Youth and parent screening measures such as the Strength and Difficulties Questionnaire [SDQ; 1, 2] or the Achenbach Systems of Empirically Based Assessments [ASEBA; 3] are easy to use and cost-effective methods to identify adolescents with psychological difficulties. Both of these instruments are highly popular among mental health practitioners and researchers and also among other child care professionals. They have been translated into many different languages and implemented in clinical processes worldwide. Mental health professionals use these screening measures to decide whether further and more detailed assessments of emotional or behavioural disorders are indicated. Researchers use these screening measures in epidemiological and clinical studies to measure the type, the extent, and the course of mental health problems. Nurses and practitioners in general hospitals and social workers in schools and juvenile justice institutions use these screening measures to decide which adolescents need more specific assessment and treatment and should be referred to mental health practitioners. However, discrepancies between multiple informants often create considerable uncertainties in delivering services to youth and drawing conclusions from research [4].

Informant discrepancies on mental health problems are one of the major challenges in child and adolescent psychiatry. A recent meta-analysis of 341 studies [5] found that modest cross-informant agreement is one of the most robust phenomena in clinical child and adolescent research (with mean correlation: r = 0.28). However, the degree of cross-informant agreement for mental disorders varies between mental health domains, different societies and cultures and also depends on the youth’s age and gender [5–8].

A number of different factors contribute to informant discrepancies on mental health problems [9, 10]. First, some mental health problems emerge only in specific situations such as school and family contexts or within peer interactions. Contextual variations occur within a variety of psychiatric domains including social anxiety, attention-deficit-hyperactivity, and conduct problems [e.g., 11–13]. Secondly, informants (e.g., parent and youth) may differ on their perceptions and awareness of mental health problems and what kinds of behaviours are within the norm. For example, parents may be worried about the adolescent’s withdrawal, whereas the adolescent perceives his behaviour as within the normal range and views the intrusiveness of the parents as the area of concern. Thirdly, informant discrepancies may result from measurement errors in regard to the frequency and severity of behavioural, emotional or hyperactivity problems.

Different strategies have been suggested for how to choose informants and how to aggregate data from multiple informant data for diagnostic decision making [12, 14]. In order to disentangle three meaningful components of psychopathology such as (1) the trait (measure of interest for youth’s psychopathology), (2) the context (factors related to the emergence and the reporting of symptoms), and (3) the informants perspective, principal component analysis and regression analyses have been proposed [15, 16]. However, these approaches are quite complex and cannot easily be implemented into clinical practice.

Two factors seem crucial for researchers and clinicians to decide whether parent or youth information is more accurate: (1) the area of mental health problems addressed (e.g., emotional vs. behavioural problems) and (2) the context in which the assessment took place (e.g., clinical vs. community assessments) [17, 18]. For detecting any mental health problems, information from both informants can be useful [19]. In a community sample, parent and youth information uniquely and indispensably contributed to later signs of maladjustment (referral to mental health services, need for professional help, and presence of a disorder) [20]. Similarly, both, self-reports and parent reports were found necessary to detect the presence of a psychiatric diagnosis in a clinical outpatient sample [17].

To explore emotional problems/disorders such as depression and anxiety, clinicians and researchers usually rely on adolescents’ self-reports from questionnaires or interviews because adolescents themselves are assumed to be the most valid source of information for this kind of problems [21]. In fact, adolescents do report significantly more internalizing symptoms than their parents in clinical samples [22, 23] and community samples [24]. Furthermore, self-information has been found accurate to predict the presence of internalizing problems/emotional disorders in community as well as in clinical samples [8, 17, 20, 21, 25–27]. However, some studies also found that the inclusion of parent information further increased the ability to detect emotional problems in community and clinical samples [17, 28].

In the exploration of externalizing problems such as attention-deficit-hyperactivity disorder (ADHD), oppositional defiant disorder (ODD), and conduct disorder (CD), parent information has been considered to be more valid than youth self-reports by mental health professionals [21]. Though on theoretical grounds, self-reports also seem important to assess conduct problems, because many of these behaviours (e.g., thefts, fire setting, physical attacks) occur in setting to which parents are not privy [22]. In community samples, adolescent self-reports show higher levels of behavioural problems than parents reports [18, 24] and adolescent self-reports were found to be valid predictors of externalizing problems, behavioural disorders and later criminal behaviours [20, 28–31]. In clinical samples, adolescents may underreport behavioural problems [18, 32] and adolescent self-reports are sometimes less accurate than parent reports in detecting behavioural disorders [17]. Some adolescents may minimize their conduct problems to avoid possible adverse consequences of full disclosure [33].

