A total of 257 Kinder-DIPS interviews were conducted at either the outpatient clinic (54.5%) of the University of Basel or at psychiatric and psychological services (45.5%) in Basel. Participants interviewed at the two sites did not differ considerably in terms of demographic information: Children in both groups were around 11 years old (M = 10.63–11.95, SD = 1.63–3.52) and parents were in their early forties (M = 41.68–45.14, SD = 4.73–6.30). The group seen at the outpatient setting (University) included 44.6% boys while of the group interviewed at the psychiatric and psychological services 55% were boys. A greater number of parents seen at the outpatient setting (82%) were married compared to parents at the psychiatric and psychological services (60%).

Demographic information largely matched when the interview partner was the mother or both parents: The participants were similar with respect to children's age (M = 10.89–11.01); SD = 2.00–2.54) and gender (boys = 47.1–50.0%). Also, in both groups, mothers respectively mother and father were in their early forties (M = 42.4–44.70, SD = 4.79–5.48). There was a difference between the two groups with respect to the marital status. 68.8% of mothers interviewed alone were married, whereas 81.4% of parents interviewed together were married.

In 147 (57.2%) interviews a diagnosis of at least one anxiety disorder as primary diagnosis was assigned. In 32 parent-child dyads no diagnosis (12.5%) was assigned. In additional 78 (30.4%) interviews another clinical disorder except anxiety disorders was assigned, however, these interviews were excluded from our analyses to maintain the focus on anxiety disorders. Interview partners were mothers (36.3%), fathers (1.7%), or both parents (56.4%). Interview data with fathers and missing information about the interview partner (5.1%) were deleted pairwise from further analyses.

The final sample consisted of 166 children (51.8% girls) between the ages of 6 and 18 years (M = 10.94; SD = 2.22). Of them 128 (77.1%) were interviewed in the outpatient clinic of the University of Basel and 38 (22.9%) at psychiatric and psychological services. Mothers' and fathers' average ages were 41.81 (SD = 4.84) and 45.13 (SD = 6.89), respectively (for an overview of sociodemographic data see Appendix A). A subgroup of participants (n = 129) filled out questionnaires to investigate parent and child related predictors on parent-child agreement.

Participants were recruited for a treatment study for SAD at the University of Basel and from psychiatric and psychological services (Kinder- und Jugendpsychiatrie Baselland, KJP; Schulpsychologischer Dienst Basel-Stadt, SPD) between May 2004 and September 2008. Participants recruited at the University of Basel were screened for anxiety disorders before they underwent the Kinder-DIPS while participants recruited at psychiatric and psychological services were not selected by a specific primary diagnosis. The Kinder-DIPS was part of the initial diagnostic procedure. All participants needed a basic level of German literacy to complete the diagnostic assessment. Exclusion criteria for all participants were a comorbid diagnosis of pervasive developmental disorder and acute drug abuse.

The studies had been approved by the local ethical committee for medical research. Prior to the interview, children and parents were informed about the interview process and gave their written informed consent. For children younger than 16 years old, parents gave their written informed consent. Parents and children were interviewed separately in random order, having different interviewers for their Kinder-DIPS interviews. For each interview current and lifetime diagnoses were derived. Interviewers were blinded to the child's referral reasons and they conducted parent and child interviews independently. All interviews were recorded and subsequently recoded by different raters to ensure interrater reliability. There were three diagnostic profiles for each child-parent(s) couple: (a) clinical diagnoses according to the child's interview, (b) clinical diagnoses according to the parents' interview, (c) a composite rating between a and b. The inter-rater reliability was calculated for the child and the parent interview separately. The diagnosis with the most severe rating (child or parent interview) was labelled as the primary diagnosis. Discrepant clinical ratings were discussed during regularly held conferences with two senior clinical supervisors until consensus was reached.

Prior to the interview, parents and children filled out several questionnaires which are described in detail in the measures section. Additionally, parents indicated how many hours per week they usually spend with their children.

The following questionnaires were filled in by the mother and the father of the child.

