Eur Child Adolesc Psychiatry DOI 10.1007/s00787-014-0638-4
ORIGINAL CONTRIBUTION
The longitudinal BELLA study: design, methods and first results on the course of mental health problems Ulrike Ravens‑Sieberer · Christiane Otto · Levente Kriston · Aribert Rothenberger · Manfred Döpfner · Beate Herpertz‑Dahlmann · Claus Barkmann · Gerhard Schön · Heike Hölling · Michael Schulte‑Markwort · Fionna Klasen · The BELLA study group
Received: 30 April 2014 / Accepted: 24 October 2014 © Springer-Verlag Berlin Heidelberg 2014
Abstract The high prevalence of mental health problems (MHP) in childhood and adolescence is a global health challenge of the 21st century. Information about age of onset, persistence and development of MHP in young people is necessary to implement effective prevention and intervention strategies. We describe the design and methods of the longitudinal BELLA study, which investigates developmental trajectories of MHP from childhood into adulthood, their determinants, and the utilisation of mental health services. First results on the developmental course of Members of the BELLA study group are: Ulrike Ravens-Sieberer and Fionna Klasen, Hamburg (Principal Investigators); Claus Barkmann, Hamburg; Monika Bullinger, Hamburg; Manfred Döpfner, Köln; Beate Herpertz-Dahlmann, Aachen; Heike Hölling, Berlin; Franz Resch, Heidelberg; Aribert Rothenberger, Göttingen; Sylvia Schneider, Bochum; Michael SchulteMarkwort, Hamburg; Robert Schlack, Berlin; Frank Verhulst, Rotterdam; Hans-Ulrich Wittchen, Dresden. Electronic supplementary material The online version of this article (doi:10.1007/s00787-014-0638-4) contains supplementary material, which is available to authorized users. U. Ravens‑Sieberer (*) · C. Otto · L. Kriston · C. Barkmann · M. Schulte‑Markwort · F. Klasen Department of Child and Adolescent Psychiatry, Psychotherapy, and Psychosomatics, University Medical Center HamburgEppendorf, Martinistr. 52, 20246 Hamburg, Germany e-mail: ravens‑[email protected] L. Kriston Department of Medical Psychology, University Medical Center Hamburg-Eppendorf, Hamburg, Germany A. Rothenberger Department of Child and Adolescent Psychiatry and Psychotherapy, University Medical Center Göttingen, Göttingen, Germany
MHP in children and adolescents are reported over a 6-year period. The BELLA study is the mental health module of the German National Health Interview and Examination Survey for children and adolescents (KiGGS). BELLA examines the mental health and well-being of children and adolescents aged 7–17 years (a representative subsample of KiGGS, n = 2,863 at baseline). Standardised screening measures served to identify MHP at baseline and at followups (1, 2, and 6 years later). Among children and adolescents participating at all measurement points (n = 1,255), 10 % showed clinically significant MHP at baseline (n = 130). Over the 6-year period, 74.3 % showed no signs of MHP (n = 933), 15.5 % had remitted (n = 194), 2.9 % showed persistent (n = 36) and 7.3 % acute or recurrent MHP (n = 92). Overall, MHP were more likely to occur between the age of 7 and 12 and after the age of 19 years. Regarding mental health service use, 33 % of the participants with acute or recurrent MHP (n = 30) and 63.9 % with persistent MHP used mental health services (n = 23). Mental health problems in children and adolescents have a
M. Döpfner Department of Psychiatry, Psychosomatics and Psychotherapy in Childhood and Adolescence, Medical Faculty, University of Cologne, Cologne, Germany B. Herpertz‑Dahlmann Department of Child and Adolescent Psychiatry, Psychosomatics and Psychotherapy, University Clinics, RWTH Aachen, Aachen, Germany G. Schön Department of Medical Biometry and Epidemiology, University Medical Center Hamburg-Eppendorf, Hamburg, Germany H. Hölling Robert Koch Institute, Berlin, Germany
13
high risk to persist into adulthood. In children and adolescents a low rate of mental health service use was observed, even among those with mental health problems. Keywords Mental health · Children and adolescents · Longitudinal study · BELLA study · Cohort · Germany
Introduction The high prevalence of mental health problems in childhood and adolescence is considered to be a global health challenge of the 21st century [1, 2]. This encompasses not only individual and familial suffering, but also societal implications, e.g., loss of productivity, increased costs in health care and education, as well as resource demands in criminal justice and welfare systems [3]. A meta-analysis including 33 cross-sectional studies (n = 72,978) found that 17.6 % of German children and adolescents show clinically relevant emotional and behaviour problems [4]. A Dutch population-based study reported that lifetime, 12-month and 30-day prevalence rates of any DSM-IV disorder were 45, 31 and 15 %, respectively [5]. The Norwegian Bergen Child study found an overall prevalence rate of 7 % for any psychiatric disorder in 8–10 year olds [6]. In the US National Comorbidity Survey the 12-months prevalence of any DSM-IV disorder in adolescents was 40.3 % [7], with most mental disorders beginning in adolescence and persisting into adulthood. The New Zealand Dunedin Multidisciplinary Health and Development Study followed a representative birth cohort of 1,037 babies until adulthood. At the age of 26 years, 73.9 % of all patients with a mental disorder had the onset before the age of 18, and 50 % before the age of 15 years [8]. Although up to 30 % of the population worldwide is expected to have mental disorders every year [9], at least two-thirds receive no treatment [10]. Mental disorders impose burden on all parties involved and have high direct and indirect costs for society. In 2010, data from 30 European countries quantified the total costs for mental and neurologic disorders at 798 billion Euros with more than 5,000 Euro per inhabitant [11]. To implement effective prevention and intervention strategies, information about onset, persistence, and developmental trajectories of mental health problems, as well as on risk and protective factors determining these trajectories is needed. In contrast to other Western countries, there is a noticeable lack of data from large-scale longitudinal studies in Germany. This paper describes the longitudinal study of the behaviour and wellbeing of children and adolescents in Germany (German: Befragung Seelisches Wohlbefinden und Verhalten)—the BELLA study. The design of the study allows analyses of developmental trajectories of mental
13
Eur Child Adolesc Psychiatry
health from childhood into young adulthood. Risk and protective factors are examined not only regarding their effects on the development of mental health problems, but also concerning their impact on persistence and remission. Additionally, the BELLA study focuses on the consequences of impaired mental health for somatic health, health-related quality of life, and psychosocial, academic, and vocational development. Finally, health care utilisation, the type of health care utilised (e.g., somatic vs. psychosocial) and obstacles to receiving the appropriate care for mental health problems are addressed. The objectives of the current paper are twofold: first, to describe the design and methods of the longitudinal BELLA study, and second to report on the developmental course of mental health problems in children and adolescents in Germany.
