Differences in Physical Activity Prevalence and Trends From 3 U.S. Surveillance Systems: NHIS, NHANES, and BRFSS

Transcription

1 Journal of Physical Activity and Health, 2009, 6(Suppl 1), S18 S Human Kinetics, Inc. Differences in Physical Activity Prevalence and Trends From 3 U.S. Surveillance Systems: NHIS, NHANES, and BRFSS Susan A. Carlson, Dianna Densmore, Janet E. Fulton, Michelle M. Yore, and Harold W. Kohl, III Background: Three U.S. surveillance systems National Health Interview Survey (NHIS), National Health and Nutrition Examination Survey (NHANES), and Behavioral Risk Factor Surveillance System (BRFSS) estimate physical activity prevalence. Methods: Survey differences were examined qualitatively. Prevalence estimates by sex, age, and race/ethnicity were assessed for comparable survey periods. Trends were examined from NHIS 1998 to 2007, NHANES 1999 to 2006, and BRFSS 2001 to Results: Age-adjusted prevalence estimates appeared most similar for NHIS 2005 (physically active: 30.2%, inactive: 40.7%) and NHANES 2005 to 2006 (physically active: 33.5%, inactive: 32.4%). In BRFSS 2005, prevalence of being physically active was 48.3% and inactive was 13.9%. Across all systems, men were more likely to be active than women; non-hispanic whites were most likely to be active; as age increased, overall prevalence of being active decreased. Prevalence of being active exhibited a significant increasing trend only in BRFSS 2001 to 2007 (P <.001), while prevalence of being inactive decreased significantly in NHANES 1999 to 2006 (P <.001) and BRFSS 2001 to 2007 (P <.001). Conclusions: Different ways of assessing physical activity in surveillance systems result in different prevalence estimates. Before comparing estimates from different systems, all aspects of data col- Carlson, Densmore, and Fulton are with the Division of Nutrition, Physical Activity, and Obesity, Centers for Disease Control and Prevention, Atlanta, GA. Yore is with Orlando Epidemiology, Orlando, FL. Kohl is with the Division of Epidemiology, University of Texas School of Public Health, Austin, TX. lection and data analysis should be examined to determine if comparisons are appropriate. Keywords: health behavior, questionnaires, Healthy People 2010, adults Public health surveillance is defined as the ongoing, systematic collection, analysis, and interpretation of data on health issues for use in planning, implementing and evaluating public health programs. 1 Because many public health concerns have behavioral determinants, health-related behaviors are routinely tracked using ongoing surveys, as part of public health surveillance and prevention efforts. 2 Data from these ongoing surveys can then be used to assess the prevalence of behaviors at a certain time and track trends in behaviors over time. Physical activity is 1 of the 10 leading health indicators in the United States, which correspond with the major health concerns of the nation. 3 Strong evidence shows that, compared with less active persons, physically active men and women have lower rates of allcause mortality, coronary heart disease, high blood pressure, stroke, type 2 diabetes, metabolic syndrome, colon cancer, breast cancer, and depression. 4 Consequently, physical activity is a health behavior that is tracked at state and national levels. Surveillance systems that measure and track changes in physical activity prevalence are necessary to identify population segments that need intervention, monitor secular trends in physical activity, and evaluate the effectiveness of population-based interventions. 5 Currently, the following 3 surveillance systems estimate and track the prevalence of physical activity for adults in the United States: National Health Interview Survey (NHIS), National Health and Nutrition Examination Survey (NHANES), and Behavioral Risk Factor Surveillance System (BRFSS). S18

