Educational paper literature review 1

Journal of Physical Activity and Health, 2013, 10, 470-479 © 2013 Human Kinetics, Inc. Official Journal of ISPAH

www.JPAH-Journal.com ORIGINAL RESEARCH

470

Liu and Sun are with the Dept of Epidemiology and Biostatis- tics, University of South Carolina, Columbia, SC. Beets is with the Dept of Exercise Science, University of South Carolina, Columbia, SC. Probst is with the Dept of Health Services and Policy Management, University of South Carolina, Columbia, SC.

Assessing Natural Groupings of Common Leisure-Time Physical Activities and Its Correlates Among US Adolescents

Jihong Liu, Han Sun, Michael William Beets, and Janice C. Probst

Objectives: We examined the natural groupings of leisure-time physical activities (LTPA) among US ado- lescents and their correlates. Methods: Data came from the 1999–2006 NHANES, restricted to 3865 boys and 3641 girls 12–19 years old. Respondents were asked to report > 40 types of moderate-to-vigorous LTPA in the past month. Latent class analyses were used to identify natural groupings of the top 10 LTPA using the proportion of each activity’s metabolic equivalents (METs) to total energy expenditure from all physical activities. Results: For each gender, 5 natural groupings of LTPA were identified. Among boys, they were basketball players and runners (72.8%), football players (9.0%), bicycle riders (7.5%), soccer players (5.8%), and walkers (4.7%). For girls, the 5 natural groupings in descending order were dancers/walkers/joggers (79.0%), aerobic exercisers (6.1%), swimmers (5.6%), volleyball players (4.9%), and soccer players (4.2%). The natural groupings of physical activities were also impacted by age, race, weight status, region, and season of interview. Conclusions: The natural groupings of LTPA reflect adolescent’s preference and these activity patterns are likely shaped by their social and physical environments. Better understanding of common LTPAs and their natural groupings is useful in the design of effective PA interventions.

Keywords: gender differences, preferred physical activities, physical education, latent class analysis

Regular physical activity in children and adolescents both promotes physical and mental health and improves cardiorespiratory fitness.1 Public health authorities have called for increased physical activity among all seg- ments of the US population and physical activity has been explicitly listed in the national health objectives.2 However, most US children and adolescents do not meet the 2008 Physical Activity Guidelines, that is, engaging in moderate- and vigorous-intensity physical activity for periods of time that add up to 60 minutes or more each day.3

Physical activity participation among adolescents is known to be related to a myriad of factors such as cul- ture, gender, socioeconomic status, race/ethnicity, health status, weather, access to sports facilities, and neighbor- hood environment.4–6 Evidence-based efforts to improve youth physical activity should build upon the preferences that naturally emerge from the confluence of these fac- tors. The existing literature on physical activity among adolescents mainly focuses on the total level of physical activity, based on the frequencies and duration of physical activities and its intensity.7,8 Studies which investigated the types of common or preferred physical activities

among boys and girls tend to focus on the frequencies of participating in those activities and their relative ranking.9–17 To our knowledge, no study has examined the natural grouping of physical activities. An individual might participate in 1 or many types of activities across a typical month, which all contribute to their total energy expenditure and are beneficial to health. Understanding the natural grouping of physical activities and correlates associated with these patterns can be helpful in designing programs that aim to promote physical activity by tailor- ing activities that are commonly reported in subgroups of the population. Accordingly, the aims of this study were 2-fold. First, we examined the top common leisure-time physical activities among US adolescent boys and girls, respectively. Second, we studied the natural groupings of common leisure-time physical activities (LTPA) among US adolescent boys and girls, and further investigated the correlates for natural groupings of these activities.

Data and Methods We used the data from the 1999–2006 continuous National Health and Nutrition Examination Survey (NHANES), an ongoing, nationally representative study conducted by the National Center for Health Statistics (NCHS). NHANES uses a complex, stratified, multistage probability sampling procedure designed to provide prevalence estimates describing the health and nutritional status of the civilian, noninstitutionalized US population. To allow for accurate estimates for subgroups, NHANES oversampled adolescents, Blacks, and Hispanics.

Natural Groupings of Common Physical Activities 471

NHANES reaches about 5000 persons each year in 15 selected counties across the US, with a new sample of counties implemented each year. Households are selected randomly within each county. The response rate for years 1999–2006 for home interviews was 81%, and 95% of respondents interviewed at home had a follow-up exami- nation at mobile examination centers (MEC). The study methods of NHANES are described in detail on the study website (http://www.cdc.gov/nchs/nhanes.htm). Data from NHANES are broadly used in standard-setting and tracking of Healthy People objectives.18

In NHANES’ home interviews, age-specific ques- tions were used to examine physical activity for children age 2–11 and 12–19 years old. For children age 2–11 years old, only a single question was used to assess physi- cal activity. Due to this, we restricted to adolescents age 12–19 years old where a series of questions were used to measure their physical activities. Respondents were first asked whether they participated in “any vigorous activities for at least 10 minutes that caused heavy sweat- ing, or large increases in breathing or heart rate” over the past 30 days. Those who reported “yes” were also asked to indicate which of 24 types of vigorous sports [the metabolic equivalent (MET) value ≥ 6] that they had engaged in, with 3 open-ended questions allowing them to write down any unlisted activities.19 Adolescents were also asked to report the frequency and duration of each selected activity. After this, all respondents were asked to report their participation in moderate activities that caused only light sweating or a slight to moderate increase in breathing or heart rate (3 ≤ MET < 6) for at least 10 minutes. Adolescents reporting moderate were again asked to indicate the frequency and duration of their participation in specific sports, from a slightly different list of 32 types of moderate recreational activities plus 3 open-ended questions.

