Psychology Help

Evaluation of the Children’s Depression Inventory—Short Version (CDI–S)

Johan Ahlen Uppsala University

Ata Ghaderi Karolinska Institutet

The Children’s Depression Inventory—Short Version (CDI–S), an abbreviated version of the widely used Children’s Depression Inventory (CDI), has been regularly used in recent research. In comparison to the original CDI, the CDI–S has not been rigorously evaluated for its psychometrics. The present study examined the dimensionality, convergent and discriminant validity, and gender differences of the CDI–S in a school-based sample of 809 children 8 –12 years of age. All children completed the CDI–S. One subsample additionally completed another measure of depression, 1 subsample completed a measure of anxiety, and 1 subsample completed the CDI–S at a second occasion, after 2 weeks. Information regarding parents’ education and household income were available for 476 children. We evaluated the dimensionality of the CDI–S in a series of exploratory factor analyses. Despite some evidence of multidimensionality, a bifactor model revealed that the variation of scores was primarily explained by variations of the general factor. Consequently, the CDI–S is most adequately interpreted as a univocal measure. The CDI–S showed high correlation to another measure of depression and a moderately high correlation to a measure of anxiety, with nonoverlapping confidence intervals. We also found that girls reported higher levels of depressive symptoms than did boys, and we found a negative correlation between depressive symptoms and socioeconomic factors for boys only. Future studies should preferably include a broader age range, to acquire a more comprehensive understanding of the validity of the CDI–S.

Public Significance Statement This study examined the reliability and validity of a widely used questionnaire assessing depressive symptoms in children. The results showed that the questionnaire could be interpreted as a one- dimensional measure of depression, that girls reported higher levels of depressive symptoms than did boys, and that boys from low-income households reported higher depressive symptoms than did boys from high-income households.

Keywords: depression, Children’s Depression Inventory, factor structure, gender differences, socioeconomic status

The Children’s Depression Inventory—Short Version (CDI–S; Kovacs, 2003) is a measure of depressive symptoms that is widely used for screening purposes, and it commonly serves as a second- ary measure in clinical trials (e.g., Hanks, McGuire, Lewin, Storch, & Murphy, 2016; Sibinga, Webb, Ghazarian, & Ellen, 2016). The 10-item CDI–S was originally developed as a one-dimensional, rapid version of the original 27-item Children’s Depression Inven- tory (CDI) for screening purposes (Kovacs, 2003). Despite its common occurrence in studies, few studies have examined its psychometric properties, and as far as we know, no study has thoroughly examined its dimensionality. For a better understand- ing, and to correctly interpret the results of the CDI–S, the present study aimed to examine test’s dimensionality, reliability, and con- vergent and discriminant validity in a large school-based sample. Additionally, the present study also aimed to examine mean dif-

ferences by gender, parent education, and household income. It should be noted that a new 28-item revised version of the full- length questionnaire (CDI–2) and a new 12-item short version (CDI–2S) derived from the revised version were developed re- cently (Kovacs, 2011). Although the new CDI–2S holds some possible advantages over the CDI–S, it has not yet been widely recognized in research and thus has been used in only a few studies.

Literature Review

Depression is the most prevalent lifetime psychiatric disorder (Kessler et al., 2005), and even though predominantly studied in adults and adolescents, depression has been found to be quite common also in children (E. J. Costello, Erkanli, & Angold, 2006). Depressive symptoms in children are quite stable over time (Cole & Martin, 2005) and are associated with impairment in school and peer functioning (Twenge & Nolen-Hoeksema, 2002). Further, depressive symptoms have been shown to predict impairment and depressive disorders in adolescence (Keenan et al., 2008), which in turn often continue into adulthood (Lewinsohn, Rohde, Klein, & Seeley, 1999) and involve an increased risk of suicidal behavior and alcohol and drug abuse (Bittner et al., 2007).

This article was published Online First December 5, 2016. Johan Ahlen, Department of Psychology, Uppsala University; Ata Gha-

deri, Department of Clinical Neuroscience, Karolinska Institutet. Correspondence concerning this article should be addressed to Johan

Ahlen, Department of Psychology, Uppsala University, P.O. Box 1225, SE-751 42 Uppsala, Sweden. E-mail: [email protected]

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

Psychological Assessment © 2016 American Psychological Association 2017, Vol. 29, No. 9, 1157–1166 1040-3590/17/$12.00 http://dx.doi.org/10.1037/pas0000419

1157

According to several studies, only a few children suffering from depression become psychiatrically assessed and hence identified. These studies generally report that only about 20% to 30% of children and adolescents with depression have received help for their condition (Allgaier et al., 2012; Bienvenu & Ginsburg, 2007; Essau, 2005; Samm et al., 2008).

To identify depression in early ages, self-reported screening questionnaires can be used to select “at risk” children for further assessment (Stevanovic, 2012; Timbremont, Braet, & Dreessen, 2004). Brief self-report questionnaires, for instance as a part of regular checkups for schoolchildren, could be a practical, efficient, and cost-effective solution (Ivarsson, Svalander, & Litlere, 2006; Samm et al., 2008; Stevanovic, 2012).

The original Children’s Depression Inventory (CDI; Kovacs, 1992) is the most frequently used questionnaire measuring depres- sive symptoms in children, and it has regularly been used in nonclinical populations (Giannakopoulos et al., 2009; Twenge & Nolen-Hoeksema, 2002). Several advantages of the CDI have been emphasized in the literature, for example its simple wordings, short completion time, and strong psychometrics properties (All- gaier et al., 2012; Craighead, Smucker, Craighead, & Ilardi, 1998; Myers & Winters, 2002; Samm et al., 2008; Twenge & Nolen- Hoeksema, 2002). However, others have critiqued the CDI for its low specificity and poor construct validity, suggesting the CDI could be better described as a measure of distress rather than depression (Myers & Winters, 2002).

The CDI was developed to measure several domains of depres- sion (Twenge & Nolen-Hoeksema, 2002), and its multidimension- ality has been extensively supported (for a review see Huang & Dong, 2014). Using a sample of 1,266 Florida schoolchildren, the constructor of the CDI originally found a five-factor model com- prising the following subscales: Negative Mood, Inter-Personal Problems, Ineffectiveness, Anhedonia, and Negative Self-Esteem (Kovacs, 1992). However, reliability coefficients of the subscales scores were not impressive, with alphas ranging from .58 to .69, and subsequent studies have yielded mixed results regarding the factor structure, ranging from one to eight factors (Huang & Dong, 2014; Kovacs, 1992).

In a meta-analysis of 24 studies examining the dimensionality of the CDI, an overall support was found for a five-factor structure (Huang & Dong, 2014). These factors, however, differed notice- ably from those reported by Kovacs (1992). The items of the original Negative Mood subscale loaded on two different factors (Somatic Concerns and Dysphoric Mood), the items of the original Inter-Personal Problems subscale and Ineffectiveness were in large part merged into a single factor (Externalizing), and the items of the original Anhedonia subscale loaded on two different factors (Somatic Concerns and Lack of Personal and Social Interest). Only the items of the original Negative Self-Esteem subscale loaded on a corresponding factor (Negative Self-Concept).

