A factor analysis-based study of trends in mental health problems among adolescents over a twenty-year period
Centre for Health Equity Studies
Master thesis in Public Health (30 credits) Spring 2014
Name: Mia Eriksson
Supervisor:
Co-supervisor:
Anton Lager
Jennie Ahrén
Abstract
Background: Research points in different directions when looking at possible increases in mental health problems among adolescents. Findings in favor of an increase are questioned due to methodological problems.
Aim: Investigating whether mental health problems among young adolescents are increasing over time in Europe and North America. If so, does the trend apply both to mean levels of symptoms and to the proportion of adolescents with substantial problems? Are the time-trends similar over sex and age-categories?
Method: A total of 401 089 adolescents from a total of 38 countries are included in the
analysis. Based on the eight health variables on self-rated health provided by the HBSC study, a measurement of mental health problems was created using factor analysis in SPSS.
Results: Increases of mental health problems were found in Europe and North America.
Increases were found both in terms of mean levels of symptoms and to the proportion of adolescents with substantial problems. Increases were seen in all age groups and among both girls and boys.
Conclusion: Reasons behind the discovered increases are not known and should be further investigated as extensive research point to severe consequences of mental health problems in adolescence for later life.
Key words
Adolescents, mental health problems, trends, self-reported health (SRH), psychological health
complaints (PHC)
Table of contents
1.0 Introduction ... 1
1.1 Trends in mental health problems among adolescents ... 1
1.1.1 Increases in Sweden ... 2
1.2 Implications for adverse mental health among adolescents ... 2
1.3 Self-reported health and morbidity ... 4
1.4 Gender- and age related trend patterns in self-reported mental health problems ... 4
1.5 Aim and research questions ... 5
2.0 Methods ... 6
2.1 Data material ... 6
2.2 Variables ... 7
2.3 External attrition ... 7
2.4 Internal attrition ... 8
2.5 Statistical analysis ... 10
2.6 Ethical approval ... 11
3.0 Results ... 12
3.1 Descriptive statistics ... 12
3.1.1 Increase of adolescents with substantial problems (ASPs) ... 12
3.1.2 Increase of factor score mean value (FSMV) ... 14
3.1.3 Additional measuring ... 16
3.1.4 Sweden ... 16
3.1.5 Additional findings ... 17
4.0 Discussion ... 18
4.1 Principal findings ... 18
4.2 Strengths and limitations ... 18
4.3 Previous research ... 19
4.4 Possible explanations for time trends in mental health problems among adolescents ... 21
4.4 Future research ... 22
Acknowledgements ... 24
Appendices ... 31
1
1.0 Introduction
1.1 Trends in mental health problems among adolescents
Numerous findings suggest that mental health problems may have become more frequent among adolescents in western countries (Fombonne in Rutter & Smith, 1995; Collishaw, Maughan, Natarajan and Pickles, 2010; Prosser & McArdle, 1996; SOU, 2006:77; West & Sweeting, 2003). Different studies are however pointing in different directions, and the findings in favor of an increase are questioned.
Collishaw et al., (2010) looked at British trends in emotional problems and compared cohorts of 16-17-year-olds from 1986 and 2006 using questionnaires and scales. They found that twice as many adolescents reported recurrent feelings of depression or anxiety in 2006 compared to 1986.
Symptoms like anxiety, irritability and fatigue increased whereas other symptoms, -like loss of enjoyment and worthlessness, remained stable. There were no differences in trends regarding socially advantaged or disadvantaged backgrounds, or among intact or non-intact families.
In a review based on several extensive datasets including prospective studies, cross-sectional studies and data from mortality and police statistics, Fombonne (1998) found an increase in suicide, depressions, eating disorders and addictive behavior among youth. However, Fombonne also mentions that the magnitude of the reported increases of depression actually is not known and probably is rather small. A large part of the studies that showed increases also had potential problems with artifacts and study-methods effects.
