While the wide range of health benefits of physical activity is undisputed in both adults and children, there is a declining trend in physical activity, particularly in developed countries. This has obvious effects on efforts to counteract the obesity pandemic. According to the WHO 2004 report, at least 60 % of the world’s population fails to achieve the minimum recommendation of 30 min of moderate-intensity physical activity daily(1). Data from the USA indicate that 61·5 % of children aged 9–13 years do not participate in any organized physical activity during non-school hours and 22·6 % do not engage in any free-time (leisure-time) physical activity(2).
Based on the 1996 WHO report, the level of leisure-time physical activity in the Israeli population is low with about 14 % of the working population reporting participation in leisure-time physical activity more than once weekly(3). In a national survey of time utilization and lifestyle in Israel conducted in 1991–2, 20 % of the population aged 14 years and over participated in leisure-time activity more than once per week (24·6 % of males and 16·5 % of females), which was found to be higher among young people (38 %) than adults (19 %)(3). From a public health perspective it is important to determine the level of physical activity and its determinants in order to provide a theoretical basis for the development of appropriate policies and programmes to enhance health and prevent the many complications attending physical inactivity, especially obesity. Intervention programmes should begin as early as possible and information on health behaviours among children is essential. Since there are many cultural determinants to lifestyle, the Israeli population offers an instructive example on how these may affect health behaviours.
The Israel Ministry of Health performed a national survey among middle- and high-school children, to study health and nutrition status and health behaviours among 11–19-year-olds. Using these data, the present paper describes the sociodemographic distribution of optimal level of physical activity and explores the interrelationships between gender, age, socio-economic status (SES), lifestyle habits and anthropometric measurements (obesity) in relation to participation in physical activity.
Research methods and procedures
Survey population, design and sampling frame
The study population included 6274 youth enrolled in the MABAT Youth Survey (First Israeli National Health and Nutrition Survey in 7th–12th grade students, 2003–2004), a cross-sectional, nationally representative, school-based study of youth in grades 7 to 12 (ages 11–19 years). The survey was undertaken from April 2003 to March 2004. Data were collected from two school year groups such that the mean age of pupils was 13·4 years for middle schools (grades 7–9) and 16·0 years for high schools (grades 10–12). The MABAT Youth Survey included both Jewish and non-Jewish (93·4 % Arabs and Druse), religious and non-religious schools. Schools from the Haredi (ultra orthodox) sector were not included, and nor were independent, private or boarding schools.
Sample size
The sample size was calculated by assuming that 40 % of students consumed 30 % or more of their daily energy intake from fat, which would give statistically significant information with power of 80 % and allowing for a 95 % confidence interval. This 40 % prevalence was found to be similar to the prevalence of optimal level of physical activity, preserving the power of the test. Overall, a sample size of 369 pupils in each category (subgroup) was used. With an additional 20 % over-sampling, the country-wide sample size was 6760. The sampling procedures are described fully elsewhere(4).
Study variables
Type, frequency and duration of physical activity, and physical inactivity (television/video viewing or listening to music, talking on telephone, working on computer), smoking data, age and ethnicity were determined from self-reported data using self-administered questionnaires whereas level of SES was taken from the Ministry of Education classification of the welfare level of schools based on location. Height (with a coil-spring tape) and weight (with portable analogue scale) were measured by a trained person.
Composite variables
BMI
BMI was calculated by dividing weight by the square of height (kg/m2). For the purpose of comparison BMI was categorized based on the age- and sex-specific cut-off values of the Centers for Disease Control and Prevention (CDC)/National Center for Health Statistics 2000 growth charts. The categories were underweight (<5th percentile), normal (5th to 85th percentile), overweight (85th to 95th percentile) and obese (>95th percentile). For multivariate analysis, BMI was classified as obese (>95th percentile) and non-obese (≤95th percentile)(5).
Level of physical activity
Optimum level of physical activity was defined as any moderate and vigorous level of physical activity which adds up to 60 min/d for four or more days a week as is recommended for schoolchildren by the CDC(6). Light physical activities, i.e. walking, playing musical instruments and gardening, were excluded from the above definition.