Despite strong support for multidimensionality, factors have shown to be strongly correlated, indicating possible higher order factors (Huang & Dong, 2014; Sun & Wang, 2015). However, possible hierarchical structure of the CDI has not been evaluated. In recent studies regarding the dimensionality of clinical question- naires, researchers have emphasized the insufficiency of merely examining one or multiple dimensionalities without investigating the partitioning of variance due to higher order and group factors (e.g., Ebesutani et al., 2012). Despite evidence of multidimension-

ality, the interpretation of subscale scores does not necessarily provide reliable information about the group factor (Brouwer, Meijer, & Zevalkink, 2013; Reise, 2012).

In comparison to the original CDI, the CDI–S has not been comprehensively evaluated. Only one study has rudimentarily examined the dimensionality of the CDI–S (Stevanovic, 2012). In a confirmatory factor analysis, a one-dimensional model provided a good fit to the data, but Stevanovic (2012) did not examine any alternative models.

Studies of the CDI (and the CDI–S for that matter) have gen- erally presented only Cronbach’s alphas when estimating the reli- ability of scale scores. Several researchers (Brunner, Nagy, & Wilhelm, 2012; Cortina, 1993) have highlighted the perfunctory use of the Cronbach’s alpha in psychometrics and the inappropri- ateness of using the Cronbach’s alpha coefficient when dealing with multidimensional and hierarchically structured constructs. In the case of multidimensional scales, Brunner et al. (2012) and Cortina (1993) proposed the use of the omega and the omega hierarchical coefficient.

In support of the convergent validity of the original CDI, large positive correlations to other self-ratings of depressive symptoms, such as the depression subscale of the Revised Child Anxiety and Depression Scale and other related constructs such as hopeless- ness, self-esteem, and locus of control, have been reported (Chor- pita, Yim, Moffitt, Umemoto, & Francis, 2000; Kazdin, Rodgers, & Colbus, 1986; Myers & Winters, 2002). However, more varied results have been reported regarding discriminant validity. First, the scores of the CDI have shown an ability to differentiate between depressed and nondepressed youths (Carey, Faulstich, Gresham, Ruggiero, & Enyart, 1987; Chorpita, Moffitt, & Gray, 2005) but not between depressed and other clinical groups (Carey et al., 1987). Further, the CDI has shown a lower correlation to measures of anxiety (.18 –.58) compared to measures of depression (.68 –.70) in some studies (Chorpita et al., 2000; Muris, Meesters, & Schouten, 2002), whereas other studies have found equal or higher correlation to measures of anxiety (.52– 63) compared to measures of depression (.58; Doerfler, Felner, Rowlison, Raley, & Evans, 1988).

Stevanovic (2012) examined the convergent validity of the CDI–S and found a large positive correlation (.60) to another self-rated questionnaire of depressive symptoms (the Short Mood and Feelings Questionnaire) and a moderate correlation (.38) to a measure of anxiety (Screen for Child Anxiety Related Disorders). However, the possible overlap of these correlation coefficients was not examined.

Regarding gender differences, slightly higher rates of major depressive disorders have been found in adolescent girls compared to adolescent boys (E. J. Costello et al., 2006). To explain gender differences emerging in adolescence, theories regarding the phys- iological changes of puberty and theories of social changes have been proposed (Twenge & Nolen-Hoeksema, 2002). Gender dif- ferences in the CDI scores have been evaluated in several studies. In a meta-analysis of 310 samples responding to the CDI, girls showed higher symptom ratings from the age of 13 (Twenge & Nolen-Hoeksema, 2002). The meta-analysis did not find any sig- nificant gender differences in children below the age of 13. How- ever, studies that more thoroughly examined gender differences in children regarding different factors of the CDI, have found higher scores for girls regarding symptoms of anhedonia and negative

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1158 AHLEN AND GHADERI

self-esteem and for boys regarding symptoms of ineffectiveness (Aluja & Blanch, 2002; Samm et al., 2008). Moreover, a study by Craighead and colleagues (1998) found higher scores for girls on Dysphoria and Negative Self-Esteem, and higher scores for boys on Ineffectiveness, Externalizing, and Social Problems. Only one study has examined gender differences for the CDI–S, and it found higher scores for girls (Houghton, Cowley, Houghton, & Kelleher, 2003).

The association between socioeconomic status (SES) and de- pression has shown ambiguous results in different studies. How- ever, a meta-analysis by Lorant et al. (2003) analyzing 51 studies found that in adults, low SES implied greater odds of depression onset and depression persistence. In this sample, both education and income showed a negative dose response relation to depres- sion.

Regarding CDI scores, no association to SES was found in the meta-analysis by Twenge and Nolen-Hoeksema (2002). However, a recent study from Greece showed that children from well-off families reported lower CDI scores (Giannakopoulos et al., 2009). To our knowledge no study has examined the association between depressive symptoms according to the CDI–S and SES factors.

The prevalence of major depressive disorders has been found to vary broadly across countries, whereas other aspects of depression such as age of onset, gender differences, and persistence have been found to be quite consistent across countries (Kessler & Bromet, 2013). Regarding the CDI, mean scores have been found to vary between studies performed in different countries, with European samples typically reporting lower mean scores than do American samples (Craighead et al., 1998; Doerfler et al., 1988; Giannako- poulos et al., 2009; Ivarsson et al., 2006; Larsson & Melin, 1992). The CDI–S has not been examined in populations as large as those tested with the original CDI; however, available results have indicated that CDI–S symptoms are rather consistent across Euro- pean and Caucasian American samples (Allgaier et al., 2012; Mata, Thompson, & Gotlib, 2010; Stevanovic, 2012; Thompson et al., 2010) but higher in Australian samples and African American and Mexican American samples (Bauman, 2008; Bennett, Sulli- van, & Lewis, 2010; Vines & Nixon, 2009).

The current study contributes to the research literature in several ways. First, in comparison to the psychometric and normative data presented by the constructor of the CDI–S (and the CDI–2S), the results in the current study are based on administration of the CDI–S in its proper short form rather than the full-length ques- tionnaire. As expressed by Smith, McCarthy, and Anderson (2000), “The key empirical evidence should not be based on a sample in which the full, long form was administered. One should show that the short form, as it will be used, performs as hypoth- esized” (p. 107). Further, the current study is the first to thoroughly examine the factor structure of the CDI–S, by using adequate analyses for a multifaceted construct such as depression. Last, the present study is the first to examine associations between scores on the CDI–S and SES factors.

The first aim of the present study was to examine the dimen- sionality of the CDI–S and the reliability of scale scores, by comparing one-dimensional, multidimensional, and bifactor mod- els. The second aim of the present study was to examine the convergent and discriminant validity of the CDI–S. The third aim was to examine gender differences and associations with SES factors.

Method

Participants

Participants were recruited from 17 schools in Stockholm, Swe- den. Written consent, signed by parents, was required for partici- pation in the study. A total of 1,262 children were asked to participate. The parents of 809 children (64%) consented, whereas 113 parents (9%) refused to consent and 340 parents (27%) did not respond to the invitation. Among the children included, 405 (50.1%) were girls and the age range was 8 –12 years (M � 9.8).

The parents of 454 children (56.1%) and 476 children (58.9%) provided information regarding household income and education, respectively.