Most of the studies that exist have encountered the task of using datasets with problematic and non-identical items (Sweeting, West, Young & Der, 2010), not having socially and
geographically comparable groups, and the lack of repeat cross-sectional surveys (Angold &
Costello, 2001). Roberts, Attkisson and Rosenblatt (1998) performed a meta-analysis on 52 studies looking at psychopathology among children in ages 1-18 years. They concluded that problems in measuring child and adolescent disorders involving “sampling, case ascertainment, case definition, data analysis and presentation” (Roberts et al., 1998, p. 715) make it
questionable whether there are in fact any increases at all. They also argue that case definition often confuses certain diagnostic criteria with the functional impairment and/or the perceived need for help.
Busfield (2012) examined evidence on the claimed increases in mental health problems among
both children and adults and concluded that while there were some findings supporting the claim,
2
numerous studies suggested no changes at all. Regarding child and adolescent depression, one meta-analysis covered twenty-six studies and 60,000 observations on children born between 1965 and 1996, i.e. aged 15 years between 1980 and 2001. Results showed that no increased prevalence of depression was found when using concurrent assessment instead of retrospective recall (Costello, Erkanli, & Angold, 2006). In an Australian setting Eckersley (2011) discusses the possibility that the claimed increase could be due to increased use of diagnoses and the medicalization of human emotions. Busfield (2012) discusses how the argued decline in mental well-being maybe can be explained by the expansion of the definition of mental illness, and by the fact that it gives critics of society some needed ammunition.
1.1.1 Increases in Sweden
Taking the Swedish context into consideration, the Swedish National Board of health and Welfare (Socialstyrelsen, 2013) reports empirical findings that seem to suggest that mental problems among children in Sweden have increased since the 1990’s. Between years 1994-2006 the number of young adults in aged 16-29 who reported severe worrying and anxiety more than doubled, from 2 % to almost 5 %. Milder forms of anxiety also increased considerably.
Sweden has also seen an increase in hospital admissions due to psychiatric problems in all ages, but the steepest incline has been found in the age group 15-24-years (Folkhälsoinstitutet &
Socialstyrelsen, 2013), which could imply that the increase in mental health problems is not just about an increase in the reporting of symptoms, but actually reflects severe problems in certain groups.
1.2 Implications for adverse mental health among adolescents
Suffering from mental health problems in adolescence has been linked to increased risk of mental health problems in adulthood (Lewinsohn, Rohde, Klein, & Seely, 1999; Fichter, Kohlboeck, Qaudflieg, Wyschkon, & Esser, 2009; Pine, Cohen, Gurley, Brook, & Ma, 1998;
Fombonne, Wostear, Cooper, Harrington & Rutter, 2001), and as parent anxiety and depression
are strong predictors of emotional disorders among children and adolescents (Merikangas,
Avenevoli, Dierker & Grillon, 1999; Rice, Harold & Thapar, 2005) there has also been concerns
for vicious cycles over generations.
3
It is estimated that about a quarter of adolescents diagnosed with depression are expected to experience continued problems with chronic recurrent depressions. This in turn, will likely increase their risk for other mental and physical disorders and premature death (Yiend, Paykel, Merritt, Lester, Doll, & Burns, 2009; Socialstyrelsen, 2013). Also, adolescents with “sub- threshold depression” have an elevated risk for depression in adult age, as well as a risk of suicidal behaviors (Fergusson, Horwood, Ridder, & Beautrais, 2005). The “sub threshold depression” is considered to share the same features as a clinical depression, meaning that depressive symptoms are manifested along a continuum and thus include more individuals than the ones that are diagnosed with depression. Hence, the number of people that are actually at risk stretches beyond the clinically diagnosed and includes other levels and patterns of depressive symptoms (Lewinsohn, Seely, Solomon, & Zeiss, 2000).
Previous studies have seen a strong association between somatic symptoms in adolescence and severe mental disorders and depression in adulthood, (Bohman, et al., 2012; Hotopf, Mayou, Wadsworth & Wesely, 1998) as well as coexistence between somatic symptoms and depression (Härmä, Kaltiala-Heino, Rimpelä & Rantanen, 2002; Larsson, 1991). Also, the severity of the depression has been shown to correlate with the numbers of somatic symptoms (Bohman et al., 2010).
