We employed data from the youth@hordaland-survey, which includes information on child and adolescent mental health, lifestyle, school performance and use of health services. Of all 19,430 adolescents born between 1993 and 1995 living in Hordaland County in Western Norway, 10,253 (53%) agreed to participate. The youth@hordaland-survey is a cross-sectional population-based study carried out during early 2012, when the adolescents ranged from 16 to 19 years of age. Participants received information by email and one school hour was used to complete the questionnaires. In addition, adolescents not going to school received the questionnaires by mail at their home address, and mental health services and other institutions were contacted to allow participation for adolescents in these settings. The questionnaires used in the youth@hordaland-survey were web-based.

We linked data from the youth@hordaland-survey to data from the National Patient Registry (NPR) through the participants’ personal identification number. The NPR is the official national registry in Norway on specialist mental health care services, and includes information on specialist mental health care use and Axis 1 psychiatric diagnoses from January 2008 to December 2011, at a time when the adolescents ranged from 12 to 18 years of age, and before youth@hordaland participation. 845 (8.2%) of the adolescents did not provide consent for merging the data from the youth@hordaland-survey with other registries, and were excluded from the analyses. The final sample therefore included 9,408 participants, of whom 853 (9.1%) had at least one registration in NPR.

The Regional Committee for Medical and Health Research Ethics (REC) in Western Norway approved the study. In accordance with the regulations from the REC and Norwegian health authorities, adolescents aged 16 years and older can make decisions regarding their own health (including participation in research). All participants of the study gave written informed consent themselves to participate in the current study. Parents/guardians have the right to be informed, and in the current study, all parents/guardians received written information about the study in advance.

All analyses were performed using STATA V.14.0 (StataCorp, 2015), with the exception that the omega internal consistency coefficient for the CRAFFT questionnaire was calculated in R Core Team (2018). First, frequencies of alcohol/drug use and problems, self-reported symptoms of mental health problems, and sociodemographic variables were examined across individuals who gave their consent to the linkage between the youth@hordaland-survey and the NPR, and those who refused to consent to this linkage (Table 1). Second, the sample was described according to age, sex, SES, and alcohol/drug use and problems among adolescents that had received specialist mental health care services compared to the adolescents that had not receiving these services during the past 4 years (Table 2). Third, psychiatric comorbidity rates within each diagnostic category were described, and all investigated psychiatric diagnoses were analyzed in terms of Spearman’s Rank correlations with other psychiatric diagnoses (Table 3). Fourth, logistic regression models were employed to calculate associations between psychiatric diagnoses received within specialist mental health care and single measures of alcohol/drug use and problems. We then adjusted for the potential confounding effects of comorbid psychiatric diagnoses, in order to determine the independence of associations between each psychiatric diagnosis and substance use. For each psychiatric diagnosis that was investigated, all other psychiatric diagnostic categories were separately entered as potential confounding variables. Finally, we adjusted the analyses for the confounding effects of comorbid psychiatric diagnoses, as well as age, sex and SES (Table 4). Finally, ordered logistic regression models were employed to calculate unadjusted and adjusted odds ratios for the associations between psychiatric diagnoses and the total degree of risky substance use (Table 5). Due to the small size of the psychotic disorders group (n = 10), and due to considerable conceptual heterogeneity within the “other diagnoses” group and the “no Axis 1 diagnosis” group, we did not perform analyses of associations between these categories and substance use.