Forty-eight interviewers (39 psychologists with Bachelor's degree; 9 psychologists with Master's degree), with a standardized training in the administration and scoring of the Kinder-DIPS conducted the interviews (see Kinder-DIPS manual for DSM-IV-TR; Neuschwander et al., 2013). All clinicians successfully completed a certification process under supervision. High inter-rater reliability for the parent (α ≥ 0.88) and child version (α ≥ 0.92) in the Kinder-DIPS for anxiety disorders were reported (Neuschwander et al., 2013).

Agreement between children and parents was computed using correlational statistics with SPSS 22.0.0.0 for Mac. All agreement statistics were calculated within both a “current” and a “lifetime” frame, as well as on three diagnostic levels: (1) diagnostic category (presence or absence of an anxiety disorder), (2) diagnoses (SAD, SoP, SP), and (3) diagnostic criteria. Kappa coefficient (k, Cohen, 1960) and Yule's Y coefficient (Yule, 1912) were calculated to examine the parent-child agreement. The received values of kappa coefficient < 0.4 indicates low agreement, 0.4–0.6 moderate, 0.6–0.8 good, and >0.8 excellent agreement (Fleiss, 1981). Statistical significance of the kappa coefficient was determined with x²-exact tests. The Yule (1912) coefficient was proposed as a chance-corrected measure of agreement that is, contrary to the kappa statistic, independent of the prevalence of the observed event (Spitznagel and Helzer, 1985). Based on the odds ratio (standardized to range from −1 to 1), the Yule's Y is comparable to other correlational coefficients (Wirtz and Caspar, 2002). The kappa and Yule's Y statistics reduce reliability if prevalence of the observed events is low resulting in symmetrically or asymmetrically skewed marginal distributions of the contingency tables (Kraemer, 1979; Maclure and Willett, 1987; Thompson and Walter, 1988; Cicchetti and Feinstein, 1990; Feinstein and Cicchetti, 1990). Correlational statistics of parent-child agreement were only reported for diagnoses with a higher base-rate than 10% (a complete table of inter-rater measurements is available upon request). Additionally, total agreement as the percentage of the number of agreement between parent and child were reported.

For the investigation of child- and parent related predictors of diagnostic agreement, the frequency of disagreements rather than agreements on the diagnostic criteria per parent-child dyad were calculated to minimize the chance agreement on the absence of symptoms. Chance agreement on the absence of symptoms is typically high since children report a relative limited number of positive diagnostic criteria.

The predictors of parent-child agreement were investigated in three steps. First, unadjusted bivariate correlations (Pearson's r, two-tailed) were calculated between the number of parent-child agreement and predictor variables. Second, if there was a significant correlation between a predictor variable and the number of disagreement, a linear regression analyses and finally a multiple regression analyses were conducted with these variables. Child related variables were gender, age, and the amount of interaction time between child and parents. Parent related variables were scores on psychopathological (BDI, ASI, BAI), parenting (PSOC), and social support questionnaires.

According to parental report, 113 (68.1%) of the children met DSM-IV-TR criteria for a current anxiety disorder, and 122 (71.3%) of the children met DSM-IV-TR criteria for any lifetime diagnosis. According to child report, 79 (47.6%) of the children met DSM-IV-TR criteria for any current diagnosis, and 89 (52.0%) of the children met DSM-IV-TR criteria for a lifetime anxiety disorder.

Overall, children and parents agreed significant, but low to moderate on current and lifetime diagnostic information (see Tables 1, 2). On the level of diagnostic category, kappa values about the presence of a current (k = 0.22) or a lifetime (k = 0.21) anxiety disorder were low. On the level of current and lifetime diagnoses, kappa values (base rate > 10%) were low to moderate for SAD (k = 0.52; k = 0.43), SP (k = 0.24; k = 0.19), and SoP (k = 0.25; k = 0.28), indicating the highest agreement for SAD (current and lifetime diagnoses). Contingency tables (Table 1) showed that parents reported more current diagnoses than their children across all current and lifetime anxiety disorders.

