Introduction

Nearly half of all mental disorders begin before the age of 18, and 62% manifest before the age of 25, with a peak age of onset for all mental disorders at 14.5 years [1]. In Europe, almost one in five children and adolescents suffer from a mental disorder [2]. Developing mental health issues before 14 years of age increases the risk of adult mental disorder [3]. Mental disorders are the leading cause of Years Lived with Disability (YLDs), and self-harm is a prominent cause of Years of Life Lost (YLLs) among young people in most European countries [4], highlighting the negative long-term impact of mental disorders at a young age on both individuals and societies.

Despite the high prevalence and significant impact of mental disorders in young people, a large proportion of them do not seek help. Help-seeking is recognized as an “adaptive process that is the attempt to obtain external assistance to deal with a mental health concern” that can be professional (e.g., mental health professionals, primary health care providers, teachers) or informal (e.g., friends, parents) [5]. In a representative school-based study across 11 European countries, 61.9% of adolescents aged 13–17 years were at risk for a mental disorder or risk behavior [6], but only 10% sought professional treatment within one year. Additionally, 4.2% reported current suicidality, yet the majority did not seek professional help within a year [7]. The COVID-19 pandemic has further exacerbated the delay and reduction in help-seeking for mental health problems [8].

Recent systematic reviews of qualitative and quantitative studies exploring barriers and facilitators of help-seeking for mental health problems from the perspective of the adolescents and their parents have identified various factors influencing help-seeking behavior that lie with the adolescents/families, the society, or the healthcare system [9,10,11]. These barriers include limited knowledge about mental health and available services, prior negative experiences with mental health professionals, negative attitudes towards mental health problems and help-seeking, stigmatization, preference for self-help or informal support, confidentiality concerns, service and indirect costs (e.g., travel costs, loss of wages), logistical barriers (e.g., long travel distances, limited availability of parents), and limited access to professional help (e.g., waiting times, difficulty in getting a referral, inflexible appointment systems) [10, 11]. However, the current evidence is constrained by selective samples (e.g., adolescents/families who accessed services), the lack of evaluated measures of mental health problems and of facilitators/barriers for help-seeking, the focus on help-seeking attitudes or intentions rather than actual help-seeking behavior, and the insufficient examination of the relative importance of different factors influencing help-seeking behavior.

To address these gaps, the current study had three main objectives. Firstly, to examine in a large school-based sample of adolescents the rates of professional and informal help-seeking for mental health problems. Secondly, to explore various sociodemographic and clinical factors associated with attitudes towards seeking assistance and actual help-seeking behavior and assess their relative importance. According to the theory of planned behavior, attitude is not truly a measure of help-seeking in the sense of an active coping attempt but influences observable behavior [5]. Thirdly, to investigate potential differences in sociodemographic and clinical factors influencing help-seeking attitudes and behavior between adolescents who exceeded a pre-defined threshold for relevant mental health problems (i.e., “the clinical group”) and those who did not (i.e., “the non-clinical group”).

Methods

Sample and procedure

The present study is a data analysis of baseline screening data collected within the ProHEAD (“Promoting Help-seeking using E-technology for ADolescents”) consortium. The consortium aims at (1) improving help-seeking behavior in young people who experience relevant mental health problems; (2) improving the selective prevention of common disorders in those who are at risk; and (3) strengthening resources to counteract the development of mental disorders in young people [12]. The consortium conducted longitudinal, school-based, online assessments of mental health problems in adolescents between 12 and 25 years old. After the completion of the initial screening assessment, participants were allocated to one of five online randomized controlled trials based on their screening results, followed by two annual follow-up assessments. The baseline screening was conducted at randomly selected schools in five regions of Germany (i.e., Heidelberg, Hamburg, Leipzig, Schwäbisch Gmünd, Marburg). The randomized selection was stratified by regional district and school-type to ensure the selection of a representative sample of schools within each recruitment area. Inclusion criteria were sufficient German skills and internet access. Eligible adolescents and their legal guardians were asked to provide written informed consent before participating. The study protocol was approved by the Ethics Committee of the Medical Faculty at the University of Heidelberg (S-086/2018; leading study site) as well as the respective institutional review boards of the additional study sites. For the present study, the data of the initial screening were used.

