher on the IPAQ than on the BRFSS, and the meets recommendations category was 0.2-9.7% higher on the IPAQ than on the BRFSS for the various subgroups. The differences were greater when the IPAQ was scored using the IPAQ protocol: the low category on the IPAQ was 0.2-13.3% higher than the inactive category on the BRFSS, and the high category on the IPAQ was 0-16.4% higher than the meets recommendations category on the BRFSS. The IPAQ prevalence estimates were higher in all subgroups except among college graduates, who had higher prevalence estimates for the BRFSS meets recommendations category than for the IPAQ high category.
Despite the differences, the findings showed fair to good agreement for physical activity classification using the 2001 BRFSS physical activity module and the short, interviewer-administered IPAQ when the IPAQ was scored using the BRFSS protocol. Kappa statistics indicating agreement between the questionnaires ranged from 0.34 to 0.49. Agreement was lowest in non-Hispanic blacks, Hispanics, and respondents with a high school education or less. Agreement was highest for college graduates, non-Hispanic whites, and respondents earning more than $50,000 per year. Interestingly, the prevalence estimates for highest levels of activity were most similar for respondents ≥ 55 yr old, women, non-Hispanic whites, college graduates, and those earning more than $50,000 per year. This observation was consistent when the IPAQ was scored using the BRFSS and the IPAQ protocols. This pattern may reflect three possibilities. First, there may be differences in cognitive understanding of the physical activity questionnaires by some demographic groups. The questionnaires, developed by physical activity professionals, underwent cognitive testing in diverse groups to minimize ambiguity in the wording and intent of the questions. However, some questionnaire items may still be ambiguous for selected demographic groups, producing differences in the activity scores (10 −1 and 30 min·d−1 ) or regular vigorous-intensity (≥ 3 d·wk−1 and 20 min·d−1 ) physical activity to meet the highest activity level, whereas the IPAQ allows a combination of moderate- and vigorous-intensity activities to meet the highest activity level. The leisure physical activity patterns of some demographic groups may have been consistent with levels of activity required for the BRFSS meets recommendation category, resulting in a greater proportion of respondents classified in the highest activity group, whereas other demographic groups may have had more diverse physical activity patterns, excluding them from the BRFSS meets recommendation group but of sufficient volume for classification in the IPAQ highest activity group. Third, inclusion of occupational and transportation activities on the IPAQ, but not on the BRFSS, would be negligible for persons engaging in sedentary occupational and transportation behaviors, resulting in similar leisure-time activities recorded on the IPAQ and the BRFSS. Alternatively, demographic groups participating in more active modes of transportation and in active occupations would be classified as more active on the IPAQ than on the BRFSS.
Table 4 presents the kappa coefficients of agreement between BRFSS and the IPAQ categories. Different questionnaires for surveillance of physical activity, within the United States and globally, often produce prevalence estimates for activity levels that vary because of differences in the physical activity questionnaire structure and content rather than actual differences in reported physical activity (12,16,19
In comparison, the IPAQ encourages participants to report occupational, leisure/exercise, walking, and transportation activities of moderate and vigorous intensity. It also provides examples of activities and physiological cues for moderate- and vigorous-intensity physical activity. Vigorous-intensity activities are described as causing breathing much harder than normal; they include activities such as heavy lifting, digging, aerobics, or fast bicycling. Moderate-intensity physical activities are described as causing breathing somewhat harder than normal; they include activities such as carrying light loads, bicycling at a regular pace, or doubles tennis. Although we were unable to determine whether being able to report occupational physical activity increased the IPAQ prevalence estimates for physical activity categories listed in Tables 4 and 5 , it is possible that this feature of the questionnaire may partially explain the higher estimates in the high category. The IPAQ includes a third question about walking activities that the BRFSS embeds into the moderate-intensity question. The added physical activity from the walking question also may contribute to the higher IPAQ high physical activity prevalence estimates shown in Table 5 .