Based on this information, we first determined the top 10 common moderate-to-vigorous physical activities (MVPA) for boys and girls based on its frequencies. This type of ranking has already been documented in previous studies9–11 and we did this to make our study comparable with existing ones. Second, to understand the natural groupings of these LTPA, we examined the contributions of each activity to the respondent’s total physical activity energy expenditure. A respondent could have answered that they did the activity, yet only participated in it for a short period of time or might have been active at light or moderate intensity. To overcome this limitation, we chose energy expenditure from an activity or all activities as a more comprehensive measure compared with the fre- quency measure. We estimated total energy expenditure from MVPA by summing the products of the metabolic equivalents of each activity, its frequencies, and duration. Then the proportion of energy expenditure from each activity was calculated as the energy expenditure from the activity divided by the total energy expenditure from all reported physical activities by the individual.

Latent class analyses (LCA) were applied to the 10 activities with the highest contribution to physical activ- ity energy expenditure for boys and girls, respectively,

to determine the natural groupings of MVPA. LCA is often called a person-oriented approach, as opposed to a variable-oriented approach, because LCA focuses on the relationships among individuals under the assumption that data were drawn from more than 1 population.20–22 LCA provides a rich understanding of behavioral patterns and the related demographic characteristics.23 This tech- nique uses maximum likelihood procedures to separate respondents into an optimal number of unobserved (ie, latent) classes characterized by meaningful and mutu- ally distinctive subgroups. Specifically, our analysis began with a one-class (k) model (ie, all adolescents share the same natural groupings of MVPA) and added an extra class (k+1) until the best fitting model was found. The Lo-Mendell-Rubin likelihood (aLMR) ratio tests24 and the Bayesian information criteria (BIC) were used to determine the optimal number of classes and the best fitting model. The aLMR is a likelihood ratio test that compares a model of k–1 class to the k class model, with a significant P-value corresponding to the rejection of the k–1 class model. The BIC is scaled so that small numbers indicate a good model with a large log-likelihood value. A plot of the BIC is evaluated for a plateau in the descending BIC values which indicates adding additional classes does not improve model fit. An additional criterion for selecting the optimal number of classes was based on the interpretability of the number of classes and whether these classes are substantively meaningful in practice or simply smaller groupings of larger classes. This was performed by examining the profile plots (ie, plots of the mean values from each class) to determine whether adding successive classes meaningfully contributed to the distinction of classes. The optimal number of classes was ultimately selected based on evaluating both statistical criteria and substan- tive interpretation of the classes. Because the choice of physical activities is gender-specific, the procedure for finding optimal number of classes was conducted for boys and girls, separately. Covariates, mainly demographic variables listed below, were also included in the models estimating class membership.

Multinomial logistic regression models, with the class with the highest proportion on the sample serv- ing as the reference group, were used to examine these covariates’ influence on class membership. The corre- lates were: age (continuous), race/ethnicity, overweight (age- and gender-specific body mass index ≥ 85th percentile), reference person’s education (< 12, = 12, > 12 years), family poverty status [≤ 130%, 131%–185%, 186%–250%, ≥ 250% of federal poverty level (FPL)], self-reported health status (good/fair/poor, very good, excellent), region (Northeast, Midwest, South, West), urban-rural residence, and season when interviewed. Race/ethnicity is self-reported and is recorded using the NHANES categories of Hispanic, non-Hispanic white, Non-Hispanic Black, and non-Hispanic Other (hereafter, Hispanic, white, black, and other). Reference person is the first household member age 18 years or older who is the person or 1 of the persons who owns or rents the dwelling unit. Without any other parental information,

472 Liu et al

this variable is used to estimate education level of the household. To preserve observations with missing infor- mation on household poverty status (4.5%), we created a special category “missing” for this variable. Urban residence is known to be associated with reduced physical activity,25,26 thus residence included as a correlate for our activity grouping. Urban-rural residence was measured at the census tract level using the Rural-Urban Commut- ing Area (RUCA) definition developed by the University of Washington’s Rural Health Research Center and the Economic Research Service at the US Department of Agriculture.27 We defined urban as RUCA codes between 1 and 3 and rural areas as RUCA codes between 4 and 10.28 Considering that physical activities would vary by season and by school calendar, season of interview time was categorized into 4 periods: Spring (March 1 to June 14), Summer (June 15 to August 14), Autumn (August 15 to November 30), and Winter (December 1 to Febru- ary 28). We accessed the data on urban-rural residence and season of interview time in NHANES through the Research Data Center at NCHS. SAS-callable SUDAAN was used to manage databases and conducted analyses for the results in Table 1 and 2. M-plus was used for latent class analyses. All results took into account of the

survey design structure and weighted percentages were presented. The study was approved by a local institutional review board.