Instruments

Sociodemographic information

Sociodemographic information included age, gender [f/m], center [Heidelberg, Hamburg, Leipzig, Schwäbisch Gmünd or Marburg], school type [Gymnasium, Realschule, Haupt- und Werkrealschulen, Gemeinschaftsschulen (Oberschule, Stadtteilschule), Berufsschule, other], and migration status (no [0], unknown [2] or yes [1] , meaning that the hild or one of the parent was not born in Germany). The socioeconomic status (SES) was assessed using the Family Affluence Scale (FAS) [13]. The FAS is a short four-item measure, with a total score ranging between 0 and 2 indicating low, between 3 and 5 medium, and between 6 and 9 high SES. The Laucht-Index [14] was applied to assess psychosocial adversity (e.g. parental unemployment) that may pose a risk on the development of a child. It comprises 10 items and distinguishes no risk [0], low risk [1, 2] and high risk [> 2]. For the current analysis, the mean scores of the FAS and the Laucht-Index were used.

Psychopathology

The KIDSCREEN-10 was used to assess health-related quality of life [15]. It contains ten items with a score ranging from 0 [never/not at all/excellent] to 4 [very much/always/bad]. A higher total score reflects greater health-related quality of life. Furthermore, the Strengths and Difficulties Questionnaire (SDQ) [16] was applied to assess mental health problems in the previous six months. It is a 25-item measure, with each item being scored from 0 [not applicable] to 2 [clearly applicable]. A greater total score indicates more psychological problems. An extended version of the Patient Health Questionnaire-9 modified for adolescents (PHQ-A) [17] was used as a severity measure for depressive symptoms within the past two weeks. This questionnaire consists of ten items scored from 0 [not at all/no] to 3 [almost daily/extremely] as well as three items scored as 0 [no] or 1 [yes]. A greater total score means more severe symptoms of depression.

Features of eating disorder pathology in the previous four weeks were assessed by using the Eating Disorder Examination-Questionnaire for Children (ChEDE-Q) [18]. The ChEDE-Q is a 22-item measure scored from 0 [no day/never/not at all] to 6 [everyday/every time/clearly uncomfortable]. The total mean score was used as an indicator of severity of eating disorder symptoms.

A brief screening for personality disorders was performed using the Self-Rated Standardized Assessment of Personality - Abbreviated Scale (SAPAS) [19]. This questionnaire consists of eight items with dichotomous answer options [no = 0/yes = 1], with a greater total score reflecting more severe personality pathology.

Risk-taking behavior and self-harm

The Alcohol Use Disorders Identification Test (AUDIT) [20], which is composed of ten items, was applied to detect signs of alcohol use disorders in the previous 12 months. Its items are coded from 0 [never/1–2 glasses/no] to 4 [four times a week or more/ten or more/yes in the last year]. A greater total score indicates more severe alcohol misuse.

Suicidal ideation and attempts were assessed using the Paykel Scale [21]. Two items measured suicidal thoughts and plans in the past 12 months, rated on a scale from 0 [never] to 5 [always]. In contrast, suicidal thoughts and plans in the last two weeks were assessed as either present [1] or absent [0]. The occurrence of suicidal ideation was defined as a score of 1 [seldom] or higher on either the suicidal thoughts or the suicidal plans item for the last 12 months period, or as a score of 1 [present] on the suicidal thoughts or suicidal plans items for the last two weeks. Additionally, one item measured the occurrence of suicidal attempts [0 = no/1 = yes].

The life-time occurrence of non-suicidal self-injury (NSSI) [0 = no/1 = yes] was assessed using the respective item of the Self-Injurious Thoughts and Behavior Interview (SITBI) [22].

Help-seeking

Attitudes towards help-seeking for mental health problems were assessed using the Inventory of Attitudes Towards Seeking Mental Health Services (IASMHS) [23]. This questionnaire comprises 24 items scored from 0 [do not agree] to 4 [agree] and can be classified into three subscales, with eight items each: (a) psychological openness, which evaluates a person’s willingness to acknowledge and seek professional help for mental health problems, (b) help-seeking propensity, which measures the extent to which a person believes he or she is capable of seeking professional help and how strongly he or she is personally inclined to do so, and (c) indifference to stigma, which assesses how much a person worries about what other people might think, if they found out, that he or she is seeking professional help.

Actual help-seeking behavior was assessed by using the Actual Help-Seeking Questionnaire (AHSQ) [24] containing 13 items. The first question asks whether the individual has ever sought help for a mental health problem [0 = no; 1 = yes, in the last 12 months; 2 = yes, but it has been longer than 12 months]. In the case of help-seeking, items 2–13 list various forms of professional and informal help that can be marked as used [1] or not used [0]. Dichotomous items [0 = no; 1 = yes, in the last 12 months or more than 12 months ago] for professional [i.e., item 2 (partner), 3 (friend), 4 (parent), 5 (other family member)] and informal [i.e., item 6 (school psychologist/social worker), 7 (psychotherapist), 8 (psychiatrist), 9 (counselling centre), 10 (telephone counselling), 11 (family doctor), 12 (teacher)] help were used for the current analysis [24, 25].

