Introduction to the ESPAD 2015 reporting

The main purpose of the ESPAD project is to collect comparable data on substance use among students of the same age in as many European countries as possible. The studies are conducted as school surveys among students turning 16 during the year of the data collection, and following a common methodology. A handbook describing methodology and reporting procedures facilitates the collection of comprehensive and comparable data.

ESPAD surveys have been performed every fourth year since 1995. This means that the sixth data collection was performed in 2015 and that results for a 20-year period are available. Each of the five previous ESPAD data collections were presented in extensive printed reports. The presentation for 2015 is, however, done differently. The main findings are presented in a shorter printed report (EMCDDA and ESPAD, 2016), while additional material is made available online.

Apart from this methodological section, the online material includes a presentation of the 2015 results country by country, further graphics that are not included in the printed report, a comprehensive result tables section and the ESPAD master questionnaire. As in previous reports and when possible, comparable data from the two non-ESPAD countries of Spain and the United States have been included in tables and graphs.

Countries participating in ESPAD 1995-2015

In total, 35 countries took part in the sixth study wave in 2015 (Albania, Austria, Belgium (Flanders), Bulgaria, Croatia, Cyprus, Czech Republic, Denmark, Estonia, the Faroes, Finland, the former Yugoslav Republic of Macedonia, France, Georgia, Greece, Hungary, Iceland, Ireland, Italy, Latvia, Liechtenstein, Lithuania, Malta, Moldova, Monaco, Montenegro, the Netherlands, Norway, Poland, Portugal, Romania, Slovakia, Slovenia, Sweden and Ukraine). Georgia was new to the project in 2015. National research teams, as well as funding agencies and supportive organisations for 2015, are listed in the acknowledgements in Appendix 3.

The 1995 ESPAD data collection covered 23 countries, while the report also included data from three more European countries with similar data (Hibell et al., 1997). In 1999 data were collected in 30 countries (Hibell et al., 2000), and in 2003 the number had increased to 35 (Hibell et al., 2004). The 2007 report also included 35 countries (Hibell et al., 2009), while five additional countries collected ESPAD data in 2008. In 2011 the number of countries contributing with results in the 2011 report was 36 (Hibell et al., 2012), while three more countries collected data in the autumn of 2011, and were presented in a digital supplement (Hibell and Guttormsson, 2013).

In total 48 countries (or entities) have participated in at least one of the data-collection waves (see Table A). Twenty-one countries have collected data in all six consecutive waves.

Table A. Countries participating in ESPAD data collections. 1995-2015

ESPAD average

The result tables and graphs make it possible to compare countries not only with each other but also with an ESPAD average. There are different ways of calculating the average for the participating countries. It could take account of the size of the target population in each participating country or it could be computed as a simple ‘average of averages’, which in practice involves assigning each country the same weighting of one. The latter means that each country influences the average to the same extent, regardless of whether it is a small or large country. Such country averages have been used in all previous ESPAD reports, and this practice has been retained also for the 2015 presentation. Country averages presented in the tables do not include Latvia, Spain and the United States (explained later).

Tables

In this methodological section references are made to tables of a methodological nature, identified by letters, while the result tables are numbered and published separately. The following symbols are used in the tables: 0 A percentage below 0.5. . No such data exist. .. Data exist but have either been considered noncomparable or are inaccessible. All percentages are calculated on the basis of valid responses for each variable. Hence, non-responses are deducted from the denominator. Internal non-response rates are given separately in the result tables.

Statistical significance

In all countries, classes (groups of students as an organisational unit) were sampled using a more or less complex procedure. Since the final sampling unit was the class, not the student, and since all students in sampled classes were supposed to take part, it is important to consider the cluster effects in any statistical calculations. This is because a group of students who make up a class (cluster) are more likely to have similar habits than a group containing the same number of students but spread across classes and schools. This affects the precision of the estimates in each country but — provided that the ESPAD guidelines are followed — in principle it should not bias the point estimate itself.

It is also important to note that a certain absolute difference in a particular variable between two surveys may be statistically significant in one country but not in another. Differences have to be tested separately from each country´s result to make it possible to decide whether a difference is significant or not. However, to be able to calculate confidence intervals and assess the statistical significance of differences, it is necessary to have access to the data, including a class variable, for all students. This was not the case in ESPAD surveys previous to 2007, which is why the figures in earlier ESPAD reports were compared between countries and over time in terms of substantive rather than statistical significance. To avoid considering too-small differences, a standardised procedure was used where a difference smaller than ± 3 percentage points was not considered as a ‘real difference’.

Since databases from the past three data collections are available, differences between countries in the trend graphs have been statistically tested to identify any significant differences from 2007 onwards. Gender differences are tested for possible statistically significant differences within countries in the graphs presenting the 2015 results. Since these calculations require inclusion in the ESPAD databases, no such tests have been carried out for the two non-ESPAD countries (Spain and the United States).

A bivariate logistic regression model was used to test whether the differences observed are significant or not. The gender differences were tested using a bivariate model with gender as the only independent variable. Differences over time were tested using the same procedure, with year as the only independent variable. When testing differences between years, the whole sample was used, i.e. boys and girls together. In the logistic regressions, school class was modelled as a cluster, thus taking into consideration that the respondents were not individually sampled. Significance was tested at the 95 % level. The average alcohol consumption during the last alcohol drinking day was tested using a regression with robust standard errors. Rather than using a t-test, this method allows adjustment for the possible effect that the cluster sampling of the students might have on the results, even though this variable is continuous.

Some countries did not perform a sample but instead included all students in the survey. Although it can be argued that testing for significance in such a case is unnecessary, for conformity reasons it was decided to do so anyway.