Results

Sample Characteristics

Our analysis is based on responses from 3865 adolescent boys and 3641 adolescent girls who had complete infor- mation on all variables mentioned in the previous sec- tion. Table 1 describes the characteristics of our sample by gender. On average, our study population was 15.4 years old. Boys and girls were similar to each other in all characteristics except self-reported health status. For both genders, most respondents described themselves as white, followed by Hispanic, black, and other. Nearly a third of responding adolescents were overweight. Nearly 40% of these children lived in households under 185% of federal poverty line, and half of them lived in the house where the adults had education level beyond high school level. Nearly 30% of them were from Midwest states, additional 40% from Western states, 10% living in Northeastern states, and 19% living in Southern states. About 22% of

Table 1 Characteristics of US Adolescents Age 12–19 Years by Gender, NHANES 1999–2006 (n = 7506)

Boys (unweighted n = 3865) Girls (unweighted n = 3641)

Characteristics % (se) % (se)

Age (mean, se, years) 15.4 (0.06) 15.4 (0.06)

Race/ethnicity

Hispanic 16.3 (1.4) 17.6 (1.6)

Non-Hispanic white 62.4 (2.0) 60.6 (1.9)

Non-Hispanic black 14.6 (1.4) 15.4 (1.4)

Non-Hispanic other 6.7 (1.0) 6.4 (0.8)

Child’s health

Good/fair/poor 43.9 (1.2) 36.3 (1.4)

Very good 31.7 (1.1) 31.0 (1.2)

Excellent 24.3 (1.0) 32.6 (1.3)***

Child’s weight status

Overweighta 32.6 (1.3) 31.0 (1.2)

Not overweight 67.4 (1.3) 69.0 (1.2)

Household poverty status

≤ 130% FPL 26.9 (1.3) 30.3 (1.6)

131–185% FPL 10.8 (0.8) 9.1 (0.7)

186–250% FPL 10.1 (0.8) 10.1 (0.8)

≥ 251% FPL 47.6 (1.5) 45.1 (1.5)

Missing 4.5 (0.5) 5.5 (0.7)

(continued)

473

Boys (unweighted n = 3865) Girls (unweighted n = 3641)

Characteristics % (se) % (se)

Reference person’s education

< 12 years 20.3 (1.1) 22.2 (1.1)

= 12 years 27.4 (1.5) 26.4 (1.4)

> 12 years 52.4 (1.7) 51.4 (1.6)

Region

Northeast 10.1 (3.2) 10.6 (3.4)

Midwest 31.5 (4.6) 30.6 (4.4)

South 19.2 (4.2) 18.8 (3.9)

West 39.2 (4.3) 40.0 (4.4)

Rural/urban residence

Rural 21.5 (3.1) 23.5 (3.1)

Urban 78.5 (3.1) 76.5 (3.1)

Season of interview time

Spring (3/1 to 6/14) 34.3 (4.5) 36.0 (4.6)

Summer (6/15 to 8/14) 18.5 (3.2) 16.3 (2.9)

Autumn (8/15 to 11/30) 31.5 (4.2) 30.7 (4.3)

Winter (12/1 to 2/28) 15.7 (2.9) 17.0 (3.0)

Note. P-values from chi-square tests of independence between gender and a characteristic were marked as * <0.05; ** <0.01; *** < 0.0001.

Abbreviations: FPL, federal poverty level. a Overweight: BMI percentile ≥ 85th gender- and age-specific percentile.

Table 1 (continued)

them lived in rural areas. Furthermore, 3 out of 5 children were interviewed in either spring or autumn season and additional 16% to 18% of them were interviewed in either summer or winter season, respectively. On average, boys expended 546.1 MET minutes (s.e. = 14.4) on MVPA in the past month, higher than that for girls (mean = 369.1, s.e. = 15.9) (data not shown in the table).

Common Physical Activities

For boys, the 10 most common physical activities, in descending order, were basketball (12.9%), running (11.3%), football (7.8%), bicycling (7.2%), walking (5.9%), weight lifting (5.9%), jogging (5.1%), swim- ming (4.3%), soccer (4.0%), and dance (3.0%). For girls, the most common 10 physical activities were running (11.9%), walking (11.0%), dance (10.5%), basketball (7.0%), jogging (6.0%), bicycling (5.7%), aerobics (4.9%), swimming (4.6%), volleyball (4.0%), and soccer (3.7%) (Table 2).

When we used the proportion of total activity energy expenditure from each activity, the ranking was slightly

different. For boys, the top 5 activities that account for the highest proportion of total energy expenditure from the reported activities were basketball (19.6%), running (11.5%), football playing (9.1%), bicycling (7.6%) and soccer (5.5%). For girls, the top 5 activities contributing to the higher energy expenditures were dancing (12.9%), walking (12.7%), running (12.6%), basketball (8.2%), and bicycling (4.9%) (Table 2).

Natural Groupings of Physical Activities and Its Correlates

The 10 activities with the highest contribution to energy expenditure for boys and girls were used to determine the natural groupings of MVPA. Table 3 presents the model fit estimates for K-class solutions from latent class analyses for boys and girls, separately. Based on the statistical criteria and evaluation of the profile plots from each model, a 5-class solution best represented the underlying natural groupings of the data for both genders.

As shown in Table 4, for boys, the most popular group was characterized as basketball players and runners

474 Liu et al

Table 3 Model Fit Estimates for K Class Solutions for Latent Class Analysis by Gender

Model fit estimates Class 2 Class 3 Class 4 Class 5 Class 6 Class 7 Class 8 Class 9

Boys

Loglikelihooda –163721.5 –160755.8 –158355.8 –156611.4 –154447.3 –154478.1 –152923.0 –149934.9

BICb 327864.7 322189.8 317645.9 314413.6 310341.7 310659.6 307805.9 302085.9

Entropyc 0.999 0.998 0.997 0.990 0.994 0.986 0.979 0.993

aLMRd P for k–1 0.2395 0.0009 0.0908 0.1812 0.7067 0.6418 0.8534 0.1455

Girls

Loglikelihooda –151639.5 –149092.5 –146589.9 –144192.8 –143009.4 –143752.4 –138372.5 –138287.3