Barriers to help-seeking behavior were surveyed using a 14-item questionnaire that was based on an extensive literature review and specifically designed for the purpose of the ProHEAD consortium. The first item asked whether the individual would seek professional help for mental health problems and was dichotomous [0 = no/1 = yes]. If “no” was indicated, items 2 to 14 were presented listing potential reasons for the lack of help-seeking behavior. These items were scored from 0 [not true] to 3 [true]. The mean score, indicating more perceived barriers to seeking help, was included in the current analysis.

Data analysis

Participants were allocated to the clinical group or non-clinical group, respectively, reflecting whether they reached a predefined threshold for relevant mental health problems or not. The threshold reflects the allocation criteria to the first RCT of the ProHEAD consortium (i.e., ProHEAD Online by Kaess et al. [26]), which are detailed in Online Resource 1. Pearson correlations between all variables revealed a high overlap between the SDQ, the KIDSCREEN-10, and the PHQ-A (see Fig. 1 in Online Resource 1). Accordingly, a composite score reflecting “general psychopathology” was created by summing up the z-standardized scores of the three scales. Notably, the KIDSCREEN-10 score was multiplied by − 1 before the composite score was built. For descriptive purposes, the number and percentages for categorical variables and the mean and standard deviations for continuous variables were calculated.

Fig. 1
figure 1

Regression coefficients and 95% confidence intervals for each predictor for help-seeking attitudes (IASMHS subscales). Continuous predictors were standardized (by one standard deviation). The predictor “center” and the subgroup “migration status unknown” (n = 25) are not included

Full size image

To explore predictors of help-seeking attitudes and behavior, three linear regression analyses were conducted for IASMHS psychological openness, help-seeking propensity, and indifference to stigma, and three logistic regression analyses for AHSQ actual help-seeking, professional help-seeking, and informal help-seeking as outcome variables, respectively. Age, gender, center, migration status, socioeconomic status (FAS), psychosocial adversity (Laucht-Index), general psychopathology (composite score derived from SDQ, KIDSCREEN-10, and PHQ-A), eating disorder psychopathology (ChEDE-Q), personality pathology (SAPAS), alcohol problems (AUDIT), suicidal ideation (Paykel), suicide attempts (Paykel), and NSSI (SITBI) were entered as predictor variables into the regression analyses. In the regression analyses with the AHSQ variables as outcomes, the IASMHS subscales and the barriers to help-seeking questionnaire were considered as additional predictors. Continuous predictor variables were z-standardized to facilitate comparison between predictors. The importance of single predictor variables was determined using the LMG formula [27]. LMG calculates the relative contribution of each predictor to the explained total variance (R2) for linear regressions and the explained Mc Fadden’s pseudo R2 for logistic regressions, respectively, with the consideration of the sequence of predictors appearing in the model. The explained variance (in % of the total variance) or pseudo R2 by each predictor variable reflecting its relative importance estimated by LMG is presented in the results below. The regression coefficients for linear regression and Odds Ratios (OR) for logistic regressions, respectively, representing the direction and strength of the relationships between predictor and outcome variables can be found in Online Resource 1. Finally, to explore whether the importance of predictor variables shows differences between adolescents who reached the threshold for relevant mental health problems and those who did not, all regression analyses were repeated including interaction terms clinical group-status × predictor for all predictor variables. The inclusion of the interaction terms allows the examination of the clinical group-status as a moderator of the associations between the predictor variables and the outcomes. As suggested by a reviewer, the regression analyses predicting help-seeking attitudes (IASMHS subscales) and actual help-seeking behavior (AHSQ subscales) were repeated with the SDQ subscales (i.e., conduct problems, emotional symptoms, hyperactivity/inattention, peer relationship problems, and prosocial behavior), the KIDSCREEN-10 and the PHQ-9 as individual predictors instead of the general psychopathology composite score. Full results of these additional analyses are provided in Online Resource 1 (Tables 3–6). The overall patterns of findings remained unchanged. All analyses were run with R [28] using r packages relaimpo and sensitivity for LMG calculations.

Results

Sample characteristics

Out of 767 invited schools, 185 (24.12%) did not respond, 378 declined (49.28%), and 195 (25.42%) agreed to participate in the study. While 45’084 adolescents were invited to take part in the study, 9954 (22.08%) initiated, and 9509 (21.09%) completed the initial screening assessment, and were, thus, included in the current study.