2 Differences in Physical Activity Prevalence S19 NHIS collects information on a variety of health measures and is used to track progress on many national health objectives, including the physical activity objectives in Healthy People 2010 (ie, 22 to 1 through 22 to 5). 3 The NHIS has been conducted continuously since 1957 and has a long history of collecting physical activity data. 6 The NHIS is conducted by the National Center for Health Statistics (NCHS) of the Centers for Disease Control and Prevention (CDC). 7 NCHS also conducts NHANES. In addition to collecting information on a variety of health measures, NHANES combines information from interviews with a physical examination. 8 The NHANES program began in the early 1960s and has been conducted as a series of surveys. Findings from NHANES can be used to determine the prevalence of risk factors for diseases, such as physical activity, and connect this information with physical examination and laboratory results. In contrast to NHIS and NHANES, BRFSS is a telephone-based survey conducted by state and territorial health departments with technical and methodological assistance provided by the National Center for Chronic Disease Prevention and Health Promotion (NCCDPHP) of CDC. Samples for BRFSS are drawn based on probability sampling from state and territory health departments. Consequently, BRFSS is the only 1 of the 3 surveillance systems that can generate state- or territorial-based estimates on a variety of health measures (eg, physical activity every other year). 9 BRFSS is particularly useful to state and territorial health departments for surveillance and planning, although national U.S. estimates from BRFSS also are possible. The BRFSS began in States and U.S. territories have used BRFSS data to identify emerging health problems, establish and track health objectives, and develop and evaluate public health policies and programs. 9 Physical activity prevalence and trends are estimated from a variety of sources in the United States; however, no study has directly compared data and trends from all sources. This study has 3 main purposes: 1) to compare how and what physical activity data are collected across the 3 surveillance systems; 2) to estimate physical activity prevalence by sex, age, and race/ethnicity for the most recent years that data are available across the 3 surveillance systems; and 3) to examine annual physical activity prevalence trends over periods in which each surveillance system used consistent physical activity questions. Survey Descriptions Methods Data examining patterns by sex, age group, and race/ ethnicity were analyzed from the most recent survey period that all 3 surveys collected physical activity data (ie, NHIS 2005, NHANES 2005 to 2006, and BRFSS 2005). Data examining trends were obtained from NHIS 1998 to 2007, NHANES 1999 to 2006, and biannually from BRFSS 2001 to Details about the survey plan, operation, and sample design of the 3 surveys are described elsewhere 7-9 and are briefly summarized here. NHIS. NHIS is a face-to-face household survey of a random sample of U.S. households; it is conducted continuously throughout the year. The survey is a multistage design that starts by selecting primary sampling units (PSUs) from a list of geographically designed PSUs that cover the 50 states and the District of Columbia. The next level of sampling involves selecting households. Basic health and demographic information are collected for all household members; additional information, such as physical activity, is collected on 1 randomly selected adult ( 18 years of age). During 1998 to 2007, final response rates (which account for household and family nonresponse) for the sample adult component ranged from 67.8% (2007) to 74.3% (2002); the response rate in 2005 was 69.0%. 7 NHANES. NHANES combines a face-to-face survey with physical examinations at a mobile examination center. It uses a stratified, multistage, probability cluster sample of the U.S. civilian, noninstitutionalized population; similar to NHIS, NHANES first selects PSUs before selecting clusters of households. For respondents 16 years of age, physical activity data are collected during the interview at their homes. Data are released in 2-year increments. From 1999 through 2006, response rates for the interview portion ranged from 79.3% ( ) to 83.9% ( ); the response rate for 2005 to 2006 was 80.4%. 8 BRFSS. BRFSS is a state-based, random-digit-dialed telephone survey of the civilian, noninstitutionalized U.S. adult population 18 years of age. It is conducted in all 50 states, the District of Columbia, Guam, Puerto Rico, and the U.S. Virgin Islands. Each state works with CDC to develop a sampling protocol to select households, and 1 adult is selected from each household. Physical activity data are collected every other year (ie, 2001, 2003, 2005, and 2007). From 2000 through 2007, the Council of American Survey and Research Organization median state response rates ranged from 50.6% in 2007% to 53.2% in 2003; the median state response rate in 2005 was 51.1%. 