BICb 303695.5 298854.6 294102.5 289561.5 287447.7 289186.9 278680.2 278762.9

Entropyc 0.997 0.997 0.996 0.994 0.987 0.988 0.989 0.993

aLMRd P for k–1 0.0052 0.0007 <0.0001 0.0003 0.3688 0.7998 0.0013 0.8269

a The larger the loglikelihood value, the better fit to the model with k number of classes. b The smaller (ie, lower) the Bayesian Information Criteria (BIC) values, the better the fit to the model with k number of classes. c Entropy is a measure of correct classification and can be viewed similar to R2 in assessing model fit, with values ranging from 0.0–1.0. High values are desirable. d Adjusted Vuong-Lo-Mendell-Rubin likelihood ratio test (aLMR) for k–1 classes indicates that if the value is significant, then the current model fits the data better than a model specified with k–1 classes.

(Class 1), which accounted for 72.8% of all adolescent boys in our sample. Other groups in descending order in terms of its prevalence were football players (9.0%), bicycle riders (7.5%), soccer players (5.8%), and walkers (4.7%). The boys in Class 1 had higher mean proportions of playing basketball and running. Compared with boys in other classes, the boys in Class 1 had small but still the highest mean proportions of doing other leisure-time activities such as swimming, weight lifting, jogging,

dancing, etc. In addition to the dominant activities for these groups, Table 4 also presents the mean proportions of engaging in the other 9 types of activities by adolescent boys in each class.

Table 5 further presents the correlates of natural groupings for boys. Relative to being basketball players and runners (Class 1), every 1-year increase in age among boys was associated with reduced odds of being football players, bicycle riders, and soccer players and higher odds

Table 2 Top 10 Common Physical Activities for US Adolescents Age 12–19 Years, by Gender, Ranked by Frequency of Activities Reported and by Calculated Energy Expenditure From Each Activity, NHANES 1999–2006

Boys Girls

Frequency Energy

expenditure Frequency Energy expenditure

Activity Rank % Rank % Activity Rank % Rank %

Basketball 1 12.9 1 19.6 Running 1 11.9 3 12.6

Running 2 11.3 2 11.5 Walking 2 11.0 2 12.7

Football 3 7.8 3 9.1 Dance 3 10.5 1 12.9

Bicycling 4 7.2 4 7.6 Basketball 4 7.0 4 8.2

Walking 5 5.9 6 5.4 Jogging 5 6.0 8 4.4

Weight lifting 6 5.9 8 4.3 Bicycling 6 5. 7 6 4.9

Jogging 7 5.1 9 2.9 Aerobics 7 4.9 5 5.8

Swimming 8 4.3 7 4.3 Swimming 8 4.6 9 4.3

Soccer 9 4.0 5 5.5 Volleyball 9 4.0 7 4.5

Dance 10 3.0 10 2.2 Soccer 10 3.7 10 4.1

Natural Groupings of Common Physical Activities 475

of being walkers. Compared with white boys, Hispanic boys had higher odds of playing soccer [adjusted odds ratio (AOR): 2.2]. Black boys had higher odds of being in football player group (AOR: 1.5), whereas the black boys had lower odds of being bicycle riders (AOR: 0.4), soccer players (AOR: 0.3), and walkers (AOR: 0.5) than white boys. Overweight boys were more likely to be in football player group (AOR: 1.8) and had lower odds of being in soccer player group. Boys living in the house- holds with less than high school education for its head of household had 1.9 times higher odds of playing soccer than those with higher education. Boys living in the South had twice higher odds of being in bicycle riders. Football

Table 4 Mean Proportion of Energy Expenditure From Each Activity to the Total Energy Expenditure by Natural Groupings of Moderate-to-Vigorous Physical Activities, NHANES 1999–2006, Mean (s.e.)

Boys All (%)

Class 1: Class 2: Class 3: Class 4: Class 5:

Basketball players & runners

Football players

Bicycle riders

Soccer players Walkers

% of boys in each class 72.8% 9.0% 7.5% 5.8% 4.7%

Basketball 19.6 24.9 (0.8) 6.6 (0.9) 4.5 (0.9) 6.1 (1.2) 4.4 (1.5)

Bicycling 7.6 2.6 (0.2) 2.1 (0. 5) 69.9 (2.9) 2.7 (0.7) 2.4 (0.9)

Dance 2.2 2.7 (0.3) 0.4 (0.1) 0.5 (0.4) 0.9 (0.2) 1.9 (0.8)

Football 9.1 3.1 (0.2) 71.4 (1.9) 2.7 (0.7) 2.1 (0.7) 1.3 (0.6)

Jogging 2.9 3.3 (0.3) 2.1 (0.6) 1.3 (0.5) 1.0 (0.2) 1.4 (0.6)

Running 11.5 13.9 (0.7) 6.2 (0.9) 3.9 (0.9) 5.0 (0.9) 3.9 (1.2)

Soccer 5.5 1.3 (0.1) 1.1 (0.4) 1.0 (0.5) 73.6 (2.0) 1.1 (0.5)

Swimming 4.3 5.3 (0.5) 1.1 (0.4) 2.6 (1.0) 1.3 (0.6) 0.9 (0.3)

Walking 5.4 2.2 (0.2) 1.1 (0.3) 2.6 (0.9) 0.5 (0.1) 75.2 (3.0)

Weight lifting 4.3 5.3 (0.4) 2.5 (0.5) 1.4 (0.5) 1.0 (0.3) 1.6 (0.8)

