Anxiety disorders are the most common psychiatric illnesses in both adulthood and childhood (Beesdo-Baum and Knappe, 2012; Kessler et al., 2012). Anxiety symptoms tend to begin early in life and are likely to become chronic and persistent (Beesdo et al., 2009). Thus, early recognition and treatment are desirable (Beidel and Turner, 2007; Davis et al., 2011; Scaini et al., 2016). However, since anxious children do not manifest behavioral difficulties as clearly as children with externalizing disorders, they often go unrecognized and underdiagnosed (Costello et al., 2005; Beesdo et al., 2009).

In light of these considerations, it is important to find new adequate and validated instruments to screen the severity of anxiety symptoms in childhood. In addition to structured psychiatric interviews, which are specific and detailed, questionnaires may be used as flexible, easy to administer, less costly, and less time consuming first screening instruments (Verhulst and Van der Ende, 2006; Garcia-Lopez et al., 2015). Several anxiety self-report questionnaires have been developed to assess anxiety symptoms in both clinical and general population. In a review on anxiety rating scales, Myers and Winters (2002) reported that the Screen for Child Anxiety Related Emotional Disorders (SCARED) is one of the best self-report instrument to evaluate anxiety symptoms. Moreover, a meta-analysis by Hale et al. (2011) suggested that the SCARED presents good psychometric properties and can be used as a cross-cultural screening instrument. The SCARED offers both self- and a parent-report version to assess childhood anxiety disorders by including subscales on General Anxiety, Separation Anxiety, Somatic/Panic, Social Anxiety, and School Phobia. Several reports have already investigated psychometric properties in community and clinical populations reporting good internal consistency, test-retest reliability, and discriminant validity (Birmaher et al., 1997, 1999; Ogliari et al., 2006; Crocetti et al., 2009; Hale et al., 2013).

A great part of the literature has confirmed the importance of both child and parent report as sources of factual information, especially for childhood emotional syndromes. Indeed, when using rating scales, it is crucial that clinicians collect information from different sources. The information provided by several raters, which see children in different contexts, is important to obtain a comprehensive picture of the child's functioning (Verhulst and Van der Ende, 2006). As reported by Verhulst and Van der Ende (2006) parents are usually familiar with their child's functioning in many situations and throughout time. However, children and adolescents' self-reports are also indispensable, especially to have a better overview of internalizing problems that are more difficult to detect (Costello et al., 2005; Beesdo et al., 2009).

Previous studies using different methods and measures indicated low levels of agreement among informants, showing higher scores for anxiety, and depressive symptoms in child report than in parent report (De Los Reyes and Kazdin, 2005; Salbach-Andrae et al., 2009).

Besides focusing on low agreement levels as a nuisance source (Verhulst and Van der Ende, 2006), in existing works it has been reported the importance of multiple informants as a potentially valid source of information (Kraemer et al., 2003). Self-reports filled in by parents and children are now regarded as being of equal weight in the diagnostic process (Angold, 2002). Several reports showed that parents perception of anxiety may be influenced by cultural factors (Gaines et al., 1997; Murphy et al., 2003; Wren et al., 2007).

Until now, the agreement of the Italian child and parent versions of SCARED has not been examined. This lack calls for a systematic analysis of parent-child agreement in the Italian culture. Moreover, the possible effect of sex and age on agreement was investigated. Better knowledge of the impact of these factors could help in shaping more effective assessment protocols and treatments. For instance, information on sex and age could encourage clinicians in considering also these variables in their clinical decisions (Rutter et al., 2008).

To the best of our knowledge, no previous studies have analyzed parent-child agreement using indices suited for ordinal variables such as κ index. Moreover, the level of agreement (taking into account age and sex effect) for each item within each SCARED subscale has not been addressed in the previous literature so far.

Indeed, all recent studies that have evaluated the psychometric properties of the SCARED questionnaire (e.g., Wren et al., 2007; Su et al., 2008; Jastrowski Mano et al., 2012) in exploratory fashion, applied a principal component analysis without specifying on which correlation matrix (Pearson, Spearman, or polychoric correlation matrix) was the analysis carried out. In the common practice, ordinal variables with at least five categories are treated as continuous variables and analyzed with a traditional factor analysis procedure, thus ignoring the ordinal nature of the outcomes. Treating nominal and ordinal data as interval or ratio data, may lead to unreliable parameter estimates as well as to poor goodness of fit, especially in presence of highly skewed non Gaussian data or polarized/bimodal distributions (Muthén and Kaplan, 1985).

As emphasized in Holgado-Tello et al. (2010) in presence of ordinal data, factor analysis should be carried out on the matrix of polychoric correlations rather than on Pearson correlation matrix since analyses based on the former produce a more accurate reproduction of the original correlation structure.

Motivated by the outlined importance of accounting for different sources of information in screening for anxiety disorders, we investigate the parent-child agreement level and examined separately the factorial structures of both parent and child versions of the SCARED. For the first time in the literature we have carried out the analysis in an Italian clinical sample.

In particular, we analyze the cross-informant agreement levels for each questionnaire item and, considering age and sex effect, for each factor-based scales. We, then inspect the differences in the number of symptoms reported by children and parents. Lastly, we examine separately the factor structures of parent- and self-report versions of SCARED questionnaire.

Although structured interviews still represent the gold standard to determine the presence and degree of severity of a clinical disorder, the use of reliable and valid self-reported questionnaires may support clinicians in the diagnosis and may be used as screening tools for children (Alfano et al., 2002; Viana et al., 2008).