On the level of current and lifetime diagnostic criteria (see Table 2), kappa values were significant and moderate for SAD (k = 0.53–0.67; k = 0.49–0.57), low for SoP (k = 0.22–0.38; k = 0.24–0.38), and low for SP (k = 0.27–0.63; k = 0.24–0.41). Again, parents indicated more criteria as positive than their children when large discrepancies were found. The criterion with the lowest agreement in SoP and SP was “symptoms interfere with the person's normal routine, or cause marked distress” (current: k = 0.22, lifetime: k = 0.24).

For the regression analyses, 85 parent-child dyads with complete interview data for current diagnostic criteria were available. The interview data conducted with mothers and both parents were calculated in one group because there was no significant difference between the number of current diagnostic criteria children and parents/mothers disagreed upon, t₍₈₃₎ = 0.78, p = 0.44.

In accordance with earlier studies (e.g., Grills and Ollendick, 2003; Comer and Kendall, 2004; Weems et al., 2011; Propp et al., 2014), parent-child agreement was low on the diagnostic category (average k = 0.22), higher but still low on specific anxiety diagnoses (average k = 0.34) and moderate on diagnostic criteria (average k = 0.52). A possible explanation for this difference might be that important diagnostic information is assessed at the level of diagnostic criteria that might get lost in the course of the diagnostic evaluation. For example, a parent-child dyad would agree on the level of criteria but not on the level of diagnoses and diagnostic category if parent and child agree on criterion A of SAD (e.g., child is extremely afraid of being separated from the parents) but not on criterion D of SAD (e.g., distressed or impaired in daily functioning). These findings highlight the importance of a combination of categorical and dimensional diagnostics since subclinical problems might be identified at the level of criteria but remain undiscovered at the level of diagnoses.

As expected, parent-child agreement on the level of diagnostic criteria varied between the types of anxiety disorder. Parents and children agreed more on criteria of SAD compared to SoP and SP replicating results of earlier studies (Jensen et al., 1999; Choudhury et al., 2003; Grills and Ollendick, 2003). Agreement on SAD was probably higher in the present sample because symptoms of SAD are separation-related (e.g., clinging, avoidance of being alone) and therefore more easily noticeable for parents than symptoms of SoP or SP. Furthermore, the prevalence rate for SAD was higher (n = 69) than for SoP (n = 42) and SP (n = 49) and therefore less susceptible for bias.

In general, parents endorsed more criteria than their children across all anxiety disorders analyzed in this study. One example therefore is the criterion E (“symptoms interfere with the person's normal routine, or cause marked distress”) of current and lifetime SP and SoP. These findings are supported by earlier studies that found that parents report generally more symptoms than their children (MacLeod et al., 1999; Martin et al., 2004). Anxious children, especially those with social phobia were identified to report fewer symptoms, particularly social avoidance than their parents (DiBartolo et al., 1998). A possible explanation might be that anxious children neglect symptoms to avoid a possible negative evaluation by seeming non-distressed. Alternatively, anxious children avoid negative feelings that might come up when they are reminded of anxiety provoking situations. Additionally, psychotherapeutic appointments are usually initiated by parents and not by children themselves. Parents might therefore have a clearer idea of concerned problems and why they are seeking help than their children which might result in lower diagnostic agreement.

The predictive analyses showed that the child's age and gender predicted not significantly the number of parent-child agreement. In spite of mixed results, the present findings replicated the results of the majority of empirical work that showed no effect of gender or age (e.g., Grills and Ollendick, 2003; Propp et al., 2014). The inconsistent findings in the existing literature might be caused by different age groups and age ranges covered in earlier studies. Some studies, for example, covered younger children (8–12, Barbosa et al., 2002; 7–14, Comer and Kendall, 2004), older children (11–17, Propp et al., 2014) or wider age ranges (7–18, Brown-Jacobsen et al., 2011; 7–17, Hoffman and Chu, 2015) with a mean age of around 11 years.