Girls All (%)

Class 1: Class 2: Class 3: Class 4: Class 5:

Dancers, walkers, & joggers

Aerobic exercisers Swimmers

Volleyball players

Soccer players

% of girls in each class 79.0% 6.1% 5.6% 4.9% 4.2%

Aerobics 5.8 1.6 (0.2) 72.6 (2.6) 1.0 (0.5) 1.5 (0.7) 0.4 (0.3)

Basketball 8.2 9.6 (0.5) 2.1 (0.7) 2.3 (0.7) 4.0 (1.1) 2.9 (0.9)

Bicycling 4.9 5.8 (0.5) 1.3 (0.5) 3.1 (0.7) 0.4 (0.2) 1.0 (0.5)

Dance 12.9 15.4 (0.8) 5.4 (1.3) 3.4 (1.2) 1.7 (0.5) 1.6 (0.5)

Jogging 4.4 5.1 (0.4) 1.4 (0.4) 1.0 (0.4) 2.1 (0.7) 2.2 (0.7)

Running 12.6 14.8 (0.7) 2.2 (0.6) 3.3 (0.8) 4.9 (0.9) 8.1 (1.7)

Soccer 4.1 1.1 (0.1) 0.1 (0.1) 0.8 (0.4) 1.9 (0.7) 72.9 (3.6)

Swimming 4.3 1.5 (0.2) 1.8 (0.8) 69.4 (3.4) 1.2 (0.7) 0.7 (0.6)

Volleyball 4.5 1.1 (0.1) 1.0 (0.6) 1.2 (0.4) 70.5 (3.4) 2.1 (0.9)

Walking 12.7 15.0 (0.8) 4.4 (0.9) 4.9 (1.2) 3.3 (0.8) 1.9 (0.5)

player group had higher odds of being interviewed in the spring, summer and autumn instead of winter.

For girls, the 5 natural groupings in descending order by prevalence were dancers/walkers/runners (Class 1, 79.0%), aerobic exercisers (6.1%), swimmers (5.6%), volleyball players (4.9%), and soccer players (4.2%) (Table 4). Table 6 presents the correlates of the natural groupings for girls. First, every 1-year increase in age was associated with increased odds of being aerobic exercisers (AOR: 1.42) and reduced odds of being soccer players (AOR: 0.87). Hispanic girls had lower odds of being in the swimmer group than white girls (AOR: 0.4). Black girls had lower odds of being in swimmer and soccer

476

Table 5 Multinomial Logistic Regression Odds Ratio (95% Confidence Intervals) for Correlates in Determining Final Natural Groupings of MVPA Among Boys

Class 2 Class 3 Class 4 Class 5

Football players Bicycle riders Soccer players Walkers

Boys OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI)

Age (1-year increase) 0.8 (0.74–0.85)* 0.82 (0.75–0.9)* 0.83 (0.76–0.91)* 1.18 (1.05–1.32)*

Race/ethnicity

Non-Hispanic white Reference Reference Reference Reference

Hispanic 0.77 (0.50–1.19) 0.70 (0.38–1.29) 2.21 (1.47–3.32)* 0.58 (0.26–1.29)

Non-Hispanic black 1.54 (1.10–2.16)* 0.37 (0.24–0.58)* 0.27 (0.15–0.47)* 0.51 (0.30–0.88)*

Non-Hispanic other 0.82 (0.38–1.76) 0.57 (0.19–1.70) 0.69 (0.27–1.76) 1.03 (0.46–2.34)

Child’s health

Good/fair/poor Reference Reference Reference Reference

Very good 1.19 (0.79–1.78) 0.97 (0.60–1.58) 1.10 (0.74–1.64) 0.76 (0.44–1.31)

Excellent 1.03 (0.67–1.56) 0.86 (0.52–1.44) 1.03 (0.65–1.62) 0.69 (0.41–1.19)

Child’s weight status

Overweightb 1.78 (1.29–2.47)* 1.08 (0.73–1.60) 0.65 (0.44–0.96)* 1.40 (0.89–2.19)

Not overweight Reference Reference Reference Reference

Household poverty

≤ 130% FPLc Reference Reference Reference Reference

131–185% FPL 1.02 (0.60–1.74) 1.41 (0.75–2.65) 0.51 (0.29–0.92)* 0.78 (0.35–1.74)

186–250% FPL 1.22 (0.69–2.17) 1.22 (0.64–2.33) 0.47 (0.22–1.01) 0.66 (0.29–1.52)

≥ 251% FPL 1.12 (0.74–1.69) 0.67 (0.38–1.16) 0.92 (0.58–1.46) 0.52 (0.27–1.00)

Missing 1.29 (0.68–2.42) 0.97 (0.47–1.98) 0.91 (0.44–1.88) 0.95 (0.45–2.03)

Reference person’s education

< 12 years 1.42 (0.92–2.20) 1.38 (0.83–2.29) 1.88 (1.17–3.01)* 1.17 (0.65–2.12)

= 12 years Reference Reference Reference Reference

> 12 years 1.07 (0.73–1.56) 0.80 (0.50–1.28) 1.04 (0.64–1.71) 0.79 (0.46–1.36)

Region

Northeast 0.70 (0.39–1.27) 1.19 (0.53–2.65) 0.71 (0.30–1.69) 0.74 (0.29–1.85)

Midwest 1.03 (0.70–1.53) 1.40 (0.85–2.31) 0.77 (0.46–1.28) 0.89 (0.51–1.53)