Procedure

A total of 804 children (99%) completed the CDI–S at their school. The remaining five children who were sick or absent at all occasions offered to complete the questionnaire. Most of the children (n � 688) were recruited within an intervention study (Ahlen, Hursti, Tanner, Tokay, & Ghaderi, 2016) and additionally completed the Spence Children’s Anxiety Scale (SCAS; Spence, 1997) on the same occasion as the completion of the CDI–S. The remaining 116 children were recruited from the same schools but from six school classes that did not participate in the intervention study. These classes were randomly divided into two subgroups, either completing the Revised Child Anxiety and Depression Scale (n � 56) on the same occasion as the CDI–S or completing the CDI–S on two occasions with 2 weeks in between (n � 60). Four of the children in the latter subgroup completed only the CDI–S on the first occasion and were therefore not included in the test–retest analysis. We chose the procedure of using different subgroups for different analyses (divergent validity, convergent validity, and test–retest reliability) due to the schools’ request to reduce the number of questionnaires completed by the children. To increase understanding, a clinical psychologist, or a master’s-level psychol- ogy student was present and read items out loud and answered any questions regarding the comprehension of items. Parents were asked to respond to a web survey containing questions about their education and income. All questionnaires (paper and web surveys) were coded to ensure confidentiality, and no IP addresses were collected from the web surveys.

Measures

Children’s Depression Inventory—Short Version (Kovacs, 2003). The CDI–S is a 10-item short form of the original 27-item CDI. It was originally developed as a one-dimensional question- naire by excluding items with low interitem correlation (Kovacs, 2003). According to the constructor, the CDI–S provides an ac- ceptable approximation of the total scale content, given a strong correlation (r � .89) between the CDI and the CDI–S. The 10 items in the CDI–S cover sadness, pessimism, self-deprecation, self-hate, crying, distress, negative body image, loneliness, lack of friends, and feeling unloved. Although the new 28-item version CDI–2 is similar to the original 27-item CDI (sharing 25 items), the new 12-item short version CDI–2S differs markedly from the CDI–S (sharing only six items). The CDI–2S was refined by

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1159CHILDREN’S DEPRESSION INVENTORY—SHORT VERSION

another method, specifically by examining the items’ ability to discriminate between depressed and nondepressed children (Ko- vacs, 2011). In an evaluation of the CDI–S in a pediatric sample in Germany, Allgaier et al. (2012) found that the CDI–S showed a predictive validity (diagnostic accuracy) as good as that of the original CDI. Kovacs (2003) found high internal consistency of scores in the CDI–S (� � .80), and a meta-analysis of Cronbach’s alpha estimates from 22 studies using the CDI–S found similar results (� � .77; Sun & Wang, 2015).

The CDI–S has not previously been subjected to an evaluation in Sweden; however, the scores of the original CDI (Ivarsson et al., 2006) have shown good internal consistency (� � .86), similar to an overall alpha coefficient of .84, estimated in a meta-analysis of 331 studies from 44 countries (Sun & Wang, 2015). Internal consistency of CDI–S scores in the current sample was .80. Total score mean and standard deviation in the current sample (M � 1.78, SD � 2.55) were comparable to means and standard devia- tions of the same age group reported in studies in Serbia (Ste- vanovic, 2012) and in Germany (Allgaier et al., 2012) and in the United States, when including predominantly Caucasian Ameri- cans (Thompson et al., 2010). Children in the current sample, however, scored lower on the CDI–S compared to the same age group in an Australian sample (Vines & Nixon, 2009) and samples of African Americans (Bennet et al., 2010) and Mexican Ameri- cans (Bauman, 2008).

Spence Children’s Anxiety Scale (SCAS; Spence, 1997). The SCAS is a questionnaire assessing anxiety symptoms. It contains 44 items, of which 38 are divided into the following subscales; separation anxiety disorder, social phobia, obsessive– compulsive disorder, panic attack and agoraphobia, physical injury fears, and generalized anxiety disorder. The remaining six items are “filter items” that serve to reduce negative response bias. An evaluation in a Swedish sample found excellent internal consis- tency of the total score (� � .93) and acceptable internal consis- tency of the subscale scores (� � .71–.76; Essau, Sasagawa, Anastassiou-Hadjicharalambous, Guzmán, & Ollendick, 2011). In- ternal consistency of scores in the current sample was .93. Total score mean and standard deviation in the current sample (M � 26.4, SD � 15.1) were comparable but somewhat higher than German norms and were comparable but somewhat lower than Australian norms (Essau, Sakano, Ishikawa, & Sasagawa, 2004; Spence, 1998 respectively).

Revised Child Anxiety and Depression Scale (RCADS; Chorpita et al., 2000). The RCADS is an adaption of the SCAS that was developed to better correspond to the dimensions of anxiety disorders and major depression, according to the Diagnos- tic and Statistical Manual of Mental Disorders (4th ed.; DSM–IV; American Psychiatric Association, 2000). In the present trial, only the depression subscale of the RCADS (RCADS-MDD) was com- pleted by children. The RCADS-MDD consists of 10 items cor- responding to the criteria of major depressive disorders in the DSM–IV. The scores of the RCADS-MDD have showed accept- able to good internal consistency (� � .76 –.87) and good test– retest reliability (intraclass correlation coefficient [ICC] � .86). Moreover, it has shown an acceptable ability to discriminate be- tween those with and those without any affective disorder (Ebe- sutani et al., 2012). The RCADS has not been psychometrically evaluated in Sweden; however, in the current sample, mean and standard deviation of the RCADS-MDD subscale (M � 5.9, SD �

4.3) and internal consistency of scores (� � .84) very much resembled the results of evaluations in the same age group in Denmark and the Netherlands (Esbjørn, Sømhovd, Turnstedt, & Reinholdt-Dunne, 2012; Muris et al., 2002 respectively). Children in the current sample, however, scored lower on the RCADS- MDD subscale compared to samples in Hawaii (Chorpita et al., 2000; Ebesutani et al., 2012).

Socioeconomic status. Parents reported total monthly house- hold income, and parents’ educational level was dichotomized into “no postsecondary education” (32%, 95% confidence interval [CI: 27%, 36%], N � 476) or “postsecondary education” (68%, 95% CI [64%, 73%]) and then labeled low education and high education. According to the Swedish National Agency for Education (2015), 62% of parents in Stockholm county had a postsecondary educa- tion, which means that the subsample providing household income and parents’ educational level had only slightly higher education than did the population the sample was drawn from.

Data Analysis

Earlier studies examining the dimensionality of the CDI have generally used principal components analysis as the reduction method (for a review, see Huang & Dong, 2014). However, several researchers have argued for the inadequacy of a principal compo- nents analysis in assessing the factor structure of questionnaires and have suggested factor analysis (e.g., Baglin, 2014). Conse- quently, to examine possible latent factors, we chose explanatory factor analysis (EFA), using the principal axis as the extraction method as recommended for nonnormally distributed data (A. B. Costello & Osborne, 2005).

Item-level data for the CDI–S are ordinal and have been shown not to fulfill the requirements of multivariate normality in non- clinical populations (Houghton et al., 2003). Factor analyses based on Pearson correlations do not provide sound results if these requirements are not met (Basto & Pereira, 2012). Correlations are affected by the similarity of the distribution of data, and using EFAs based on Pearson correlations for nonnormally distributed ordinal data could produce factors that are based solely on item distribution similarity (Basto & Pereira, 2012). To overcome this problem, we conducted EFAs based on polychoric correlations.