Methods Design The BELLA study is the mental health module of the German Health Interview and Examination Survey for children and adolescents (KiGGS) conducted by the Robert Koch Institute (RKI, Federal Public Health Institute of Germany) [12]. Both, the KiGGS and BELLA study, are designed as prospective longitudinal studies, representative for Germany. Sample The BELLA study sample is a representative subsample of the KiGGS survey, with KiGGS being the first nationwide survey that collects comprehensive data on the health status of children and adolescents in Germany. Participants in the KiGGS study at baseline were enrolled in two steps: first, 167 sampling units in cities and communities distributed nationwide were selected by means of a stratified random process; second, subjects were randomly chosen from the official registers of local resident offices [13]. The target sample size was 26,787. At an overall response rate of 66.6 %, the final KiGGS study population included 17,641 children and adolescents between birth and the age of 17. Data suggests a slight over-representation of middle compared to low socio-economic status in participating families, but there are hardly any differences for the health-related variables between responders and non-responders. Further details on the design and methods as well as the response of the KiGGS study are published elsewhere [13, 14]. For the BELLA study, a randomly selected subsample of 2,942 families with children aged 7–17 years (randomly selected for each age category) was drawn from the KiGGS sample. These families were informed about the BELLA
Eur Child Adolesc Psychiatry
study at their first visit to the KiGGS examination centre. They were then contacted directly by the BELLA team at baseline and asked to participate in the BELLA study. Finally, 2,863 (97.3 % of n = 2,942) families gave their written informed consent and were included in the BELLA study as participants [15]. The BELLA baseline sample was uniformly distributed with respect to the age and gender of the participants: a number of 1,129 families had children between the ages of 7 and 10 years, a number of 778 families had children aged 11–13 years, and 956 families had children between the ages of 14 and 17 years. A total of 48.5 % of the participating children and adolescents were girls, 51.5 % were boys. The BELLA study was able to draw on very comprehensive information when conducting its nonresponse analyses. Since data from the KiGGS core survey was available for all families that were invited to participate in the BELLA module, non-responders could be described in terms of sociodemographics and health indicators. An analysis of the responders and non-responders showed that the families who agreed to participate form a representative sample with respect to the German population. The sample deviated only slightly from the KiGGS sample structure regarding socio-economic status (slight over-representation of middle compared to low socio-economic status), but no differences for health-related variables were observed. A weighting procedure was applied to correct for deviations from the socio-demographic and socio-economic structure of the target population. The baseline data (n = 2,863) were weighted to correct for deviation of the sample from the age-, gender-, regionaland citizenship-structure of the German population (reference data 31.12.2004). In the BELLA study data were not collected in different languages, thus included families with migration background cannot be treated as representative of migrant families in Germany. Details on demographic data of the BELLA baseline sample are presented in Table 1 (including weighted data). Further details on the design, sample and methods of the BELLA baseline study including more information on representativeness and the weighting procedure are presented in Ravens-Sieberer and Kurth [12]. Approval for the BELLA study was obtained from the ethics committee of the University Hospital Charité in Berlin and the Federal Commissioner for Data Protection in Germany. Follow‑up Baseline data from both, the KiGGS survey and the BELLA mental health module were jointly collected between 2003 and 2006. The BELLA study sample was followed up in three additional measurement points, conducted between 2004 and 2007 (BELLA 1-year follow-up), between 2005 and 2008 (BELLA 2-year follow-up), and between 2009
and 2012 (BELLA 6-year follow-up). The third BELLA follow-up was again collected jointly with the first follow-up of the KiGGS survey (KiGGS 6-year follow-up). Information on study conduct, sample design and response of the first follow-up of the KiGGS survey have recently been published [16]. Children and adolescents participating in the BELLA study at baseline (7–17 years old) were re-invited for the BELLA 6-year follow-up (then being 13– 23 years old, n = 1,429). Of the 2,863 participants included in the BELLA study at baseline, 2,423 (84.6 %) participated in the 1-year follow-up, 2,190 (76.5 %) participated in the 2-year followup and 1,429 (49.9 %) in the 6-year follow-up (see Fig. 1). At 1-year and 2-year follow-up, all BELLA baseline participants were re-invited directly by the BELLA study team (n = 2,863). However, BELLA participants for the 6-year follow-up were re-invited by the KiGGS study team and— due to new German data protection regulations—needed to give an additional second consent to participate in BELLA, therefore numbers of re-invited and successfully reached BELLA participants were smaller for the last measurement point (n = 1,679 and n = 1,429; see Fig. 1). Besides re-assessing the cohort, the BELLA measurement point 2009–2012 additionally included new BELLA participants aged 3–26 (n = 2,411) to compensate for loss to follow-up and re-establish a cross-sectional representativeness of the sample 2009–2012. To recruit these new BELLA participants, a subsample of families was randomly drawn from the KiGGS survey and asked for willingness to participate in the BELLA study. Information on age of BELLA participants at each measurement point and of new participants are presented in the supplementary material (Supplementary Table 1). Drop-out for our analysis was defined as no mental health data being available at a corresponding measurement point of the BELLA study. Attrition was mainly due to participants reporting lack of time, no interest in the study, or had moved and could thus not be contacted because a new address was not available (for the 6-year follow-up, 7.4 % (n = 125) of families re-invited for the BELLA study could not be contacted at all). Logistic regression analysis predicting drop-out (i.e. missing data at a specific measurement point) indicated that at all measurement points, participants with lower SES (1-year follow-up: OR = 1.07; 95 % CI = 1.02–1.11; 2-year follow-up: OR = 1.06; 95 % CI = 1.02–1.10; 6-year follow-up: OR = 1.08; 95 % CI = 1.05–1.11) and those with a migration background (1-year follow-up: OR = 2.55; 95 % CI = 1.22–5.34; 2-year follow-up: OR = 2.12; 95 % CI = 1.08–4.15; 6-year follow-up: OR = 2.26; 95 % CI = 1.11–4.60) were lost significantly more often compared to participants with higher SES, or non-migrants. Neither gender, community size, region (East/West Germany), nor parent-reported general