Some of the symptoms that have been linked with adverse mental health later in life are headache and musculoskeletal pain (Egger, Costello, Erkanli, & Angold, 1999), stomach ache and backache (Härmä, Kaltiala-Heino, Rimpelä & Rantanen, 2002), sleeping problems (van Lang, Ferdinand & Verhulst, 2007), and fatigue and irritability (Fichter et al., 2009).
Bohman et al. (2012) did a 15-year follow-up study of Swedish 16,-17-year-olds (n= 2465) with depression and healthy controls, and found that somatic symptoms in adolescence predicted severe mental health disorders (suicidal attempts, bipolar disorders, psychotic disorders, post- traumatic stress disorder, and depression) in adulthood. This was a relationship that was even more pronounced when the somatic symptoms coincided with depression, but it was also present in cases without depression. Interestingly, having stomach ache or excessive transpiration could better predict later life depression than all DSM-V depressive symptoms.
Furthermore, according to Socialstyrelsen (2013) adverse mental health early in life has been
associated with several problems in adulthood: psychiatric disorders, suicide attempts and other
injuries and accidents, as well as with future incomes and family formation. Adolescents in ages
4
14-16 years who have been hospitalized due to psychiatric problems are at high risk of being re- hospitalized short term, and are more often using psychopharmacological drugs and need more hospital or specialist care later in life.
Socialstyrelsen (2013) also reports research showing that young adults (16-24) with mild and severe anxiety had a 40 percent heightened risk of having achieved merely primary education at the age of 29, as compared to young adults who have no experience of anxiety. However, the rate of young adults with anxiety who achieved post-secondary education was only to a certain degree smaller, and this was not of statistical significance. Looking at the group with only mild worrying, this group had a slightly higher degree of post-secondary education compared to the ones who didn’t experience any anxiety.
1.3 Self-reported health and morbidity
The suitableness of self-rated health as a measure of health is debated. One of the critiques brought forward is that different groups of people might have different notions of what poor health implicates, leading to misperceptions about their own health status (Fritzell & Lundberg, 2007; Sen, 2002). For example, in communities with few medical facilities and many different types of health issues, conditions might be considered normal even though they are in fact preventable. Nonetheless, there are extensive findings on the accuracy of how well self-rated health predicts future risk of death (Graham, 2007; Idler & Benyamini, 1997) and morbidity (WHO, 2012). Moreover, reporting feelings of nervousness, uneasiness and anxiety has been shown to be a strong predictor for suicide attempts and psychiatric disease during five or ten years after reporting (Ringbäck, Weitoft & Rosén, 2005). The latter study also showed that self- reported health problems were a better predictor for all-cause mortality, suicide attempt and hospital care as compared to longstanding illness, low educational achievement and smoking.
1.4 Gender- and age related trend patterns in self-reported mental health problems
Various studies have confirmed that girls report health complaints more frequently than boys (Wiklund, Malmgren-Olsson, Öhman, Bergström, Fjellman-Wiklund, 2012; West & Sweeting, 2003; Currie et al., 2008; Hetland, Torsheim & Arro, 2002). The reporting is more frequent among older adolescents, and in particular among older girls (Haugland, Wold, Stevenson, Aaroe, & Woynarowska, 2001; Currie et al., 2004; WHO, 2012; Statens Folkhälsoinstitut, 2011).
Differences in self-rated health between girls and boys are not very pronounced at age 11, but
5
become evident at ages 13 and 15. In the latter age group the differences are greater than 10 % in about half of the countries and regions in Europe (WHO, 2012).
Increases over time concerning gender-differences in self-reported health have not been well documented. However, West and Sweeting (2003) compared cohorts of 15 year-olds in years 1987 and 1999 and found that psychological distress had risen from 19 percent to 33 percent among girls, while the increase among boys was much lower; 13 percent to 15 percent. The increase was mostly seen in girls from non-manual and skilled manual backgrounds. Worries concerning unemployment decreased, but worries about family relationships increased, possibly explained by the improved youth employment and the increased divorce rates, respectively.