South 0.81 (0.47–1.40) 2.00 (1.14–3.52)* 1.05 (0.54–2.05) 0.70 (0.34–1.43)

West Reference Reference Reference Reference

Residence

Rural 1.08 (0.68–1.71) 1.10 (0.69–1.76) 0.71 (0.42–1.20) 0.69 (0.38–1.27)

Urban Reference Reference Reference Reference

Season of interview

Spring Reference Reference Reference Reference

Summer 2.85 (1.57–5.20)* 1.29 (0.81–2.06) 1.48 (0.80–2.75) 1.68 (0.93–3.04)

Autumn 7.28 (4.43–11.95)* 0.93 (0.58–1.49) 1.42 (0.82–2.45) 1.14 (0.65–1.98)

Winter 2.72 (1.52–4.86)* 0.75 (0.37–1.53) 1.78 (0.90–3.51) 0.95 (0.47–1.92)

Note. Class 1 (basketball players and runners) is reference.

Abbreviations: MVPA, moderate-to-vigorous physical activity; FPL, federal poverty level.

477

Table 6 Multinomial Logistic Regression Odds Ratio (95% Confidence Intervals) for Correlates in Determining Final Natural Groupings of MVPA Among Girls

Class 2 Class 3 Class 4 Class 5

Aerobic exercisers Swimmers Volleyball players Soccer players

Girls OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI)

Age (1-year increase) 1.42 (1.28–1.57)* 0.97 (0.87–1.08) 0.92 (0.83–1.02) 0.87 (0.78–0.96)*

Race/ethnicity

Non-Hispanic white Reference Reference Reference Reference

Hispanic 0.66 (0.37–1.21) 0.37 (0.17–0.80)* 1.07 (0.62–1.86) 1.35 (0.77–2.35)

Non-Hispanic black 1.01 (0.64–1.60) 0.37 (0.22–0.63)* 0.64 (0.39–1.05) 0.37 (0.19–0.71)*

Non-Hispanic other 1.65 (0.75–3.63) 0.37 (0.13–1.05) 0.39 (0.10–1.52) 0.74 (0.29–1.91)

Child’s health

Good/fair/poor Reference Reference Reference Reference

Very good 1.74 (1.02–2.96)* 0.73 (0.38–1.41) 1.07 (0.63–1.80) 1.42 (0.82–2.44)

Excellent 1.32 (0.77–2.27) 1.28 (0.74–2.21) 0.92 (0.53–1.61) 1.20 (0.68–2.15)

Child’s weight status

Overweight 1.23 (0.77–1.96) 1.27 (0.80–2.02) 1.08 (0.68–1.71) 1.14 (0.71–1.84)

Not overweight Reference Reference Reference Reference

Household poverty

≤ 130% FPL Reference Reference Reference Reference

131–185% FPL 0.29 (0.15–0.58)* 1.55 (0.63–3.83) 0.43 (0.16–1.18) 1.55 (0.73–3.27)

186–250% FPL 1.11 (0.53–2.32) 0.73 (0.29–1.80) 1.62 (0.78–3.37) 1.29 (0.58–2.87)

≥ 251% FPL 1.03 (0.62–1.73) 1.31 (0.65–2.68) 1.60 (0.97–2.65) 1.41 (0.76–2.62)

Missing 0.86 (0.34–2.15) 2.28 (0.97–5.40) 0.70 (0.16–3.03) 2.30 (0.86–6.16)

Reference person’s education

< 12 years 0.79 (0.42–1.5) 0.73 (0.32–1.69) 1.17 (0.65–2.11) 1.03 (0.57–1.85)

= 12 years Reference Reference Reference Reference

> 12 years 0.92 (0.57–1.49) 1.12 (0.68–1.84) 0.82 (0.47–1.43) 0.91 (0.52–1.58)

Region

Northeast 1.18 (0.60–2.35) 0.58 (0.26–1.31) 0.50 (0.20–1.22) 0.98 (0.35–2.78)

Midwest 0.86 (0.50–1.46) 1.04 (0.62–1.73) 0.39 (0.19–0.81)* 1.24 (0.68–2.25)

South 0.97 (0.51–1.86) 0.55 (0.27–1.11) 1.27 (0.67–2.42) 0.92 (0.43–1.96)

West Reference Reference Reference Reference

Residence

Rural 0.98 (0.56–1.70) 1.61 (0.98–2.66) 0.94 (0.54–1.63) 1.22 (0.65–2.30)

Urban Reference Reference Reference Reference

Season of interview

Spring Reference Reference Reference Reference

Summer 0.72 (0.35–1.48) 4.40 (2.62–7.38)* 1.09 (0.52–2.28) 1.15 (0.50–2.66)

Autumn 1.11 (0.65–1.91) 1.17 (0.62–2.23) 1.19 (0.64–2.21) 1.87 (0.94–3.70)

Winter 1.89 (1.10–3.26)* 0.67 (0.26–1.71) 0.73 (0.36–1.50) 2.08 (0.98–4.38)

Note. Class 1 (Dancers, walkers & runners) is reference.

Abbreviations: MVPA, moderate-to-vigorous physical activity; FPL, federal poverty level.

478 Liu et al

player groups. Girls living in poor households (131% to 185% FPL) were less likely to be in the aerobic exerciser group (AOR: 0.3). Girls living in the Midwest had lower odds of being in the volleyball player group than those living in the West (AOR: 0.39). Finally, girls interviewed in the winter had higher odds of being aerobic exercisers (AOR: 1.9) and those interviewed in the summer had 4.4 times higher odds of being in swimmer group instead of dancers / walkers / runners group.