To examine how many factors to retain, we chose two different methods. First, the Kaiser criterion, which includes keeping all factors with an eigenvalue larger than 1. The Kaiser criterion has been extensively used in research. Because recent studies have suggested that the Kaiser criterion might under- or overestimate the number of factors to retain (e.g., Zwick & Velicer, 1986), we also used another method: parallel analysis. Parallel analysis re- tains factors that explain more variance than do factors generated from random data (O’Connor, 2000).

Different dimensions of depressive symptoms have been repeat- edly shown to be strongly correlated. Therefore, an oblimin rota- tion was used to uniquely delineate factors. To examine one- or multidimensionality, we performed the Schmid–Leiman bifactor orthogonalization (Schmid & Leiman, 1957). Compared to the second-order, or correlated, factor models, the Schmid–Leiman bifactor model estimates an item’s factor loadings on an uncorre- lated general factor (second-order factor) and group factors (Reise, Moore, & Haviland, 2010). The Schmid–Leiman estimation method therefore more adequately enables an analysis of how the

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1160 AHLEN AND GHADERI

variance is apportioned to the general versus the group factors and clarifies the viability of group factors.

To assess the dimensionality, we first calculated common vari- ance attributable to the general and group factors (explained com- mon variance). To assess reliability, we calculated the coefficient omega for the general and any possible group factors. Coefficient omega estimates the degree of measurement precision when all sources of common variance (general and group factors) are thought to underlie the score. In the bifactor model, we also calculated the coefficient omega hierarchical, where only the com- mon variance of the factor of interest (general or group factor) is thought to underlie the score, thus treating other sources of com- mon variances as error variance (Brunner et al., 2012; Reise, 2012; Reise et al., 2010). Last, we calculated two fit indices for an evaluation of the goodness of fit applicable to ordinal data: good- ness of fit index (GFI) and root-mean-square residual (RMSR; Jöreskog & Sörbom, 1981). GFI values over .95 are interpreted as a good fit; RMSR values below .05 are interpreted as a good fit, and values below .10 are considered to imply an acceptable fit as recommended by Schermelleh-Engel, Moosbrugger, and Müller (2003).

To explore the convergent and discriminant validity of the CDI–S, we examined correlations to other self-report measures. We also compared magnitudes of correlations by examining the presence or absence of overlapping confidence intervals based on bootstrap standard errors.

To explore any possible gender differences, and to examine any differences between high and low education for parents, we ex- amined mean differences by calculating 95% confidence intervals based on bootstrap standard errors. To examine the association between scores of the CDI–S and the RCADS-MDD, the SCAS, and household income, we calculated Pearson correlation coeffi- cients with 95% confidence intervals based on bootstrap standard errors. Last, to compare the subgroups completing different ques- tionnaires, we compared continuous outcomes in bootstrap analy- ses of variance (ANOVAs) and count data using the Fisher’s exact test. We considered mean differences and correlations as signifi- cant if the 95% confidence intervals did not include 0 and ANO- VAs and Fisher’s exact tests when p values were less than .05. All bootstrap analyses included 1,000 bootstrap samples, resampled by replacement from the original sample. All statistical analyses were performed in the R software program (R Core Team, 2015) using the “psych” package (Revelle, 2015), the “boot” package (Davi- son, & Hinkley, 1997), and the “car” package (Fox, Friendly, & Weisberg, 2013).

Results

Preliminary Analyses

A Fisher’s exact test showed no significant difference on gender distribution between the subgroup completing the SCAS, the sub- group completing the RCADS-MDD, and the subgroup complet- ing the CDI–S on two occasions (p � .08). Further, a bootstrap ANOVA showed no significant difference on CDI–S scores be- tween the three subgroups, F(2, 801) � .53, p � .69. However, a bootstrap ANOVA showed a significant difference in age between the three subgroups, F(2, 798) � 216.43, p � .001. The subgroup completing the CDI–S on two occasions was older than the sub-

group completing the RCADS-MDD (MD � .89, 95% CI [.59, 1.21], n � 116) and the subgroup completing the SCAS (MD � 1.62, 95% CI [1.45, 1.78], n � 748). Additionally, the subgroup completing the RCADS-MDD was also older than the subgroup completing the SCAS (MD � 0.73, 95% CI [.46, .98], n � 744).

A bootstrap mean difference test showed no significant differ- ence in CDI–S scores between children whose parents completed the web survey and children whose parents did not (MD � .31, 95% CI [�.05, .66], N � 809). The test–retest correlation coeffi- cient between the two occasions of the subsample was .74 (95% CI [.63, .84], N � 56), indicating an adequate test–retest reliability of total scores.

Explanatory Factor Analysis

Table 1 presents factor loadings, proportions of total variance, proportions of variance explained, reliability indices, and goodness of fit statistics for the one-factor model, the two correlated factors model, and the bifactor model for the CDI–S.

In the first EFA, two factors had an eigenvalue greater than 1 and should, according to the Kaiser criterion, be retained in anal- yses. However, according to parallel analyses, the recommenda- tion is to retain only one factor. We therefore decided to perform EFAs of a one-factor model, a two-factor model, and a bifactor model with one general factor and two group factors. No other models were examined.

The one-factor model. The one-factor model showed a good fit to the data according to the GFI and a fairly acceptable fit according to the RMSR index. All items showed high loadings (�.30) on the factor (mean of loadings � .69, range � .62–.84). The omega coefficient was .90, indicating a good reliability of total scale scores.

The two correlated factors model. The two correlated fac- tors model also showed a good fit to the data according to the GFI and a more acceptable fit according to the RMSR index compared to the one-factor model. All items loaded high on a factor. One item cross-loaded, meaning it loaded high on both factors. The first factor included seven items (mean of loadings � .67, range � .51–.86) and was labeled Negative Self-Concept and Loneliness (NS & L). This factor explained about 62% of the common variance. The second factor included three items (mean of load- ings � .78, range � .64 –.86) and was labeled Sadness and Distress (S & D). This factor explained about 38% of the common variance. The omega coefficients of the factors were .86 and .83, respectively, indicating a good reliability of scale scores.

The bifactor model. The bifactor factors model showed a good fit to the data according to the GFI and an acceptable fit according to the RMSR index in similarity to the two correlated factors model. First, all items loaded high on the Depression factor (mean of loadings � .61, range � .53–.72), which however, was somewhat lower than in the one-factor model, indicating that at least some variance in item responses was influenced by group factors. The general factor explained about 67% of the common variance. Second, all items also showed high loadings on one of the group factors. There were no cross-loadings. The first factor included the same seven items as in the two correlated factors model (mean of loadings � .40, range � .30 –.51). This factor explained about 20% of the common variance after controlling for the general factor. The second factor included the same three items

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1161CHILDREN’S DEPRESSION INVENTORY—SHORT VERSION

as in the two correlated factors model (mean of loadings � .46, range � .38 –.51). This factor explained about 13% of the common variance after controlling for the general factor. The omega coef- ficient of the general factor was .92, and the omega coefficients of the group factors were .88 and .84, respectively, indicating a good reliability of scale scores. However, the omega hierarchical coef- ficient was .73 for the general factor and .28 for both group factors, Thus, despite high omega coefficients (.82–.88) of group factors most of the variance of the scores was due to the general factor, and only 28% of the variance of scores was explained by the content specific to the group factor, beyond the content of the general factor. In the following analyses, we examined the total scale scores, the NS & L factor scores, and the S & D factor scores.

