Leisure-time physical activity and associated factors among adolescents of Pernambuco , Brazil : From 2006 to 2011

Leisure-time physical activity (LTPA) has important implications in promoting physical activity among young people. The present study aimed to analyze variations occurred in the 2006-2011 period in LTPA and associated factors among high school boys and girls in the state of Pernambuco, Northeastern Brazil. A cross-sectional schoolbased and statewide survey was conducted in 2006 (n = 4,207) and 2011 (n = 6,264), and samples of adolescents (14-19 years) were compared. Data were collected through a questionnaire. The practice of LTPA was obtained by the question “Do you perform regularly, some kind of physical activity in your free time?” Poisson regression was used (confidence interval (CI) 95%) in crude and adjusted analyses stratified by gender. From 2006 to 2011, there was stability in the proportion of adolescents practicing LTPA both among boys (77.5% versus 78.9%) and girls (51.2%, versus 54.0%). The weekly frequency of physical education classes was directly associated with LTPA both in girls and boys, regardless of the survey year. Age, marital status, grade and computer use were significantly associated with LTPA, but there were distinctions between genders and surveys. In conclusion, there was a temporal variation between 2006 and 2011, but the reduction or increment in LTPA practice varied according to stratification of subgroups of adolescents. In general, physical education class was a factor that remained associated with LTPA in the investigated period.


INTRODUCTION
Regular physical activity provides many benefits to the health of adolescents both in the short and long term.Increased cardiorespiratory fitness and muscle strength, weight control and reduced risk of metabolic diseases, depression symptoms and mortality rates from cardiovascular disease and diabetes are the main benefits 1 .Although many health problems do not manifest at early age 2 , current recommendations from the World Health Organization 1 and Center for Disease Control and Prevention 3 emphasize the need for an active lifestyle in all cycles of life, including childhood and adolescence.
In adolescence, leisure-time physical activities appear to be especially important, with positive association with attention span and inversely related to the occurrence of insomnia and depression 4 .However, this concern with the options of activities available and choices of young people in leisure time results from the set of observational studies that showed high percentages of physical inactivity and preference for sedentary leisure activities in this population 5 .In addition, it turns out that girls show higher prevalence of sedentary leisure habits compared to boys 6 .In addition, the stability of physically active behaviors from adolescence to adulthood seems to be dependent on intrinsic and extrinsic factors that are more easily observed in leisure time than in other domains such as domestic tasks and work 7 .
Monitoring the practice of leisure-time physical activity (LTPA) occurs more frequently in high-income countries where trends observed in these contexts are different from those in low-and middle-income countries 8 and between 9 .In Brazil, epidemiological surveillance of this behavior is still incipient.For example, the National Survey of Students' Health 10 includes only students from the ninth grade of elementary school of Brazilian state capitals and is not focused on LTPA.However, the assessment of changes and population subgroups most likely to reduce LTPA can contribute to the development of public actions to stimulate PA among young people, especially in leisure time, but there are few studies on time trends of physical activity in young people, specifically addressing the leisure domain [9][10][11] , and if any, are focused on adolescent populations from the Southern and Southeastern regions of Brazil 12 .
In 2001 and 2006, the first Brazilian cross-sectional school-based and statewide surveys were conducted in the states of Santa Catarina 13 and Pernambuco 14 , respectively, in order to analyze the practice of PA in high school students.In 2011, these surveys were reapplied in their respective target populations, enabling the study of changes in the prevalence and factors associated with health behaviors in these regions.The present study aimed to analyze separately for boys and girls, the temporal variations occurred in five years (2006-2011) in the practice of LTPA and factors associated with these variations in high school students in the state of Pernambuco.

