The participants of the present study were part of a larger study on Affective Dysregulation on Optimizing Prevention and Treatment (ADOPT), which aimed at developing assessment tools for diagnosing AD, analyzing the epidemiology, and investigating the efficacy of treatment approaches for children with AD [10]. Within the subproject ADOPT Epidemiology, a large community sample (n = 9759) was recruited in four German cities through residents’ registration offices (for detailed information, see [10, 28].

After the initial screening with the DADYS-Screen, all families within the highest 10% raw scores on the DADYS-Screen (samplehighAD; n = 287) were invited to participate in a comprehensive assessment including clinical, parent, and child ratings. Participating families were subsequently randomized to receive either treatment as usual or an AD-specific psychotherapeutic treatment. The cut-off of 10% was chosen as an approximation to epidemiological studies, which found prevalence rates of up to 9% [24]. For the low AD comparison group, a random sample of families within the lowest 10% raw scores (samplelowAD, n = 184) was invited to participate in the same comprehensive assessment including clinical, parent, and child ratings. To be able to display the full range of AD, we additionally employed comprehensive parent-rated questionnaires in a randomly drawn sample of families within the middle 80% of the raw score distribution (samplemoderateAD; n = 643). Thus, the total sample for the current study comprised 1,114 families. We chose an age range of 8–12 years for the children since the focus of the study was on AD in childhood and we wished to include children’s self-ratings for several questionnaires. Additional inclusion criteria for the comprehensive assessment were IQ above 80, no current behavioral therapy focusing on AD, and no autism spectrum disorder. We did not exclude children with comorbid disorders such as ODD, CD, ADHD, depression, or anxiety as we wished to analyze potential symptom overlaps. The assessment was completed either online via the REDcap electronic data capture tool, hosted at the Clinical Trials Centre Cologne, or offline in paper-and-pencil format. The average time between screening and comprehensive assessment was 15.73 weeks (SD = 11.46). Randomization to the treatment condition for the samplehighAD was conducted after the completion of the comprehensive assessment.

The DADYS [12, 18, 28] is a comprehensive assessment tool including the screening questionnaire (DADYS-Screen, 12 items), a diagnostic interview for parents (DADYS-PI, 13 items) and children (DADYS-CI, 10 items), and a questionnaire for parents (DADYS-PQ, 36 items) and children (DADYS-CQ, 26 items). As mentioned above, the DADYS comprises symptoms of irritability, impulsivity, temper outbursts, anger, and mood swings (e.g., “is easily annoyed by others”, “exhibits wide mood swings”). Items are rated on a 4-point Likert scale ranging from 0 (not present) to 3 (very strong). For each questionnaire/interview, the mean item score was calculated. In the current sample, the internal consistency of each questionnaire/interview was good to excellent (86 ≤ α ≤ 0.96). Additionally, DMDD diagnosis was evaluated based on the DADYS-PI as 0 (no) and 1 (yes).

As a second measure of AD, we included the Dysregulation Profile [1] of the Child Behavior Checklist in its German version (CBCL/6-18R; [11]. The items of the subscales anxious/depressed (13 items), attention problems (10 items,) and aggressive behavior (18 items) were rated by parents on a 3-point Likert scale ranging from 0 (not true) to 2 (very true or often true). We calculated the mean item score for each subscale. For the CBCL-DP scale, we subsequently summed the scores, resulting in a range from 0 to 6, weighing each scale by the number of items [25]. In the current sample, we analyzed the anxious/depressed subscale and the CBCL-DP scale, which both demonstrated good to excellent internal consistency (α = 0.80-0.94).

We assessed emotion regulation strategies using the Questionnaire for the Regulation of Frustration in children (FRUST; [13], Junghänel [19]). The FRUST comprises the subscales adaptive and maladaptive emotion regulation strategies in the parent (adaptive: 10 items, maladaptive: 4 items) and child rating (adaptive: 33 items, maladaptive: 7 items). Adaptive strategies include, e.g., problem-solving or social support while maladaptive strategies include, e.g., rumination or avoidance. Items were rated on a 5-point Likert scale ranging from 0 (hardly ever) to 4 (almost always). The mean item score of each subscale was calculated. In the current sample, the internal consistency of the subscales was sufficient to excellent (0.78 ≤ α ≤ 0.94).