In accordance with Propp et al. (2014), maternal depression predicted parent-child agreement on diagnostic criteria of anxiety disorders. The analysis of social support indicated a correlational result. Lower social support was associated with lower parent-child agreement. However, this relation was not significant after controlling for maternal depression. It can therefore be assumed that maternal depression is a strong predictor of parent-child agreement. It is still open to discussion whether depressed parents report more accurately or are biased about the symptoms of their children (Richters, 1992; Moore and Fresco, 2012). However, clinical suggestions about the relation between maternal psychopathology and parent-child agreement can be provided. It can be expected that parent-child agreement on anxiety symptoms decrease when mothers report (1) depressive mood and (2) low levels of social support. Furthermore, a potential buffering effect of social support on maternal depression can be suggested due to the correlational relation between these predictors leading to higher levels of parent-child agreement. This idea is supported by earlier findings that social support is a moderator between maternal depression and behavioral problems in children (Lee et al., 2006).

We were able to show for the first time that, analogue to mothers, paternal lower social support predicted lower parent-child agreement. No significant correlations were found for parental sense of competence, anxiety (sensitivity), and depression. So far, the impact of paternal variables on parent-child agreement is less studied than maternal variables (Grills and Ollendick, 2003). It has been suggested that mothers spend more time with their children and provide therefore more information about the child's behavior (De Los Reyes and Kazdin, 2004). This might be an explanation why fathers' psychopathology did not predict parent-child agreement in the present study. However, the positive correlation between reported social support of mothers and fathers indicated that this might be a shared predictor for parent-child agreement. Other than psychopathological variables, social support seemed to affect the whole family leading to a lower parent-child agreement.

Furthermore, contrary to our expectations, parental sense of competence had no predicting value for parent-child agreement in the present study. An association between parental depression, self-efficacy, and social support has been reported by different studies (Herwig et al., 2004; Lee et al., 2006). In view of the high predicting value of maternal depression on parent-child agreement, it could be expected that mothers with a high level of parental self-efficacy would show higher levels of parent-child agreement due to increased parental capacity including responsive caretaking and more attention to the child's signals (Coleman and Karraker, 2000). This hypothesis was supported in the present study. Additionally, no significant effect was found for the amount of time parents and children spent together. A possible explanation might be that not the total amount of time but the quality of parent-child interaction predicts the extent to which parents and children agree upon behavioral and emotional difficulties (Grills and Ollendick, 2002). A further investigation of a potential moderating effect of different aspects of parenting (e.g., quality of communication) might be of value.

Despite the high number of participants, not all anxiety disorders (e.g., panic disorder, GAD) included in the structured diagnostic interview were represented because of the low base rates in this age group. Low base rates can lead to a limited interpretation of kappa because of its base rate dependency. Either a very low or a very high prevalence of the observed event leads to an underestimation of the kappa coefficient (Cicchetti and Feinstein, 1990; Sim and Wright, 2005). Another limitation is the limited generalizability of the present finding regarding parent- and child related predictors. Because of feasibility reasons we were only able to examine this research question in the sample of children with anxiety disorders treated in the university-related outpatient clinic. These results might therefore not be generalizable to inpatient clinical samples due to evidence that parent-child agreement differs between inpatient and outpatient clinical settings (e.g., Martin et al., 2004; van der Ende and Verhulst, 2005).

In conclusion, based on the existing research on parent-child agreement in structured diagnostic interviews, parent-child agreement in anxiety disorders can be expected to be low to moderate as replicated in the present study. Additionally, the present findings indicated that diagnostic agreement varies (1) between the diagnostic levels and (2) as a function of type of anxiety disorder. Higher parent-child agreement can be expected if the diagnostic symptoms are assessed on the level of diagnostic criteria rather than on diagnoses. Clinicians might benefit from a closer look at the anxiety symptoms parents and children agree upon in respect to diagnostic and therapeutic processes (Achenbach et al., 1987; Reyes, 2011).

There is additional growing evidence that parental psychopathology, and more specifically maternal depressive mood decreases the level of parent-child agreement. Until now, it is an open question whether maternal psychopathology increases the validity of mothers' diagnostic report. Additionally, low social support of both parents was associated with lower parent-child agreement. The assessment of maternal psychopathology and familial social support might be of additional value in interpreting diagnostic parent-child agreement.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.