9 Measures Specific information about how each survey assessed physical activity is in Table 1. Respondents were classified as active if they had an equivalent amount of physical activity that would categorize them as having met the Healthy People 2010 objective 3 of at least 30 minutes a day of moderate-intensity activity on 5 or more days a week or at least 20 minutes a day of vigorousintensity activity on 3 or more days a week. Respondents were classified as physically inactive if they

3 S20 Carlson et al Table 1 Characteristics of Physical Activity (PA) Assessments and Physical Activity Levels Category NHIS NHANES BRFSS Survey years PA data were collected a , 2003, 2005, 2007 Recall period Respondent selects recall Past 30 days Usual week period b Domain(s) of PA assessed Leisure-time physical activity Leisure-time physical activity c Nonoccupational physical activity d Self-reported Yes Yes Yes List of specific activities No Yes No Assesses moderate-intensity PA Yes, but includes lightintensity Yes Yes Assesses vigorous-intensity PA Yes Yes Yes Which intensity level is asked about Vigorous Vigorous Moderate first? Definition for moderate-intensity of physical activity Light sweating or a slight to moderate increase in breathing or heart rate Light sweating or a slight to moderate increase in breathing or heart rate Small increases in breathing or heart rate Definition for vigorous-intensity of physical activity Definition of active based on the Healthy People 2010 objective Definition of inactive based on Healthy People 2010 objective Heavy sweating or large increases in breathing or heart rate Light or Moderate 5 times/ week and 30 min/time or Vigorous 3 times/week and 20 min/time No reported light to moderate or vigorous intensity activity for at least 10 minutes Heavy sweating or large Large increases in breathing increases in breathing or or heart rate heart rate Moderate 20 days/month and 600 min/month or Vigorous 12 days/ month and 240 min/month No reported moderate or vigorous intensity activity for at least 10 minutes Moderate 5 days/week and 30 min/day or Vigorous 3 days/week and 20 min/day No reported moderate or vigorous intensity activity for at least 10 minutes Abbreviations: BRFSS, Behavioral Risk Factor Surveillance System; NHIS, National Health Interview Survey; NHANES, National Health and Nutrition Examination Survey. a Includes years in which the same physical activity question was asked of respondents. NHIS asked a slightly different physical activity question in the first half of 1997, which included a minimum duration of at least 20 minutes. This changed midyear in 1997 to at least 10 minutes and has remained unchanged ever since. b NHIS physical activity questions allow respondents to select the recall period. To define physical activity levels, the average number of times/week (rounded to the nearest time) were calculated for those respondents who selected monthly or yearly time periods. c NHANES has separate questions about active transportation and moderate household activities that are not included as part of this analysis. d BRFSS has a separate question about monthly participation (yes/no) in any physical activities or exercises such as running, calisthenics, golf, gardening, or walking for exercise that was not included as part of this analysis. reported no moderate- or vigorous-intensity physical activity lasting at least 10 minutes at a time. Respondents who did not engage in enough activity to be classified as active but reported some activity were classified as insufficiently active. The survey questions can be accessed from each system s website. 7 9 Statistical Analysis We analyzed data from each of the 3 surveillance systems separately to estimate the age-adjusted prevalence and 95% confidence intervals (95% CI) of physical activity (active and inactive level) for adults ( 18 years of age). Except for estimates that were stratified by age, estimates were age-adjusted to the 2000 U.S. population using 5 age groups (18 24, 25 34, 35 44, 45 64, and 65). 