Introduction

Over the past two decades, the field of child psychology has made significant strides toward developing evidence-based methods and instruments for assessing anxiety and its related disorders [1]. Following from the development of traditional measures, such as the Revised Children’s Manifest Anxiety Scale [2], new multidimensional assessments have been established to measure different types of childhood anxiety disorder symptoms within the DSM [3]. For this purpose, three standardized instruments have been developed: the Spence Children’s Anxiety Scale (SCAS) [4], the screen for child anxiety related emotional disorders (SCARED) [5], and the Multidimensional Anxiety Scale for Children (MASC) [6].

As with the MASC [7] and the SCARED [8], the SCAS has been translated into various languages and used in several countries, including the Netherlands [9], Belgium [10], Germany [11], South Africa [12], and Japan [13].

Ishikawa et al. [13] investigated anxiety symptoms among Japanese children and adolescents using the SCAS Japanese version. Factor analysis using a multi-group method based on data from 1,046 children and 1,182 adolescents supported a 6-factor structure with a single higher-order factor reflecting the original factor structure [4, 14]. Scores for overall anxiety symptoms and each subscale were similar when comparing the Japanese sample with samples from other countries. For instance, girls reported more anxiety symptoms than boys, and adolescents presented with fewer anxiety symptoms than younger children.

The items most frequently endorsed by Japanese students differed from those endorsed by Western students. Specifically, the most common item endorsed by Japanese children was related to obsessive–compulsive disorder (“I have to keep checking that I have done things right, like the light switch is off, or the door is locked”); this item was not common for German children [11] or Australian adolescents [15]. Conversely, the least frequent anxiety symptoms were nearly the same for the Japanese sample and Western samples from previous studies.

One limitation of Ishikawa et al. [13] study was that it only relied on youth self-reports. The SCAS also has a parent version: the Spence Children’s Anxiety Scale for Parents (SCAS-P) [16]. The SCAS-P items correspond with the child version, and a confirmatory factor analysis on total 745 children (484 children in anxiety-disordered group and 261 children in normal control group) supported the 6-factor structure consistent with the SCAS. While child self-reports provide useful information, parental reports are also important in the cross-cultural study of childhood anxiety disorders.

Children with anxiety symptoms sometimes present with socially desirable behaviors [17]. Within Asian countries, such behavior is likely to be viewed positively. For example, shy and sensitive children and adolescents are likely to be more accepted by their peers, parents, and teachers in China [18]. In the Japanese instructional system, teachers expect discipline from students until they reach early adolescence [13]. Moreover, if not too excessive, socially anxious behaviors, such as being non-assertive, humble, and inhibited, are viewed as favorable in Japan. Specifically, children who are recognized by teachers as obedient and following rules (i.e., compliant) showed higher anxiety symptoms than children who show lower social skills and problem behaviors [19]. Nevertheless, Japanese children with anxiety symptoms reported high subjective distress in relationship between friends and academic activities in school [20]. In light of these indications, adults who are in contact with a child who has anxiety symptoms are likely to overlook the underlying problem, even if the child subjectively feels distress. Therefore, more information is necessary regarding parental reports of childhood anxiety symptoms among Asian countries.

The purpose of the current study was to examine the psychometric properties of the SCAS-P in a sample of parents who had a child in elementary school. This was part of the process of developing the Japanese version of the SCAS-P. Second, we examined the factor structure of the SCAS-P in an Asian sample. Third, internal consistency and concurrent validity of the SCAS-P were evaluated. Finally, the relationship between child self-reports and parental reports of anxiety symptoms were investigated.