The placement of the vigorous-intensity items before the moderate-intensity items on physical activity questionnaires is believed to increase the proportion of participants who report engaging in vigorous PA. On the IPAQ, the vigorous-intensity questions precede the moderate questions and may contribute to the higher prevalence estimates for the IPAQ high category. The order of moderate- and vigorous-intensity questions are reversed in the BRFSS. Ham et al. (10
Some investigators have noted high prevalence estimates from the IPAQ high physical activity category (4 18
The differences in the physical activity prevalence estimates (see Table 4 ) also may be due to differences in the scoring protocols for the BRFSS and the IPAQ. In the BRFSS, the inactive category implies an absence of physical activity. In the IPAQ, however, the low category reflects a level of physical activity that is insufficient to meet the moderate category. Also, in the BRFSS, the meets recommendations category is based on the 1995 U.S. Centers for Disease Control and American College of Sports Medicine (CDC-ACSM) guidelines for the frequency and duration of physical activity needed to reduce chronic disease risks (13 22 total physical activity and includes transportation, occupational, and leisure-time physical activities. The questionnaire also queries walking behaviors. Thus, the IPAQ captures more activities than the BRFSS, giving respondents an opportunity to accumulate more minutes per day in various activities. The more comprehensive nature of the IPAQ necessitates the use of higher physical activity cut points than the BRFSS to meet the moderate and high categories. This is reflected in the IPAQ by requiring respondents to accumulate a set volume of activity, expressed as MET minutes per week, in addition to meeting the BRFSS moderate- or vigorous-intensity frequency and duration levels.
In a comparison of activity prevalence estimates obtained from the Active Australia and the BRFSS PA surveys, Brown and Miller (5 −1 and ≥ 30 min·d−1 or vigorous-intensity activity for ≥ 3 d·wk−1 and ≥ 20 min·d−1 , but not a combination of both. In the Active Australia survey, adults are classified as active if they participate in moderate- or vigorous-intensity activities, or a combination of both intensities, for ≥ 150 min·wk−1 . Using the Active Australia scoring protocol, the prevalence of adults classified as active was 58.3%, whereas the prevalence of adults classified as active using the BRFSS scoring protocol was 43.3%. As with the Active Australia survey, the IPAQ scoring protocol has several ways an individual can meet the moderate and high physical activity categories. To meet the high criteria, one can either engage in vigorous-intensity physical activity for 3 d·wk−1 and 20 min·d−1 and accumulate at least 1500 MET·min·wk−1 in moderate-intensity, vigorous-intensity, or walking activities, or they can engage in any combination of moderate-intensity, vigorous-intensity, or walking physical activities totaling 5 d·wk−1 with an accumulation of at least 3000 MET·min·wk−1 . Fifteen hundred MET-minutes are equal to about 60 min of moderate-intensity physical activity performed 5.5 d·wk−1 (60 min × 4.5 METs × 5.5d·wk−1 ). Vigorous-intensity physical activity performed for 3 d·wk−1 and 20 min per session expends 420 MET·min·wk−1 (20 min × 7.0 METs × 3 d·wk−1 ). The remaining 1080 MET·min can be accumulated by performing 30-40 min·d−1 of moderate-intensity physical activity or walking. Similarly, 3000 MET·min·wk−1 is equal to about 2 h of moderate-intensity physical activity for 5 d·wk−1 . Given that the IPAQ allows participants to include leisure/exercise, walking, occupation, and transportation physical activity in their responses, and the scoring protocol allows greater flexibility in how physical activity is accumulated, it is understandable that the prevalence estimates for the IPAQ high activity category exceeded 50% in this study sample.
In conclusion, this analysis compared two physical activity measurement tools used in public health surveillance and research. The demographic patterns for the physical activity prevalence estimates observed in this study were consistent on the BRFSS and IPAQ and were similar to findings for the U.S. BRFSS physical activity surveillance data ((11,12 http://www.cdc.gov/pcd/issues/2005/apr/04_0114.htm
The authors appreciate the assistance of Dennis Shepard and Maxine Mixon from the University of South Carolina Prevention Research Center, Jennifer Muncil and James Dayton from ORC Macro International, and Heidi Blank, Deborah Galuska, Mary Serdula, and Laura Kettle-Kahn from the U.S. Centers for Disease Control and Prevention for their efforts toward a successful completion of the NPAWLS study. This research was funded by a cooperative agreement from the CDC (U48/CCU409664). Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the CDC.
REFERENCES
1. Abu-Omar, K., A. Rutten, and J. M. Robine. Self-rated health and physical activity in the European Union.
Soz Praventivmed 49:231-232, 2004.
2. Abu-Omar, K., A. Rutten, and V. Lehtinen. Mental health and physical activity in the European Union.
Soz Praventivmed 49:301-309, 2004.
3. Belander, O., M. K. Torstveit, and J. Sundgot-Borgen. Are young Norwegian women sufficiently active?
Tidsskr Nor Laegeforen 124:2488-2489, 2004.