10 Significant differences by sex, race/ethnicity, and year within each survey were assessed using pair- wise t tests. Linear and quadratic trends were tested using orthogonal polynomial contrasts. SUDAAN version 9.0 software (Research Triangle Institute, Research Triangle Park, NC) was used for data analysis. Results Characteristics of survey respondents are notably different across the 3 surveillance systems (Table 2). The unweighted percentage of men in the 3 surveillance systems varied from 38.3% in BRFSS 2005% to 48.1% in NHANES 2005 to Distributions by age group also differed by surveillance system, with the most pronounced variation in the youngest age groups (18 24 years of age: 5.1% in BRFSS 2005% to 20.1% in NHANES 2005 to 2006). The unweighted percentage of non-hispanic whites in the 3 surveillance systems varied

4 Table 2 Characteristics of Sample, NHIS 2005, NHANES , and BRFSS 2005 NHIS 2005 NHANES BRFSS 2005 a Characteristics of responders Weighted Weighted Weighted Sample Sample size % % 95% CI b size % % 95% CI Sample size % % 95% CI Sex Men 13, , , , , 48.9 Women 17, , , , , 51.8 Age group (yrs) , , , , , , , , , , , , , , , , , , , , , 16.9 Race/ethnicity White, non-hispanic 20, , , , , 70.5 Black, non-hispanic , , , , 9.8 Other , , , , 20.6 Abbreviations: NHIS, National Health Interview Survey; NHANES, National Health Nutrition Examination Survey; BRFSS, Behavioral Risk Factor Surveillance System. a Respondents from the U.S. territories (n = 6211) were excluded; estimates stratified by race/ethnicity excluded persons with unknown race/ethnicity (n = 3349). b 95% CI = 95% confidence interval. S21

5 Table 3 Age-Adjusted Prevalence a and 95% Confidence Intervals (CI) of Being Physically Active and Inactive National Health Interview Survey (NHIS) 2005, National Health and Nutrition Survey (NHANES) , and BRFSS 2005 b Characteristic NHIS 2005 NHANES BRFSS 2005 Active Inactive Active Inactive Active Inactive % 95% CI % 95% CI % 95% CI % 95% CI % 95% CI % 95% CI Overall , , , , , , 14.1 Sex Men 31.8 c 30.8, c 38.0, , , c 49.3, c 13.0, 13.7 Women , , , , , , 14.6 Age group (yrs) , , , , , , , , , , , , , , , , , , , , , , , , d 20.1, d 52.5, e 20.7, e 42.7, e 39.2, d 23.7, 24.8 Race / ethnicity White, non- Hispanic 33.8 f 32.8, f 34.4, , f 24.6, f 50.5, f 11.2, 11.6 Black, non- Hispanic 21.4 g 19.7, g 52.0, , g 33.6, g 39.1, g 20.6, 22.4 Other 22.3 g 20.9, g 50.3, g 25.3, g 36.1, f,g 42.4, f,g 17.9, 19.8 Abbreviations: NHIS, National Health Interview Survey; NHANES, National Health and Nutrition Examination Survey; BRFSS, Behavioral Risk Factor Surveillance System. a Estimates are age-adjusted to the 2000 U.S. standard population, except estimates stratified by age group. b Respondents for whom physical activity data were missing have been excluded for our analysis (NHIS 2005: 915, NHANES : 3, BRFSS 2005: 24788). In addition, for BRFSS estimates stratified by race/ethnicity, those with unknown race/ethnicity (n = 3349) were excluded. c Significantly different than women (P <.05). d Both a significant linear and quadratic effect (P <.05). e Significant linear effect (P <.05). f Significantly different than non-hispanic blacks (P <.05). g Significantly different than non-hispanic whites (P <.05). S22

6 Differences in Physical Activity Prevalence S23 Figure 1 Age-adjusted prevalence of being physically active, National Health Interview Survey (NHIS) 1998 to 2007, National Health and Nutrition Examination Survey (NHANES) 1999 to 2006, and Behavioral Risk Factor Surveillance System (BRFSS) 2001 to a Significant quadratic effect (P <.05). b Significant linear effect (P <.05). from 47.3% in NHANES 2005 to 2006% to 80.3% in BRFSS After applying the appropriate statistical weights, weighted distributions were similar, although the weighted distribution for BRFSS 2005 by race/ethnicity appeared to be slightly different from NHIS 2005 and NHANES 2005 to 2006, with BRFSS 2005 having a slightly lower percentage of non-hispanic blacks. In NHIS 2005 and BRFSS 2005, men were significantly more likely to be physically active than women. Although men were slightly more physically active than women in NHANES 2005 to 2006, this difference was not significant (Table 3). However, in previous years, NHANES has shown significant differences in being physically active between men and women, with the most pronounced difference in 2001 to 2002 [men: 35.5% (95% CI: 30.8, 40.6), women: 30.1 (95% CI: 27.1, 33.3), P =.006]. Across the 3 surveillance