Planning and sample selection
The sample design including the sample size and the strategy adopted to select participants followed similar methodological procedures.However, the parameters used to establish the minimum sample size were different, and in 2006, a larger sample design effect value was adopted (deff = 4.0 in 2006; deff = 2.0 in 2011), while that in the survey conducted in 2011, lower maximum tolerable error was adopted (3% in 2006 and 2% in 2011).These adjustments resulted in minimum sample sizes of 4,217 and 4,770 in 2006 and 2011, respectively.In 2006, the parameters and procedures adopted for sample designing are described in previous publications 15 , while compared to the survey conducted in 2011, these were similar to those reported by Silva et al. 16 .
All public schools of Pernambuco were considered eligible for the study and were arbitrarily classified into three levels: small size (schools with less than 200 students), midsize (200-499 students) and large (≥500 students).
The sample selection was performed by two stages.In the first, schools were selected (primary sampling unit) and secondly the classes (secondary sampling unit).In the first stage, the proportionality criterion by region and size was adopted to make the draw of schools where data collection would be performed, defining, therefore, that a school would be selected from the stratum with lower proportional representation.
Then, considering the average number of students per high school class, the number of classes that should be drawn was defined, so that it reached the minimum sample size.In 2006, to reach the sample size previously defined, 234 high school classes were randomly selected 14 , while in 2011 323 were randomly selected.The draw of classes was random, considering the proportionality of classes per shift (day and night).All students in selected classes were invited to participate in the study, excluding those older than the target age group (14-19 years).

Instrument
Data were collected through an adapted version of the questionnaire "Global School-based Student Health Survey" proposed by the World Health Organization 17 .Prior to carrying out both surveys, pilot studies (test-retest after a week) were performed to test the instrument.Data for the pilot study in 2006 were collected in two public schools of Recife, with a sample of 138 adolescents aged 14-19 years (59 girls).The 2011 pilot study was conducted with 86 adolescents in the same age group in a school of the State Public Network located in the Metropolitan Region of Recife.Kappa concordance indexes ranged from 0.52 to 1.00 in 2006 and 0.63 to 0.98 in 2011.Data collection was conducted by previously trained researchers (theory and practice) and the questionnaire was applied to students in classroom, without the presence of teachers.
The practice of LTPA was assessed using the following question: Do you perform regularly some type of physical activity in your free time, like exercise, sports, dance or martial arts (yes/no).Adolescents who answered positively were considered physically active during leisure time.Demographic (gender, age and marital status) and economic variables (maternal education, housing zone, occupational status, ownership of computer and geographic mesoregion) and those related to school (grade, shift and participation in physical education classes) were collected and categorized as shown in Table 1.
Data tabulation was performed using the EpiData software, version 3.1.Typing was performed in duplicate on separate data files, which were compared ("check" function) in order to identify and correct typing errors.After cleaning data files of 2006 and 2011, these were stored together in a single file so that analyses proposed in this study could be performed.

Data analysis
Absolute (n) and relative (%) frequencies and confidence intervals of 95% (CI 95%) were described for LTPA.The differences between proportions were calculated in percentage deltas.The proportions of students physically active during leisure time in both surveys were compared using the Pearson chi-square test.Poisson regression, with crude and adjusted analysis was used to identify factors associated with LTPA.Independent variables were organized into three hierarchical levels based on the model of Dumith et al. 18 .Level 1 included demographic variables (geographical mesoregion, housing zone, age and marital status).Socioeconomic variables (employment status, ownership of computer and maternal schooling) were considered at level 2. Level 3 included variables related to school (grade, school shift and physical education classes).The association of a variable with LTPA was controlled by the variables of the same level and previous hierarchical levels.
To control possible confounding factors, all variables were taken to multivariate analysis, regardless of the level of significance in the bivariate analysis and respecting their respective conceptual level.In the final analysis, 5% significance level was considered to identify variables associated with the outcome.In all analyses, procedures for studies with complex methodologies were adopted (sampling by conglomerates, multiple stages) incorporating to the syntax "svyset" prefix, feature available in STATA.All analyses were stratified by sex.