MET calculations
Each type of activity was given a MET (metabolic energy equivalent task) score using the compendium of Ainsworth et al.(Reference Ainsworth, Haskell, Leon, Jacobs, Montoye, Sallis and Paffenbarger7). MET-h were computed by multiplying the MET score by the hours performed at each activity, which was again multiplied by the frequency of the activity to obtain MET-h/week. For children involved in more than one activity, the sum of each activity’s MET-h/week was taken as the total MET-h/week.
Classification of intensity of physical activity
A modified Borg perceived exertion scale(Reference Borg8) was used to classify the intensity of physical activity based on perceived difficulty of exercise: light (if the subject reported the level of difficulty of the exercise as ‘light’ in all activities); moderate (if the subject reported the level of difficulty of the activity as ‘moderate’ at least in one of the main activities); and vigorous (if the subject reported as the level of difficulty as ‘difficult’ in at least in one of the main activities).
Physical inactivity
Time spent on television and video viewing and listening to music was categorized as <2 h or ≥2 h, based on the recommendations of the American Academy of Pediatrics Committee on Public Education(9). Although there are no data on recommended duration of working at the computer and talking on the telephone, we classified these activities similar to that for watching television/videos.
Statistical analyses
All statistical analyses were performed using the Statistical Package for the Social Sciences statistical software package version 12 (SPSS Inc., Chicago, IL, USA). The χ 2 test and the crude odds ratio were used to compare prevalence rates of optimal levels of physical activity. Differences in mean MET-h/week values were determined using one-way pairwise ANOVA and the Scheffe post hoc procedure significant at the 5 % level. The nature of the MABAT questionnaire and data collection was such that the answers were best summarized as dichotomous distributions. This was because most of the answers were not continuous variables, but rather in four or five categories of frequency choice and not simple numerical answers. Examples of this are, for hours viewing television: up to half an hour/half an hour to two hours/more than two hours/not at all, and similarly for smoking and alcohol habits. In the logistic regression analyses, binary logistic regression was used to examine the association between SES, lifestyle habits, BMI classification and level of physical activity (Tables 3), with the level of physical activity acting as the dependent variable. Dummy variables were created to compute odds ratios for these factors. In all analyses, odds ratios and associated 95 % confidence intervals are presented for each level of the lifestyle variable in comparison with the referent level. For each independent variable the odds ratio was adjusted for all of the other variables included in the model.
Because lifestyle, anthropometric and physical activity data were missing from some subjects, odds ratios were determined for only 76·5 % of Jewish boys, 75·6 % of non-Jewish boys, 76·8 % of Jewish girls and 78·3 % of-non Jewish girls included in the survey. Age, school level, SES and BMI values of subjects with missing values were not significantly different from those with no missing values (P > 0·05).
Response rate
The response rate for the schools was 81·4 % for the Jewish schools and 96·4 % for Arab schools. From the Jewish sector 85·7 % of eligible students responded while 94·0 % took part in the survey from the Arab sector. A comparison of health parameters and risk behaviour carried out between the questionnaires of those present and those absent did not reveal any differences. There was no way of checking those students who were present but refused to participate. Differences between schools that agreed to participate and those that did not were also examined. In the Jewish sector, seventeen schools refused to take part. They were more likely to be from the state religious stream (59 % among those refusing, 48 % among schools surveyed). Similarly, the average welfare level among the schools refusing was 4, while among the participating schools the average was 6. Thus, schools that refused were generally of a higher socio-economic classification.
Validity of questionnaire
The MABAT youth questionnaires are similar, although not identical, to other self-reporting questionnaires that have been used and validated in other large-scale epidemiological studies among adults and adolescents(Reference Andersen, Crespo, Bartlett, Cheskin and Pratt10–Reference Pate, Heath, Dowda and Trost13). However, they have yet to be evaluated for use among younger children.
Results
A total of 6274 children were enrolled in the study, of whom 55·1 % (3458) were girls while 44·9 % (2816) were boys. Demographic, BMI and lifestyle characteristics of the participants are shown in Tables 1 and 2.