In terms of factor structure of the SCAS-P, we hypothesized five models. The first model was a single factor model (Model 1). In this model, all symptoms were predicted to load on a single factor, which was related to general anxiety vulnerability. The model hypothesized that anxiety symptoms among children (rated by their parents) were constructed as a single dimension rather than various subtypes of anxiety. For example, behavioral genetic researches showed genetically based vulnerability related to a broad range of anxiety symptoms, like behavioral inhibition, neuroticism, and trait anxiety [21]. The second model (Model 2) had 6 uncorrelated factors, which were consistent with anxiety disorder subtypes within the DSM-IV-TR [3]: separation anxiety disorder, social phobia, generalized anxiety disorder, panic attack and agoraphobia, obsessive–compulsive disorder, and specific phobia. In this model, childhood anxiety symptoms were experienced as distinct and independent clusters as reported by the parents.

The third model (Model 3) proposed that anxiety symptoms would cluster into 6 correlated dimensions. The model hypothesized that parents would report their children’s anxiety symptoms as multiple and intercorrelated clusters corresponding with the DSM. Considering high comorbidity of childhood and adolescent anxiety disorders [22], each anxiety disorders subgroup can be related one another. In terms of goodness of fit indices, a 6-correlated model for the SCAS was a better fit for the original samples [4, 14]. The fourth model (Model 4), which had 6 factors, could be explained by one higher order factor. In light of general anxiety vulnerability and high comorbidity of childhood and adolescent anxiety disorders, hierarchical model which includes a general vulnerability and subgroup of anxiety disorders symptoms can be plausible. The original studies [4, 14] have also supported the validity of a higher-order factor structure by child self-reports. Furthermore, the following study with a Japanese sample has replicated a higher-order factor structure for the SCAS [13]. The higher-order model examined the degree to which the intercorrelation between factors could be explained by a single, second-order factor representing a general dimension of anxiety problems [15].

In the original studies validating the SCAS-P, 5 factors, and generalized anxiety as one higher order factor, were also evaluated as another plausible model [16]. In previous studies, it has been suggested that generalized anxiety disorder might be viewed as the ‘basic’ anxiety disorder [23]. For instance, the original study of the development of the SCAS [14] found that most of the variance in generalized anxiety was explained by one higher-order factor of anxiety in general. Thus, we also examined this model as the fifth model (Model 5). Finally, several studies suggested translated scales had the culturally-specific factor structure as well as the original six-factor structure [11, 12]. The previous study of development of the Japanese version of the SCAS supported the five factors model: separation anxiety disorder, panic (panic/generalized anxiety disorder), worry (generalized anxiety disorder/social phobia/separation anxiety disorder), specific phobia (fears of physical injury), obsessive–compulsive disorder, and one higher-order factor [13]. Therefore, the study also examined the Japanese-specific model as the sixth model (Model 6).

Methods

Participants

Participants for this study were caregivers or guardians (mostly parents) who had a child aged 9–12 years old. Caregivers were selected from six public elementary schools in the suburban Prefectures of Gunma and Miyazaki, Japan. In total, 714 caregivers consented to participate, and 677 caregivers and children (94.82 %) completed the questionnaires (caregivers’ mean age = 40.44, SD = 5.50). Their children consisted of 341 boys and 336 girls (mean age = 10.70, SD = .95). Most respondents were mothers (n = 568, 83.90 %) with 59 fathers, 7 grandmothers, and 1 uncle completing the questionnaires.

This study was approved by the IRB from the third and fifth author’s university. Based upon Japan’s IRB requirements, information about SES could not be collected; however, most respondents were middle-class, and there were very few families with a low SES based on demographics of the school districts. In addition, most participants were Japanese, and all participants could read/write Japanese.