Convergent and Divergent Validity

There was a positive and significant correlation between the CDI–S and the RCADS-MDD, r(54) � .72, 95% CI [.62, .88].

Further, there was also a positive and significant correlation be- tween the CDI–S and the SCAS, r(686) � .53, 95% CI [.47, .59]. The correlation between the two depressive questionnaires was considered significantly larger than the correlation between the CDI–S and the anxiety questionnaire, because estimated confi- dence intervals based on bootstrap standard errors did not overlap.

Gender Differences

See Table 2 for means, standard deviations, and test statistics regarding gender differences on scale scores. Three bootstrap mean difference tests showed significant differences between gen- ders, with girls reporting higher scores on the total scale (d � .21, 95% CI [.08, .35]), the NS & L factor (d � .15, 95% CI [.02, .29]), and the S & D factor (d � .27, 95% CI [.13, .47]). All differences were small or trivial, when interpreted according to guidelines by Cohen (1987).

Table 1 Factor Loadings, Reliability Indices, and Fit Statistics for One-, Two-, and Bifactor Exploratory Factor Analyses Models for the CDI–S

Variable One-factor model (D)

Two correlated factors model Bifactor model

NS & L S & D D NS & L S & D

Item Sadness .68 �.02 .86 .67 �.01 .51 Pessimism .66 .56 .14 .56 .33 .08 Self-deprecation .70 .81 �.08 .59 .48 �.05 Self-hate .84 .86 .03 .72 .51 .02 Crying .67 �.01 .84 .66 �.01 .50 Distress .66 .14 .64 .62 .08 .38 Negative body image .63 .70 �.04 .53 .42 �.03 Loneliness .81 .55 .33 .70 .33 .20 Lack of friends .63 .72 �.06 .53 .43 �.04 Feeling unloved .62 .51 .16 .53 .30 .09

Total variance (%) 47.6 33.0 20.2 37.8 11.7 7.1 Common variance (%) 62.0 38.0 66.8 20.6 12.6 � .90 .86 .83 .92 .88 .84 �h .73 .28 .28 GFI .98 .99 .99 RMSR .09 .06 .06

Note. Italic font indicates the factor on which the item loaded most heavily according to the model examined. CDI–S � Children’s Depression Inventory—Short Version; D � Depression factor; NS & L � Negative Self-Concept and Loneliness factor; S & D � Sadness and Distress factor; �h � omega hierarchical coefficient; GFI � goodness of fit index; RMSR � root-mean-square residual.

Table 2 Descriptive Statistics on the Total Scale and Subscales for the Total Sample and Divided by Gender, and a Bootstrap Mean Difference Test by Gender

Scale

Total (N � 804)

Girls (n � 402)

Boys (n � 402) Girls/boys

M SD M SD M SD MD 95% CI df

Total scale 1.78 2.55 2.05 2.75 1.51 2.31 .54� [.19, .92] 803 NS & L 1.36 1.95 1.51 2.04 1.21 1.85 .30� [.02, .58] 802 S & D .42 .96 .55 1.13 .30 .74 .25� [.11, .39] 801

Note. CI � confidence interval; NS & L � Negative Self-Concept and Loneliness factor; S & D � Sadness and Distress factor. � p � .05.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1162 AHLEN AND GHADERI

Education

See Table 3 for means, standard deviations, and test statistics regarding differences in CDI–S between high and low education in parents. A series of bootstrap mean difference tests showed no significant differences in CDI–S total scale or subscale scores between high and low education in parents. When divided by gender, we found no significant differences in the CDI–S total scale or subscale scores between girls with parents with high or low education. However, we found a significant difference on the CDI–S total scores between boys with parents with low versus high education (d � .32, 95% CI [.05, .59]) and on the NS & L factor scores between boys with parents with low versus high education (d � .34, 95% CI [.07, .62], but no differences regarding the S & D factor scores.

Household Income

See Table 4 for correlation coefficients, as well as confidence intervals between the CDI–S and household income. In the total sample, we found a significant small negative correlation between total scale and household income, and a small negative correlation between the NS & L factor and household income. We found no significant correlation between the S & D factor and household income. Rerunning the analyses divided by genders, we found no significant correlations between the CDI–S and household income for girls for neither the total scale nor the subscales. However, for boys we found a significant small negative correlation between total scale and household income, a small negative correlation between the NS & L factor and income, and a small negative correlation between the S & D factor and income.

Discussion

The aim of the present study was to examine the dimensionality and the convergent and discriminant validity of the CDI–S, a commonly used measure of depressive symptoms in children, and

to examine possible gender differences and associations with SES factors.

Although the concept of depression is multifaceted and the items of the CDI–S reflect several dimensions of depression, the CDI–S should above all, and as intended by its author, be inter- preted as a univocal measure. Although the two correlated factors model (presented in this article) seemingly demonstrates a sound model with a good fit to data, the variation of scores in these group factors are better interpreted as variations in the general factor than variations stemming from the specific content of the group factors. The three items forming the S & D group factor in the current study have been found to cluster in the same factor in earlier studies (Huang & Dong, 2014), labeled Dysphoric Mood or Neg- ative Affect. The seven items forming the NS & L group factor in the current study have generally loaded on different factors in earlier studies. However, the majority of these items (pessimism, self-hate, negative body image, feeling unloved) have generally clustered in the same factor in other studies, labeled Negative Self-Concept or Negative Self-Esteem (Huang & Dong, 2014). A practical conclusion of these results is that the scores of the total scale are adequately interpreted as a measure of depression,

Table 3 Descriptive Statistics for the Total Scale and Subscales, Divided by Gender and Educational Level, and a Bootstrap Mean Difference Test by Education

Sample and scale

High education Low education High education/low education

M SD n M SD n MD 95% CI df

Total D 1.55 2.18 325 1.89 2.72 151 .35 [�.17, .86] 474 NS & L 1.18 1.57 325 1.50 2.26 151 .32 [�.10, .68] 474 S & D .37 .90 325 .40 .83 151 .03 [�.14, .20] 474

Girls D 1.97 2.60 154 2.03 2.91 73 .05 [�.92, .43] 225 NS & L 1.47 1.80 154 1.56 2.50 73 .09 [�.68, .42] 225 S & D .51 1.13 154 .47 .87 73 .04 [�.10, .36] 225

Boys D 1.16 1.62 171 1.77 2.55 78 .61� [ .11, 1.60] 247 NS & L .92 1.27 171 1.44 2.03 78 .51� [ .11, 1.26] 247 S & D .24 .60 171 .33 .78 78 .09 [�.07, .41] 247

Note. CI � confidence interval; D � Depression factor; NS & L � Negative Self-Concept and Loneliness factor; S & D � Sadness and Distress factor. � p � .05.

Table 4 Pearson Correlation Coefficients Between the CDI–S and Household Income, With 95% Confidence Intervals Based on Bootstrap Standard Errors

CDI–S scale

Total sample (N � 454) Girls (n � 215) Boys (n � 239)

Total scale �.11� [�.20, �.02] �.06 [�.12, .14] �.17� [�.35, �.10] NS & L �.11� [�.21, �.02] �.06 [�.13, .13] �.17� [�.35, �.09] S & D �.07 [�.15, .02] �.03 [�.11, .13] �.11� [�.27, �.03]

Note. CDI–S � Children’s Depression Inventory—Short Version; NS & L � Negative Self-Concept and Loneliness factor; S & D � Sadness and Distress factor. � p � .05.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1163CHILDREN’S DEPRESSION INVENTORY—SHORT VERSION

whereas the scores of the S & D and the NS & L factors could be only cautiously interpreted.