Discussion and Conclusions Different from existing studies which examined the types of common or preferred physical activities among boys and girls,9–17 our study provides unique perspectives on the natural groupings of physical activities among US adolescents. Among boys, more than 70% engaged in a group of activities anchored by basketball (24.9%) and running (13.9%), while other activities contributed to lower proportions of energy expenditure. The second nat- ural grouping among boys was characterized by football players (9.0% of all boys), followed by groups in which bicycle riders, soccer players, and walkers, respectively, were most common. Among girls, nearly 80% fell into the group characterized as dancers/walkers/runners. These findings suggest that future exercise programs might be more effective if they can tailor groups of naturally emerging patterns of activities as found in this study, such as dancing/walking/running for girls.29 Targeting multiple activities also has the potential of increasing total physical activity while maintaining it over time.

Consistent with existing literature,10 we found that minority adolescents reported different types of pre- ferred activities than their white counterparts. Hispanic boys liked playing soccer more than white boys. Black boys were more interested in playing football and less interested in bicycle riding, soccer playing, and walk- ing. Both Hispanic and black girls were less interested in swimming than white girls, and black girls were also less interested in soccer. As boys grow older, they were less likely to engage in group activities while they were more likely to do individual activities such as walking. For girls, older age was associated with increased interest in aerobic exercise and decreased interest in soccer play- ing. After controlling for other individual characteristics, differences associated with family socioeconomic status were less striking than we anticipated. The Southern boys reported riding bicycles more than those living in the west. Girls living in the Midwest were less interested in volleyball playing than those living in the West. Finally, seasonality did play a role in the selection of leisure-time activities, which is not surprising to us. For example, girls interviewed in winter were more likely to report aerobic exercise, while those interviewed in summer had higher odds of reporting swimming. All those findings suggest that effective interventions in ethnically diverse popula- tions should pay attention to the preferred activities for the racial groups and its differences by age, geographic

region, and seasonality. Our study specifically provided evidence that multiple subgroups exist within a popula- tion that exhibit different patterns of participation in physical activities. This information is important in the context of designing physical activity interventions, in that a “one size fits all” approach to the selection of activities is unlikely to appeal to the majority of the sample.

We found that activities among American adoles- cent boys and girls conformed to common sex-typed activities, which are consistent with existing literature to certain degrees. For instance, team sports were more frequent among boys than girls (eg, basketball, foot- ball).12,17 Adolescent boys liked basketball, running, football, and weight lifting.11,12,17 Girls liked running, dancing, walking, basketball, bicycling, swimming, etc.11–14,16,17,30 When you considered the total energy expenditure for each activity, the ranking for top 4 activities were consistent with the ranking based on the frequency only. However, soccer playing as a vigorous activity increased its ranking from #9 in its frequency to #5 in terms of its contribution to energy expenditure. Our list of common activities for boys were similar to that of Harrell and colleagues11 using the data from 5 middle schools in 3 rural counties in North Carolina, except that Harrell et al identified baseball as the fifth popular activity. However, our top 5 activities for girls were different from the top 5 activities for girls found by Harrell (eg, talking, running, walking, bicycling and dancing). The differences might indicate the differences in study population, geographic region, and the types of questions asked in the survey.

The Task Force on Community Preventive Services recommended school-based physical education as 1 inter- vention strategy for increasing physical activity among adolescents.31 Thus, this study provides useful ideas and information for various school-based physical education programs, programs occurring in the after school hours, and intervention programs about the approach to offer- ing a wide range of activities and that some youth may be more inclined to participate in a given activity due to numerous factors (eg, gender, race, age, season, geo- graphic locations, etc.). Thus, interventions should focus on providing access to, and participation in, a continu- ously changing and adaptive menu of physical activities that appeal to boys and girls. To choose the preferred activities for adolescents with certain characteristics will likely make the exercise an enjoyable experiences for them, thus it will improve the quality and effectiveness of these programs. Ultimately this careful tailoring of preferred activities will help them establish and maintain physically active lifestyles throughout adolescence and into adulthood.

Acknowledgments

This work was funded by a grant award from the Office of Rural Health Policy, Health Resources and Services Administration (No 6 U1CRH03711-06-00). Liu had full access to all the data

Natural Groupings of Common Physical Activities 479

in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. The results were previ- ously presented at the 43rd Annual Meeting of the Society of Epidemiologic Research as a poster in Seattle, Washington in June 2010.

References 1. Strong WB, Malina RM, Blimkie CJ, et al. Evi-

dence based physical activity for school-age youth. J Pediatr. 2005;146:732–737. PubMed doi:10.1016/j. jpeds.2005.01.055

2. US Department of Health and Human Services. Healthy People 2020. Washington, DC: US Government Printing Office 2010.

3. U.S. Department of Health and Human Services. 2008 Physical Activity Guidelines for Americans. Available at http://www.health.gov/paguidelines/guidelines/default. aspx

4. Van Der Horst K, Paw MJ, Twisk JW, et al. A brief review on correlates of physical activity and sedentari- ness in youth. Med Sci Sports Exerc. 2007;39:1241–1250. PubMed doi:10.1249/mss.0b013e318059bf35

5. Ferreira I, van der Horst K, Wendel-Vos W, et al. Envi- ronmental correlates of physical activity in youth—a review and update. Obes Rev. 2007;8:129–154. PubMed doi:10.1111/j.1467-789X.2006.00264.x