From 1999 there was a visible gender difference in worries about school performance, where females worried more than males.
1.5 Aim and research questions
The aim of this master thesis is to answer the three following questions;
1. Are mental health problems among young adolescents increasing over time in Europe and North America?
2. Does that trend, if any, apply both to mean levels of symptoms and to the proportion of adolescents with substantial problems?
3. Are the time-trends similar over sex and age-categories?
These questions are approached with the help of data on self-reported mental health problems among children (aged 11, 13 and 15 years) in 39 European and North American countries over 20 years (1986-2006) by using data from the Health Behavior in School-Aged Children (HBSC) study.
To my knowledge, the HBSC has not yet been used to this end despite the fact that it posits
unique strengths: with the cross-national and repeated cross-sectional design (which is one of its
kind for the included countries and the time period at hand) the HBSC includes an instrument
with eight mental health symptoms, allowing the fitting of factor analyses, theoretically reducing
6
measurement error (due e.g. to the meaning of separate items changing over time, or meaning slightly different things in different languages) to a minimum.
Previous research has pointed in different directions in regards to a possible increase in mental health problems. Studies that have shown increases have had problems with retrospective recall (recall bias), cohort-dependent non-response, different locations (home and in school) for replies that could interfere with how the mental health is perceived (methods effect), and meta-studies have included studies with various types of assessment methods. This poses the question whether we can know that mental health problems among adolescents actually have increased or not. In this regard the HBSC study, lacking many of the above-mentioned problems, provides a valuable and incomparable possibility to follow the trend by supplying more reliable data.
2.0 Methods
2.1 Data material
This study uses data material from the Health Behavior in School-aged Children: WHO
Collaborative Cross-National survey/study (HBSC), which consists of a cross-national alliance of researchers and the regional WHO Regional Office for Europe. The HBSC-study currently involves 43 countries in Europe and North America. The study’s object is to understand the health of young people from the perspective of their social context -“where they live, at school, with family and friends”. The HBSC project began in 1982, when researchers in England, Norway, and Finland decided to start a shared monitoring of school children. In 1983, WHO Regional Office for Europe decided to adopt the study and make it collaborative (HBSC, 2014).
Every fourth year, data is being collected on adolescents in the mean-ages 11.5, 13.5 and 15.5 years. Since classes can include individuals that have advanced or have been held back sometimes sampling is performed not just in one single class but also across grades. The
sampling of classes is random, and countries may choose to also stratify their samples in order to guarantee representation of a certain kind, ethnic groups and school types for example. The
recommended sample size is 1, 500 students per age group (Roberts, et al., 2009).
The data concerns health and well-being, social environments, and health behaviors and is
collected through self-completion questionnaires that are administered in the classroom. The data allows cross-national comparisons and, as repeated cross-sectional studies are made,
observations of trends on both national and cross-national level.
7
The data used in this study stems from surveys collected in years 1985/86, 1993/94, 1997/98, 2001/02, and 2005/06 (see appendix, table 14, for participating countries per year).
2.2 Variables
Mental health problems were measured by using the HBSC Symptom Checklist (HBSC-SCL) as the dependent variables. HBSC-SCL consists of eight subjective health complaints: “headache”,
“stomach ache”, “backache”, “feeling low”, “bad temper”, “feeling nervous”, “difficult to sleep”, and “feeling dizzy”. Participants could report frequency of the problems on a five-point scale with the alternatives every day, more than once a week, every week, every month, seldom or never.
The independent variables used were country, gender, age and time. By using the variable time, each country acts as its own control in a way, as comparing single cross-sectional values from different countries can measure anything from problems with the translation of the questionnaire to all the things that differ between countries and in turn is related to how you respond to a mental health questionnaire.
2.3 External attrition
In the wave of 1985/86, 13 countries participated. The numbers of countries received for this study was seven (n=34 211), whereof three were excluded in analysis. Switzerland (n= 4793) was missing information on the variable sleeping difficulties, and Finland (n=3216) and Hungary (n=4461) only had four value labels instead of five like the rest of the countries.
