Trends from 2002 to 2010 in Daily Breakfast Consumption and its Socio-Demographic Correlates in Adolescents across 31 Countries Participating in the HBSC Study

Breakfast is often considered the most important meal of the day and children and adolescents can benefit from breakfast consumption in several ways. The purpose of the present study was to describe trends in daily breakfast consumption (DBC) among adolescents across 31 countries participating in the HBSC survey between 2002 to 2010 and to identify socio-demographic (gender, family affluence and family structure) correlates of DBC. Cross-sectional surveys including nationally representative samples of 11–15 year olds (n = 455,391). Multilevel logistic regression analyses modeled DBC over time after adjusting for family affluence, family structure and year of survey. In all countries, children in two-parent families were more likely to report DBC compared to single parent families. In most countries (n = 19), DBC was associated with family affluence. Six countries showed an increase in DBC (Canada, Netherland, Macedonia, Scotland, Wales, England) from 2002. A significant decrease in DBC from 2002 was found in 11 countries (Belgium Fr, France, Germany, Croatia, Spain, Poland, Russian Federation, Ukraine, Latvia, Lithuania and Norway), while in 5 countries (Portugal, Denmark, Finland, Ireland, Sweden) no significant changes were seen. Frequency of DBC among adolescents in European countries and North America showed a more uniform pattern in 2010 as compared to patterns in 2002. DBC increased significantly in only six out of 19 countries from 2002 to 2010. There is need for continued education and campaigns to motivate adolescents to consume DBC. Comparing patterns across HBSC countries can make an important contribution to understanding regional /global trends and to monitoring strategies and development of health promotion programs.

Large surveys document that many children and adolescents do not regularly eat breakfast. An earlier publication from the 2005/06 Health Behaviour in School-aged Children (HBSC) study documented that in Europe between 33% and 75% of [11][12][13][14][15] year old adolescents reported DBC. Still, in only four countries (Netherlands, Portugal, Denmark, and Sweden) more than 70% of adolescents reported DBC [18]. US data showed that 20% of 9-13-year-olds and 32% of 14-18-year-olds in the 1999-2006 National Health and Nutrition Examination Survey (NHANES) did not eat breakfast [19]. Also, studies examining trends over time report that breakfast skipping among children and adolescents has increased over the past decades in the USA [20,21]. In contrast, an overall increase in DBC was found in trend analyses of the Scottish HBSC study [22]. To our knowledge no study has examined trends in DBC over time across multiple countries using the same standardized methods of data collection, particularly with nationally representative samples.
Numerous factors influence breakfast consumption including socio-economic status (SES), family structure and gender. Being a child or adolescent of low SES is associated with irregular breakfast habits. This relationship exists for a range of different SES indicators, such as parental education [23,24], parental occupation [1][2][3], family affluence [4] and area level economic indicators [5,6]. Moreover, while an overall increase in DBC was found in the Scottish HBSC study [22], a decrease was observed over time for older children and those from single parent families [22]; other studies have also shown DBC to be higher among two-parent families [1,[6][7][8][9][10][11][12][13]. Gender differences demonstrate that boys consume breakfast on a daily basis more often than girls [5]. Information on socio-economic correlates of breakfast consumption among children and adolescents is important for identifying adolescents and families in need of intervention and for planning initiatives that enable frequent breakfast consumption.
Studies of time trends in DBC across countries are important for identifying trends in adolescent breakfast consumption and for informing strategies for promoting breakfast consumption cross-nationally. Promoting breakfast consumption for everyone is important but those most at risk may need additional or indeed different supports. Thus, a better understanding of how young people's breakfast habits are distributed across socio-demographic groups is important for identifying at risk groups and targeting breakfast promotion initiatives.
The purpose of this study was to analyze trends in DBC from 2002 to 2010 in adolescents aged 11 to 15 years across 31 countries participating in the HBSC survey and to identify sociodemographic (gender, family affluence and family structure) correlates of DBC.
America. The overall aim of the study is to gain insight into adolescents' health and health behaviors. The populations selected for sampling are 11, 13 and 15 year olds attending school.
In total, 455,391 adolescents from national representative samples within 31 countries or regions in the three sampling waves were included. Each country follows a standardized international research protocol to ensure consistency in survey instruments, data collection and processing procedures. A clustered sampling design, with the initial sampling unit being either school class or school was used. Participating countries were required to include a minimum of 95% of the eligible target population within their sample frame. The recommended sample size for each of the three age groups was approximately 1,500 students, assuming a 95% confidence interval of +/-3 percent around a proportion of 50 per cent and allowing for the clustered nature of the samples. More detailed information about the study is provided elsewhere [25,26].
The survey instrument is an internationally standardized self-report questionnaire that is administered in the classroom by trained personnel, teachers, or school nurses and whose completion takes approximately 50min. Parental written informed consent to participate was obtained before administration. Student participation was voluntary and anonymity and confidentiality of the data were ensured. In Italy ethical approval was granted by the Italian Higher Institute of Health. All participating countries within the HBSC network must adhere to ethical guidelines and principles as described in the study protocol. Adherence to protocol requirements is managed by the HBSC International Coordinating Centre in Bergen, Norway [26].