The present study represents the first contribution that evaluates cross-informant agreement of the Italian child and parent versions of SCARED. Moreover, statistical approaches specifically suited to ordinal data have been applied.

Our results confirmed hypotheses that children report significant information about anxiety symptoms and thus contribute with a different perspective to the assessment of the disorder.

Our analysis shows a moderate to strong mother-child agreement. When accounting for age, we found that the level of cross-informant agreement for the total SCARED score increased with age. This can be interpreted in the following way: as children grow up, they become more able to communicate and explain their symptomatology to parents who in turn become more aware of the child's anxiety. Our results suggested that, independently on age, the subscale with the higher rate of agreement is SAD. From a clinical perspective, this reflects the fact that SAD symptoms emerge clearly in the family since the child would often ask for help and assurances. In addition, SAD symptoms compromise family life and increase parental stress limiting the activities of siblings and parents (Fischer et al., 1999). In fact, SAD's essential feature is an excessive anxiety concerning separation of the child from its home or parents (American Psychiatric Association, 2013). Differently from other anxiety disorders, the fear of separation causes great distress to the child and may interfere with both the child's and the family's normal activities (Fischer et al., 1999). A child suffering from SAD may often show avoidance and oppositional behavioral patterns in an attempt to prevent separation, or may refuse to participate in activities requiring separation from the attachment figure (Fischer et al., 1999).

Moreover, current studies (e.g., Scaini et al., 2012) suggest that separation anxiety develops from multiple and interrelated factors that include genetic vulnerabilities and environmental factors, such as negative family experiences (Chorpita, 2001; Kearney, 2008).

Our results appear consistent with recent research (Wren et al., 2004; Su et al., 2008; Weitkamp et al., 2010; Jastrowski Mano et al., 2012) that reports agreement values higher than the ones previously reported by Birmaher et al. (1997, 1999).

Despite evidence of good cross-informant agreement in the present study, strong differences in number of reported symptoms and their severity emerged from the analyses.

When accounting for the age effect, we observe that children reported on average more symptoms of Separation Anxiety at 8–10 years and of General Anxiety at 16–18 years. Moreover, mean scores in General Anxiety scale increase with age.

The bulk of excessive symptoms reported by children was for Somatic/Panic Anxiety and General Anxiety and came from girls. These results are consistent with other reports investigating parent-child agreement (Wren et al., 2004) and with epidemiological studies that reported higher rates of anxiety symptoms for girls (Kessler et al., 2005; Ogliari et al., 2006; Beesdo et al., 2009). However, we found similar mother-child agreement when stratifying by gender. This suggests that mother-son and mother-daughter accord is similar.

Following the approach of Birmaher et al. (1997) we separately analyzed parent- and self-report versions of the SCARED.

Differently from all the recent studies that have evaluated psychometric properties of the SCARED, an exploratory factor analysis for ordinal data was carried out on polychoric correlation matrices calculated on SCARED items. This approach has been recommended in the literature to properly analyze Likert type ordinal items (Muthén and Kaplan, 1985; Gilley and Uhlig, 1993; Jöreskog and Moustaki, 2001) with factor analysis procedures. Simulation studies showed that the solutions obtained using polychoric correlations provide a more accurate reproduction of the measurement model used to generate the data (Holgado-Tello et al., 2010).

Two different factorial solutions emerged for parent and child SCARED versions. Analyses on parent version revealed a 5 factor structure that approximately reproduces the original structure proposed by Birmaher et al. (1997, 1999) and by other authors (e.g., Jastrowski Mano et al., 2012) for parent version. Based on our findings, a 6 factor solution was chosen as best in the child version, with the first 5 factors reproducing “Birmaher” factors, whereas the 6th include the two items of Separation Anxiety subscale associated with a possible danger for parents (16. I have nightmares about something bad happening to my parents/31. I worry that something bad might happen to my parents). This result suggests that, in children, anxiety concerning of “being alone” or “a possible danger for children” emerged separately from anxiety about a possible danger for parents.

A limitation of the present study concern sample size. Indeed, for factor analysis, we considered SCARED questionnaire completed by 126 mothers and 147 children. However, when analyzing age effects in the evaluation of mother-child agreement, sample size within each age category is unbalanced and some categories are not well-represented. In particular, descriptive statistics for the SCARED scales are actually based on the information provided by 49 children in the age-class category 8–10 yrs, 53 children in the age-class category 11–13 yrs, 34 children in the age-class category 14–15 yrs and 35 in the last category 16–18.

In future work, the current analysis could be enhanced by addressing the following issues.

Data on parent reports were gathered from mother alone: analyses integrating father ratings may yield more informative results thus providing greater generalizability.

Future research should focus on longitudinal study designs, thus allowing test-retest reliability evaluations.

Hence, to conclude, our findings could help clinicians to determine which information, and from which rater, must be accounted for in evaluating treatment decisions. Moreover, due to the emerging differences in reported symptoms and cross-informant agreement depending on respondent's age, clinicians should consider potential age effect in interpreting SCARED results during assessment process.

SS and CB conceived of the study, participated in its design and coordination and drafted the manuscript; CB performed the statistical analysis; AO, CD, and LB participated in the design, coordination of the study and helped to draft the manuscript; LD contributed to literature collection and helped with data analysis. All authors read and approved the final manuscript.

This research has been funded by the FIRB Project RBFR12VHR7 entitled “Interpreting emotions: a computational tool integrating facial expressions and biosignals based on shape analysis and Bayesian networks.”