The wave of 1989/90 has been entirely excluded from the study since only one country (Austria) had variable-values for the eight health-complaints used that ranged between one and five (every day, more than once a week, every week, every month, seldom or never) while the rest had four values (often, sometimes, seldom never), in contrast to all the other years that had five values. A total of 16 countries participated in the study.
In the wave of 1993/94, 26 countries participated. In the case of Flemish Belgium and French Belgium, and Scotland and Wales respectively, they have been considered as belonging to the same country: Belgium and UK, respectively. Consequently, 24 countries participated.
Switzerland and Netherlands were however not received from HBSC. Out of the remaining 22
8
countries (n= 102 799), Czech Republic (n=2207) and Spain (n=3051) have been excluded due to missing all information on the eight health variables, leaving the final number of countries to 20 (n=102 799).
The wave of 1997/98 consisted of 30 countries. Again, Flemish and French Belgium have been considered as one country (Belgium), as well as England, Scotland, Wales and Northern Ireland (United Kingdom). Spain and Netherlands were not received from HBSC, and Canada (n= 6567) lacked comparable values on the variable “grade”, which leaves 23 countries (n= 119 165).
The wave of 2001/02 consisted of 36 countries. Flemish and French Belgium have been merged into one country. England, Scotland and Wales occur as separate participants in the information on HBSC’s website but are merged into one country in the received data. Northern Ireland is not mentioned in the information supplied by HBSC, thus whether they are merged in “UK” is not known. This leaves 33 countries, of which Slovakia was not included in the dataset received from HBSC. Greece (n= 3807) lacked comparable values on the variable “class”. The remaining 32 countries consist of 162 305 adolescents.
The wave of 2005/06 consisted of 41 countries (n=205 938). As with prior waves, Flemish and French Belgium have been merged into one country. England, Scotland and Wales occur as separate participants in the information on HBSC’s website but are merged into one country in the received data. Northern Ireland is not mentioned in the information supplied by HBSC, thus whether they are merged in “UK” is not known. Czech republic (n= 4782) was excluded from analysis due to lacking information on the variable “year of birth”, and Bulgaria (n= 4854), Iceland (n=9540), Luxemburg (n=4387), Romania (n=4684), and Turkey (n=5639) were excluded in analysis due to the fact that they participated only in the last wave end therefore couldn’t contribute to the trend level (their levels are presented in tables 2-13 in the appendix).
The remaining countries consist of 172 052 adolescents.
2.4 Internal attrition
Participants were divided into groups depending on their age. For example, adolescents
participating in the wave of 1985/86 were born between 1969-1976. Since the questionnaires are
handed out per class, and a class can contain children who have been held back or have started
earlier in school, one age group has been decided to constitute one single year of birth. Thus, in
1985/86 15-year-olds were stipulated to contain people born only in 1970, 13-year-olds to be
9
born in 1972, and 11-year-olds to be born in 1974. All other birth years have been excluded from analysis. This has been done for each subsequent wave of data.
It is also a way to ensure that the measurement can be considered reliable in so far as problems connected to certain parts of adolescence; hormonal changes for example, shouldn’t interfere in the measuring as the development of a 15-year-old can be quite different from a 16-year old as well as differences between 11-, 12-, and 13-year-olds.
From the wave of 1985/86 the number of students excluded due to birth years was 11 240, and another 255 had missing information on birth year. This makes up a total internal attrition of 11 495 students. There was no attrition due to missing information on gender. There was a 1,8 percent
1(n= 243) attrition due to some students not answering all questions connected to the eight health variables. The remaining share of valid respondents were 13 287.
From the wave of 1993/94, the number of students excluded due to birth years was 26 415, and another 405 had no information on birth year. This makes up a total internal attrition of 26 820.
There was no attrition due to missing information on gender. There was a 7,6 percent
1(n= 5369) attrition due to some students not answering all questions connected to the eight health variables.
The remaining share of valid respondents were 65 350.