4. Brown, W., A. Bauman, T. Chey, S. Trost, and K. Mummery. Comparison of surveys used to measure physical activity.
Aust. NZ J. Public Health 28:128-134, 2004.
5. Brown, W. J., and Y. D. Miller. International differences in management of physical activity data: can they explain some of the difference in prevalence estimates?
J. Phys. Act. Health 2:460-469, 2005.
6. Cohen, J. A coefficient of agreement for nominal scales.
Educ. Psychol. Meas. 20:37-46, 1960.
7. Craig, C. L., A. L. Marshall, M. Sjostrom, et al. International Physical Activity Questionnaire: 12-country reliability and validity.
Med. Sci. Sports Exerc. 35:1381-1395, 2003.
8. DeBourdeaudhuij, I., J. F. Sallis, and B. E. Saelens. Environmental correlates of physical activity in a sample of Belgian adults.
Am. J. Health Promotion 18:83-92, 2003.
9. Hallal, P. C., C. G. Victoria, J. C. Wells, and R. C. Lima. Physical inactivity: prevalence and associated variables in Brazilian adults.
Med. Sci. Sports Exerc. 35:1894-1900, 2003.
10. Ham, S. A., C. A. Macera, D. A. Jones, B. E. Ainsworth, and K.M. Turczyn. Considerations for physical activity research: Variations on a theme.
J. Phys. Act. Health 1:98-113, 2004.
11. Macera, C. A., D. A. Jones, M. M. Yore, et al. Prevalence of physical activity, including lifestyle activities among adults-United States, 2000-2001.
Morb. Mort. Weekly Rep. 52:764-769, 2003.
12. Macera, C. A., and M. Pratt.
Public health surveillance of physical activity.
Res. Q. Exerc. Sport 71(2 Suppl):S97-S103, 2000.
13. Pate, R. R., M. Pratt, S. N. Blair, et al Physical activity and
public health : a recommendation from the Centers for Disease Control and Prevention and the American College of Sports Medicine.
JAMA 273:402-407, 1995.
14. Rutten, A., A. Vuillemin, W. T. Ooijendijk, et al. Physical activity monitoring in Europe. The European Physical Activity Surveillance System (EUPASS) approach and indicator testing.
Pub. Health Nutr. 6:377-384, 2003.
15. Rutten, A., H. Ziemainz, F. Schena, et al. Using different physical activity measurements in eight European countries. Results of the European Physical Activity Surveillance System (EUPASS) time series survey.
Pub. Health Nutr. 6:371-376, 2003.
16. Rutten, A., and K. Abu-Omar. Prevalence of physical activity in the European Union.
Soz Praventivmed 49:281-289, 2004.
17. Rutten, A., and K. Abu-Omar. Perceptions of environmental opportunities for physical activity in the European Union.
Soz Praventivmed 49:310-317, 2004.
18. Rzewnicki, R., Y. Vanden Auweele, and I. De Bourdeaudhuij. Addressing overreporting on the International Physical Activity Questionnaire (IPAQ) telephone survey with a population sample.
Pub. Health Nutr. 6:299-305, 2003.
19. Stephens, T., D. R. Jacobs JR., and C. C. White. A descriptive epidemiology of leisure-time physical activity.
Public Health Rep. 100:147-158, 1985.
20. Strath, S. J., D. R. Bassett JR., S. A. Ham, and A. M. Swartz. Assessment of physical activity by telephone interview versus objective monitoring.
Med. Sci. Sports Exerc. 35:2112-2118, 2003.
21. Taylor, H. L., D. R. Jacobs Jr., B. Schucker, J. Knudsen, A. S. Leon, and G. Debacker. A questionnaire for the assessment of leisure time physical activities.
J. Chronic Dis. 31:741-755, 1978.
22. U.S. Department of Health and Human Services.
Healthy People 2010. 2nd ed. Washington, DC: U.S. Government Printing Office, pp. 22-1-22-38, 2000.
23. World Health Organization.
The World Health Report 2002: Reducing Risks, Promoting Healthy Life. Geneva, Switzerland: World Health Organization, p. 61.
24. Yore, M. M., S. A. Ham, B. E. Ainsworth, C. A. Macera, D. A. Jones, H. Kohl III. Occupational physical activity: Reliability and Comparison of activity levels.
J. Phys. Act. Health 2:358-365, 2005.