Measures

Spence Children’s Anxiety Scale for Parents (SCAS-P)

The SCAS-P [16] is a 38-item parent-report measure of anxiety symptoms for children and adolescents. The Japanese version of the SCAS-P was developed to assess children’s anxiety symptoms on the basis of a parental report. The items from the original version of the SCAS-P were formulated, as closely as possible, to the corresponding items from the child version of the SCAS [4]. There were 38 items in the SCAS-P scored on a scale from 0 (never) to 3 (always). The SCAS-P consisted of 6 subscales: separation anxiety disorder (SAD; 6 items), social phobia (SoPh; 6 items), generalized anxiety disorder (GAD; 6 items), panic attack and agoraphobia (Panic/Ag; 9 items), obsessive–compulsive disorder (OCD; 6 items), and physical injury fears (PhInj; 5 items). The all 38 items are summed to produce an overall anxiety score. Internal reliabilities of the subscales were satisfactory for the community and clinical samples (Cronbach’s α with corrected Spearman Brown coefficients: .80–.92). Convergent and divergent validity were also confirmed by correlations between other parental and child reports based on the data from Australia and the Netherlands.

Child Behavior Checklist (CBCL)

The CBCL [24] was designed to capture information across a broad range of children’s behavioral and emotional problems. This scale is a standardized measure where parents report on their children’s problems. The version used for children aged 4–18 years (CBCL/4–18) includes 118 items. In terms of children’s behavioral and emotional problems, parents were asked to evaluate each description on a scale from 0 (not true) to 2 (very true or often true). The Japanese version of the CBCL has also been established, and its reliability and validity has been confirmed in several studies [25, 26]. Considering participants’ burden, the 14 items dealing with anxiety/depression were used for the current study.

Spence Children’s Anxiety Scale (SCAS)

The SCAS [4] is a 38-item self-report measure of anxiety symptoms designed for children and adolescents. The SCAS has 6 factors that correspond with the SCAS-P: SAD, SoPh, GAD, Panic/Ag, OCD, and PhInj. The Japanese version of the SCAS [13] has sufficient reliability coefficients: .94 and .92 for the full-scale scores among children and adolescents, respectively. In addition, the scale has sufficient test–retest reliabilities of 2–4 weeks: r = .76 for children and r = .86 for adolescents (ps < .001). Each item was rated on a 4-point scale in terms of its frequency ranging from 0 (never) to 3 (always). The anchors for the 4-point scales were completely consistent with those of the SCAS-P.

Procedure

The main aims and methods of the present study were explained to the school principals and teachers. After the school gave their approval, the questionnaires were distributed to the children. Children received an explanation of confidentiality and only those who consented to participate to the study completed the SCAS as a homeroom activity. The children then brought the questionnaires and a consent form home for their caregivers or guardians. Caregivers only completed the questionnaires when they consented to participate. Students handed in the completed questionnaires, which were sealed by their caregivers before submission. This study required a caregiver who took mainly charge of child care to complete the questionnaires. In Japan most of the principal nurturers are mothers. Since most of the caregivers could be mothers, this study focused on the mother’s data for the main analyses.

The SCAS-P was translated according to widely accepted guidelines for the successful translation of instruments in cross-cultural research [27]. One bilingual translator, who was a native Japanese speaker or understood Japanese culture, blindly translated the questionnaire from the original English version into Japanese. Another bilingual translator back-translated the questionnaire into English. Differences in the original and the back-translated versions were discussed and resolved by joint agreement of both translators.