13
Table 1 Socio-demographic characteristics of the BELLA sample at baseline and of the analysed sample of children and adolescents participating in all four measurement points of the BELLA study
SD Standard deviation a Data weighted to correct for deviation of the sample from the age-, gender-, regional- and citizenship-structure of the German population (reference data 31.12.2004), n = 2,863; source: Ravens-Sieberer et al. [15] b Children and adolescents participating at all four measurement points with no missing data on mental health measures on any measurement point of the BELLA study
Eur Child Adolesc Psychiatry BELLA baseline samplea (n = 2,863)
Gender Male Female Region Former German Democratic Republic Former Federal Republic of Germany Age (years) 7–10 11–13 14–17 Migration background Yes No No information Family structure Living with both biological parents Living with a single biological parent Living with a biological parent and a new partner Living in another family structure No information Socio-economic status Low Middle High No information
health or parent-reported mental health scores of children and adolescents at baseline were significantly associated with drop-out status. A higher drop-out at 6-year follow-up is likely to be due to the fact that study participants were concerned that participation might be too time consuming, because participation in BELLA was only possible for those who had participated in KiGGS, and due to the fact that new data protection regulation required an additional second consent for BELLA at two different points in time, one gathered by the KiGGS study team and one gathered by the BELLA study team. Longitudinally consistently participating children and adolescents In the present analyses, we investigated a total of 1,255 children and adolescents (43.84 % from baseline sample, Table 1). This subsample was generated including only participants who participated at all four BELLA measurement
13
Mean (SD)
Longitudinally consistently participating children and adolescentsb (n = 1,255)
n
%
%
1,467 1,396
51.2 48.8
607 48.4 648 51.6
489 2,374
17.1 82.9
435 34.7 820 65.3
950 746 1,167
33.2 26.0 40.8
587 46.8 342 27.3 326 26.0
347 2,513 3
12.1 87.8 0.1
70 5.6 1,185 94.4
2,192 375 256
76.5 13.1 9.0
1,026 81.8 120 9.6 91 7.3
18 22
0.6 0.8
716 1,330 789
25.0 46.5 27.6
27
0.9
n
12.29 (3.191)
Mean (SD)
11.24 (3.092)
12 6
1.0 0.8
196 15.6 631 50.3 428 34.1
points with no missing data on mental health measures at any measurement point. This generated subsample includes 51.6 % girls, the mean age of participants being 11.24 years (SD 3.09), and the majority of families has a middle or high SES (84.4 %). Thus, our analysed subsample differs in terms of gender (2.8 % more girls), age (1.05 years younger), and SES (10.3 % more families with a middle or high SES) from the representative BELLA baseline sample. Measurements An overview of instruments that were utilised at the different measurement points of the BELLA study is provided in the supplementary material (Supplementary Table 2). The BELLA study used standardised instruments to assess different constructs relating to mental health problems, mental well-being and their determinants. In addition, data on known protective and risk factors as well as on health care utilisation were collected (Supplementary Table 2). Since
Eur Child Adolesc Psychiatry
Fig. 1 Numbers of invited and participating children and adolescents in the BELLA study
participants in the BELLA study also had participated in the KiGGS survey, a large number of indicators of somatic health (e.g., vaccination status, medication use, anthropometry, blood pressure, heart rate, whole blood, serum, urine) were available and linked to mental health indicators from the BELLA study. In the present analyses of the longitudinal BELLA study, the following measures were used: to assess mental health problems both the parent- and self- report of the Strengths and Difficulties Questionnaire (SDQ) [17, 18] were administered along with the SDQ Impact supplement at baseline and at every follow-up. For children younger than 11 years the parent-report of the SDQ was administered and analysed, while for older participants the selfreport of the SDQ was used. Standardised instruments were used in the BELLA study for assessing specific mental health problems, such as anxiety (assessed with the Screen for Child Anxiety Related Emotional Disorders (SCARED) [19, 20]), depressive symptoms (assessed with the Center for Epidemiological Studies Depression Scale for Children (CES-DC) [21] and the Patient Health Questionnaire-8 for young adults (PHQ) [22, 23]), conduct disorders (assessed with the Child Behaviour Checklist (CBCL) [24, 25]), and attention deficit-hyperactivity disorder (assessed with
the Conners‘ Rating Scales-Revised (CRS-R) [26, 27]). Similar to the administration of the SDQ, we used parentreports of the SCARED, CES-DC and CRS-R for children younger than 11 years, while self-reported data of these measures were analysed for older participants. Concerning the CBCL, only parent-reported data were assessed. For more detail see Supplementary Table 2 and Ravens-Sieberer and Kurth [12]. In addition, the BELLA 6-year follow-up included the Diagnostic Interview for Mental Disorders in Children and Adolescents (DIPS) [28, 29], which allows identification of mental disorders based on DSM-IV and ICD-10 criteria. Mental health service utilisation was assessed via self- and parent-reports at baseline. The child’s or adolescent’s utilisation of any physician(s), psychologist(s) or therapist(s) in the last 12 months (e.g. dentist, surgeon, general practicioner, psychiatrist, psychologist, psychotherapist, speech therapist) was assessed from both, the children and adolescents themselves, as well as from their parents. Visiting a psychiatrist, psychologist or psychotherapist was considered mental health service use. Concerning demographic variables, SES was assessed using the Winkler-index [30]. This index gathers data according to education, profession and income of both
13