In unadjusted models, depression was among the psychiatric diagnoses with the most consistent positive associations with alcohol/drug use and problems. Specifically, we found positive associations between depression and all single measures of alcohol/drug use and problems except high-level alcohol consumption. These associations remained significant after adjustment for psychiatric comorbidity. However, the only independent association between depression and alcohol/drug use and problems, after the additional adjustment for sociodemographic factors, was in relation to a positive CRAFFT-score. These findings lend some support to previous studies that have reported positive associations between depression and self-reported adolescent alcohol-related problems (Harrell et al., 2009) and general alcohol/drug use (Crum et al., 2008; Maslowsky and Schulenberg, 2013). Our study adds to previous findings that report positive associations between depression and alcohol intoxication (Crum et al., 2008; Pedersen, 2013), as our findings suggest that the association between depression and frequent alcohol intoxication was significant in the unadjusted model and after the adjustment of psychiatric comorbidity. However, this association was non-significant after the additional adjustment of sociodemographic factors. Moreover, depression was associated with the total degree of risky substance use, and this association was independent from psychiatric comorbidity and sociodemographic factors. In sum, these findings suggest that depression is a robust indicator for the total degree of risky substance use among adolescents.

We found that anxiety was positively associated with illicit drug use in unadjusted models. However, this association were non-significant after adjustment for psychiatric comorbidity. This finding suggests that no unique association between anxiety and illicit drug use was present in our data. Interestingly, anxiety was negative associated with frequent alcohol intoxication after adjustment for psychiatric comorbidity, but this association was no longer significant after the additional adjustment for sociodemographic variables. In sum, the present study lends little support to an independent association between anxiety and alcohol/drug use or problems among adolescents, and adjustment of psychiatric comorbidity tended to reduce the positive direction of the associations between anxiety and alcohol/drug use and problems, while the additional adjustment of sociodemographic variables did not affect the estimates in a consistent direction. Importantly, the previous literature is characterized by highly inconsistent findings, pointing to both negative (Schmits et al., 2015, 2016; Nelemans et al., 2016; Savage et al., 2016) and positive associations between anxiety and adolescent alcohol/drug use (Pardee et al., 2014; Birrell et al., 2015; Stapinski et al., 2016). Some studies have suggested that different anxiety disorders (Wu et al., 2010; Ohannessian, 2014) and different anxiety typologies within a given disorder (Tomlinson et al., 2013) yield different prediction of alcohol/drug use and problems, and anxiety may also have a role on adolescent alcohol/drug use and problems through interactions with other diagnoses (Lansford et al., 2008). Future studies should expand the scope to also distinguish between subtypes of anxiety disorders (e.g., social anxiety and generalized anxiety) and their independent associations with alcohol/drug use and problems during adolescence. Also of note, future studies could expand with measures of substance use/problems related to drugs with anxiolytic effects (e.g., alcohol and benzodiazepines), as our study did not include measures of alcohol-problems in particular (as CRAFFT is related to potential alcohol and/or drug-related problems) or benzodiazepine use/problems.

In the present study, autism was negatively associated with frequent alcohol intoxication after the adjustment of both psychiatric comorbidity and sociodemographic variables. In addition, all other single measures of alcohol/drug use and problems tended to go in a negative direction; however, these associations were non-significant. A limited number of previous studies have to our knowledge explored associations between autism and alcohol/drug use and problems among adolescents (Ramos et al., 2013; Mangerud et al., 2014), suggesting a low alcohol/drug use in this group. The present study support these findings by indicating that adolescents with autism had significantly lower odds for frequent alcohol intoxication compared with the general adolescent population. A possible explanation for this finding is that the social skills deficits of autistic adolescents may keep them out of the social situations and/or peer relations typically associated with alcohol consumption in adolescence (Santosh and Mijovic, 2006). However, autism may be associated with alcohol/drug use and problems beyond the adolescent years (Butwicka et al., 2017).

Our findings suggest that eating disorders were positively associated with frequent alcohol intoxication and a positive CRAFFT-score in unadjusted analyses. However, eating disorders were no longer significantly associated with any alcohol/drug use and problems measures after the adjustment for psychiatric comorbidity, while eating disorders were positively associated with frequent alcohol intoxication after the additional adjustment for demographic variables. This finding supports previous studies linking eating disorders with specific patterns of alcohol use characterized by a loss of control, such as frequent intoxication (Mustelin et al., 2016) and binge drinking (Khaylis et al., 2009). Previous studies indicate that elevated alcohol/drug use and problems are more likely amongst those with bulimia nervosa and binge eating disorders relative to those with anorexia nervosa (e.g., Gadalla and Piran, 2007). The present study did not, however, distinguish between different subtypes of eating disorders, something that should be considered in future research.