Results

Preliminary Analysis

Since most of the caregivers were mothers as we hypothesized, we only used data from mothers (n = 568) for the following analyses.Footnote 1 There were no significant differences between fathers’ and mothers’ scores on the SCAS-P, SAD: mother: M = 2.20, SD = 2.52, father: M = 2.19, SD = 2.29, F(1, 625) = .12; SoPh: mother: M = 2.05, SD = 2.32, father: M = 1.58, SD = 1.74, F(1, 625) = 2.33; GAD: mother: M = 2.15, SD = 1.97, father: M = 2.10, SD = 1.80, F(1, 625) = .03; Panic/Ag: mother: M = .52, SD = 1.45, father: M = .68, SD = 1.20, F(1, 625) = .63; OCD: mother: M = 1.12, SD = 1.77, father: M = 1.17, SD = 1.93, F(1, 625) = .05; PhInj: mother: M = 3.23, SD = 2.77, father: M = 3.10, SD = 2.26, F(1, 625) = .12; and the total score: mother: M = 11.27, SD = 9.53, father: M = 10.81, SD = 8.22, F(1, 625) = .13. In addition, in terms of mothers’ data, there were no differences between scores for the two Prefectures, SAD: Miyazaki: M = 2.27, SD = 2.57, Gunma: M = 2.11, SD = 2.44, F(1, 566) = .55; SoPh: Miyazaki: M = 1.90, SD = 2.20, Gunma: M = 2.27, SD = 2.47, F(1, 566) = 3.49; GAD: Miyazaki: M = 2.23, SD = 1.80, Gunma: M = 2.03, SD = 2.17, F(1, 566) = 1.33; Panic/Ag: Miyazaki: M = .43, SD = .96, Gunma: M = .65, SD = 1.94, F(1, 566) = 3.12; OCD: Miyazaki: M = 1.08, SD = 1.55, Gunma: M = 1.16, SD = 2.05, F(1, 566) = −.25; PhInj: Miyazaki: M = 3.22, SD = 2.62, Gunma: M = 3.25, SD = 2.97, F(1, 566) = .02; and the total score: mother: M = 11.13, SD = 8.34, father: M = 11.47, SD = 10.98, F(1, 566) = .18.

Confirmatory Factor Analysis

According to previous studies of the SCAS-P [4, 14, 16], we examined 6 models: (1) one factor structure, (2) an uncorrelated 6-factor structure, (3) a correlated 6-factor structure, (4) one higher order factor with a 6-factor structure, (5) 5 factors with generalized anxiety as a higher order factor, and (6) one higher order factor with Japanese-specific 5-factor structure (26 items). We used EQS 6.1 to evaluate these models with an unweighted least-squares estimation (ULS) method. Because the children who participated in this study were from a community sample, many of the questionnaire items had positive kurtosis and skewness values. While it is conventional to use maximum likelihood (ML) and generalized least squares (GLS) method in SEM analysis, both of these estimations require the assumption of a multivariate normal distribution. Studies have shown that parameter estimates in moderately sized samples remain robust against the violation of this prerequisite [28]; however, the standard errors and test statistics become biased [29]. The asymptotically distribution-free (ADF) method has been developed to overcome these shortcomings [28], but ADF requires a very large sample to obtain reliable weight matrices. This was not suitable considering the cross-validation process we employed in the selection of the alternative models. Since ULS discrepancy function is free from distribution types and approximates to the ML estimator when the variances of the observed variables are in the same range [30], we opted for the ULS estimation for the present analysis. We used the goodness of fit index (GFI), adjusted goodness of fit index (AGFI), root mean square residual (RMR), normed fit index (NFI), comparative fit index (CFI), and root mean square error of approximation (RMSEA) for the evaluation of each model. The upper limits of the GFI, AGFI, NFI, and CFI are 1.0 and the closer to 1.0 means the model fit the data adequately. The lower limits of the RMR and RMSEA are .0 and the closer to the lower limits indicates a good fit for the data. Especially, whether the RMSEA is lower than .05 (probability of CLOSE fit: PCLOSE) is one of the useful goodness of fit indices [31].

For the Model 1, confirmatory factor analysis revealed that 15 items had a loading in excess of .40 on the single factor, 8 items had a loading lower than .35, and the remaining 15 items’ loadings were >.30. As shown in Table 1, although goodness of fit indices for the model was satisfactory, there was room of further examination.

Table 1 Goodness of fit indices for the 5 models
Full size table

All items for the Model 2 loaded on the hypothesized factors with factor loadings ranging from .23 to .75. Although only 6 items had loadings lower than .35 (as compared to 8 items in Model 1), the goodness of fit indices (GFI, AGFI, and NFI) were lower than .90. Thus, this model was not a good fit for the data.