RESULTS
The percentage of participation in the study was 98.1% and 95.7% in 2006 and 2011, respectively.In 2006, 6,112 students in selected classes attended the school during the period of data collection and were invited to participate in the study, but 83 students refused to participate; 1,819 questionnaires were excluded from the study because students were older than 19, and the other 03 for being under 14 years.Thus, the final sample was composed of 4,207 adolescents.In 2011, they 7,467 students of selected classes attended school in the data collection period, but 282 students refused to participate; 930 questionnaires were excluded because students aged above 19 years, resulting in a final sample of 6,264 students.The final sample was slightly lower in 2006 and higher than the sample dimension previously established in 2011.Table 1 shows the demographic and economic characteristics and those related to school of participants in both surveys.
In both surveys, there was a higher proportion of single female students aged 16-17 years who reported not working and living in urban areas.It was also observed that, from 2006 to 2011, there was a rise in the proportion of adolescents who had computer (from 10.7% to 41.1%; p <0.001), students in the morning shift (from 57.5% to 72.0%; p <0.001) and those who participated in physical education classes (35.1% to 74.1%; p <0.001).
Overall, comparing the findings of both surveys, no statistically significant changes were observed in the proportion of adolescents physically active during leisure, both among boys (77.5% in 2006, CI 95%: 75.5-79.6;and 78.9% in 2011, CI 95%: 77.2-80.6)and among girls (51.2% in 2006, CI 95%: 48.7-53.6, and 54.0% in 2011, CI 95%: 51.8-56.2).However, it was found that among students aged 16-17 years and those living in rural areas, there was an increase in the proportion of leisure-time physically active students.In contrast, the prevalence of leisure-time physically active students decreased among adolescents who participated in two or more weekly physical education classes (Table 2).Especially among girls, the proportion of leisure-time physically active students increased from 2006 to 2011 among those not working, among those who reported not having computer, among those whose mothers have not completed elementary school, among those resident in the sertão of São Francisco, among students of the second grade of high school and among those of the evening shift (Table 2).In the adjusted analysis, an inverse association between LTPA and age was found among boys in 2011 but not in 2006.There was a direct association between LTPA and participation in physical education classes both in 2006 as in 2011.After adjustments, boys who had marital status different from single (other) were more likely to perform LTPA.The other variables were not statistically associated with LTPA among boys (p> 0.05), both in 2006 as in 2011 (Table 3).
In analysis adjusted for girls, it was found that high grades were associated with lower prevalence of LTPA practice in 2006 and 2011.The frequency of physical education classes had positive association with the practice of LTPA in both surveys.Finally, girls who had no computer were more likely to perform LTPA only in 2011.The remaining variables were not statistically associated with LTPA among women (p> 0.05), both in 2006 as in 2011 (Table 4).The increase in the practice of LTPA in turn, was observed in some subgroups of adolescents students.This may be due to the reduction of social inequalities that occurred from 2006 to 2011, especially because the state of Pernambuco presented a growth rate higher than that of Brazil as a whole 19 .In addition, this period was also marked by expansion of some public programs to encourage the practice of physical activity specifically focused on young populations, such as the Programa Segundo Tempo 20 and Programa Saúde na Escola 21 .
Girls continue to show lower prevalence of leisure-time physical activity compared to boys.The explanations for this gender difference have been explored in some studies 22,23 and seem to be mainly related to social and cultural factors.For example, girls are encouraged from childhood to become involved with less physically active recreational activities 23 .