Previous studies have linked both childhood ADHD symptoms (Lee et al., 2011; Heradstveit et al., 2018) and ADHD symptoms during adolescence (August et al., 2006) with adolescent alcohol/drug use and problems. However, several researchers have highlighted that most studies have not controlled for associated psychopathology (August et al., 2006; Lee et al., 2011; Bidwell et al., 2014), therefore leaving doubt on the independence of this association. Our findings were that ADHD was associated with illicit drug use alone; however, this association was non-significant after the adjustment of psychiatric comorbidity as well as after the additional adjustment of sociodemographic variables. These findings suggest that no unique association between ADHD and alcohol/drug use and problems were present in our data, lending support to a study by August et al. (2006). They demonstrated that adolescent ADHD was positively associated with illicit drug use only for individuals with a comorbid externalizing disorder, primarily oppositional defiant disorder. A recent literature review similarly indicated that ADHD does not increase the risk of illicit drug use beyond the effect of conduct-related disorders (Serra-Pinheiro et al., 2013). In sum, our results suggest that ADHD was not independently associated with alcohol/drug use and problems among adolescents, and that comorbid psychiatric disorders may be important confounders in association between ADHD and illicit drug use. However, future studies are encouraged to test whether certain features of ADHD (e.g., impulsivity) might show more unique relations to alcohol/drug use and problems.

A range of previous studies have pointed to positive associations between externalizing problems and alcohol/drug use and problems (e.g., Heron et al., 2013; Hopfer et al., 2013; Cerda et al., 2016; Heradstveit et al., 2018; Pedersen et al., 2018), while studies exploring the independence of these associations with respect to comorbidity are more limited. In the present study, we found that conduct disorders were associated with illicit drug use, a positive CRAFFT score, and the total degree of risky substance use in unadjusted models, while we also found independent associations between conduct problems and illicit drug use and the total degree of risky substance use. Hence, conduct disorders appeared to be an important indicator for alcohol/drug use and problems in this study. Also of note, conduct disorders were the only specific psychiatric diagnosis that had an independent association with illicit drug use, highlighting the importance of this diagnosis in relation to illicit drug use among adolescents. This finding lends support to the deviancy proneness model that highlights deviant behaviors as an important risk factor for illicit drug use among adolescents (Tarter et al., 2006).

In unadjusted models, trauma-related disorders were positively associated with all single measures of alcohol/drug use and problems, while they were positively associated with all single measures except illicit drug use after the combined adjustment for psychiatric comorbidity and sociodemographic variables. This latter finding suggest that psychiatric comorbidity and sociodemographic variables should be taken into account as potential confounders in the associations between trauma-related disorders and illicit drug use. Additionally, trauma-related disorders were independently associated with the total degree of risky substance use. This finding supports previous studies that report positive associations between trauma-related disorders and alcohol/drug use and problems among adolescents (Giaconia et al., 2000; Haller and Chassin, 2014). The mechanisms behind associations between trauma-related problems and alcohol/drug-related problems are complex. A longitudinal, community-based study by Haller and Chassin (2014) found that PTSD symptoms increased the risk for later alcohol/drug-related problems among adolescents, and the authors argued that these results supported a self-medication hypothesis. However, other mechanisms may potentially be at work. For example, early alcohol/drug-related problems often involves chaotic and violent lifestyles, which could possibly increase the risk for trauma exposure (Deykin and Buka, 1997).