From the wave of 1997/98, the number of students excluded due to birth years was 28 623, and another 654 had no information on birth year. This makes up a total internal attrition 29 277.
There was no attrition due to missing information on gender. There was a 2,35 percent
1(n=
2116) attrition due to some students not answering all questions connected to the eight health variables. The remaining share of valid respondents were 87 772.
From the wave of 2001/02, the number of students excluded due to birth years was 37 098, and another 737 had missing information on birth year. This makes up a total internal attrition of 37 835. There was no attrition due to missing information on gender. There was a 2,73 percent
1(n=
3401) attrition due to some students not answering all questions connected to the eight health variables. The remaining share of valid respondents were 121 069.
From the wave of 2005/06, the number of students excluded due to birth years was 40 494, and another 779 had missing information on birth year. This makes up a total internal attrition of 41 273. There was no attrition due to missing information on gender. There was a 3,20 percent
1(n=
1
Country-level attrition; see appendix Table 1.
10
4176) attrition due to some students not answering all questions connected to the eight health variables. The remaining share of valid respondents were 126 603.
2.5 Statistical analysis
All analyses have been performed with SPSS 21.0. Factor analysis was chosen as method for analyzing the data, and the total number of individuals included in the sample were 414 081.
Factor analysis allows investigating whether there are any underlying factors behind the chosen variables (in this case the eight psychosomatic complaints), and if so: which variables constitute this/these factor/s. In this study, such a factor will not only have some kind of explanatory purpose, eg how is mental health problems constructed, but it will also create a measurement that makes it possible to study time trends in its prevalence.
Mental health problems include a variety of symptoms. Using the eight health variables,
containing both physical and psychological factors, therefore creates a measurement that is more reliable than a single variable. Single symptoms, like back pain or feeling dizzy for example, could be related to something completely different than mental health problems, for instance having a muscle fever or having a cold. When exploring the correlation between variables, in this case a psychosomatic health complaint and the outcome “mental health problems”, factor
analysis allows the investigation of whether there are latent variables that explains the
relationship between the measured variables. If one or several variables do exist they are called factors. Factor analysis is the way in which these factors are extracted and made visible.
As a first step, called exploratory factor analysis (EFA), the factor analysis identifies the numbers of factors, and it also finds out which variables that are connected to which factor (Brace, Kemp & Sneglar, 2003). In a second step, an idea of what factors that may be underlying the variables is tested in confirmatory factor analysis (CFA). This is done by suggesting a
hypothesis on the factor structure, which in this study has been tested using maximum likelihood
(ML). ML is a commonly used method in order to fit confirmatory factor analysis models (Liu,
Rubin, 1998), as it “provides standard errors (SEs) for each parameter estimate, which are used
to calculate p-values (levels of significance), and confidence intervals, and its fitting function is
used to calculate many goodness-of-fit indices” (Harrington, 2008, p. 28-29). A three-component
solution was chosen as a first step, with the varimax rotation. As this showed, all health variables
loaded positively in a one-component solution. Thus, a second confirmatory factor analysis with
a single factor component was performed. This model had good fit (Approx. Chi-
11 square=266645,652, df=28, p < 0,0005).
Thirdly, the estimated model was used to predict a factor score for each individual. The values on this factor vary between 0 and 192.
For simplicity, a month was stipulated to contain 28 days (i.e. four weeks), and sum-scores were calculated for each reply. The answer “every day” was interpreted to signify 5-7 days per week, which makes up 20-28 days per month. The mean of these is 24 and this was therefore the sum score for “every day”. A similar way of calculating sum score was applied to the remaining answering alternatives: “more than once a week” was estimated to correspond to 12-16 days a month (3-4 days per week), “every week” to 4-8 days a month (1-2 days per week), “every month” to 1-3 days a month, and “never” to 0.
