Study population

Cross-sectional epidemiological data from N = 1,180 participants aged 14 to 21 years of the Behavior and Mind Health (BeMIND) study were used [20]. In 2015, an age- and sex-stratified random sample was drawn from the population registry of Dresden, Germany, and assessed between November 2015 and December 2016 (response/participation rate: 21.7%, cooperation rate: 42.8%). Participation was higher among females and those with higher education; lack of time and lack of interest were the most frequently given reasons for non-participation. For more details on sampling and recruitment procedures, please see Beesdo-Baum et al. [20].

Procedures

In the study center at the Technische Universität Dresden, two assessment days were conducted approximately one week apart including a standardized diagnostic interview, self-report questionnaires, cognitive paradigms, and bio-sampling (blood and hair samples). Ecologic Momentary Assessment in everyday life and an online questionnaire assessment took place between these two in-person appointments. Participants received 50 Euro as incentive.

Measures

An updated version of the fully standardized computer-assisted Munich-Composite International Diagnostic Interview (DIA-X/M-CIDI [21, 22]) was conducted face-to-face by trained clinical interviewers accompanied by tablet-based self-administered lists and questionnaires (DIA-X-5/D-CIDI [23]). Reliability and validity of the instrument have been established previously [22,23,24].

Lifetime mental disorders

The following diagnostic categories (including the respective mental disorders) were assessed according to DSM-5 criteria [25]: Substance Use Disorders (including tobacco use disorder, alcohol use disorder, any illegal substance use disorder), Possible Psychotic Disorder, Bipolar Disorders (bipolar I and II disorder), Depressive Disorders (major depressive disorder, persistent depressive disorder/ dysthymia), Anxiety Disorders (agoraphobia, social anxiety disorder, panic disorder, generalized anxiety disorder, specific phobias, other phobic anxiety disorder (situational type), separation anxiety disorder), Obsessive–Compulsive Disorder (OCD), Trauma-related disorders (acute stress disorder, post-traumatic stress disorder), Somatic Symptom or related disorders (somatic symptom disorder, illness anxiety disorder), Eating Disorders (anorexia nervosa, bulimia nervosa, binge eating disorder), Attention-Deficit Hyperactivity Disorder (ADHD; predominantly inattentive type, predominantly hyperactive type, mixed type), and Disruptive, Impulse-Control or Conduct Disorders (oppositional defiant disorder, intermittent explosive disorder, conduct disorder (childhood-onset and adolescent-onset type), antisocial personality disorder). For the present analyses, comorbidity was operationalized as having diagnoses in two or more diagnostic categories.

Health service use

Lifetime health service use for mental disorders was assessed in section Q of the DIA-X-5 interview by asking the following gate question while viewing a list of service institutions on the tablet screen: ‘Have you ever visited/contacted any of the health service institutions [as listed on the tablet screen] because of mental health, psychosomatic, or substance use problems, either by yourself, or through recommendation of others, e.g., medical doctors, relatives, or your partner?’. Health service use included inpatient treatment (e.g., psychiatric, neurological, psychotherapeutic, or psychosomatic hospitals), mental health outpatient treatment (e.g., delivered by psychiatric/psychotherapeutic outpatient clinics, resident psychiatrist/neurologist/psychotherapist/psychologist, and/or primary care physician), and complementary health services (e.g., different counseling centers). Individuals endorsing the gate question were further asked if they had been treated with cognitive-behavioral therapy (CBT), other psychotherapy, and/ or medication. Individuals who denied the gate question were asked if they had ever thought about using health services because of mental health, psychosomatic, or substance use problems.

Sociodemographic factors

The DIA-X-5/D-CIDI section A (see above) and self-report questionnaires delivered information on sociodemographic and psychological variables. Participants completed one part of the questionnaires on a tablet during on-site sessions and another part online between sessions. Education was dichotomized in high education (terminated or current education in higher secondary school, high school, or university) vs. low/middle/other education (terminated or current education in regular secondary school or complemented elementary school) for the means of the present analyses. The self-assessment of social class (lowest, lower middle, middle, upper middle, upper) was collapsed into three categories for the present analyses (low, middle, high) to achieve cell sizes no smaller than 5%. Migration background was categorized as present when one or both parents were not born in Germany [26].

Psychological correlates

The following psychological constructs were assessed using self-report questionnaires:

Self-esteem: The Single-Item Self-Esteem Scale (SISE) is a one-item measure of global self-esteem [27] that was translated into German language for the present study. It was rated on a 7-point Likert scale (1 = ”not very true of me”, 7 = ”very true of me”). Retest reliability and construct validity have been established [27].

Locus of control: Locus of control was assessed by the Internal–External Locus of Control-4 (IE-4) [28] with its’ two subscales Internal Control (Cronbach’s alpha in the current sample: alphaBeMIND = 0.66) and External Control (alphaBeMIND = 0.51). Each subscale had two items that were answered on a 5-point Likert scale (1 = ”strongly disagree”, 5 = ”fully agree”). The factor structure, reliability, and construct validity have been established by previous research [28].

Emotion regulation: The Emotion-Regulation Skills Questionnaire (ERSQ; German version: [29]) was used. Its 27 items assess nine competencies relevant for a successful emotion regulation. These were answered on a 5-point Likert scale (0 = ”not at all”, 4 = ”(almost) always”, alphaBeMIND = 0.94). The reliability and validity of the ERSQ have been confirmed [29,30,31].

Social support: The seven item-version of the Social Support Questionnaire (F-SozU; [32]) was used to assess the subjectively perceived or anticipated social support from the social network. It was answered on a 5-point Likert scale (1 = ”strongly disagree”, 5 = ”fully agree”, alphaBeMIND = 0.88). The factor structure, reliability, and construct validity of the F-SozU have been established [33].

Stigma: Personal stigma associated with mental disorders was assessed with one item (“I would be ashamed if I had a mental disorder”) which was answered on a visual analog scale with labeled endpoints (0 = ”that’s not at all like me”, 10 = ”that’s very much like me”).

Subjective physical health: One item (“How would you describe your physical health in general?”) which was answered on a 5-point Likert scale (1 = ”excellent”, 5 = ”bad”) was used to assess subjective physical health.

Statistical analyses

Data were weighted to improve representativeness regarding sex and age (for details on weighting and representativeness of the sample see Beesdo-Baum et al. [20]). Only absolute frequencies are reported unweighted. Psychological scale variables were z-standardized. Predictors of health service use were examined using logistic regression analyses with weighted data. We report univariate analyses to inform about the effects of predictors by themselves and multiple analyses that included all variables in a single model to examine the contribution of each predictor when adjusted for all other, non-collinear variables as covariates. Odds ratios (ORs) and 95% confidence intervals [95% CI] are given.

All analyses were carried out for the whole sample and for both sexes separately as differences in health service use between sexes and within age groups have been reported before. The calculations of weighted row percentages (%wrow) and logistic regression analyses were based on cases with complete information on the respective predictor variables (for sociodemographic and psychological variables, min. 11 and max. 78 individuals with any mental disorder did not provide data, for resulting sample sizes in the respective analyses, please refer to Table 2). Missing data for health service use were counted as no occurrence (twelve individuals with any mental disorder had not completed DIA-X-5/D-CIDI section Q on health service use). Significance was set at α = 0.05. As our epidemiological study is of exploratory nature, no multiplicity adjustment was applied [34]. All analyses were carried out with Stata version 15.1 [35].