systems, patterns in prevalence estimates of being physically active and inactive were similar by age group, with persons 18 to 24 years of age most likely to be physically active and those 65 least likely to be physically active; the opposite was observed in the prevalence of being inactive (Table 3). Among racial/ethnic groups, non-hispanic whites had the highest prevalence of being physically active and the lowest prevalence of being inactive across all 3 surveillance systems. However, in NHANES 2005 to 2006, the difference in the prevalence of being physically active between non-hispanic whites and blacks was not significant (P =.085). NHIS 2005 and NHANES 2005 to 2006 showed no difference in prevalence of being physically active or inactive between non-hispanic blacks and persons in the other race/ethnicity category, but in BRFSS 2005 non-hispanic blacks had a significantly lower prevalence of being physically active and a significantly higher prevalence of being inactive than persons in the other race/ethnicity category. When examining physical activity trends over time, the surveillance systems differ in the number of data collection points at which the same physical activity questions were asked (Figures 1 and 2). BRFSS provided 4 data points during 2001 to 2007, and there was a significant linear increase in being physically active (P for trend < 0.001), with a difference of 2.7 percentage points (95% CI: 2.4, 3.1) between 2001 and BRFSS 2001 to 2007 exhibited a significant linear decrease in being physically inactive (P for trend < 0.001), with a difference of 3.7 percentage points (95%

7 S24 Carlson et al Figure 2 Age-adjusted prevalence of being physically inactive, National Health Interview Survey (NHIS) 1998 to 2007, National Health and Nutrition Examination Survey (NHANES) 1999 to 2006, and Behavioral Risk Factor Surveillance System (BRFSS) 2001 to a Significant quadratic effect (P <.05). b Significant linear effect (P <.05). CI: 4.2, 3.2) between 2001 and Using data from NHANES (4 data points during 1999 to 2006), the linear trend was on the margin of significance for being physically active (P for trend = 0.051). However, the prevalence of being physically inactive significantly decreased from 1999 to 2006 (P for trend < 0.001), and the difference between 1999 to 2000 and 2005 to 2006 was 5.0 percentage points (95% CI: 8.2, 1.2). When NHANES data were limited to 2001 to 2006 to match the period covered by BRFSS data, no significant linear trend was observed for being physically active or inactive. NHIS data were available for the longest period (10 data points during 1998 to 2007), and no significant linear trend was observed in the prevalence of being physically active or inactive. However, significant quadratic trends were observed (indicating a statistically significant, but nonlinear trend in the data over time, eg, a leveling off). When NHIS data were limited to 2001 to 2007 to match the period covered by BRFSS data, the quadratic effect was no longer significant for being physically active or inactive. There was a significant linear decrease in being physically active (P =.015), however the percentage point difference between 2001 and 2007 was not significant [ 1.0 percentage points (95% CI: 2.5, 0.5)]. There was also a significant linear increase in being physically inactive (P =.001), with the difference between 2001 and 2007 on the margin of significance [1.3 percentage points (95% CI: 0.0, 2.5)]. When NHIS data were limited to 1999 to 2006, the same period covered by NHANES data, the trends by year were the same as those found using the full-time period ( ). Discussion Ongoing and accurate public health surveillance data are crucial to understanding the scope and trends of public health problems. Physical activity is a health behavior that is related to many diseases and conditions; 4 consequently, it should be included as part of public health surveillance activities. In this study, recent prevalence estimates for physical activity levels in the United States differed by surveillance system. Variations in questionnaires and data collection procedures may explain these differences. Even with these differences, patterns by sex, age, and race/ethnicity are generally similar across the 3 surveillance systems, although the statistical significance levels of these differences varied.