Having answered “every day” to all eight health variables questions thus produces a maximum score of 192 (24 times 8), and this is also the maximum value for the latent factor predicted in the study. To be able to describe the prevalence and development over time of a group of adolescents with substantial problems (ASPs), a cut-off of 97,5 was introduced. This means that adolescents with a sum score of 97,5 or higher are considered to have substantial health
problems. Research has shown that major depressive disorder (MDD) has a global prevalence of 3,2-5,5 percent (Ferrari et al., 2013). Also, a score of 96 or above is half of the total amount (192), indicating that problems are experienced more than half of the time. Many individuals in this group are likely to have very clear mental health problems.
Regression has been used as a statistical test of time trends. The regressions have been
performed as a “test of trend”, i.e. a linear regression based on the available five observations in time.
2.6 Ethical approval
No new data are collected in this study. The data that are used are completely de-identified, and
do not include names of participating schools or students. Teachers do not receive reports about
their own school. For ethical considerations on the intrusion of children’s integrity that the actual
data collection has meant, the reader is referred to each national team.
12 3.0 Results
3.1 Descriptive statistics
3.1.1 Increase of adolescents with substantial problems (ASPs)
The proportion of adolescents with substantial problems (ASPs) increases over time in all six groups over sex and age. The highest relative increase is seen among 15-year old boys (graph 1), where the increase is 180 percent (p< 0.0005) between 1985 (1.50 percent) and 2005 (4.20 percent). The increase in ASPs among 15-year old girls (graph 2) is 165,68 percent (p< 0.0005), for which the proportion went from 3.70 percent to 9.83 percent. Among 13-year old girls (graph 3) the increase is 146.30 percent (p< 0.0005), starting at 3 percent in 1985 and increasing to 7.39 percent in 2005. Boys aged 11 (graph 4) have the fourth highest increase, 78 percent (p=
0.008), from 2.19 percent to 3.90 percent. The increase among boys aged 13 is 53.50 percent (p<
0.0005), from 2.28 to 3.50 percent. Girls aged 11 have the smallest increase of 15.85 percent (p=0.026), starting at 5.18 percent and increasing to 6 percent.
Graph 1. 15-year old boys; proportion of adolescents with substantial problems (ASPs).
0%
2%
4%
6%
8%
10%
12%
1985/1986 1989/1990 1993/1994 1997/1998 2001/2002 2005/2006
Austria Belgium Canada Croatia Czech Republic Denmark Estonia Finland France Germany Greece Greenland Hungary Ireland Israel Italy Latvia Lithuania Malta Netherlands Norway Poland Portugal Russian Federation Slovakia Slovenia Spain Sweden Switzerland Ukraine MKD UK United States All Linear (All)
13
Graph 2. 15-year old girls; proportion of adolescents with substantial problems (ASPs).
Graph 3. 13-year old girls;
proportionof adolescents with substantial problems (ASPs).
0%
2%
4%
6%
8%
10%
12%
14%
16%
18%
20%
1985/1986 1989/1990 1993/1994 1997/1998 2001/2002 2005/2006
Austria Belgium Canada Croatia Czech Republic Denmark Estonia Finland France Germany Greece Greenland Hungary Ireland Israel Italy Latvia Lithuania Malta Netherlands Norway Poland Portugal Russian Federation Slovakia Slovenia Spain Sweden Switzerland Ukraine MKD UK United States ALL Linear (ALL)
-1%
1%
3%
5%
7%
9%
11%
13%
15%
1985/1986 1989/1990 1993/1994 1997/1998 2001/2002 2005/2006
Austria Belgium Canada Croatia Czech Republic Denmark Estonia Finland France Germany Greece Greenland Hungary Ireland Israel Italy Latvia Lithuania Malta Netherlands Norway Poland Portugal Russian Federation Slovakia Slovenia Spain Sweden Switzerland Ukraine MKD UK United States All Linear (All)
14
Graph 4. 11-year old boys,
proportionof adolescents with substantial problems (ASPs)
3.1.2 Increase of factor score mean value (FSMV)
The highest increase of the FSMV is seen among 15-year old girls (graph 5), with a statistically significant increase (p< 0.0005) from 26.35 in 1985 to 39.67 in 2005. Cohen’s effect size value (d= 0.36) suggested a small to moderate practical significance.