In the Model 3, 33 items had a loading in excess of .35 on each hypothesized factor. Specifically, only 3 items, item 22 “When my child has problems, (s)he feels shaky (GAD),” item 24 “My child has to think special thoughts (like numbers and words) to stop bad things from happening (OCD),” and item 25 “My child feels scared if (s)he has to travel in the car, or on a bus or train (Panic/Ag),” had lower factor loadings. The GFI, AGFI, and NFI were in excess of .90, and the RMR was satisfactory. Thus, the 6-correlated factor model was used as the preferred model.Footnote 2 The factor loadings of each item are shown in Table 2.

Table 2 Confirmatory factor analysis of the SCAS-P
Full size table

The Model 4 had 6 factors and one higher order factor. Confirmatory factor analysis also revealed that only three items (items 22, 24, and 25) had a loading lower than .35 on the generalized anxiety disorder, obsessive–compulsive disorder, and panic attack and agoraphobia factors, respectively. As shown in Table 1, the goodness of fit indices were satisfactory.

The fifth model (Model 5) supported in the original study had 5 factors and generalized anxiety as one higher order factor [16]. As shown in Table 1, the goodness of fit indices was similar to a single factor model. Thus, we concluded that the model did not adequately describe the data as compared to the original sample.

The best fitting model was Model 6, where 26 of the items that loaded the highest onto the Japanese 5-factor model were included. These results clearly warrant a further investigation into the similarities and differences in the factor structure of anxiety between Japan and other countries; however, at this early stage of research, we opted to retain all items to enable cross-cultural comparison of the scale.

There were relatively small differences among goodness of fit indices in Model 3 and Model 4. Higher order factors are merely attempting to explain the covariation among first order factors in a more parsimonious way. Consequently, even when the higher order model is able to effectively explain the factor covariations, the goodness of fit of the higher order model can never be better than that of the corresponding first-order model.

Thus, based on previous studies [14, 15], we calculated target coefficients (the ratio of the χ 2 of the first order model to the χ 2 of the higher order model) to determine whether the higher order model provides a satisfactory explanation for the covariance between the first order factors [32]. The target coefficient has an upper limit of 1.00, which would be possible only if the covariance between first order factors could be totally explained by the second order factor. A target coefficient >.90 suggests that the second order factor provides a good explanation for the covariance between factors. Comparison of the χ 2 values of the six correlated factor model and the higher order model produced a target coefficient of .96.

In order to confirm the robustness of Model 3 and Model 4, a cross validation was examined by randomly selecting two samples from the original data. There were no significant differences between the first (n = 288) and second samples (n = 280) in terms of child age, child gender, mother age, and the mean score on the SCAS-P, t(566) = 1.67; χ 2 (1) = .45; t(566) = .17; and F(1, 566) = .53, respectively. In terms of the first sample, there were no clear differences between the two models: GFI = .948 versus .946, AGFI = .941 versus .940, NFI = .967 versus .966, and RMR = .026 versus .027, respectively. Meanwhile, Model 3 provided a slightly better fit to the data in light of the goodness of fit indices for the second sample: GFI = .919 versus .913, AGFI = .908 versus .902, NFI = .961 versus .958, and RMR = .029 versus .030, respectively. These findings suggest that the higher order model also provide a satisfactory explanation for the covariance between higher order factors and an adequate fit to the data.

Psychometric Characteristics of the Parental Reports of Anxiety Symptoms

Table 3 shows the mean value and standard deviations by age and gender for mothers’ reports of 568 children. A MANOVA revealed no significant effects for gender or an age × gender interaction for either the total scale score or any of the factor scores. However, a significant age effect for obsessive–compulsive disorder was found, F(1, 562) = 3.03, p < .001; multiple comparisons found no significant differences in terms of each age group.