Measures
Outcome variable. To assess DBC, adolescents were asked to indicate how many days they generally have breakfast (defined as having more than a glass of milk or fruit juice) on schooldays and on weekends. Response categories were "never" to "five days" for schooldays, and "never" to "two days" for the weekend. These responses were summed to a total range of days eating breakfast (0-7 days a week) and dichotomized into "daily breakfast consumption" (7 days) versus "less than daily" (0-6 days).
Explanatory variables. Socio-economic position: The Family Affluence Scale (FAS) was used to assess socioeconomic position. A sum score was constructed from the following four items: "Does your family own a car, van or truck" (No/Yes one, Yes two or more (0-2 points)). "Do you have your own bedroom for yourself? (No/Yes (0-1 points)). "During the past twelve months, how many times did you travel away on holiday (vacation) with your family?" (Not at all/ Once / Twice / More than twice (0-3 points respectively)) and "How many computers does your family own? (None/ One/ Two/ More than two (0-3 points). For each country, the FAS score was divided into low (0-3 points), medium (4-6 points) and high FAS (7-9 points) [26].
Family structure: Based on student reports of who they live with most of the time, family structure was categorized into living with two parents, one parent or with "others".
Statistical analyses. Multilevel logistic regression analyses for the binary outcome variable "daily versus non-daily breakfast consumption" were conducted separately by country on data pooled across surveys. The independent variables in the model were family affluence, family structure and year of survey. Pseudo likelihood estimation methods, the Binomial probability distribution and the Logit link function were used. Only countries with complete information and association daily-daily 50% were included. This means that the predictive ability of the model to correctly classify a person who eats breakfast every day is more than 50%. Otherwise it would mean that the model fails to properly classify these subjects and thus the model is not considered good.
The estimates were performed for each country separately with adolescents nested within classes and classes within schools (three-level random intercept model) and adjusted for age category. The Bonferroni's sequential test was used to compare DBC between categories of the variables considered. OR and CI (95%) were calculated and the Wald test was used to identify significant parameter estimates; the value of Bayesian Information Criterion (DIC) was used as a measure of model fit. Fixed and random parameter estimates for the model were tabulated, where fixed estimates were defined as the average effect across the entire population of schools, classes and individuals, while the random estimates described how these varied at each level (school and classes) [27]. All independent variables are presented as dummy indicator variables, contrasted against a base category. P-values < 0.05 were considered significant. The statistical package "Generalized Linear Mixed Models" in SPSS software (v.22.0) was used for all analysis.

Results
Data on 455,391 adolescents in the 31 countries or regions were included (Table 1). In total, boys constituted 49.1% and girls 50.9% with only small differences between countries. DBC ranged between 37.8% (Slovenia) to 72.6% (The Netherlands). Among boys, DBC ranged from 39.3% (Slovenia) to 75.6% (Portugal) and among girls between 36.4% (Slovenia) to 70.7% (The Netherlands) ( Table 1 and Fig 1).
The mean FAS score ranged from 3.90 (SD = 1.80) in Ukraine to 6.68 (SD = 1.65) in Norway. The proportion of children living in high FAS homes ranged from 15.9% (Norway) to 28 For family structure, the proportion of children living with both parents ranged from 50.8% (Greenland) to 88.8% (Macedonia) while between 5.5% (Macedonia) and 31.5% (Greenland) lived in a single parent household. Distribution of DBC by family structure showed that in all countries, children in two-parent families were more likely to report DBC compared to single parent families. Children in "other" types of family structures in most countries had rates of DBC in between children from two parent and single parent families. In two-parent families, DBC ranged from 44.4% (United States) to 75.5% (the Netherlands); while in single-parent families, the range was from 34.5% (Slovenia) to 66.4% (Portugal). Reported DBC for children in "other" family structure ranged from 38.7% (United States) to 68.7% (Portugal).
Distributions of DBC by survey year revealed that in many countries the proportion of adolescents reporting DBC was lower in 2010 compared to 2002. In 2002, DBC ranged from 36.7% (Slovenia) to 67.6% (Denmark); in 2006 it ranged from 38.6% (United States) to 70.6% (The Netherlands); and in 2010 from 40.5% (Slovenia) to 74.1% (the Netherlands) ( Table 2).
Associations between DBC and socio-demographic factors. A total of 22 countries or regions had complete data to examine associations between DBC and socio-demographic factors. DBC was associated with being a child of a two-parent family with an OR between 1.17 (95% CI: 1.08-1.26) in the Russian Federation to 1.78 (955 CI: 1.63-1.93) in Norway, compared to being a child of single parent family. In most countries (n = 19), DBC was associated with being a child living in a family with high FAS score compared to being a child living in a family with low FAS status. OR ranged from 1.12 in Ukraine (95% CI: 1.02-1.24), to 1.72 (95% CI: 1.54-1.91) in Germany. In Macedonia and Russia no association between DBC and family affluence was found (Table 3).  (Table 3).