8 Differences in Physical Activity Prevalence S25 The 3 surveillance systems yielded slightly different physical activity trends. When examining trends using the full-time period available from each system, the prevalence of being physically active exhibited a small but significantly increasing linear trend only in BRFSS 2001 to 2007, and for being inactive there was a significant decreasing linear trend in NHANES 1999 to 2006 and BRFSS 2001 to In NHIS 1998 to 2007, there was a significant quadratic trend for being physically active and inactive. Perhaps the most notable difference among the surveillance systems is that each uses completely different questions to assess physical activity. Some of these differences include the following: domains assessed, recall period, definition of moderate- and vigorousintensity activity (in addition, NHIS includes lightintensity activity with moderate-intensity activity), and order of assessment of moderate- and vigorous-intensity activity. It is not surprising that the prevalence estimates for being physically active are highest in the BRFSS because this system includes more domains of physical activity (ie, leisure, household, transport). Respondents of the BRFSS reported the accumulated time that they participated in all nonoccupational physical activity, while respondents of the NHIS and NHANES reported the activity done only during their leisure-time. In 2000, BRFSS respondents reported the frequency and duration of the 2 most common leisure-time physical activities that they participated in during the preceding month, and a total of 26.2% of adults were considered physically active (ie, meeting the Healthy People 2010 objective). 11 In 2001, the BRFSS physical activity surveillance questions were changed to include all nonoccupational activity (ie, more than leisure-time physical activity), and a total of 45.3% of adults were considered active. 11 This demonstrates that the inclusion a larger number of activities and more domains of activity may result in higher prevalence estimates. There is some concern that survey questions asking respondents to recall participation across multiple domains of physical activity in a single response may be prone to overestimation. In addition, the current evidence for health benefits associated with physical activity mainly comes from studies examining the leisure-time domain. 4 Future research using measurements of physical activity across multiple domains may wish to examine the interplay across domains and the association of the different domains with health outcomes. To allow for the tracking of physical activity over time, it is important that systems maintain question consistency as much as possible to ensure that trends are monitored consistently. The large changes observed in physical activity prevalence estimates that were associated with the introduction of new BRFSS questions in 2001 demonstrates the importance of this issue. 11 NHIS currently offers the longest time period for which physical activity levels (ie, active and inactive levels) can be monitored. In 1984, 35 states included on their BRFSS core a question about monthly participation (yes/no) in any physical activities or exercises such as running, calisthenics, golf, gardening, or walking. By 1996, all states had incorporated this question as part of the BRFSS core. This question allows states the opportunity to track physical inactivity as defined by the question over a longer period of time than the BRFSS series of moderate- and vigorous-intensity physical activity questions; however, at a national level physical inactivity is tracked as part of Healthy People 2010 using the series of questions from the NHIS. When physical activity data are collected using different questions, either on the same system in different years or on different systems, estimates should not be directly compared and goals should be set using data from the same question and system that will be used for monitoring progress. Different data collection characteristics of the surveillance systems such as sample design, mode of survey administration, and characteristics of respondents also may be responsible for prevalence estimate differences. Other research has shown that even when survey questions are identical, significant differences in prevalence estimates for fair/poor health status, asthma, and cigarette smoking have been observed across BRFSS, NHANES, and NHIS. 12 Therefore, the different data collection techniques used by the surveillance systems should be examined to determine their potential effect on physical activity prevalence estimates. Briefly, both NHIS and NHANES start their sampling plan with households in geographical areas. However, BRFSS begins with a state-based random-digitdialed sample. The BRFSS sampling procedure insures that all states and territories are represented and allows for state and territorial-specific estimates; however the sampling frame excludes households without landline telephones. Some researchers have concluded that noncoverage of households without landline telephones results in minimal bias for general population telephone surveys of adults. 13,14 However, if the prevalence of households without landline telephones continues to grow, or if adults using mostly wireless devices rarely (if ever) answer their household landline telephones, landline telephone surveys may experience increasing issues with bias from noncoverage and nonresponse. 15 If this noncoverage or nonresponse is related to physical activity, estimates of physical activity prevalence could be biased. The physical activity questionnaire for NHIS and NHANES are conducted in person, whereas BRFSS is a telephone survey, and different modes of survey administration may produce different results The surveillance systems also have different response rates. Although BRFSS uses a different method to calculate the response rate, the differences between the response rates among the 3 systems is pronounced, with NHANES having response rates of approximately 80% compared with the BRFSS response rates of approximately 50%. The effect of higher nonresponse rates on physical activity prevalence is not known. Some researchers have concluded