Table 3 Means and standard deviations (in parentheses) of the SCAS-P
Full size table

Intercorrelations of the SCAS-P subscales ranged from weak to moderate, r = .29–.54 (Table 4). The highest correlations were found among separation anxiety disorder, social phobia, and generalized anxiety disorders, as well among generalized anxiety disorders and panic attack and agoraphobia. In terms of correlations with the total scale score, the correlation coefficients ranged from .80 to .65.

Table 4 Intercorrelations of the SCAS-P subscales
Full size table

Following previous studies [13, 15], an estimate of the common anxiety symptoms was indicated by the percentage of mothers that rated each item as either 3 “often” or 4 “always” (Table 5). The most prevalent symptoms were physical injury fears (dark, insects/spiders, and dogs), separation anxiety disorders related to sleeping alone and being on his/her own at home, and social phobia related to making a fool of oneself in front of people and poor performance at school. The least frequently reported items by mothers corresponded to panic attack and agoraphobia.

Table 5 Most common anxiety symptoms (rated as “often” or “always”) for Japanese children
Full size table

Reliability and Validity of the SCAS-P

The internal consistency was computed for the SCAS-P total scale score as well as each of the subscales. Since αs are largely dependent on scale length, corrected reliability coefficients were computed by the Spearman Brown coefficients based on the original study [16]. Cronbach’s α (and Cronbach’s α with corrected Spearman Brown coefficients in parentheses) for the SCAS-P total score was .88 (.96). The internal consistency of each subscale was moderate: .65 (.85) for separation anxiety disorder, .65 (.85) for social phobia, .58 (.81) for generalized anxiety disorder, .75 (.90) for panic attack and agoraphobia, .66 (.85) for obsessive–compulsive disorder, and .61 (.82) for physical injury fears. In order to examine convergent validity of the SCAS-P, scores were compared with those obtained from the anxiety/depression subscale of the CBCL. The correlation between the SCAS-P total score and the CBCL anxiety/depression score was .51 (p < .001). Each subscale also correlated significantly with the CBCL anxiety/depression score: r = .38 for separation anxiety disorder, r = .44 for social phobia, r = .48 for generalized anxiety disorder, r = .33 for panic attack and agoraphobia, r = .38 for obsessive–compulsive disorder, and r = .27 for physical injury fears (all ps < .001).

Relationship Between Parental and Child Self-reportsFootnote 3

In terms of the paired data from 568 child-parent dyads, correlation analysis revealed that most correlation coefficients between parental and child self-reports were significant but relatively low (Table 6). Even among the subscales, which were evaluated as the same conceptual, anxiety subtypes, the relationship between self- and parental-reports were very weak in the current sample. There was no significant correlation within the panic attack and agoraphobia subscale; remaining coefficients within the same subscales were lower than .20. We also directly compared the SCAS-P scores with the SCAS. As shown in Table 7, mothers’ scores were lower than child self-report scores for the full scale score, t(567) = 14.22, p < .001, separation anxiety disorder, t(567) = 8.32, p < .001, social phobia, t(567) = 11.92, p < .001, generalized anxiety disorder, t(567) = 7.12, p < .001, panic attack and agoraphobia, t(567) = 14.99, p < .001, obsessive–compulsive disorder, t(567) = 20.33, p < .001, and physical injury fears, t(567) = 3.76, p < .001.

Table 6 Correlations between the SCAS-P and the SCAS
Full size table
Table 7 Means and standard deviations of the SCAS and the SCAS-P
Full size table

Discussion

The current study developed a Japanese version of the SCAS-P, examined its reliability and validity, and investigated the relationship between child and parental reports of anxiety symptoms among Japanese children. Confirmatory factor analysis based on mothers’ ratings revealed that the parent scale could be satisfactorily explained by a 6-factor structure corresponding to current diagnostic criteria. Moreover, the Japanese original factor structure showed the highest goodness of fit indices based on 26 items. Thus, our results suggest that the factor structure of the parental measure was consistent with that of the child version of the SCAS [13].