Child internalizing and externalizing symptoms were assessed using the DISYPS-III [9]. We used the therapist-rated diagnostic screening checklist for internalizing (19 items) and externalizing symptoms (9 items) based on a parent interview, the parent and child-rated symptom checklists for ADHD (20 items) and disruptive disorders—including DMDD, ODD, and CD—(28 items), and the parent-rated symptom checklist post-traumatic stress disorder (PTSD; 19 items). Items were rated on a 4-point Likert scale ranging from 0 (age-typical) to 3 (very strong). The mean item score across all items was calculated per checklist, with the exception of the checklist for disruptive disorders, where we calculated subscales for ODD, and CD. Note that the PTSD scale was only assessed if the child had experienced at least one potentially traumatizing event (n = 588). In the current sample, internal consistency was good to excellent (0.79 ≤ α ≤ 0.94), with the exception of the CD scale (α = 0.61) due to the diverse behaviors assessed in this scale.

Additionally, we evaluated diagnoses of ADHD, disruptive disorders (ODD and/or CD), and depression as 0 (no) and 1 (yes) based on the DISYPS parent interviews. If parents reported symptoms of ADHD, disruptive disorders, or depression on the screening checklist, the respective comprehensive checklist from the DISYPS-III was employed to confirm the diagnosis.

HRQoL was assessed using the KIDSCREEN questionnaire (The KIDSCREEN Group Europe, 2006), which assesses subjective health and well-being in children and adolescents. We used the child-rated KIDSCREEN-10 Index (10 items) and the parent-rated short version KIDSCREEN-27 (27 items). Items were rated on a 5-point Likert scale ranging from 1 (never/not at all) to 5 (always/very strong). The mean item score was calculated. In the current sample, internal consistency was good to excellent (0.81 ≤ α ≤ 0.91).

Statistical analyses were performed using SPSS 20 (IBM Corp, 2011). Missing data were imputed using expectation maximization (EM) if at least 70% of the items per scale were available. Items of each respective scale were used for imputation.

Differences in sample characteristics between the subsamples were examined using χ2 tests for categorical variables and Kruskal–Wallis tests for continuous variables. As measures of effect size, we used Cramer’s V for χ2 tests (0.10 ≤ ϕc < 0.30 small, 0.30 ≤ ϕc < 0.50 moderate, 0.50 ≤ ϕc large) and Pearson correlations for Kruskal–Wallis tests (0.10 ≤ r < 0.30 small, 0.30 ≤ r < 0.50 moderate, 0.50 ≤ r large; [4].

For the correlation analyses (concurrent, convergent, and divergent validity), we calculated partial rank correlations controlling for age and gender between the DADYS-Screen and comprehensive measures of AD, emotion regulation strategies, externalizing and internalizing symptoms, and measures of HRQoL. Note that we report the correlations of the DADYS-Screen with the DADYS-PQ both for the total scale and for a reduced scale excluding items which were also part of the DADYS-Screen. To avoid item overlap between the validators, we did not calculate the DISYPS-III DMDD subscale due to item overlap with the DADYS-PQ/-CQ. Furthermore, we excluded three items of the ODD subscale in our analyses, which were part of the DMDD subscale and the DADYS-PQ/-CQ. As measures of effect size, the correlation coefficients were interpreted as mentioned above [4]. We classified correlations accounting for at least 50% of the variance (r > 0.70) as very large. Additionally, we compared correlation coefficients of the different conceptualizations of AD, externalizing symptoms, and internalizing symptoms according to Meng et al. [26] if at least two measures for the same rater in the same (sub)sample were present.

The AUC (concurrent validity) and the optimal cut-off of the DADYS-Screen were determined using ROC analyses. The ROC curve displays the relation between sensitivity and 1-specificity values. We compared the DADYS-Screen with two measures of the DADYS-PI: a) total score cut-off: total score of at least 1 (yes/no) and (b) DMDD: diagnosis (yes/no). For the total score cut-off, each item must be fulfilled at least mildly. This cut-off is comparable to that used in the Multimodal Treatment Study of Children with ADHD (MTA) study [35]. To quantify the discriminatory power of the DADYS-Screen, we analyzed the AUC for both (a) and (b) [20]. The AUC score ranges from 0.50 (at random) to 1 (perfect). We used the following interpretations of AUC scores: 0.50 ≤ AUC < 0.70 poor, 0.70 ≤ AUC < 0.80 acceptable, 0.80 ≤ AUC < 0.90 excellent, 0.90 ≤ AUC outstanding [15]. To determine the optimal cut-off for the DADYS-Screen, we employed the Youden Index [38], which aims to maximize the distance between the line of equal sensitivity and specificity (diagonal line) and the point farthest from this line [20]. Thus, the highest score demonstrates the best cut-off. Based on this cut-off, we analyzed sensitivity and specificity.

Additionally, we tested differences between the DADYS-Screen scores of children with and without a diagnosis of DMDD, disruptive disorders, ADHD, depression, or any of these diagnoses using Mann–Whitney U tests (discriminant validity). To account for differences in age and gender, we implemented a case–control matching for each child with a diagnosis. As a measure of effect size, we used Pearson correlations.