parents resulting in a sum-score ranging from 3 to 21. Participants were categorised into groups with low (scores from 3 to 8), middle (scores from 9 to 14) and high SES (scores from 15 to 21) [31]. Migration background was assessed by means of five items offered to children and adolescents and parents, respectively, at baseline. Country of birth of children and adolescents as well as of both parents was assessed. Additionally, data on the nationalities of the parents and the language spoken at home were collected. A participant was classified as migrant if he/she as well as at least one parent were not born in Germany or if both parents were not born in Germany and had no German nationality [32]. In general, missing data in the instruments administered were handled following the corresponding manuals. If no information was provided for a measure, 30 % missing items per scale were tolerated and replaced using the Expectation–Maximization algorithm for calculating scale scores (see Little et al. [33]). Procedures After the overall study procedure had been fully explained, written consent was obtained from adolescents (aged 14 years and older) and from parents at all measurement points. Hereafter, trained and supervised interviewers conducted a computer-assisted telephone interview (CATI) to assess mental health with children and adolescents aged at least 11 years (self-report) and one parent (proxy-report) at each measurement point. Standard operating procedures ensured the quality of the interviews at all steps (establishing contact, conducting and ending of interviews). A child and adolescent psychiatrist supervised the interviewers and was available to provide psychological support throughout all points of data collection, although such support was never solicited. In addition to the interviews, participants were asked to complete a mailed questionnaire. To assure the quality of the data, experts from the Robert Koch Institute monitored the data collection, data entry, data preparation, and data processing. A scientific advisory board of experts in the fields of epidemiology, pediatrics, health surveys and ethics supervised the study [12]. Case definition: mental health problems A participating child or adolescent was defined as suffering from a mental health problem if (a) significant symptoms of mental health disorders and (b) impairment due to the symptoms were present in self- or parent-reports. Significant symptoms of mental health disorders in children and adolescents were operationalised by means of the total difficulties score of the SDQ, reported symptoms of anxiety based on the SCARED, depression based on the CES-DC,
13
Eur Child Adolesc Psychiatry
ADHD based on the CRS-R, conduct disorder based on the CBCL and/or due to reported mental health diagnoses by a physician. Impairment was defined using the SDQ Impact score for children and adolescents. For the total difficulties score of the SDQ we used available international cut-offs (www.sdq.org) [18, 34]. Additionally, we used published international cut-offs to dichotomise total scale scores of the SCARED [19], CES-DC [35], CRS-R [36] and for the externalising behaviours scale of the CBCL [24]. For young adults (participants aged 18 years or older at 6-year follow-up), significant symptoms of mental health disorders were given if any psychiatric disorder was given in the DIPS-interview and/or if the cutoff for the PHQ depression was exceeded [37, 38]. Further, specific questions were administered to young adults to assess impairment (“In the past 4 weeks, how often have you been bothered by mental health problems, e.g. anxiety, low mood or irritability?” and “In the past 4 weeks, to what degree have your personal or mental health problems hindered you to carry out your normal activities at work, at school, at university or similar settings of everyday life?”). Based on these data, participants were categorised as cases or non-cases for baseline as well as for each follow-up. Based on the case definition, participants were categorised into four groups to describe transitions between states with and without mental health problems over time in the BELLA study. Young people were considered to have persistent mental health problems if they were categorised as cases at three or all measurement points including the 6-year follow-up. Acute or recurrent mental health problems were assumed, if mental health problems were reported at one or two points of data collection including the 6-year follow-up. If no mental health problems were reported at the 6-year follow-up, but at one to three former measurement points, we assumed remitted mental health problems. Young people were considered to be mentally healthy (i.e., to have no mental health problems), if no significant mental health problems had been reported at any point of assessment. Data analysis Data were investigated based on the above defined case and group definitions describing the proportion of cases for each measurement point, describing the defined transition categories across the measurement points of the study, comparing the groups according to transitions between states due to mental health problems and investigating the course of risk of mental health problems across age using multilevel models. Groups were compared concerning gender, age and socio-economic status at baseline as well as regarding the reported health care utilisation by a psychiatrist, psychologist or psychotherapist at any measurement
Eur Child Adolesc Psychiatry
point using analysis of variance (ANOVA) for dimensional and chi-squared test for categorical measures. The course of the risk of mental health problems between the age of 7 and 24 years (age groups with available data from at least 50 participants) was investigated using generalised random-effects linear models using a logit link. In this analysis, linear, quadratic, and cubic growth models were tested with random intercepts and random linear slopes, including tests for a possible moderating effect of gender. The bestfitting model was selected using Akaike’s information criterion (AIC) [39]. To sustain consistency between the applied methods, we performed only complete case analyses including participants with information at all measurement points. Given the lack of a reasonable basis for estimating the large proportion of missing data found at the 6-year follow-up, any data imputation or weighting would have been likely to lead to results that are driven more strongly by the assumptions about the missing data mechanisms than by the actually observed pieces of information themselves. Our strategy may slightly impair representativeness of findings due to selective drop-out, but is expected to be internally unbiased.1 All analyses were performed with PASW version 20 and R version 2.15.3 using the lme4 package [40].
Results Cases with mental health problems and state transitions over time in the BELLA study Of the young people who participated at all four BELLA measurement points (n = 1,255), 10.4 % (n = 130) showed clinically significant mental health problems at baseline, 11.2 % (n = 141) at the 1-year follow-up, 10.6 % (n = 133) at the 2-year follow-up, and 10.2 % (n = 128) at the 6-year follow-up, respectively. Transitions of participants between states with and without mental health problems across the measurement points are presented in Fig. 2. Of the participants with mental health problems at baseline (n = 130), 51.5 % also had problems at the 1-year follow-up, 50 % of the baseline cases had problems at the 2-year follow-up and 31.5 % of the cases at baseline reported mental health problems also at the 6-year follow-up. From those reporting mental health problems at the 1-year follow-up (n = 141), 42.6 % also reported mental health problems 1 year later at the 2-year follow-up, and from those reporting mental health
1 Other analyses of the BELLA data within this ECAP special issue used various procedures to deal with missing data, so that robustness of the findings can easily be assessed.