9 S26 Carlson et al that large differences in response rates have only minor effects on estimates, 19,20 but more analytical research is needed to examine how physical activity prevalence is affected by lower response rates in national surveys. Finally, for the 3 surveillance systems, the population size and composition differ by age, sex, and race/ ethnicity, both unweighted and weighted. The 3 systems vary greatly in their population size; BRFSS had the largest population size and NHANES had the smallest. This difference may explain why some differences in the survey results are statistically significant in BRFSS, but not in NHANES or NHIS. Differences in the unweighted population composition across the 3 surveillance systems may be caused by a number of potential factors, which include oversampling of certain groups in NHIS (eg, Blacks, Hispanics, Asians [ only]) and NHANES (eg, persons 60 and older, African Americans, and Mexican Americans) and differing levels of survey nonresponse or noncoverage of certain groups. Differences in weighting procedures help explain why the weighted data population composition differs slightly by race/ethnicity. To create population controls to be used to develop weights, NHIS is based on census projections, NHANES on the Current Population Survey, and BRFSS on population estimates from Claritas, a marketing information company (Claritas Inc., San Diego, CA). These sources for population control totals vary in their analysis of the U.S. population by sex, age, and race/ethnicity. In addition, NHIS and NHANES account for differences by race/ethnicity in their weighting procedure; BRFSS adjusts weights by race/ethnicity in some states. Differences in survey coverage, survey response, and weighting procedures that are related to correlates of physical activity may result in different physical activity estimates across surveillance systems. The recently published 2008 Physical Activity Guidelines for Americans 21 suggests that health benefits for adults occur when they participate in moderate-intensity physical activity for at least 150 minutes per week, vigorous-intensity activity for at least 75 minutes, or an equivalent combination of moderate- or vigorousintensity physical activity. This new guideline leads to different definitions of physical activity than we used in our analysis, which is based on the Healthy People 2010 objectives. In a study that analyzed BRFSS 2007 data, according to the activity levels in the 2008 guidelines, the prevalence of being physically active was 15.7 percentage points higher than when the definitions from the Healthy People 2010 objectives were used. 22 We performed our analysis using the definitions from the Healthy People 2010 objectives because these were the appropriate definitions for the period covered in our study. In the future, researchers may wish to examine differences in prevalence estimates in the surveillance system using physical activity level definitions based on the 2008 Guidelines. There are inherent and ongoing challenges to all forms of public health surveillance. To be cost-effective and practical, behavioral surveillance systems must be used to assess multiple health behaviors in one survey. This strategy creates difficulties in obtaining detailed information about any particular behavior. Physical activity is a complex and difficult behavior to measure, yet the physical activity survey question sets used in surveillance systems must be short to meet space and time constraints. In 2003, accelerometers were introduced as part of NHANES to offer an alternative to selfreported information. Studies with NHANES 2003 to 2004 data have found that physical activity levels measured with accelerometers are substantially lower than those measured using self-report. 23 Although estimates of physical activity prevalence based on self-report may suffer from reporting bias, 24 one issue with the use of accelerometers to categorize individuals persons into groups based on public health objectives or guidelines, is that the objectives and guidelines originate from epidemiological studies that examined the association between self-reported physical activity and health benefits. 21 More work is needed to understand further the differences in self-reported estimates among different surveillance systems, and between self-reported data and data collected from accelerometers. In addition, more studies are needed to examine the validity and reliability of current survey questions, although the validity and reliability of the current BRFSS physical activity questions suggests that this instrument is sufficient to categorize adults into physical levels, as defined by Healthy People Conclusions The 3 surveillance systems used in the United States assess physical activity in different ways. Not surprisingly, their respective prevalence estimates differ. Differences in questionnaires and data collection procedures may explain these differences, although the percentage of variance explained by each of these factors cannot be determined. However, patterns by sex, age group, and race/ethnicity are relatively similar across the 3 systems. Before comparing prevalence estimates from different surveillance systems, all aspects of data collection and data analysis must be examined; it is also important to understand the strengths and limitations inherent in each system to determine if comparisons are appropriate. In addition, ensuring estimate comparability over time is essential to setting and monitoring state- or national goals related to physical activity. Therefore, baseline and tracking estimates should be from the same surveillance system and use the same survey questions that appropriately reflect health objectives. Author s Note The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.