Discussion
This study adds to the existing literature by being the first study to compare DBC trend data for 31 countries across two continents. It updates previous HBSC work [4] and adds new information by studying trends in DBC over nearly a decade in a multinational context using the same standardized methods. This study also adds to the literature on socio-economic correlates with DBC by studying associations based on pooled data covering three survey cycles of the HBSC study.  [20][21][22]. This is the first study to investigate trends across many countries. The differences in time trends in DBC across countries are not easily explained, but there may be some overarching/common factor(s) that require further investigation. We know that especially in Western countries, the availability of foods outside the home at all times of the day has increased [28]. This higher availability of foods especially snack foods outside the home might contribute to the decrease observed in DBC [29,30]. If left unchecked, these changes in food environments will continue to contribute to poor food habits, which over time, may further exacerbate rates of obesity and diabetes.
Public health agencies have tried to increase DBC. However, they have had less impact on low affluence families as our study and others demonstrate that DBC is generally higher among adolescents from high affluent families. Also in line with the existing literature, we found that in a majority of countries the proportion of adolescents consuming breakfast daily were generally higher in two-parent families [1,[6][7][8][9][10][11][12][13] and among boys [5]. The gender differences in breakfast consumption have been attributed to weight concerns among adolescent girls [2,3,31]. Additional factors that may impact on daily breakfast consumption include a limited knowledge about nutrition and health [32], lack of time to eat or prepare breakfast [33] and unavailability of foods for breakfast [2]. Generally the findings highlight the importance of the family environment for influencing the dietary behaviours of young people [18,[34][35][36]. Socialisation of health-related behaviours occurs within the family, with parents' beliefs, attitudes and behaviours substantially affecting children's health behaviours [37]. Consistently, within the literature it is documented that parental eating behaviours are positively associated with both unhealthy [36] and healthy [14,23,[38][39][40][41][42][43][44][45] dietary behaviours of children and adolescents. The differences in DBC observed among boys and girls however indicate that family-related processes affecting adolescents' DBC may operate differently across gender.
When interpreting the results a number of study limitations should be considered. We investigated daily versus less than daily breakfast consumers and it could be argued that skipping breakfast on just one day does not have an effect on adolescent health, although others have found that breakfast consumption shows dose-response association with overweight [7]. Our measures of breakfast frequency has been validated and Kappa statistics comparing daily consumption of breakfast to diary measures in the Flemish population were fair for weekends (0.34) and moderate for weekdays (0.47). Further, the frequency measures have been validated in a Danish agreement study comparing daily consumption of breakfast with 24 hour recall during one week and the kappa statistics showed good agreement for the dichotomized item (0.62 weekdays and 0.46 weekend) [46].
The HBSC study is based on a frequency measure and defines breakfast as having more than a glass of milk or fruit juice. This measure precludes any assessment of the nutritional quality of the meal. Also, we did not distinguish between breakfast consumption during the school week and at weekends, which is pertinent when attempting to explain the rates of breakfast consumption over time and the settings where interventions are most needed. Indeed, there are numerous school-based breakfast programmes in operation across many countries although the results of our study should not be affected if meal programmes remained unchanged within countries. However, it was beyond the scope of this study to map school food programmes across countries. Yet our data, up to 2010 indicate that DBC remains low for many students. In part this could be due to less structured meal times at weekends for adolescents and their families.
Another issue relates to the use of FAS over time. There is a risk that the classification by FAS may not be uniform over time (e.g. by more families owing computers by increasing survey year) and it may be hypothesized that misclassifications of families in less affluent groups into more affluent groups may therefore increase over time. In such case there is a risk that the patterns of social inequality is increasingly being underestimated by each survey year.
This HBSC study offers an opportunity for cross-national and time trend analyses on adolescent health and the study of DBC covers three survey rounds spanning nine years across 31 countries. Other strengths include the large sample size and the representative samples of adolescents from 31 countries across two continents, completed by a standardized protocol for data collection ensuring internationally comparable data. DBC should be encouraged within the context of each country and family [17]. Increased attention to DBC is especially necessary during the transition from childhood to adolescence and especially in girls and young people from disadvantaged families. Both the home and school setting needs attention to reduce social inequalities in breakfast consumption. yield benefits on cognitive performance [10,11]. Further research should explore the countries that have experienced an increase in DBC over time and associated changes in policies, strategies and programmes. In particular, it would be valuable in a multilevel analytical design to describe and compare national characteristics and guidelines and their implementation taking cultural and normative practices into account.