A total of 1606 participants (16.9%) showed clinically relevant mental health problems (i.e., the clinical group). Of those adolescents, 895 (55.7%) reported having sought help (in a lifetime), with higher rates observed for informal (n = 834, 51.9%) compared to professional (n = 411, 25.6%) sources. Detailed sociodemographic and clinical data of the whole sample as well as of the clinical and non-clinical groups are provided in Table 1.

Table 1 Sociodemographic and clinical characteristics of the sample (N = 9509)
Full size table

Correlates of attitudes toward seeking mental health services

All sociodemographic and clinical factors together explained 15.84% of the variance of indifference to stigma, 9.91% of the variance of help-seeking propensity, and 6.84% of the variance of psychological openness. The respective proportions of the variances explained by each predictor reflecting their relative importance are shown in Table 2. The regression coefficients (and their 95% confidence intervals) are presented in Table 1 in Online Resource 1 and illustrated in Fig. 1.

Table 2 Respective proportions (%) of the total variance (R2) of psychological openness, help-seeking propensity, and indifference to stigma (IASMHS) explained by each predictor
Full size table

Correlates of actual help-seeking behavior

Sociodemographic, clinical, and attitudinal factors as well as perceived barriers had pseudo R2 values of 20.44% for actual help-seeking behavior, 22.62% for professional help-seeking, and 18.75% for informal help-seeking. The respective proportions of the pseudo R2 explained by each predictor reflecting their relative importance are given in Table 3. The OR (and their 95% confidence intervals) are presented in Table 2 in Online Resource 1 and illustrated in Fig. 2.

Table 3 Respective proportions (%) of the pseudo R2 for actual help-seeking, professional help-seeking, and informal help-seeking (AHSQ) explained by each predictor
Full size table
Fig. 2
figure 2

Odds ratios (OR) and 95% confidence intervals for each predictor for help-seeking behavior (AHSQ subscales). Continuous predictors were standardized (by one standard deviation). The predictor “center” and the subgroup “migration status unknown” (n = 25) are not included

Full size image

Moderation by clinical group-status

The additional amount of variance or pseudo R2 for the outcome variables explained by the interaction terms clinical group status × predictor (for all variables) ranged between 0.003 and 0.007 (see Table 4). This indicates that the moderation of the associations between the sociodemographic and clinical variables and attitudes towards or actual help-seeking behavior, respectively, by the clinical group status was negligible.

Table 4 Total explained variance or pseudo R2, respectively, of attitudes towards seeking mental health services (IAMHS subscales) and actual help-seeking behavior (AHSQ subscales) by (A) all predictors, and (B) when the interactions “clinical group status x predictor” (for all variables) were additionally considered
Full size table

Discussion

In a large sample of German adolescents aged 12–25, 16.9% showed clinically relevant mental health symptoms according to the results of an online self-report screening. This matches the recently reported prevalence rates of mental disorders among children and adolescents in Europe [2]. Also, 17.6% reported lifetime NSSI, which is consistent with the meta-analytic finding that approximately one in five adolescents engage in deliberate self-harm over time [29]. Moreover, 34.3% of the adolescents reported suicidal ideation and 2.9% a suicide attempt in the previous 12 months. These 12-month prevalence rates are somewhat lower than the reported life-time prevalence rates for suicidal thoughts and attempts in young people [30]. Strikingly, while 55.7% of the young people with clinically relevant mental health symptoms reported having sought help (lifetime), substantially more did so from informal (51.9%) than from professional (25.6%) sources, confirming the previously reported low professional help-seeking behavior of this age group [6, 7].