Table 1 Background characteristics and anthropometric measurements of study subjects, Israel, 2003–2004
†Number of observations for age: 1916 in Jewish boys, 2378 in Jewish girls, 773 in non-Jewish boys, 1031 in non-Jewish girls.
‡Number of observations for BMI: 1725 in Jewish boys, 2078 in Jewish girls, 821 in non-Jewish boys, 964 in non-Jewish girls.
Table 2 Physical and recreational activity of study subjects, Israel, 2003–2004
MET, metabolic energy equivalent task.
†Number of observations for MET index: 1669 in Jewish boys, 2042 in Jewish girls, 649 in non-Jewish boys, 859 in non-Jewish girls.
‡Number of observations for number of cigarettes smoked daily: 130 in Jewish boys, 130 in Jewish girls, 52 in non-Jewish boys, 14 in non-Jewish girls.
Among the Jewish students, 12·8 % of the boys and 12·0 % of girls were found to be overweight (BMI between 85th and 95th percentile) compared with 13·3 % of boys and 15·2 % of girls among the non-Jewish population. The prevalence of obesity was found to be higher among Jewish boys than Jewish girls (6·8 % v. 3·7 %), and was 9·3 % and 5·1 % among non-Jewish boys and girls, respectively.
The majority of Jewish boys (45·5 %) reported involvement in ball games (including tennis, football, basketball and volleyball) followed by aerobic activities (30·9 %) as their most common type of physical activity. However, most Jewish girls reported participating most commonly in either walking (31·7 %) or aerobic activities (36·2 %). Among non-Jews, more than 80 % of boys reported participating in either ball games or aerobic activities as their most frequent activity, while girls reported mainly aerobic activities (50·9 %) followed by walking (2·8 %). The difference in the type of most frequent physical activity by ethnicity (Jews and non-Jews) was found to be significant in both boys (χ 2 = 26·78, P = 0·001) and girls (χ 2 = 119·4, P < 0·001).
Ethnicity was also found to be associated with intensity of physical activity. Non-Jews were more involved in physical activity of vigorous intensity, while Jews were more involved in moderate-intensity physical activity, in both boys and girls (see Table 2). This difference was also found to be very significant (χ 2 = 121·07, P < 0·001 among boys; χ 2 = 99·96, P < 0·001 among girls).
These data indicate that 36 % and 57 % of Jewish girls and boys respectively, and 40 % and 58 % of non-Jewish girls and boys respectively, were optimally active. Boys were more likely to be optimally active among both Jews (OR = 2·30; 95 % CI 2·02, 2.63) and non-Jews (OR = 2·07; 95 % CI 1·69, 2·51, adjusted for age) compared with girls.
Crude and adjusted odd ratios for optimal level of physical activity by different characteristics are presented in Table 3. Such description might lead to over-adjustment because of multi-collinearity and some underestimation of the relationships found. Comparison with the crude ratios provides a basis for comparison of such possible effects.
Table 3 Crude and adjusted† odds ratios (95 % confidence interval) of being optimally physically active by gender, ethnicity and selected lifestyle factors among schoolchildren, Israel, 2003–2004
†Adjusted for all other variables included in the table.
Obese non-Jewish boys were found to be 1·82 (95 % CI 1·08, 3·06) times more likely to be active compared with non-obese non-Jewish boys, but this association disappeared when adjusted for all other variables. However, among Jewish and non-Jewish girls, no association was found. Running the regression equations with overweight (>85th percentile) did not affect the lack of relationship between overweight/obesity and physical activity (data not shown). Another assessment of activity was performed using MET-h/week. When mean MET-h/week values were compared in BMI categories using one-way ANOVA and the Scheffe post hoc procedure, there were no significant differences among boys or non-Jewish girls. However, there was a significant difference among Jewish girls wherein underweight girls (BMI < 5th percentile) were significantly less active than obese girls (BMI > 95th percentile) (Table 4).
Table 4 Mean MET-h/week in different BMI categories by sex and ethnicity among schoolchildren, Israel, 2003–2004
MET, metabolic energy equivalent task.
*Significantly different categories using the Scheffe post hoc test.