problems at the 2-year follow-up, 30.1 % reported also mental health problems at the 6-year follow-up. At 1-year follow-up, 48.5 % of those showing significant mental health problems at baseline had recovered. Developmental trajectory of the probability of experiencing mental health problems The best-fitting model describing the developmental trajectory of the presence of mental health problems included statistically significant linear, quadratic and cubic slopes (with a random intercept and a random linear slope). No differences between genders could be identified, but differences between individuals regarding the initial probability of mental health problems and the linear change of this probability were statistically significant and substantial. The average estimated developmental trajectory of the probability of developing mental health problems between the age of 7 and 24 years is reported in Fig. 3; however, due to the large individual differences the displayed course should be seen as a crude approximation and interpreted with caution. The figure displays the estimated probability that a child/adolescent shows mental health problems between his/her 7th and 24th life years. Probabilities are displayed as decimals on the y axis. Circles indicate observed proportions for mental health problems for each age group, respectively. The size of the circles corresponds to the number of available data points (ranging from 57 to 438). The displayed trajectory line is estimated from a cubic generalised linear regression model. The findings indicate a doubling of the probability of developing mental health problems between the age of 7 and 12 years with an increase from around 6 to 12 %. Following a decrease to around 9 % at the age of 19, the probability rises continuously and exceeds 20 % at the age of 24. Groupings based on persistence of mental health problems Among all participating children and adolescents at all points of data collection (n = 1,255), approximately twothirds (74.3 %) had no mental health problems (i.e., were mentally healthy) over the 6-year period, 7.3 % had acute or recurrent mental health problems, 2.9 % had persistent mental health problems, and 15.5 % had remitted mental health problems (see Table 2). Groups differed significantly according to gender (p = 0.028), age (p = 0.033) and mental health service use (p
ORIGINAL CONTRIBUTION
The longitudinal BELLA study: design, methods and first results on the course of mental health problems Ulrike Ravens‑Sieberer · Christiane Otto · Levente Kriston · Aribert Rothenberger · Manfred Döpfner · Beate Herpertz‑Dahlmann · Claus Barkmann · Gerhard Schön · Heike Hölling · Michael Schulte‑Markwort · Fionna Klasen · The BELLA study group
Received: 30 April 2014 / Accepted: 24 October 2014 © Springer-Verlag Berlin Heidelberg 2014
Abstract The high prevalence of mental health problems (MHP) in childhood and adolescence is a global health challenge of the 21st century. Information about age of onset, persistence and development of MHP in young people is necessary to implement effective prevention and intervention strategies. We describe the design and methods of the longitudinal BELLA study, which investigates developmental trajectories of MHP from childhood into adulthood, their determinants, and the utilisation of mental health services. First results on the developmental course of Members of the BELLA study group are: Ulrike Ravens-Sieberer and Fionna Klasen, Hamburg (Principal Investigators); Claus Barkmann, Hamburg; Monika Bullinger, Hamburg; Manfred Döpfner, Köln; Beate Herpertz-Dahlmann, Aachen; Heike Hölling, Berlin; Franz Resch, Heidelberg; Aribert Rothenberger, Göttingen; Sylvia Schneider, Bochum; Michael SchulteMarkwort, Hamburg; Robert Schlack, Berlin; Frank Verhulst, Rotterdam; Hans-Ulrich Wittchen, Dresden. Electronic supplementary material The online version of this article (doi:10.1007/s00787-014-0638-4) contains supplementary material, which is available to authorized users. U. Ravens‑Sieberer (*) · C. Otto · L. Kriston · C. Barkmann · M. Schulte‑Markwort · F. Klasen Department of Child and Adolescent Psychiatry, Psychotherapy, and Psychosomatics, University Medical Center HamburgEppendorf, Martinistr. 52, 20246 Hamburg, Germany e-mail: ravens‑[email protected] L. Kriston Department of Medical Psychology, University Medical Center Hamburg-Eppendorf, Hamburg, Germany A. Rothenberger Department of Child and Adolescent Psychiatry and Psychotherapy, University Medical Center Göttingen, Göttingen, Germany
MHP in children and adolescents are reported over a 6-year period. The BELLA study is the mental health module of the German National Health Interview and Examination Survey for children and adolescents (KiGGS). BELLA examines the mental health and well-being of children and adolescents aged 7–17 years (a representative subsample of KiGGS, n = 2,863 at baseline). Standardised screening measures served to identify MHP at baseline and at followups (1, 2, and 6 years later). Among children and adolescents participating at all measurement points (n = 1,255), 10 % showed clinically significant MHP at baseline (n = 130). Over the 6-year period, 74.3 % showed no signs of MHP (n = 933), 15.5 % had remitted (n = 194), 2.9 % showed persistent (n = 36) and 7.3 % acute or recurrent MHP (n = 92). Overall, MHP were more likely to occur between the age of 7 and 12 and after the age of 19 years. Regarding mental health service use, 33 % of the participants with acute or recurrent MHP (n = 30) and 63.9 % with persistent MHP used mental health services (n = 23). Mental health problems in children and adolescents have a
M. Döpfner Department of Psychiatry, Psychosomatics and Psychotherapy in Childhood and Adolescence, Medical Faculty, University of Cologne, Cologne, Germany B. Herpertz‑Dahlmann Department of Child and Adolescent Psychiatry, Psychosomatics and Psychotherapy, University Clinics, RWTH Aachen, Aachen, Germany G. Schön Department of Medical Biometry and Epidemiology, University Medical Center Hamburg-Eppendorf, Hamburg, Germany H. Hölling Robert Koch Institute, Berlin, Germany
13
high risk to persist into adulthood. In children and adolescents a low rate of mental health service use was observed, even among those with mental health problems. Keywords Mental health · Children and adolescents · Longitudinal study · BELLA study · Cohort · Germany
Introduction The high prevalence of mental health problems in childhood and adolescence is considered to be a global health challenge of the 21st century [1, 2]. This encompasses not only individual and familial suffering, but also societal implications, e.g., loss of productivity, increased costs in health care and education, as well as resource demands in criminal justice and welfare systems [3]. A meta-analysis including 33 cross-sectional studies (n = 72,978) found that 17.6 % of German children and adolescents show clinically relevant emotional and behaviour problems [4]. A Dutch population-based study reported that lifetime, 12-month and 30-day prevalence rates of any DSM-IV disorder were 45, 31 and 15 %, respectively [5]. The Norwegian Bergen Child study found an overall prevalence rate of 7 % for any psychiatric disorder in 8–10 year olds [6]. In the US National Comorbidity Survey the 12-months prevalence of any DSM-IV disorder in adolescents was 40.3 % [7], with most mental disorders beginning in adolescence and persisting into adulthood. The New Zealand Dunedin Multidisciplinary Health and Development Study followed a representative birth cohort of 1,037 babies until adulthood. At the age of 26 years, 73.9 % of all patients with a mental disorder had the onset before the age of 18, and 50 % before the age of 15 years [8]. Although up to 30 % of the population worldwide is expected to have mental disorders every year [9], at least two-thirds receive no treatment [10]. Mental disorders impose burden on all parties involved and have high direct and indirect costs for society. In 2010, data from 30 European countries quantified the total costs for mental and neurologic disorders at 798 billion Euros with more than 5,000 Euro per inhabitant [11]. To implement effective prevention and intervention strategies, information about onset, persistence, and developmental trajectories of mental health problems, as well as on risk and protective factors determining these trajectories is needed. In contrast to other Western countries, there is a noticeable lack of data from large-scale longitudinal studies in Germany. This paper describes the longitudinal study of the behaviour and wellbeing of children and adolescents in Germany (German: Befragung Seelisches Wohlbefinden und Verhalten)—the BELLA study. The design of the study allows analyses of developmental trajectories of mental
13
Eur Child Adolesc Psychiatry
health from childhood into young adulthood. Risk and protective factors are examined not only regarding their effects on the development of mental health problems, but also concerning their impact on persistence and remission. Additionally, the BELLA study focuses on the consequences of impaired mental health for somatic health, health-related quality of life, and psychosocial, academic, and vocational development. Finally, health care utilisation, the type of health care utilised (e.g., somatic vs. psychosocial) and obstacles to receiving the appropriate care for mental health problems are addressed. The objectives of the current paper are twofold: first, to describe the design and methods of the longitudinal BELLA study, and second to report on the developmental course of mental health problems in children and adolescents in Germany.