Previous studies testing the informant validity of parent and adolescent self-ratings reported conflicting findings and were limited by the use of either just community or just clinical samples and by a paucity of validation measures, (e.g., relying on clinicians’ diagnoses of unclear reliability). Furthermore, previous studies did not consider impact measures as additional information to detect psychiatric disorders. Some adolescents find it hard to report psychological symptoms and may find it easier to describe specific impairments in school, family and peer group. Given the previous findings on the validity of the SDQ impact scales [34], we predicted that impact measures in addition to symptoms scores would make a useful contribution to the assessment of mental health disorders.

The present study intended to confirm and expand previous findings by analysing data collected in a community and an outpatient sample. The ability of parent and youth SDQ scales measuring problems and impact were analysed in order to predict mental health problems/disorders across several mental health domains (any disorder, emotional disorders, behavioural disorders), as validated against two contrasting indices of validity derived from the Development and Well-Being Assessment, DAWBA (see method section below): One approach used the empirically developed multi-informant DAWBA bands (ordinal measures) based on a computer algorithm to aggregate parent and/or youth information from structured interview questions, while the other approach used ICD-10 diagnosis generated by expert DAWBA raters, i.e., experienced clinicians who rated the presence of an ICD-10 disorders after reviewing the answers to closed and open-ended questions. Because the DAWBA is a well validated multi-informant based instrument [35, 36], the current study may overcome some methodological limitations of diagnoses derived from single informants or unstructured clinical evaluations.

Based on the existing literature, we hypothesized that in multivariate analyses (1) the youth and parent SDQ total scores would both be highly associated with any problems/disorders in both samples, (2) the youth SDQ total score would be more strongly associated with emotional problems/disorders than the parent SDQ total score in both samples, (3) parent and youth SDQ total scores would be associated with behavioural problems/disorders in the community sample, (4) but only parent SDQ total score would be associated with behavioural problems/disorders in the clinical sample. Hypotheses 3 and 4 were established a posteriori in accordance with findings from previous studies. We further assumed that youth and parent SDQ impact scores would supplement the predictive power of symptoms scores in the prediction of any problems/disorders, emotional problems/disorders, and behavioural problems/disorders in both samples.

In addition, we tested the ability of the SDQ conduct and emotional problem scales in the prediction of emotional and behavioural problems/disorders in both samples. Further supplemental analyses of parent and youth SDQ hyperactivity and conduct problem scales in the prediction of ODD, CD and ADHD were performed in the clinic sample only (because of the low prevalence rates of these disorders in the community sample).

The current study adds to previous findings on the validity of multi-informant assessments of mental disorders in youth [5, 19]. Unlike earlier studies, the present investigation is based on internet-based instruments only. The DAWBA has previously been used to identify mental health disorders with similar properties to traditional diagnostic interviews such as the Diagnostic Interview Schedule for Children (DISC) and the Child and Adolescent Psychiatric Assessment (CAPA) [44]. However, the DAWBA was a more conservative measure, generating fewer diagnoses than the other two measures [44]. In the present study, two different approaches to validation were used in parallel across multiple mental health domains: First, validation against an empirically derived computerized algorithm (the DAWBA bands) and, secondly, validation against ICD-10 diagnoses by clinical experts. Overall, the two validation approaches generated similar results supporting the likely robustness of the findings. Based on the rather low prevalence rates of affective and anxiety disorders, the corresponding correlations of DAWBA bands and expert ratings were only modest in the community sample. This finding may also reflect the rather moderate agreement of different diagnostic approaches when assessing affective and anxiety disorders in youth [45]. Correlation coefficients between parent and youth SDQ scales were similar to findings from previous studies [6, 7]. However, the correlations between all reported subscales were highly significant in the clinical sample, but the total score was not. There is no clear and easy explanation to this sample-dependent finding that is in need of more detailed studies. Furthermore and in contrast to our and previous findings in community samples [34], youth and parents in the clinic sample did not agree on the level of distress and impairment caused by mental health problems. Also this finding needs further studies aiming at some clarification of the origins of these discrepant views.

The current findings illustrate the importance of considering motivation and the nature of behavioural and emotional problems in self-report ratings. Clinical practitioners should keep in mind that adolescents may display problem behaviours only in specific settings but also have limited ability to report behavioural and hyperactivity problems. The “Operations Triad Model” [OTM; 5, 10] is a conceptual frame-work on how to use and interpret multi-informant assessments which is guided by evidence based information on the divergence and convergence of informants’ reports. OTM guides clinicians (a) to hypothesize about patterns of convergence and divergence among informants reports and (b) to develop personalized assessments that directly test these hypotheses. To do this, practitioners may rely on information on the context in which the problems emerge as well as the informant’s ability to report mental health problems across different domains. The current findings may guide clinicians to choose which kind of information should be collected from which informants and how to aggregate that information in order to decide on further assessment and treatment.