Lower level of parental education (high school completed or less) was associated with lower level of physical activity among Jewish girls, with an odds ratio of 0·69 (95 % CI 0·58, 0·84) at crude level when compared with children with higher father’s educational status. This association remained significant when adjusted for all other variables included in the logistic regression model (OR = 0·74; 95 % CI 0·58, 0·96). However, there was no association among Jewish boys or non-Jewish girls and boys. Similarly, lower maternal educational level was associated with a lower prevalence of optimal level of physical activity among both Jewish girls (OR = 0·80; 95 % CI 0·67, 0·96) and non-Jewish girls (OR = 0·50; 95 % CI 0·37, 0·68) at crude level. However, this association remained significant only among non-Jewish girls when adjusted for other variables included in the model (OR = 0·46; 95 % CI 0·31, 0·67). There was no association detected for boys.
Jewish boys who habitually slept for <6 h per night were found to be 0·56 (95 % CI 0·43, 0·73) times less likely to be optimally active compared with those who slept longer. This association was not found among the other sex and ethnic groupings.
Jewish and non-Jewish boys who smoked hookah were found to be 0·78 (95 % CI 0·61, 1·00) and 0·64 (95 % CI 0·43, 0·96) times less likely to be optimally active compared with their non hookah-smoking counterparts. However, this association disappeared among Jewish boys and just failed to be significant among non-Jewish boys (OR = 0·59; 95 % CI 0·35, 1·01) when adjusted for all other variables included in the logistic regression model.
History of alcohol consumption one month prior to interview was negatively associated with optimal level of physical activity among Jewish boys (OR = 0·82; 95 % CI 0·68, 0·99) only at the crude level.
Jewish boys from lower social status schools were found to be 1·29 (95 % CI 1·06, 1·56) times more likely to be optimally active than Jewish boys from higher SES schools, which stayed significant when adjusted for all other variables (OR = 1·30; 95 % CI 1·02, 1·65).
Middle-school students were more likely to be optimally active than high-school students among Jewish boys (OR = 1·36; 95 % CI 1·13, 1·66) and girls (OR = 1·53; 95 % CI 1·28, 1·84) and non-Jewish girls (OR = 1·39; 95 % CI 1·07, 1·82). However, this association stayed significant only among girls (OR = 1·53; 95 % CI 1·23, 1·91 among Jews, OR = 1·89; 95 % CI 1·36, 2·63 among non-Jews) when adjusted for all other variables included in the model.
However, being a cigarette smoker and watching television/videos or listening to music for ≥2 h/d had no significant effect on level of optimal physical activity (Table 3).
Discussion
Sociodemographic characteristics and physical activity
Although the prevalence of optimal level of physical activity (57–58 % in boys, 36–38 % in girls) is not satisfactory, it still is higher than in other developed countries such as the USA(2). Age-adjusted logistic regression analysis showed boys twice as likely to be optimally active compared with girls. This gender difference has also been observed in many other studies which showed more boys reporting higher levels of physical activity than girls(Reference Pate, Heath, Dowda and Trost13–16). These findings could be explained by less social desirability of physical activity for girls v. boys and may be due to less accessibility of structured activity for girls, which need further investigation.
In addition to gender, age demonstrated a relatively consistent pattern, with older (high school) children reporting less physical activity than younger children, although once again this relationship seemed more pronounced among females. Such decline in the prevalence of optimal physical activities with increasing age was also found in other studies among both boys and girls(Reference Sallis, Berry, Broyles, McKenzie and Nader17, Reference Biddle, Sallis and Cavill18) and among girls(Reference Goran, Gower, Nagy and Johnson19), although no studies have been done investigating the behavioural impact of age on physical activity.
Ethnicity is also an additional factor presumed to influence physical activity among children in terms of self-reported activity(Reference Pate, Heath, Dowda and Trost13, Reference Anderssen and Wold15, 16). Our findings also showed that Jews were less likely to be classified as optimally active compared with non-Jews, with age-, sex- and social class-adjusted OR of 0·86 (95 % CI 0·76, 0·97).