The study described here has several strengths. First, the sample consists of a well-defined population-based sample of adolescents aged 16 to 19 years, and was sufficiently large to enable a detailed investigation of the associations of interest. Second, a unique linkage with official registry data on specialist mental health care services was utilized, facilitating an investigation of formal psychiatric diagnoses independently determined according to the ICD-10 by professional mental health practitioners. However, the diagnoses are based on clinical judgment and no inter-rater reliability between the raters exists. Third, we investigated alcohol/drug use and problems across a broad range of diagnostic groups, which is rare in previous research (Mangerud et al., 2014), and enabled us to evaluate the independence of associations with alcohol/drug use and problems across several psychiatric diagnoses. Finally, due to the relatively large sample and comprehensive information about psychiatric diagnoses, we were able to adjust our analyses for psychiatric comorbidity, SES, sex, and age.

Some limitations require consideration when drawing inferences from this study. First, although the measures of psychiatric diagnoses preceded those of alcohol/drug use and problems, the study does not have a stringent longitudinal design, and it is not possible to draw conclusions on the causality between psychiatric diagnoses and alcohol/drug use and problems in this study, since some substance use may have predated the mental health care contacts. Second, the response rate in the population-based sample was 53% and included a relatively low proportion of adolescents with self-reported low SES, who in previous studies are found to have higher levels of mental health problems (e.g., Bøe et al., 2012). Official Norwegian statistics indicate that in 2012, 92% of all adolescents in Norway aged 16 to 18 years of age attended high school, compared with 98% in the youth@hordaland-sample (Hysing et al., 2015). The sample may therefore not have been fully representative of adolescents with psychiatric diagnoses due to selective participation. However, a previous publication from the Bergen Child Study indicated that although non-participation in the survey affected the estimated frequency of mental health problems, it did not affect patterns of associations between sociodemographic characteristics and mental health problems (Heiervang and Goodman, 2011). Nevertheless, our findings on associations between the broad range of psychiatric diagnoses and alcohol/drug use and problems require replication within more comprehensively ascertained clinical samples. Third, we did not differentiate between subtypes of psychiatric diagnoses. Fourth, psychiatric comorbidity was significantly higher within some of the diagnostic categories, particularly anxiety, depression, and conduct disorders. We may therefore have underestimated the independent associations between these psychiatric diagnoses and alcohol/drug use and problems. Fifth, alcohol/drug use and problems were measured by self-report, and does not imply the presence of diagnosable SUDs. In addition, the summed variable of the total degree of risky substance use included both substance use (illicit drug use, high-level alcohol consumption, frequent alcohol intoxication) and substance problems (a positive CRAFFT score). The use of a compound measure can be considered as a limitation, as it is difficult to differentiate between substance use and substance problems. Moreover, this summed variable did not specifically include prescription drug misuse as well as tobacco use. The results may therefore not be generalized to all types of drug use. Sixth, although adolescents that did not consent to the linkage between the youth@hordaland-survey and the NPR were overall similar to those that consented to this linkage, they reported somewhat higher frequency of high-level alcohol consumption, self-reported conduct problems, and higher age. Hence, this limitation may affect the generalizability of our findings. Seventh, the present study included many sets of analyses of associations. Multiple testing might therefore be an issue to consider when interpreting the results. Finally, an important limitation related to the generalizability of the findings from the present study is that individuals with untreated mental health problems in the general youth@hordaland-population were not identified. Psychiatric diagnoses in the present study was restricted to individuals that had received specialist mental health care services during the past 4 years. However, some adolescents in the general population may be suffering from undetected and/or untreated psychiatric disorders, or may have received treatment for psychiatric disorders elsewhere other than through specialist mental health care services (e.g., in private practice or through community-based psychologist services). A range of factors may potentially affect specialist mental health care use, such as functional impairment levels (Hintzpeter et al., 2015) and sociodemographic characteristics (Zwaanswijk et al., 2003). Also, a former wave of the Bergen Child Study concluded that specialist mental health care use differed considerably across psychiatric diagnoses, in which children with emotional disorders were underrepresented in mental health care services (Heiervang et al., 2007). Therefore, our findings on associations between psychiatric diagnoses and alcohol/drug use and problems should be interpreted with caution, particularly in relation to anxiety and depression disorders.