6. Sallis JF, Prochaska JJ, Taylor WC. A review of correlates of physical activity of children and adolescents. Med Sci Sports Exerc. 2000;32:963–975. PubMed

7. Ward DS, Dowda M, Trost SG, et al. Physical activ- ity correlates in adolescent girls who differ by weight status. Obesity (Silver Spring). 2006;14:97–105. PubMed doi:10.1038/oby.2006.12

8. Webber LS, Catellier DJ, Lytle LA, et al. Promoting physical activity in middle school girls: trial of Activity for Adolescent Girls. Am J Prev Med. 2008;34:173–184. PubMed doi:10.1016/j.amepre.2007.11.018

9. Bradley CB, McMurray RG, Harrell JS, et al. Changes in common activities of 3rd through 10th graders: the CHIC study. Med Sci Sports Exerc. 2000;32:2071–2078. PubMed doi:10.1097/00005768-200012000-00017

10. Dowda M, Pate RR, Felton GM, et al. Physical activities and sedentary pursuits in African American and Caucasian girls. Res Q Exerc Sport. 2004;75:352–360. PubMed doi :10.1080/02701367.2004.10609168

11. Harrell JS, Pearce PF, Markland ET, et al. Assessing physi- cal activity in adolescents: common activities of children in 6th-8th grades. J Am Acad Nurse Pract. 2003;15:170–178. PubMed doi:10.1111/j.1745-7599.2003.tb00259.x

12. Dowda M, Saunders RP, Hastings L, et al. Physical activ- ity and sedentary pursuits of children living in residential children’s homes. J Phys Act Health. 2009;6:195–202. PubMed

13. Kuo J, Schmitz KH, Evenson KR, et al. Physical and social contexts of physical activities among adolescent girls. J Phys Act Health. 2009;6:144–152. PubMed

14. Grieser M, Vu MB, Bedimo-Rung AL, et al. Physical activ- ity attitudes, preferences, and practices in African Ameri- can, Hispanic, and Caucasian girls. Health Educ Behav. 2006;33:40–51. PubMed doi:10.1177/1090198105282416

15. Reeder AI, Stanton WR, Langley JD, et al. Adolescents’ sporting and leisure time physical activities during their 15th year. Can J Sport Sci. 1991;16:308–315. PubMed

16. Barr-Anderson DJ, Young DR, Sallis JF, et al. Struc- tured physical activity and psychosocial correlates in middle-school girls. Prev Med. 2007;44:404–409. PubMed doi:10.1016/j.ypmed.2007.02.012

17. Wilson DK, Williams J, Evans A, et al. Brief report: a qualitative study of gender preferences and motivational factors for physical activity in underserved adolescents. J Pediatr Psychol. 2005;30:293–297. PubMed doi:10.1093/ jpepsy/jsi039

18. National Center for Health Statistics. About National Health and Nutrition Examination Survey.

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. 2000;32:S498–S504. PubMed doi:10.1097/00005768-200009001-00009

20. McCutcheon AL. Latent class analysis. Beverly Hills, CA: Sage Publications 1987.

21. Muthen L, Muthen B. Mplus User’s Guide, Release 3.5. Los Angeles: Muthen & Muthen 2004.

22. Muthen B, Muthen LK. Integrating person-centered and variable-centered analyses: growth mixture modeling with latent trajectory classes. Alcohol Clin Exp Res. 2000;24:882–891. PubMed doi:10.1111/j.1530-0277.2000. tb02070.x

23. Beets MW, Foley JT. Comparison of 3 different analytic approaches for determining risk-related active and seden- tary behavioral patterns in adolescents. J Phys Act Health. 2010;7:381–392. PubMed

24. Lo Y, Mendell NR, Rubin DB. Testing the number of com- ponents in a normal mixture. Biometrika. 2001;88:767– 778. doi:10.1093/biomet/88.3.767

25. Liu J, Bennett KJ, Harun N, et al. Urban-rural differences in overweight status and physical inactivity among US children aged 10-17 years. J Rural Health. 2008;24:407– 415. PubMed doi:10.1111/j.1748-0361.2008.00188.x

26. Springer AE, Hoelscher DM, Castrucci B, et al. Prevalence of physical activity and sedentary behaviors by metropoli- tan status in 4th-, 8th-, and 11th-grade students in Texas, 2004-2005. Prev Chronic Dis. 2009;6:A21. PubMed

27. Hart LG, Larson EH, Lishner DM. Rural defini- tions for health policy and research. Am J Public Health. 2005;95:1149–1155. PubMed doi:10.2105/ AJPH.2004.042432

28. Economic Research Service. Measuring Rurality: Rural- Urban Commuting Area Codes. United States Department of Agriculture: 2005.

29. Collins LM, Murphy SA, Bierman KL. A concep- tual framework for adaptive preventive interventions. Prev Sci. 2004;5:185–196. PubMed doi:10.1023/ B:PREV.0000037641.26017.00

30. O’Neill JR, Pate RR, Liese AD. Descriptive epidemiology of dance participation in adolescents. Res Q Exerc Sport. 2011;82:373–380. PubMed doi:10.1080/02701367.2011 .10599769

31. Lee SM, Burgeson CR, Fulton JE, et al. Physical education and physical activity: results from the School Health Poli- cies and Programs Study 2006. J Sch Health. 2007;77:435– 463. PubMed doi:10.1111/j.1746-1561.2007.00229.x

Copyright of Journal of Physical Activity & Health is the property of Human Kinetics Publishers, Inc. and its

content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's

express written permission. However, users may print, download, or email articles for individual use.