The original study [16] showed no gender differences but did find age differences; specifically, younger children showed significantly higher scores for separation anxiety disorder, generalized anxiety disorder, and physical injury fears. However, the current study indicated no age and gender differences within the SCAS-P except for obsessive–compulsive disorder. These results might be related with the age range within each study. For the current study, we included children between the ages of 9 and 12, but the original study covered an age range between 6 and 18. On the other hand, the mean score in the current study was slightly lower than that of the original study [16], but similar with the US study [33]. In terms of Japanese data, the mean score of physical injury fear was higher than that in other countries. Therefore, similar to the child version, it is important to collect data from diverse countries and indicate culture norms of the SCAS-P. And also, internal consistencies of the subscale in this study were different from those of other countries. Specifically, the internal consistency of the GAD was relatively lower whereas overall corrected reliability coefficients were at the same level of those of the original study. Previous studies applied the translated versions of the SCAS, SCARED, and MASC also showed diverse internal consistency depending on each subscale [10, 34].

Furthermore, the Japanese five-factor model also fitted the data well. The results of the current and the previous study [13] in Japan suggested consistently that the uniqueness factor structure of Japanese provides a strong fit to the data using the 26 items the common anxiety structure also fits the data adequately. Especially, the Japanese specific model did not comprise the items representing “social phobia” as an independent factor. As mentioned before, if not too excessive, shy and social withdrawal behaviors are accepted in Japanese culture. Thus, regardless of children’s severity of anxiety disorder symptoms, Japanese parents might respond the items of social anxiety developed from Western culture norms. In Japan, a culture specific interpersonal phobic disorder known as “taijin kyofusho,” and several studies have discussed similarities and differences between social phobia and “taijin kyofusho.” Therefore, it is an important future issue to examine factorial invariance as well as culture-specific models of childhood anxiety disorders in comparison with other samples from Western countries.

With regard to the prevalence of each item within the SCAS-P, the most common anxiety symptoms dealt with specific fears of the dark and insects/spiders. These items were also ranked highly among children reporting anxiety symptoms in the previous Japanese study [13]. Besides, a previous cross-cultural study also reported that the fear of spiders, as well as other fear-evoking animals, was the most frequent elicitor of anxiety among Japanese children as compared to children in other countries [35]. An item for “checking” as related to obsessive–compulsive disorders was the most frequently endorsed anxiety symptom reported by Japanese children (39.0 % for children [13]). Yet, the corresponding item in the SCAS-P was ranked as less frequent. Items of obsessive–compulsive disorders (item 37) was a relatively common anxiety symptom in the SCAS (25.9 %) but not common in the SCAS-P. Ishikawa et al. [13] pointed out that instructions of ascertainment (e.g. “Check if the door is closed” or “Turn off the light if you turn it on”) are commonly given in Japanese elementary and middle schools.

Furthermore, an item represents worry about other people’s views (item 26) was only a frequent symptom for the SCAS (26.1 %). Although fear of public situations (item 9) was a highly ranked anxiety symptom in the SCAS and the SCAS-P, there were measurable differences in terms of frequency rates (31.0 % for children, 10.9 % for parents). As mentioned before, socially anxious behaviors are viewed as favorable in Japan if they are not extreme. Therefore, there are some differences in frequency of anxiety symptoms between child self-reports and parent-reports in Japan.

The most striking result of the current study was the relationship between child and parental reports. The correlation coefficients for this relationship were very low or non-significant. In the original study [16], with a non-clinical child sample, at least weak to moderate correlations were found between subscales from the SCAS and SCAS-P. Coefficients in the original study ranged from .60 to .27 among corresponding subscales; correlations were lower than .20 in the current study. Such low positive correlations between children’s self- and mothers’ reports suggests that when a child reports anxiety symptoms, the mother also recognizes her child has some anxiety symptoms; however, the mother might sometimes overlook her child’s symptoms, or her recognition is not at the same level as her child’s.