The results of this study are in line with those of a recent study of a depressive measure for adults, in which the total score of the general factor was the only one that could be reliably interpreted (Brouwer et al., 2013). Earlier studies of the CDI have not exam- ined the hierarchical structure when considering correlated factors, which is why it is hard to say anything about how our results are related to the factor structure of the original CDI.

The current study sample was recruited in Sweden, and its scores were comparable to those of European and Caucasian American samples. However, higher scores on the self-report measures used in the current study have been found in Australian and non-Caucasian samples in the United States. These differences generate some uncertainty regarding the overall generalizability of the results and motivate further evaluations of the CDI–S in samples with a broader ethnicity.

We found support for the convergent and discriminant validity of the CDI–S, in that the scores of the CDI–S showed a high correlation (r � .72) with a measure of depression and a moder- ately high but lower correlation (r � .53) to a measure of anxiety. Worth noting is that the correlation coefficients between CDI–S and RCADS-MDD, and between the CDI–S and SCAS presented in our study, highly resemble coefficients presented in evaluations of the original CDI and the same measures (r � .70 and r � .48, respectively; Chorpita et al., 2000; Spence, 1998).

The present study further showed that girls reported higher symptoms of depression. This result is interesting because it is seemingly not in line with previous research, where gender differ- ences on total scale have been shown only above the age of 13. One possible explanation is the fact that most items in the CDI–S reflect the constructs of sadness and negative self-esteem, where girls have shown higher scores than boys have also in younger ages. The CDI–S does not contain the same number of items reflecting constructs of externalization or ineffectiveness where boys generally score higher.

We also found a negative association between scores of the CDI–S and SES (parents’ education and household income) for boys, indicating higher symptoms of depression among boys whose parents have low education and low income. The associa- tion was most apparent between depressive symptoms and house- hold income, and this seems to be primarily linked to the items of the NS & L factor. The associations are significant but small. For example, income explains only about 3% of the variation in depressive symptoms in boys. Researchers such as Nolen- Hoeksema (2001), Piccinelli and Wilkinson (2000), and Keenan and Hipwell (2005) have examined possible precursors of depres- sion in girls and have suggested factors such as increased stress reactivity and more empathy and compliance, as well as the idea that girls may have less power and therefore, to a larger extent, experience adverse events, harassment, lack of respect, and con- strained choices. The present study suggests that low SES is associated with depressive symptoms for only boys, in the 8 –12 age level. Further, the primary link between SES and the NS & L factor, and the partial absence of association between SES and the S & D factor, could also imply that the real relationship exists between SES and the specific content of the group factor, as opposed to a relationship between SES and the general factor.

Worth noting is that the procedure that founded the new short version CDI–2S might reasonably mean that it has a stronger predictive validity than does the CDI–S, and future studies should preferably evaluate the CDI–2S in an independent sample, and in its actual form, to provide further evidence of its validity and reliability of scale scores.

There are a number of limitations regarding the present study that must be taken into consideration when interpreting the results. First, the study sample included only children between 8 and 12 years, who overall reported few depression symptoms. Therefore, similar studies with a broader age span, including adolescents (among whom depressive symptoms are more prevalent and the difference in depressive symptoms between genders increases) would be of much interest to more comprehensively examine the dimensionality, gender differences, and associations with SES. Second, although efforts were made to recruit a representable population, the sample was recruited in an urban environment and therefore has a limited external validity. Third, we used different subsamples for different indices of psychometric criteria, which is problematic, especially because the age of the subsamples differed. Further, the correlations between the CDI–S and the RCADS- MDD, and between the CDI–S and the SCAS, may be inflated due to shared method variance. Therefore, further validity studies of the CDI–S should include additional methods to self-report, pref- erably structured diagnostic interviews. Fourth, there was also a rather sizable group of parents who did not report their SES, which limited the internal validity of the associations between scores of the CDI–S and SES. With these limitations in mind, we conclude that only the total scale score of the CDI–S could be reliably interpreted and that gender differences regarding symptom ratings and associations with SES are found at young ages.

References

Ahlen, J., Hursti, T., Tanner, L., Tokay, Z., & Ghaderi, A. (2016). Pre- vention of anxiety and depression in Swedish school children: A Cluster- randomized effectiveness study. [Manuscript submitted for publication].

Allgaier, A. K., Frühe, B., Pietsch, K., Saravo, B., Baethmann, M., & Schulte-Körne, G. (2012). Is the Children’s Depression Inventory Short Version a valid screening tool in pediatric care? A comparison to its full-length version. Journal of Psychosomatic Research, 73, 369 –374. http://dx.doi.org/10.1016/j.jpsychores.2012.08.016

Aluja, A., & Blanch, A. (2002). The Children Depression Inventory as predictor of social and scholastic competence. European Journal of Psychological Assessment, 18, 259 –274. http://dx.doi.org/10.1027// 1015-5759.18.3.259

American Psychiatric Association. (2000). Diagnostic and statistical man- ual of mental disorders (4th ed.). Washington, DC: Author.

Baglin, J. (2014). Improving your exploratory factor analysis for ordinal data: A demonstration using FACTOR. Practical Assessment, Research & Evaluation, 19, 1–15.

Basto, M., & Pereira, J. M. (2012). An SPSS R-menu for ordinal factor analysis. Journal of Statistical Software, 46, 1–29. http://dx.doi.org/10 .18637/jss.v046.i04

Bauman, S. (2008). The association between gender, age, and accultura- tion, and depression and overt and relational victimization among Mex- ican American elementary students. Journal of Early Adolescence, 28, 528 –554. http://dx.doi.org/10.1177/0272431608317609

Bennett, D. S., Sullivan, M. W., & Lewis, M. (2010). Neglected children, shame-proneness, and depressive symptoms. Child Maltreatment, 15, 305–314. http://dx.doi.org/10.1177/1077559510379634

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1164 AHLEN AND GHADERI

Bienvenu, O. J., & Ginsburg, G. S. (2007). Prevention of anxiety disorders. International Review of Psychiatry, 19, 647– 654. http://dx.doi.org/10 .1080/09540260701797837

Bittner, A., Egger, H. L., Erkanli, A., Costello, E. J., Foley, D. L., & Angold, A. (2007). What do childhood anxiety disorders predict? Jour- nal of Child Psychology and Psychiatry, 48, 1174 –1183. http://dx.doi .org/10.1111/j.1469-7610.2007.01812.x

Brouwer, D., Meijer, R. R., & Zevalkink, J. (2013). On the factor structure of the Beck Depression Inventory-II: G is the key. Psychological As- sessment, 25, 136 –145. http://dx.doi.org/10.1037/a0029228

Brunner, M., Nagy, G., & Wilhelm, O. (2012). A tutorial on hierarchically structured constructs. Journal of Personality, 80, 796 – 846. http://dx.doi .org/10.1111/j.1467-6494.2011.00749.x

Carey, M. P., Faulstich, M. E., Gresham, F. M., Ruggiero, L., & Enyart, P. (1987). Children’s Depression Inventory: Construct and discriminant validity across clinical and nonreferred (control) populations. Journal of Consulting and Clinical Psychology, 55, 755–761. http://dx.doi.org/10 .1037/0022-006X.55.5.755

Chorpita, B. F., Moffitt, C. E., & Gray, J. (2005). Psychometric properties of the Revised Child Anxiety and Depression Scale in a clinical sample. Behaviour Research and Therapy, 43, 309 –322. http://dx.doi.org/10 .1016/j.brat.2004.02.004

Chorpita, B. F., Yim, L., Moffitt, C., Umemoto, L. A., & Francis, S. E. (2000). Assessment of symptoms of DSM–IV anxiety and depression in children: A Revised Child Anxiety and Depression Scale. Behaviour Research and Therapy, 38, 835– 855. http://dx.doi.org/10.1016/S0005- 7967(99)00130-8

Cohen, J. (1987). Statistical power analysis for the behavioral sciences. Hillside, NJ: Erlbaum.