The second highest increase is seen among girls aged 13 (graph 6), who have a statistically significant increase (p= 0.002) from 24.54 to 34. Cohen’s effect size value (d= 0.27) suggested a small to moderate practical significance. The FSMV for boys aged 15 (graph 7), also statistically significant (p= 0.019), increased from 18.38 to 25. Cohen’s effect size value (d= 0.22) suggested a small to moderate practical significance. Among boys aged 13 a marginally significant increase (p= 0.08) was found; 20.16 to 23.67. Cohen’s effect size value (d= 0.12) suggested a small practical significance. No significant changes were found in FSMV for boys and girls aged 11.
0%
2%
4%
6%
8%
10%
12%
14%
16%
18%
1985/1986 1989/1990 1993/1994 1997/1998 2001/2002 2005/2006
Austria Belgium Canada Croatia Czech Republic Denmark Estonia Finland France Germany Greece Greenland Hungary Ireland Israel Italy Latvia Lithuania Malta Netherlands Norway Poland Portugal Russian Federation Slovakia Slovenia Spain Sweden Switzerland Ukraine MKD UK United States All Linear (All)
15
Graph 5. 15-year old girls, Factor score mean value (FSMV).Graph 6. 13-year old girls, Factor score mean value (FSMV).
20 25 30 35 40 45 50 55 60
1985/1986 1980/1990 1993/1994 1997/1998 2001/2002 2005/2006
Austria Belgium Canada Croatia Czech Republic Denmark Estonia Finland France Germany Greece Greenland Hungary Ireland Israel Italy Latvia Lithuania Malta Netherlands Norway Poland Portugal Russian Federation Slovakia Slovenia Spain Sweden Switzerland Ukraine MKD UK United States All Linear (All)
20 25 30 35 40 45 50
1985/1986 1989/1990 1993/1994 1997/1998 2001/2002 2005/2006
Austria Belgium Canada Croatia Czech Republic Denmark Estonia Finland France Germany Greece Greenland Hungary Ireland Israel Italy Latvia Lithuania Malta Netherlands Norway Poland Portugal Russian Federation Slovakia Slovenia Spain Sweden Switzerland Ukraine MKD UK United States All Linear (All)
16
Graph 7. 15-year old boys, Factor score mean value (FSMV).3.1.3 Additional measuring
As each wave has an increasing number of participating countries, an analysis limited to countries that have taken part during the whole period was performed.
Limiting the four countries from the first wave produces a fairly similar picture as presented above. Among 15-year old girls the ASPs increase by 122.43 percent (from 3.70 to 8.23
percent), and FSMV by 35.90 percent (from 26.35 to 35.81). Among 15-year old boys the ASPs increase by 124.67 percent (from 1.50 to 3.37 percent), and FSMV by 27.20 percent (from 18.38 to 23.38). 13-year-old girls increase their rates of ASPs by 90 percent (from 3 to 5.70 percent), and their FSMV by 17.93 percent (from 24.54 to 28.94). 13-year old boys increase their rate of ASPs by 13.60 percent (from 2.28 to 2.59 percent). The ASPs among 11-year-old boys decrease by 6.85 percent (from 2.19 to 2.04 percent), while 11-year old girls have a decrease of 10.62 percent (from 5.18 to 4.63 percent).
3.1.4 Sweden
The upward trend of mental health problems is also evident in Sweden. Looking at the rate of ASPs, the relative increase is strongest among male 15-year-olds, who have a massive increase of 2283 percent, (from 0.18 to 4.29 percent). The increase among girls in the same age group is
10 15 20 25 30 35 40 45 50
1985/1986 1989/1990 1993/1994 1997/1998 2001/2002 2005/2006
Austria Belgium Canada Croatia Czech Republic Denmark Estonia Finland France Germany Greece Greenland Hungary Ireland Israel Italy Latvia Lithuania Malta Netherlands Norway Poland Portugal Russian Federation Slovakia Slovenia Spain Sweden Switzerland Ukraine MKD UK United States All Linear (All )