Social class was found to have a significant positive association with optimal level of physical activity among Jewish boys. This might be due to the difference in accessibility to facilities between schools or children of high social class. Similar findings were also reported among Scottish schoolchildren, where adolescents from lower SES groups consistently reported lower levels of vigorous physical activity than those from higher SES groups and the differences were more pronounced among girls(Reference Inchley, Currie, Todd, Akhtar and Currie20).
While activity among Jewish girls was associated more with their father’s educational status, activity among non-Jewish girls was associated more with their mother’s educational status. This difference could be explained by the traditional difference in relationships between parents and girls among Jews and non-Jews.
Relationship between overweight and physical activity
There was no association seen between obesity and level of physical activity both at the crude and the adjusted level, except among non-Jewish girls where there was a positive association at the crude level. This lack of association between obesity and level of physical activity is supported by the findings from a similar study utilizing a cross-sectional analysis of a national sample in 14- to 16-year-old girls(Reference Dowda, Ainsworth, Addy, Saunders and Riner21). Moreover, several other national studies(Reference Andersen, Crespo, Bartlett, Cheskin and Pratt10, Reference Dietz and Gortmaker22), one trial(Reference McKenzie, Nader and Strikmiller23) and two other studies(Reference Wolf, Gortmaker, Cheung, Gray, Herzog and Colditz24, Reference Hernandez, Gortmaker, Colditz, Peterson, Laird and Parra-Cabrera25) found no relationship between physical activity and body composition, including obesity.
The associations (or lack of them) found in the present study could be explained by the cross-sectional nature of the study rendering it impossible to determine which comes first: the exercise or the overweight. Another potential explanation could be differential over-reporting of physical activity by obese or overweight children in a manner similar to their under-reporting of food consumption(Reference Pikholz, Swinburn and Metcalf26, Reference Heitmann and Lissner27), even to the extent of reversing the positive association between high fat intake and obesity(Reference Macdiarmid, Vail, Cade and Blundell28).
Sedentary lifestyle and physical activity
Although many studies have found a negative association between level of physical activity and time spent watching television, in Australian adults(Reference Salmon, Ball, Crawford, Booth, Telford, Hume, Jolley and Worsley29) as well as among US children(Reference Andersen, Crespo, Bartlett, Cheskin and Pratt10, Reference Dietz and Gortmaker22, Reference Moore, Ngyen, Rothman, Cupples and Ellison30, Reference Berkey, Rockett, Field, Gillman, Frazier, Camargo and Colditz31), the present study showed no association between time spent viewing television or videos or listening to music with levels of physical activity when adjusted for all other independent variables. This might be partially due to the adding together of the time spent watching television with that spent listening to music, which might in fact not reflect physical inactivity. Similarly, there was no significant association between the time spent at the computer or talking on the telephone.
Duration of sleep and level of physical activity
Previous studies have shown a negative association between obesity and length of sleep among children(Reference Hui, Nelson, Yu, Li and Fok32, Reference Sekine, Yamagami, Hamanishi, Handa, Saito, Nanri, Kawaminami, Tokui, Yoshida and Kagamimori33), but there has been no study on the association between sleep and physical activity. The above two studies put shorter duration of sleep as a risk factor for obesity. The present study too showed less duration of sleep to be associated with a low prevalence of optimal physical activity among Jewish boys, with a non-significant similar pattern seen among Jewish girls and non-Jewish boys too. However, owing to the design of our study, at this stage it is impossible to determine whether inactivity causes lack of sleep or vice versa; this needs to be studied in prospective research in the future.
Cigarette smoking and physical activity
The negative impact of smoking on physical activity among adults is well documented(Reference Hedblad, Gren, Isacsson and Janzon34, Reference Pederson, Poulin, Lefcoe, Donald and Hill35), but there have been few studies among children and adolescents. In our study, hookah smoking (also called ‘narghile’ or water pipe, which has similar composition to cigarettes except for higher levels of nicotine and carbon monoxide in the hookah(Reference Barry and Yona36)) but not cigarette smoking was found to be associated with less optimal level of physical activity among boys, although it was significant only among non-Jewish boys. Lack of association between cigarette smoking and physical activity might be due to the low prevalence of cigarette smoking or the higher dose of nicotine in hookah as compared with cigarettes, or other confounding factors like use of additional drugs such as hashish. This was supported by a study done in Israel where 6 % of the respondents in schools reported adding drugs or alcohol to the tobacco(Reference Barry and Yona36).