... continues
The first study conducted in Brazil covering time trends of physical activity in adolescent students (10-19 years) was conducted in southern Brazil 24 .The results were similar to those found in this study; however, the proportion of students classified to be physically active during leisure time was very different between studies, being higher in this study (61.8% in 2006 versus 64.0% in 2011) than in the survey conducted by Coll et al. 24  Compared to the findings of international studies, the results of this study were similar to those found in American students in the period from 1991 to 2007 25 , but different from those observed among Catalan adolescents 26 .Among Catalan adolescents (10-17 years) who were followed from 1993 to 2003, the prevalence of leisure time physical activity increased from 66% to 78.1% among boys and from 45% to 52.8% among girls.
The trend in the reduction of prevalence of LTPA among students for greater participation in physical education classes was found between boys and girls from Pernambuco.This finding was somehow unexpected, as results of cross-sectional studies have shown positive relationship among variables 27,28 , and longitudinal and intervention studies have shown that participation in physical education classes can positively influence physical activity 29 .
It is noteworthy that the considerable increase in the participation of students in physical education classes in the state did not contribute, within five years, for the increase in the practice of LTPA among adolescents.So, it could not be said that this trend began with the action of the physical education teacher at school, providing physical and sporting activities, encouraging and guiding students for a physically active and healthy lifestyle, which would be ideal.On the contrary, perhaps traditional physical education classes (specific sports, practices with low motor variation) 30 could be a factor that discourages leisure-time physical activity, especially among high school students.This may also be partly explained by the entry of older adolescents into the labor market and / or their withdrawal from other activities to devote exclusively to technical training courses and / or preparation for entry into the university.
The main strengths of this study are the performance of extensive fieldwork with the use of previously tested tool and with data collection in samples sufficiently large for the proposed analyses.The methodological similarity of both surveys and their scope, including the participation of students of the night shift and those living in rural areas are also a strong point.The main limitation of the study was the use of a tool that only allows obtaining self-reported measures.

CONCLUSION
The findings indicate the occurrence of temporal variation in prevalence of LTPA in different subgroups of adolescents, with reduced LTPA among those who participate in two or more physical education classes and increased among residents in rural areas.The weekly frequency of physical education classes was directly associated with the practice of LTPA among girls and boys, both in 2006 as in 2011.Age, marital status, grade and the ownership of computer were also significantly associated with LTPA, but with distinctions between genders and surveys.
However, it is necessary to review the strategies of physical activity programs to adapt to the different ages of adolescence.Changes in the structure of the physical education classes seem to be essential in the LTPA promotion among adolescents.Interventions are needed to prevent the negative impact of urbanization and technological growth in the lifestyle of young people, especially the excessive use of computer / video games during leisure time.
This study compared data from two epidemiological, cross-sectional, school-based and statewide surveys called "Projeto Atitude".The first had data collection carried out from April to October 2006 and the second from May to November 2011.Both surveys were conducted in a target population of students (359,897 in 2006 and 367,813 in 2011) from state public schools of Pernambuco in the age group 14-19 years.This limitation is justified by the need to obtain data for this population in order to support the development of policies for public schools in the state of Pernambuco.In addition, students enrolled in public schools (668 in 2006 and 769 in 20110 represented 80.2% in 2006 and 85.6% in 2011 of the total number of high school students in the state of Pernambuco (448,653 in 2005 and 429,451 in 2010), while the rest were distributed as follows: 13.2% in 2005 and 11.2% in 2010 in private schools, 5.9% in 2005 and 1.3% in 2010 in municipal schools and 0.72% in 2005 and 1.82% in 2010 in federal schools.All requirements regarding ethical aspects were adopted with the research protocol of survey conducted in 2006 was approved by the Ethics Committee in Research with Human Beings of the "Agamenom Magalhães" Hospital (Recife, PE) while protocol of survey conducted in 2011 was approved by Ethics Research Committee of the University of Pernambuco (CAAE: 0158.0.097.000-10).

Table 1 .
Characteristics of samples of adolescents aged 14-19 years, high school students from the public education network of Pernambuco in 2006 and 2011, by sex.

Table 2 .
Comparison of prevalence in 2006 versus 2011 of leisure-time practice of physical activity among high school students of the Pernambuco according to demographic, economic variables and those related to school, by sex.

Table 3 .
Crude and adjusted analysis of leisure-time physical activity in 2006 * and 2011 **, among high school male students of Pernambuco according to demographic, economic variables and those related to school.

Table 4 .
Crude and adjusted analysis of leisure-time physical activity in 2006 * and 2011 **, among high school female students of Pernambuco according to demographic, economic variables and those related to school.
, 26.3 % in 2005 versus 28.1% in 2012.The National Survey of Students' Health from 2009 to 2012 10 does not bring data on time trends of PA, total or during leisure.