As mentioned above, Japanese adults tend to accept anxious behaviors as obedient and overlook their impairments even if the child subjectively feels distress [17]. Cross-cultural personality study of behavioral genetics indicated that Japanese were the least influenced from genetics compared with Germans or Canadians. That is, it was suggested that cultural characteristics in Japan inhibited hereditary behavioral tendencies [36]. It is rational that as a child grows up in one culture, cultural influence has a high impact on him/her. In light of our results, including item analysis and comparison between the SCAS and the SCAS-P, we might presume that Japanese parents tend not to see their child’s anxiety symptoms as serious as does the child because of such a cultural influence. In addition, many previous reports show that Japanese affirm the presence of somatic symptoms (e.g. stomachache) rather than affective ones such as anxiety [37, 38]. The SCAS-P has several items of physical symptoms related to panic attacks, but does not have those of functional somatic symptoms (FSS). The FSS are consistently associated with anxiety and depressive symptoms and disorders in childhood and adolescence [39]. Considering cultural specific expressions of anxiety, it is possible that several symptoms are not covered by the current measure. Therefore, future research should address the potential factors contributing to the inconsistency in child and parental reports of anxiety symptoms within Japanese samples.

There were some limitations to our study. First, we failed to include data from fathers into our entire analysis. Second, questionnaires focusing on adult psychopathology were not included in this study. These variables might influence parents’ reports of their own child’s psychopathology. Third, we did not include measures of other internalizing disorders, such as depression (or other externalizing disorders) for children. Thus, it is still unclear whether the SCAS-P can assess negative affect or anxiety specifically. Given this third limitation, it was impossible for us to examine divergent validity of the SCAS-P. Fourth, the study did not include any behavioral observations. A third, objective variable could be useful for examining the accuracy of parental reports as an indicator of children’s anxiety symptoms. Fifth, we failed to assess younger children and older adolescents, as well a child clinical sample. Furthermore, an evaluation of discriminant validity by comparing children with and without anxiety disorders is important for future research. Finally, it is necessary to determine whether the SCAS-P can distinguish among specific anxiety disorders.

In conclusion, the Japanese version of the SCAS-P shows adequate reliability and validity as a parental instrument for assessing anxiety symptoms among children. This is especially the case when the SCAS-P is used in conjunction with the SCAS. The factor structure of the Japanese version of the SCAS-P was consistent with the previous studies. Although the most frequently endorsed items within the SCAS-P were similar to those of the SCAS, some items (which could be influenced by cultural norms) were not endorsed to the same extent by parents. However, more data are needed to support the use of the SCAS-P alone to screen for the presence of clinically significant anxiety in Japanese youth. In Japan, the consistency between child and parental reports was lower than the original SCAS-P validation study [16]. A low level of concern or insufficient knowledge of childhood anxiety symptoms by Japanese parents could reflect this result. Therefore, future studies should examine the relationships to functional impairment and anxiety symptoms considering expression of psychological disorders in Asian population.

Summary

The current study investigated parent-report of their child’s anxiety symptoms in a sample of Japanese children as part of the process of developing a Japanese version of the SCAS-P. Participants were 677 parents and their children (aged 9–12 years). Confirmatory factor analysis on 568 mothers supported that the SCAS-P had a 6-factor structure corresponding to current diagnostic criteria as well as the original factor structure of the Japanese version. Therefore, future research should examine factorial invariance as well as culture-specific models of childhood anxiety disorders. The scale showed satisfactory internal consistency and good convergent validity. There were no significant gender or age differences except for on the obsessive-compulsive disorder subscale. Among Japanese children, the most prevalent symptoms reported by parents were related to fear of the dark and insects/spiders. The most striking result of this study was the relationship between child and parental reports. That is, the correlation coefficients for this relationship were very low. A low level of concern or insufficient knowledge of childhood anxiety symptoms by Japanese parents could reflect this result. Finally, we briefly discuss the utility of the SCAS-P for screening child anxiety symptoms and review suggestions for future research.