Methods Design The BELLA study is the mental health module of the German Health Interview and Examination Survey for children and adolescents (KiGGS) conducted by the Robert Koch Institute (RKI, Federal Public Health Institute of Germany) [12]. Both, the KiGGS and BELLA study, are designed as prospective longitudinal studies, representative for Germany. Sample The BELLA study sample is a representative subsample of the KiGGS survey, with KiGGS being the first nationwide survey that collects comprehensive data on the health status of children and adolescents in Germany. Participants in the KiGGS study at baseline were enrolled in two steps: first, 167 sampling units in cities and communities distributed nationwide were selected by means of a stratified random process; second, subjects were randomly chosen from the official registers of local resident offices [13]. The target sample size was 26,787. At an overall response rate of 66.6 %, the final KiGGS study population included 17,641 children and adolescents between birth and the age of 17. Data suggests a slight over-representation of middle compared to low socio-economic status in participating families, but there are hardly any differences for the health-related variables between responders and non-responders. Further details on the design and methods as well as the response of the KiGGS study are published elsewhere [13, 14]. For the BELLA study, a randomly selected subsample of 2,942 families with children aged 7–17 years (randomly selected for each age category) was drawn from the KiGGS sample. These families were informed about the BELLA
Eur Child Adolesc Psychiatry
study at their first visit to the KiGGS examination centre. They were then contacted directly by the BELLA team at baseline and asked to participate in the BELLA study. Finally, 2,863 (97.3 % of n = 2,942) families gave their written informed consent and were included in the BELLA study as participants [15]. The BELLA baseline sample was uniformly distributed with respect to the age and gender of the participants: a number of 1,129 families had children between the ages of 7 and 10 years, a number of 778 families had children aged 11–13 years, and 956 families had children between the ages of 14 and 17 years. A total of 48.5 % of the participating children and adolescents were girls, 51.5 % were boys. The BELLA study was able to draw on very comprehensive information when conducting its nonresponse analyses. Since data from the KiGGS core survey was available for all families that were invited to participate in the BELLA module, non-responders could be described in terms of sociodemographics and health indicators. An analysis of the responders and non-responders showed that the families who agreed to participate form a representative sample with respect to the German population. The sample deviated only slightly from the KiGGS sample structure regarding socio-economic status (slight over-representation of middle compared to low socio-economic status), but no differences for health-related variables were observed. A weighting procedure was applied to correct for deviations from the socio-demographic and socio-economic structure of the target population. The baseline data (n = 2,863) were weighted to correct for deviation of the sample from the age-, gender-, regionaland citizenship-structure of the German population (reference data 31.12.2004). In the BELLA study data were not collected in different languages, thus included families with migration background cannot be treated as representative of migrant families in Germany. Details on demographic data of the BELLA baseline sample are presented in Table 1 (including weighted data). Further details on the design, sample and methods of the BELLA baseline study including more information on representativeness and the weighting procedure are presented in Ravens-Sieberer and Kurth [12]. Approval for the BELLA study was obtained from the ethics committee of the University Hospital Charité in Berlin and the Federal Commissioner for Data Protection in Germany. Follow‑up Baseline data from both, the KiGGS survey and the BELLA mental health module were jointly collected between 2003 and 2006. The BELLA study sample was followed up in three additional measurement points, conducted between 2004 and 2007 (BELLA 1-year follow-up), between 2005 and 2008 (BELLA 2-year follow-up), and between 2009
and 2012 (BELLA 6-year follow-up). The third BELLA follow-up was again collected jointly with the first follow-up of the KiGGS survey (KiGGS 6-year follow-up). Information on study conduct, sample design and response of the first follow-up of the KiGGS survey have recently been published [16]. Children and adolescents participating in the BELLA study at baseline (7–17 years old) were re-invited for the BELLA 6-year follow-up (then being 13– 23 years old, n = 1,429). Of the 2,863 participants included in the BELLA study at baseline, 2,423 (84.6 %) participated in the 1-year follow-up, 2,190 (76.5 %) participated in the 2-year followup and 1,429 (49.9 %) in the 6-year follow-up (see Fig. 1). At 1-year and 2-year follow-up, all BELLA baseline participants were re-invited directly by the BELLA study team (n = 2,863). However, BELLA participants for the 6-year follow-up were re-invited by the KiGGS study team and— due to new German data protection regulations—needed to give an additional second consent to participate in BELLA, therefore numbers of re-invited and successfully reached BELLA participants were smaller for the last measurement point (n = 1,679 and n = 1,429; see Fig. 1). Besides re-assessing the cohort, the BELLA measurement point 2009–2012 additionally included new BELLA participants aged 3–26 (n = 2,411) to compensate for loss to follow-up and re-establish a cross-sectional representativeness of the sample 2009–2012. To recruit these new BELLA participants, a subsample of families was randomly drawn from the KiGGS survey and asked for willingness to participate in the BELLA study. Information on age of BELLA participants at each measurement point and of new participants are presented in the supplementary material (Supplementary Table 1). Drop-out for our analysis was defined as no mental health data being available at a corresponding measurement point of the BELLA study. Attrition was mainly due to participants reporting lack of time, no interest in the study, or had moved and could thus not be contacted because a new address was not available (for the 6-year follow-up, 7.4 % (n = 125) of families re-invited for the BELLA study could not be contacted at all). Logistic regression analysis predicting drop-out (i.e. missing data at a specific measurement point) indicated that at all measurement points, participants with lower SES (1-year follow-up: OR = 1.07; 95 % CI = 1.02–1.11; 2-year follow-up: OR = 1.06; 95 % CI = 1.02–1.10; 6-year follow-up: OR = 1.08; 95 % CI = 1.05–1.11) and those with a migration background (1-year follow-up: OR = 2.55; 95 % CI = 1.22–5.34; 2-year follow-up: OR = 2.12; 95 % CI = 1.08–4.15; 6-year follow-up: OR = 2.26; 95 % CI = 1.11–4.60) were lost significantly more often compared to participants with higher SES, or non-migrants. Neither gender, community size, region (East/West Germany), nor parent-reported general
13
Table 1 Socio-demographic characteristics of the BELLA sample at baseline and of the analysed sample of children and adolescents participating in all four measurement points of the BELLA study
SD Standard deviation a Data weighted to correct for deviation of the sample from the age-, gender-, regional- and citizenship-structure of the German population (reference data 31.12.2004), n = 2,863; source: Ravens-Sieberer et al. [15] b Children and adolescents participating at all four measurement points with no missing data on mental health measures on any measurement point of the BELLA study
Eur Child Adolesc Psychiatry BELLA baseline samplea (n = 2,863)