Cole, D. A., & Martin, N. C. (2005). The longitudinal structure of the Children’s Depression Inventory: Testing a latent trait-state model. Psychological Assessment, 17, 144 –155. http://dx.doi.org/10.1037/ 1040-3590.17.2.144

Cortina, J. M. (1993). What is coefficient alpha? An examination of theory and applications. Journal of Applied Psychology, 78, 98 –104. http://dx .doi.org/10.1037/0021-9010.78.1.98

Costello, A. B., & Osborne, J. W. (2005). Best practices in exploratory factor analysis: Four recommendations for getting the most from your analysis. Practical Assessment, Research & Evaluation, 10, 1–9.

Costello, E. J., Erkanli, A., & Angold, A. (2006). Is there an epidemic of child or adolescent depression? Journal of Child Psychology and Psy- chiatry, 47, 1263–1271.

Craighead, W. E., Smucker, M. R., Craighead, L. W., & Ilardi, S. S. (1998). Factor analysis of the Children’s Depression Inventory in a community sample. Psychological Assessment, 10, 156 –165. http://dx .doi.org/10.1037/1040-3590.10.2.156

Davison, A. C., & Hinkley, D. V. (1997). Bootstrap methods and their applications. Cambridge University Press: Cambridge, UK. http://dx.doi .org/10.1017/CBO9780511802843

Doerfler, L. A., Felner, R. D., Rowlison, R. T., Raley, P. A., & Evans, E. (1988). Depression in children and adolescents: A comparative analysis of the utility and construct validity of two assessment measures. Journal of Consulting and Clinical Psychology, 56, 769 –772. http://dx.doi.org/ 10.1037/0022-006X.56.5.769

Ebesutani, C., Reise, S. P., Chorpita, B. F., Ale, C., Regan, J., Young, J., . . . Weisz, J. R. (2012). The Revised Child Anxiety and Depression Scale-Short Version: Scale reduction via exploratory bifactor modeling of the broad anxiety factor. Psychological Assessment, 24, 833– 845. http://dx.doi.org/10.1037/a0027283

Esbjørn, B. H., Sømhovd, M. J., Turnstedt, C., & Reinholdt-Dunne, M. L. (2012). Assessing the Revised Child Anxiety and Depression Scale (RCADS) in a national sample of Danish youth aged 8 –16 years. PLoS ONE, 7, e37339. http://dx.doi.org/10.1371/journal.pone.0037339

Essau, C. A. (2005). Frequency and patterns of mental health services utilization among adolescents with anxiety and depressive disorders. Depression and Anxiety, 22, 130 –137. http://dx.doi.org/10.1002/da .20115

Essau, C. A., Sakano, Y., Ishikawa, S., & Sasagawa, S. (2004). Anxiety symptoms in Japanese and in German children. Behaviour Research and Therapy, 42, 601– 612. http://dx.doi.org/10.1016/S0005-7967(03)00164-5

Essau, C. A., Sasagawa, S., Anastassiou-Hadjicharalambous, X., Guzmán, B. O., & Ollendick, T. H. (2011). Psychometric properties of the Spence Child Anxiety Scale with adolescents from five European countries. Journal of Anxiety Disorders, 25, 19 –27. http://dx.doi.org/10.1016/j .janxdis.2010.07.001

Fox, J., Friendly, M., & Weisberg, S. (2013). Hypothesis tests for multi- variate linear models using the car package. R Journal, 5, 39 –52.

Giannakopoulos, G., Kazantzi, M., Dimitrakaki, C., Tsiantis, J., Kolaitis, G., & Tountas, Y. (2009). Screening for children’s depression symptoms in Greece: The use of the Children’s Depression Inventory in a nation- wide school-based sample. European Child & Adolescent Psychiatry, 18, 485– 492. http://dx.doi.org/10.1007/s00787-009-0005-z

Hanks, C. E., McGuire, J. F., Lewin, A. B., Storch, E. A., & Murphy, T. K. (2016). Clinical correlates and mediators of self-concept in youth with chronic tic disorders. Child Psychiatry and Human Development, 47, 64 –74. http://dx.doi.org/10.1007/s10578-015-0544-0

Houghton, F., Cowley, H., Houghton, S., & Kelleher, K. (2003). The Children’s Depression Inventory short form (CDI:S) in an Irish context. Irish Journal of Psychology, 24, 193–198.

Huang, C., & Dong, N. (2014). Dimensionality of the Children’s Depres- sion Inventory: Meta-analysis of pattern matrices. Journal of Child and Family Studies, 23, 1182–1192. http://dx.doi.org/10.1007/s10826-013- 9779-1

Ivarsson, T., Svalander, P., & Litlere, O. (2006). The Children’s Depres- sion Inventory (CDI) as measure of depression in Swedish adolescents: A normative study. Nordic Journal of Psychiatry, 60, 220 –226. http:// dx.doi.org/10.1080/08039480600636395

Jöreskog, K. G., & Sörbom, D. (1981). LISREL V: Analysis of linear structural relationships by maximum likelihood and least squares meth- ods (Research Report 81– 8). Uppsala, Sweden: Department of Statistics, University of Uppsala.

Kazdin, A. E., Rodgers, A., & Colbus, D. (1986). The Hopelessness Scale for children: Psychometric characteristics and concurrent validity. Jour- nal of Consulting and Clinical Psychology, 54, 241–245. http://dx.doi .org/10.1037/0022-006X.54.2.241

Keenan, K., & Hipwell, A. E. (2005). Preadolescent clues to understanding depression in girls. Clinical Child and Family Psychology Review, 8, 89 –105. http://dx.doi.org/10.1007/s10567-005-4750-3

Keenan, K., Hipwell, A., Feng, X., Babinski, D., Hinze, A., Rischall, M., & Henneberger, A. (2008). Subthreshold symptoms of depression in preadolescent girls are stable and predictive of depressive disorders. Journal of the American Academy of Child & Adolescent Psychiatry, 47, 1433–1442. http://dx.doi.org/10.1097/CHI.0b013e3181886eab

Kessler, R. C., Berglund, P., Demler, O., Jin, R., Merikangas, K. R., & Walters, E. E. (2005). Lifetime prevalence and age-of-onset distributions of DSM–IV disorders in the National Comorbidity Survey Replication. Archives of General Psychiatry, 62, 593– 602. http://dx.doi.org/10.1001/ archpsyc.62.6.593

Kessler, R. C., & Bromet, E. J. (2013). The epidemiology of depression across cultures. Annual Review of Public Health, 34, 119 –138. http:// dx.doi.org/10.1146/annurev-publhealth-031912-114409

Kovacs, M. (1992). Children’s Depression Inventory. North Tonawanda, NY: Multi-Health System.