10 Differences in Physical Activity Prevalence S27 References 1. Thacker SB, Berkelman RL. Public health surveillance in the United States. Epidemiol Rev. 1988;10: Thacker SB, Stroup DF. Public health surveillance. In: Brownson RC, Petitti DB, eds. Applied Epidemiology: Theory to Practice. 2nd ed. New York, NY: Oxford University Press; US Department of Health and Human Services. Healthy People nd ed. Washington, DC: US Government Printing Office; Physical Activity Guidelines Advisory Committee. Physical Activity Guidelines Advisory Committee Report, Washington, DC: US Department of Health and Human Services; Dishman RK, Washburn RA, Heath GW. Physical Activity Epidemiology. Champaign, IL: Human Kinetics; Division of Health Interview Statistics. National Center for Health Statistics (NCHS). Adult Physical Activity Information in the National Health Interview Survey. Published Accessed January 20, National Center for Health Statistics (NCHS). National Health Interview Survey (NHIS). nchs/nhis.htm. Accessed November 3, Centers for Disease Control and Prevention. National Health and Nutrition Examination Survey (NHANES). Accessed November 3, Centers for Disease Control and Prevention. Behavioral Risk Factor Surveillance System (BRFSS): Turning Information into Health. Updated March 19, Accessed November 3, Klein RJ, Schoenborn CA. Age adjustment using the 2000 projected U.S. population. Healthy People 2000 Stat Notes. 2001(20): Centers for Disease Control and Prevention. Prevalence of physical activity, including lifestyle activities among adults United States, MMWR Morb Mortal Wkly Rep. 2003;52(32): Fahimi M, Link M, Mokdad A, Schwartz DA, Levy P. Tracking chronic disease and risk behavior prevalence as survey participation declines: statistics from the behavioral risk factor surveillance system and other national surveys. Prev Chron Dis. 2008;5(3):A80. cdc.gov/pcd/issues/jul/07_0097.htm. 13. Blumberg SJ, Luke JV, Cynamon ML. Telephone coverage and health survey estimates: evaluating the need for concern about wireless substitution. Am J Public Health. 2006;96(5): Ford ES. Characteristics of survey participants with and without a telephone: findings from the third National Health and Nutrition Examination Survey. J Clin Epidemiol. 1998;51(1): Blumberg SJ, Luke JV. Wireless Substitution: Early Release of Estimates from the National Health Interview Survey, July-December Rockville, MD: US Department of Health and Human Services, Centers for Disease Control and Prevention; Greenfield TK, Midanik LT, Rogers JD. Effects of telephone versus face-to-face interview modes on reports of alcohol consumption. Addiction. 2000;95(2): Dillman DA, Christian LM. Survey mode as a source of instability in responses across surveys. Field Methods. 2005;17(1): Tourangeau R, Smith TW. Asking sensitive questions the impact of data collection mode, question format, and question context. Public Opin Q. 1996;60(2): Curtin R, Presser S, Singer E. The effects of response rate changes on the index of consumer sentiment. Public Opin Q. 2000;64(4): Keeter S, Miller C, Kohut A, Groves RM, Presser S. Consequences of reducing nonresponse in a national telephone survey. Public Opin Q. 2000;64(2): US Department of Health and Human Services Physical Activity Guidelines for Americans. Washington, DC: US Department of Health and Human Services; Carlson S, Fulton J, Galuska D, Kruger J, Lobelo F, Loustalot F. Prevalence of self-reported physically active adults United States, MMWR Morb Mortal Wkly Rep. 2008;57(48): Troiano RP, Berrigan D, Dodd KW, Mâsse LC, Tilert T, McDowell M. Physical activity in the United States measured by accelerometer. Med Sci Sports Exerc. 2008;40(1): Sallis JF, Saelens BE. Assessment of physical activity by self-report: status, limitations, and future directions. Res Q Exerc Sport. 2000;71(2, suppl):s1 S Yore MM, Ham SA, Ainsworth BE, et al. Reliability and validity of the instrument used in BRFSS to assess physical activity. Med Sci Sports Exerc. 2007;39(8):