Gender Male Female Region Former German Democratic Republic Former Federal Republic of Germany Age (years) 7–10 11–13 14–17 Migration background Yes No No information Family structure Living with both biological parents Living with a single biological parent Living with a biological parent and a new partner Living in another family structure No information Socio-economic status Low Middle High No information
health or parent-reported mental health scores of children and adolescents at baseline were significantly associated with drop-out status. A higher drop-out at 6-year follow-up is likely to be due to the fact that study participants were concerned that participation might be too time consuming, because participation in BELLA was only possible for those who had participated in KiGGS, and due to the fact that new data protection regulation required an additional second consent for BELLA at two different points in time, one gathered by the KiGGS study team and one gathered by the BELLA study team. Longitudinally consistently participating children and adolescents In the present analyses, we investigated a total of 1,255 children and adolescents (43.84 % from baseline sample, Table 1). This subsample was generated including only participants who participated at all four BELLA measurement
13
Mean (SD)
Longitudinally consistently participating children and adolescentsb (n = 1,255)
n
%
%
1,467 1,396
51.2 48.8
607 48.4 648 51.6
489 2,374
17.1 82.9
435 34.7 820 65.3
950 746 1,167
33.2 26.0 40.8
587 46.8 342 27.3 326 26.0
347 2,513 3
12.1 87.8 0.1
70 5.6 1,185 94.4
2,192 375 256
76.5 13.1 9.0
1,026 81.8 120 9.6 91 7.3
18 22
0.6 0.8
716 1,330 789
25.0 46.5 27.6
27
0.9
n
12.29 (3.191)
Mean (SD)
11.24 (3.092)
12 6
1.0 0.8
196 15.6 631 50.3 428 34.1
points with no missing data on mental health measures at any measurement point. This generated subsample includes 51.6 % girls, the mean age of participants being 11.24 years (SD 3.09), and the majority of families has a middle or high SES (84.4 %). Thus, our analysed subsample differs in terms of gender (2.8 % more girls), age (1.05 years younger), and SES (10.3 % more families with a middle or high SES) from the representative BELLA baseline sample. Measurements An overview of instruments that were utilised at the different measurement points of the BELLA study is provided in the supplementary material (Supplementary Table 2). The BELLA study used standardised instruments to assess different constructs relating to mental health problems, mental well-being and their determinants. In addition, data on known protective and risk factors as well as on health care utilisation were collected (Supplementary Table 2). Since
Eur Child Adolesc Psychiatry
Fig. 1 Numbers of invited and participating children and adolescents in the BELLA study
participants in the BELLA study also had participated in the KiGGS survey, a large number of indicators of somatic health (e.g., vaccination status, medication use, anthropometry, blood pressure, heart rate, whole blood, serum, urine) were available and linked to mental health indicators from the BELLA study. In the present analyses of the longitudinal BELLA study, the following measures were used: to assess mental health problems both the parent- and self- report of the Strengths and Difficulties Questionnaire (SDQ) [17, 18] were administered along with the SDQ Impact supplement at baseline and at every follow-up. For children younger than 11 years the parent-report of the SDQ was administered and analysed, while for older participants the selfreport of the SDQ was used. Standardised instruments were used in the BELLA study for assessing specific mental health problems, such as anxiety (assessed with the Screen for Child Anxiety Related Emotional Disorders (SCARED) [19, 20]), depressive symptoms (assessed with the Center for Epidemiological Studies Depression Scale for Children (CES-DC) [21] and the Patient Health Questionnaire-8 for young adults (PHQ) [22, 23]), conduct disorders (assessed with the Child Behaviour Checklist (CBCL) [24, 25]), and attention deficit-hyperactivity disorder (assessed with
the Conners‘ Rating Scales-Revised (CRS-R) [26, 27]). Similar to the administration of the SDQ, we used parentreports of the SCARED, CES-DC and CRS-R for children younger than 11 years, while self-reported data of these measures were analysed for older participants. Concerning the CBCL, only parent-reported data were assessed. For more detail see Supplementary Table 2 and Ravens-Sieberer and Kurth [12]. In addition, the BELLA 6-year follow-up included the Diagnostic Interview for Mental Disorders in Children and Adolescents (DIPS) [28, 29], which allows identification of mental disorders based on DSM-IV and ICD-10 criteria. Mental health service utilisation was assessed via self- and parent-reports at baseline. The child’s or adolescent’s utilisation of any physician(s), psychologist(s) or therapist(s) in the last 12 months (e.g. dentist, surgeon, general practicioner, psychiatrist, psychologist, psychotherapist, speech therapist) was assessed from both, the children and adolescents themselves, as well as from their parents. Visiting a psychiatrist, psychologist or psychotherapist was considered mental health service use. Concerning demographic variables, SES was assessed using the Winkler-index [30]. This index gathers data according to education, profession and income of both
13
parents resulting in a sum-score ranging from 3 to 21. Participants were categorised into groups with low (scores from 3 to 8), middle (scores from 9 to 14) and high SES (scores from 15 to 21) [31]. Migration background was assessed by means of five items offered to children and adolescents and parents, respectively, at baseline. Country of birth of children and adolescents as well as of both parents was assessed. Additionally, data on the nationalities of the parents and the language spoken at home were collected. A participant was classified as migrant if he/she as well as at least one parent were not born in Germany or if both parents were not born in Germany and had no German nationality [32]. In general, missing data in the instruments administered were handled following the corresponding manuals. If no information was provided for a measure, 30 % missing items per scale were tolerated and replaced using the Expectation–Maximization algorithm for calculating scale scores (see Little et al. [33]). Procedures After the overall study procedure had been fully explained, written consent was obtained from adolescents (aged 14 years and older) and from parents at all measurement points. Hereafter, trained and supervised interviewers conducted a computer-assisted telephone interview (CATI) to assess mental health with children and adolescents aged at least 11 years (self-report) and one parent (proxy-report) at each measurement point. Standard operating procedures ensured the quality of the interviews at all steps (establishing contact, conducting and ending of interviews). A child and adolescent psychiatrist supervised the interviewers and was available to provide psychological support throughout all points of data collection, although such support was never solicited. In addition to the interviews, participants were asked to complete a mailed questionnaire. To assure the quality of the data, experts from the Robert Koch Institute monitored the data collection, data entry, data preparation, and data processing. A scientific advisory board of experts in the fields of epidemiology, pediatrics, health surveys and ethics supervised the study [12]. Case definition: mental health problems A participating child or adolescent was defined as suffering from a mental health problem if (a) significant symptoms of mental health disorders and (b) impairment due to the symptoms were present in self- or parent-reports. Significant symptoms of mental health disorders in children and adolescents were operationalised by means of the total difficulties score of the SDQ, reported symptoms of anxiety based on the SCARED, depression based on the CES-DC,