Alcohol consumption and physical activity
There is no consensus on the association between alcohol consumption and physical activity. A study done among college students in the USA showed a positive association between binge drinking and participation in physical activity(Reference Nelson and Wechsler37). Our results show that although there was a negative association between alcohol consumption and level of physical activity among Jewish boys at the crude level, history of alcohol consumption was found to have no association with optimal level of physical activity when adjusted for other variables. This could due to the fact that the effect of alcohol depends on the dose, or possibly that these children do not consume large amounts of alcohol.
Conclusions
Overall, gender and age were the most significant factors contributing to differences in physical activity participation. Females were less active than males, and activity levels decreased with age, which was reflected by the lower prevalence of optimal level of physical activity among high-school v. middle-school students. Moreover, sleep duration, parental educational status and social status were found to be positively associated with prevalence of optimal level of physical activity, while level of alcohol consumption and hookah smoking were found to be negatively associated.
Limitations of the study
The findings in the present study are subject to a number of limitations. First, the MABAT Youth Survey is a school-based survey and does not include children out of school or children in private or independent schools. Second, data were self-reported and subject to error, including respondent over-reporting of socially desirable responses (social desirability response bias). Third, duration and frequency of physical activity might not be reliable because it could be difficult for children in middle schools to understand the questions and be able to aggregate minutes of physical activity accurately. In addition, there may be a problem in assessing average physical activity when there may be seasonal variations from the vagaries of the weather. Fourth, it is difficult to rely on the adjusted odds ratio since many of the variables are interrelated to each other, underestimating associations.
Recommendations
Based on our findings, we recommend promotion and advocacy of physical activity to focus mainly on girls in general, high-school students (particularly girls), Jewish boys from lower social class schools, non-Jewish girls from higher social class schools, and girls according to parent’s education. Furthermore, smoking (including hookah) needs to be discouraged among school students (particularly among non-Jewish boys) for reasons of general health promotion and because it decreases physical fitness.
Means of doing this could include comprehensive physical education provided at school – an ideal way to reach and encourage activity and develop active lifestyles among children. In addition, interventions such as providing incentives e.g. for subsidized memberships/entry fees for sports facilities, increased emphasis on sport at school, encouragement to walk at all opportunities and specially dedicated cycling paths, could all assist in improving levels of activity, particularly among the economically disadvantaged.
Moreover, we recommend undertaking qualitative research among optimally active children to find their enabling factors (‘positive deviance’) through which to help promote physical activity in their less active peers. Further studies on hookah smoking and sleep duration in relation to physical activity are also warranted.
We also recommend research to test the social desirability bias by taking a sub-sample and comparing subjective fitness assessment (questionnaires) with objective measures of exercise tolerance. This can be done using the Harvard step test to measure the fitness of the child(Reference Norton and Atherton38), using doubly labelled water to measure energy expenditure(Reference Livingstone, Robson, Black, Coward, Wallace, McKinley, Strain and McKenna39) or using pedometers to measure level of activity over a specified period of time(Reference Beets, Patton and Edwards40).
Acknowledgements
There are no conflicts of interest.
Sources of funding: Israeli Ministry of Health.
Author contributions: D.N.K. – study design, writing manuscript; G.M. – data analysis, writing manuscript; R.G. – study design, data analysis; T.S. – data analysis; E.M.B. – data analysis, writing manuscript.
Acknowledgements: This work was presented as part of the requirements for the degree of MPH in the Hebrew University, Jerusalem. We are very grateful to Tali Bodlah Abraham for helping with data analysis. We would also like to thank Professor Manfred Green, Mr Chava Mayer, Orit Magled Ben-Arieh, the MABAT team and the Israel Center for Disease Control for their dedication and hard work, and the Ministry of Health for encouragement and support.