Kovacs, M. (2003). Children’s Depression Inventory (CDI): Technical manual update. North Tonawanda, NY: Multi-Health Systems.

Kovacs, M. (2011). Children’s Depression Inventory 2nd edition (CDI 2): Technical manual. Toronto, Ontario, Canada: Multi-Health Systems.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1165CHILDREN’S DEPRESSION INVENTORY—SHORT VERSION

Larsson, B., & Melin, L. (1992). Prevalence and short-term stability of depressive symptoms in schoolchildren. Acta Psychiatrica Scandi- navica, 85, 17–22. http://dx.doi.org/10.1111/j.1600-0447.1992 .tb01436.x

Lewinsohn, P. M., Rohde, P., Klein, D. N., & Seeley, J. R. (1999). Natural course of adolescent major depressive disorder: I. Continuity into young adulthood. Journal of the American Academy of Child & Adolescent Psychiatry, 38, 56 – 63. http://dx.doi.org/10.1097/00004583-199901000- 00020

Lorant, V., Deliège, D., Eaton, W., Robert, A., Philippot, P., & Ansseau, M. (2003). Socioeconomic inequalities in depression: A meta-analysis. American Journal of Epidemiology, 157, 98 –112. http://dx.doi.org/10 .1093/aje/kwf182

Mata, J., Thompson, R. J., & Gotlib, I. H. (2010). BDNF genotype moderates the relation between physical activity and depressive symp- toms. Health Psychology, 29, 130 –133. http://dx.doi.org/10.1037/ a0017261

Muris, P., Meesters, C., & Schouten, E. (2002). A brief questionnaire of DSM–IV defined anxiety and depression symptoms among children. Clinical Psychology & Psychotherapy, 9, 430 – 442. http://dx.doi.org/10 .1002/cpp.347

Myers, K., & Winters, N. C. (2002). Ten-year review of rating scales: II. Scales for internalizing disorders. Journal of the American Academy of Child & Adolescent Psychiatry, 41, 634 – 659. http://dx.doi.org/10.1097/ 00004583-200206000-00004

Nolen-Hoeksema, S. (2001). Gender differences in depression. Current Directions in Psychological Science, 10, 173–176. http://dx.doi.org/10 .1111/1467-8721.00142

O’Connor, B. P. (2000). SPSS and SAS programs for determining the number of components using parallel analysis and Velicer’s MAP test. Behavior Research Methods, Instruments & Computers, 32, 396 – 402. http://dx.doi.org/10.3758/BF03200807

Piccinelli, M., & Wilkinson, G. (2000). Gender differences in depression: Critical review. The British Journal of Psychiatry, 177, 486 – 492. http:// dx.doi.org/10.1192/bjp.177.6.486

R Core Team. (2015). R: A language and environment for statistical computing. Available at https://www.R-project.org/

Reise, S. P. (2012). The rediscovery of bifactor measurement models. Multivariate Behavioral Research, 47, 667– 696. http://dx.doi.org/10 .1080/00273171.2012.715555

Reise, S. P., Moore, T. M., & Haviland, M. G. (2010). Bifactor models and rotations: Exploring the extent to which multidimensional data yield univocal scale scores. Journal of Personality Assessment, 92, 544 –559. http://dx.doi.org/10.1080/00223891.2010.496477

Revelle, W. (2015). psych: Procedures for personality and psychological research (R Package Version 1.5.1). Available at http://personality- project.org/r

Samm, A., Värnik, A., Tooding, L. M., Sisask, M., Kölves, K., & von Knorring, A. L. (2008). Children’s Depression Inventory in Estonia: Single items and factor structure by age and gender. European Child & Adolescent Psychiatry, 17, 162–170. http://dx.doi.org/10.1007/s00787- 007-0650-z

Schermelleh-Engel, K., Moosbrugger, H., & Müller, H. (2003). Evaluating the fit of structural equation models: Tests of significance and descrip- tive goodness-of-fit measures. Methods of Psychological Research On- line, 8, 23–74.

Schmid, J., & Leiman, J. M. (1957). The development of hierarchical factor solutions. Psychometrika, 22, 53– 61. http://dx.doi.org/10.1007/ BF02289209

Sibinga, E. M., Webb, L., Ghazarian, S. R., & Ellen, J. M. (2016). School-based mindfulness instruction: An RCT. Pediatrics, 137: e20152532. http://dx.doi.org/10.1542/peds.2015-2532

Smith, G. T., McCarthy, D. M., & Anderson, K. G. (2000). On the sins of short-form development. Psychological Assessment, 12, 102–111. http:// dx.doi.org/10.1037/1040-3590.12.1.102

Spence, S. H. (1997). Structure of anxiety symptoms among children: A confirmatory factor-analytic study. Journal of Abnormal Psychology, 106, 280 –297. http://dx.doi.org/10.1037/0021-843X.106.2.280

Spence, S. H. (1998). A measure of anxiety symptoms among children. Behaviour Research and Therapy, 36, 545–566. http://dx.doi.org/10 .1016/S0005-7967(98)00034-5

Stevanovic, D. (2012). Childhood depression and anxiety disorders in Serbia: A psychometric study of four screening questionnaires. Epide- miology and Psychiatric Sciences, 21, 111–116. http://dx.doi.org/10 .1017/S2045796011000655

Sun, S., & Wang, S. (2015). The Children’s Depression Inventory in worldwide child development research: A reliability generalization study. Journal of Child and Family Studies, 24, 2352–2363. http://dx .doi.org/10.1007/s10826-014-0038-x

The Swedish National Agency for Education. (2015). SIRIS. Retrieved from http://siris.skolverket.se/siris/f?p�Siris:1:0

Thompson, R. J., Mata, J., Jaeggi, S. M., Buschkuehl, M., Jonides, J., & Gotlib, I. H. (2010). Maladaptive coping, adaptive coping, and depres- sive symptoms: Variations across age and depressive state. Behaviour Research and Therapy, 48, 459 – 466. http://dx.doi.org/10.1016/j.brat .2010.01.007

Timbremont, B., Braet, C., & Dreessen, L. (2004). Assessing depression in youth: Relation between the Children’s Depression Inventory and a structured interview. Journal of Clinical Child and Adolescent Psychol- ogy, 33, 149 –157. http://dx.doi.org/10.1207/S15374424JCCP3301_14

Twenge, J. M., & Nolen-Hoeksema, S. (2002). Age, gender, race, socio- economic status, and birth cohort differences on the Children’s Depres- sion Inventory: A meta-analysis. Journal of Abnormal Psychology, 111, 578 –588. http://dx.doi.org/10.1037/0021-843X.111.4.578

Vines, L., & Nixon, R. D. (2009). Positive attributional style, life events and their effect on children’s mood: Prospective study. Australian Journal of Psychology, 61, 211–219. http://dx.doi.org/10.1080/ 00049530802579507

Zwick, W. R., & Velicer, W. F. (1986). Comparison of five rules for determining the number of components to retain. Psychological bulletin, 99, 432.

Received March 4, 2016 Revision received October 12, 2016

Accepted October 17, 2016 �

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1166 AHLEN AND GHADERI