13
Eur Child Adolesc Psychiatry
ADHD based on the CRS-R, conduct disorder based on the CBCL and/or due to reported mental health diagnoses by a physician. Impairment was defined using the SDQ Impact score for children and adolescents. For the total difficulties score of the SDQ we used available international cut-offs (www.sdq.org) [18, 34]. Additionally, we used published international cut-offs to dichotomise total scale scores of the SCARED [19], CES-DC [35], CRS-R [36] and for the externalising behaviours scale of the CBCL [24]. For young adults (participants aged 18 years or older at 6-year follow-up), significant symptoms of mental health disorders were given if any psychiatric disorder was given in the DIPS-interview and/or if the cutoff for the PHQ depression was exceeded [37, 38]. Further, specific questions were administered to young adults to assess impairment (“In the past 4 weeks, how often have you been bothered by mental health problems, e.g. anxiety, low mood or irritability?” and “In the past 4 weeks, to what degree have your personal or mental health problems hindered you to carry out your normal activities at work, at school, at university or similar settings of everyday life?”). Based on these data, participants were categorised as cases or non-cases for baseline as well as for each follow-up. Based on the case definition, participants were categorised into four groups to describe transitions between states with and without mental health problems over time in the BELLA study. Young people were considered to have persistent mental health problems if they were categorised as cases at three or all measurement points including the 6-year follow-up. Acute or recurrent mental health problems were assumed, if mental health problems were reported at one or two points of data collection including the 6-year follow-up. If no mental health problems were reported at the 6-year follow-up, but at one to three former measurement points, we assumed remitted mental health problems. Young people were considered to be mentally healthy (i.e., to have no mental health problems), if no significant mental health problems had been reported at any point of assessment. Data analysis Data were investigated based on the above defined case and group definitions describing the proportion of cases for each measurement point, describing the defined transition categories across the measurement points of the study, comparing the groups according to transitions between states due to mental health problems and investigating the course of risk of mental health problems across age using multilevel models. Groups were compared concerning gender, age and socio-economic status at baseline as well as regarding the reported health care utilisation by a psychiatrist, psychologist or psychotherapist at any measurement
Eur Child Adolesc Psychiatry
point using analysis of variance (ANOVA) for dimensional and chi-squared test for categorical measures. The course of the risk of mental health problems between the age of 7 and 24 years (age groups with available data from at least 50 participants) was investigated using generalised random-effects linear models using a logit link. In this analysis, linear, quadratic, and cubic growth models were tested with random intercepts and random linear slopes, including tests for a possible moderating effect of gender. The bestfitting model was selected using Akaike’s information criterion (AIC) [39]. To sustain consistency between the applied methods, we performed only complete case analyses including participants with information at all measurement points. Given the lack of a reasonable basis for estimating the large proportion of missing data found at the 6-year follow-up, any data imputation or weighting would have been likely to lead to results that are driven more strongly by the assumptions about the missing data mechanisms than by the actually observed pieces of information themselves. Our strategy may slightly impair representativeness of findings due to selective drop-out, but is expected to be internally unbiased.1 All analyses were performed with PASW version 20 and R version 2.15.3 using the lme4 package [40].
Results Cases with mental health problems and state transitions over time in the BELLA study Of the young people who participated at all four BELLA measurement points (n = 1,255), 10.4 % (n = 130) showed clinically significant mental health problems at baseline, 11.2 % (n = 141) at the 1-year follow-up, 10.6 % (n = 133) at the 2-year follow-up, and 10.2 % (n = 128) at the 6-year follow-up, respectively. Transitions of participants between states with and without mental health problems across the measurement points are presented in Fig. 2. Of the participants with mental health problems at baseline (n = 130), 51.5 % also had problems at the 1-year follow-up, 50 % of the baseline cases had problems at the 2-year follow-up and 31.5 % of the cases at baseline reported mental health problems also at the 6-year follow-up. From those reporting mental health problems at the 1-year follow-up (n = 141), 42.6 % also reported mental health problems 1 year later at the 2-year follow-up, and from those reporting mental health
1 Other analyses of the BELLA data within this ECAP special issue used various procedures to deal with missing data, so that robustness of the findings can easily be assessed.
problems at the 2-year follow-up, 30.1 % reported also mental health problems at the 6-year follow-up. At 1-year follow-up, 48.5 % of those showing significant mental health problems at baseline had recovered. Developmental trajectory of the probability of experiencing mental health problems The best-fitting model describing the developmental trajectory of the presence of mental health problems included statistically significant linear, quadratic and cubic slopes (with a random intercept and a random linear slope). No differences between genders could be identified, but differences between individuals regarding the initial probability of mental health problems and the linear change of this probability were statistically significant and substantial. The average estimated developmental trajectory of the probability of developing mental health problems between the age of 7 and 24 years is reported in Fig. 3; however, due to the large individual differences the displayed course should be seen as a crude approximation and interpreted with caution. The figure displays the estimated probability that a child/adolescent shows mental health problems between his/her 7th and 24th life years. Probabilities are displayed as decimals on the y axis. Circles indicate observed proportions for mental health problems for each age group, respectively. The size of the circles corresponds to the number of available data points (ranging from 57 to 438). The displayed trajectory line is estimated from a cubic generalised linear regression model. The findings indicate a doubling of the probability of developing mental health problems between the age of 7 and 12 years with an increase from around 6 to 12 %. Following a decrease to around 9 % at the age of 19, the probability rises continuously and exceeds 20 % at the age of 24. Groupings based on persistence of mental health problems Among all participating children and adolescents at all points of data collection (n = 1,255), approximately twothirds (74.3 %) had no mental health problems (i.e., were mentally healthy) over the 6-year period, 7.3 % had acute or recurrent mental health problems, 2.9 % had persistent mental health problems, and 15.5 % had remitted mental health problems (see Table 2). Groups differed significantly according to gender (p = 0.028), age (p = 0.033) and mental health service use (p