A positive attitude towards professional help-seeking, reflected in higher help-seeking propensity scores, emerged as the most important factor explaining variability in professional help-seeking behavior, followed by perceived barriers to mental health help-seeking and clinical factors (i.e., increased levels of general psychopathology, NSSI, and suicidal ideations in the last 12 months). This aligns with previous findings that attitudes towards help-seeking and perceived barriers and facilitators significantly influence professional help-seeking [31, 32]. Clinical factors had, overall, a greater impact compared to sociodemographic factors on both, help-seeking attitudes, and actual help-seeking behavior, potentially indicating that the level of psychopathological distress is a key determinant of help-seeking. Therefore, initiatives aimed at increasing help-seeking behavior, especially among young people with high levels of psychopathological distress, should primarily focus on fostering positive attitudes towards professional help-seeking and reducing perceived barriers. This approach is crucial for minimizing the duration of untreated illness, which is associated with poor outcomes [33,34,35]. Consistent with previous findings [11, 36,37,38], higher SES, female gender, and older age were related to more positive attitudes towards professional help-seeking, and with an increased likelihood to seek professional help. The influence of SES and age may be attributed to differences in mental health literacy (which tend to increase with SES and age) [39], while the gender effect could be explained by gender-specific norms and socialization patterns [36]. Interestingly, higher levels of psychopathology (i.e., general, eating disorder, and personality pathology, NSSI, and suicidal ideation) were related to more negative attitudes (as reflected in lower levels of psychological openness, help-seeking propensity, and indifference to stigma), but increased likelihood for professional help-seeking behavior. These findings raise the question whether previous negative experiences with mental health services have contributed to the negative perception about professional help, as previously reported [40,41,42]. For instance, experiencing prior treatment as stigmatizing or unhelpful may exacerbate feelings of hopelessness and distrust toward mental health professionals. Finally, the results of the moderation analyses indicate that the relevance of sociodemographic and clinical predictors for both attitudes towards help-seeking and actual help-seeking behavior were comparable for adolescents with clinically relevant mental problems (i.e., the clinical group) and those without (i.e., the non-clinical group). However, the interpretation of this finding remains complex, as the clinical group-status overlaps with clinical variables (i.e., general and eating disorder psychopathology, and alcohol problems) that were identified as important predictors of attitudes towards help-seeking and actual help-seeking behavior.

Several limitations must be considered: First, the representativeness of the sample is limited, because recruitment took place in five pre-selected regions in Germany and was dependent on the agreement of schools and students to participate. The Covid-19 pandemic as well as the low participation rate at student level made it necessary to invite more schools than originally planned (e.g., also vocational schools). A recent analysis conducted at one of the study centers identified that a primary reason for non-participation was a lack of concern about the topic of “mental health” [43]. This suggests that students who chose not to participate might have experienced fewer mental health problems compared to those who did participate in the study. Alternatively, these findings may reflect a fear of stigma related to mental health, which presents a major barrier to help-seeking. The representativeness of the sample is further constrained by including only male and female genders, which limits the generalizability of the study results to non-binary adolescents. Future research should explore gender diversity as a predictor of help-seeking attitudes and behaviors. Second, the clinical group-status of participants was based on self-report, which is not equivalent to a psychiatric diagnosis as obtained through a structured clinical interview with an adolescent and their parents. In addition, the determination of the clinical group-status in this study is somewhat arbitrary. The chosen criteria align with the allocation criteria to the first RCT of the ProHEAD consortium [26]. Notably, the initial allocation criteria were adjusted after a preliminary analysis of 10% of the sample data revealed that they were too inclusive. The adequacy of these adjusted criteria is supported by the fact that the frequencies of mental health problems in the current study correspond well with prevalence rates reported previously [2]. Third, recent social and economic changes may have affected the FAS’ ability to accurately measure adolescent SES. These changes include climate change, which may alter travel patterns (potentially influencing the car ownership item), the COVID-19 pandemic, which led to travel restrictions and a shift to online education and remote work (potentially influencing the holiday and computer items), or technological advances making personal computers more affordable and therefore less suitable as an indicator for adolescent SES [44, 45] Future research should consider multiple SES indicators when predicting help-seeking attitudes and behaviors. Fourth, the predictor variables considered in the current analyses explained less variance in help-seeking attitudes (IASMHS subscales) compared to help-seeking behavior (AHSQ subscales). This suggests that other variables not assessed in the current study, such as peer influence or exposure to mental health education, may play a significant role in shaping attitudes towards seeking professional mental health help. These factors could be important areas for future research. Finally, as the current study was cross-sectional, correlates of mental health help-seeking attitudes and behavior were examined, which does not allow for causal conclusions.

To conclude, the findings of the current study confirm attitudinal aspects, perceived barriers, as well as clinical and sociodemographic characteristics as relevant factors for the understanding of why some young people with mental health problems do seek professional help, while others do not [9,10,11]. They extend previous research by demonstrating that the individual propensity and capacity to seek professional help is most relevant, followed by severity of psychopathology and perceived barriers (e.g., lack of knowledge of mental health services or lack of time resources, travel distances, and confidentiality concerns), while sociodemographic factors such as SES, age, and gender are of minor relevance. These findings overall carry noteworthy socio-political implications, as they emphasize that action is needed to enable and empower young people to seek professional help (e.g., through classroom-based interventions to increase mental health literacy), and to reduce stigma (e.g., through mental health campaigns) and structural barriers to mental health treatment [9, 26].