Journal of Affective Disorders 164 (2014) 19–27

Contents lists available at ScienceDirect

Journal of Affective Disorders journal homepage: www.elsevier.com/locate/jad

Research report

Obsessive–compulsive symptoms in a normative Chinese sample of youth: Prevalence, symptom dimensions, and factor structure of the Leyton Obsessional Inventory—Child Version Jing Sun a,b,n, Mark J. Boschen a,c, Lara J. Farrell a,c, Nicholas Buys a, Zhan-Jiang Li d,nn a

Griffith Health Institute, Griffith University, Parkland, Gold Coast 4222, QLD, Australia School of Medicine, Griffith University, Parkland, Gold Coast 4222, QLD, Australia c School of Applied Psychology and Griffith Health Institute, Griffith University, Gold Coast, QLD, Australia d Department of Clinical Psychology/Beijing Anding Hospital, Capital Medical University, Beijing 100085, China b

art ic l e i nf o

a b s t r a c t

Article history: Received 11 November 2013 Received in revised form 31 March 2014 Accepted 2 April 2014 Available online 15 April 2014

Background: Chinese adolescents face life stresses from multiple sources, with higher levels of stress predictive of adolescent mental health outcomes, including in the area of obsessive–compulsive disorders (OCD). Valid assessment of OCD among this age group is therefore a critical need in China. This study aims to standardise the Chinese version of the Leyton short version scale for adolescents of secondary schools in order to assess this condition. Methods: Stratified randomly selected adolescents were selected from four high schools located in Beijing, China. The Chinese version of the Leyton scale was administered to 3221 secondary school students aged between 12 and 18 years. A high response rate was achieved, with 3185 adolescents responding to the survey (98.5 percent). Exploratory factor analysis (EFA) extracted four factors from the scale: compulsive thoughts, concerns of cleanliness, lucky number, repetitiveness and repeated checking. The four-factor structures were confirmed using Confirmatory Factor Analysis (CFA). Results: Overall the four-factor structure had a good model fit and high levels of reliability for each individual dimension and reasonable content validity. Invariance analyses in unconstrained, factor loading, and error variance models demonstrated that the Leyton scale is invariant in relation to the presence or absence OCD, age and gender. Discriminant validity analysis demonstrated that the fourfactor structure scale also had excellent ability to differentiate between OCD and non-OCD students, male and female students, and age groups. Limitations: The dataset was a non-clinical sample of high school students, rather than a sample of individuals with OCD. Future research may examine symptom structure in clinical populations to assess whether this structure fits into both clinical and community population. Conclusions: The structure derived from the Leyton short version scale in a non-clinical secondary school sample of adolescents, suggests that a four-factor solution can be utilised as a screening tool to assess adolescents' psychopathological symptoms in the area of OCD in mainland Chinese non-clinical secondary school students. & 2014 Elsevier B.V. All rights reserved.

Keywords: OCD Obsessive–compulsive Assessment Chinese Youth LOI-CV.

1. Introduction Paediatric obsessive–compulsive disorder (OCD) is a debilitating neurobehavioral anxiety disorder affecting between 1 to

n

Corresponding author. Tel.: þ 61733821198; fax: þ617 55527889. Corresponding author. Tel.: þ 86 1370121 8860. E-mail addresses: j.sun@griffith.edu.au (J. Sun), m.boschen@griffith.edu.au (M.J. Boschen), l.farrell@griffith.edu.au (L.J. Farrell), n.buys@griffith.edu.au (N. Buys), [email protected] (Z.-J. Li). nn

http://dx.doi.org/10.1016/j.jad.2014.04.004 0165-0327/& 2014 Elsevier B.V. All rights reserved.

4 percent of children and youth (Douglass et al., 1995; Shaffer et al., 1996; Valleni-Basile et al., 1994; Zohar, 1999). During childhood, this condition is frequently associated with significant impairment and disruption to family relationships (Barrett et al., 2001; Cooper, 1996), school performance and peer relationships (Allsopp and Verduyn, 1990). In Chinese samples of adults with OCD, the clinical phenomenology is largely consistent with that described in Western samples (Chia, 1981; Lo, 1967), with the exception that there has been differences noted in the gender ratio, with more men experiencing OCD in Chinese samples relative to Western samples whereby the gender distribution in

20

J. Sun et al. / Journal of Affective Disorders 164 (2014) 19–27

adulthood is generally equal (although in childhood there is a male predominance) (Swedo et al., 1989). Despite the profoundly negative impact OCD has on peoples' lives, there is an alarming delay of up to 17 years from the age of onset of OCD in childhood, to the provision of adequate treatment (Hollander et al., 1996). Given the chronicity of this disorder and the knowledge that OCD frequently leads to lifelong suffering if left untreated (Amir et al., 2000), there is a pressing need to improve early detection of this disorder, and increase access to evidence based treatment. There are likely to be numerous reasons for the delay in identification and treatment of childhood OCD, including the highly secretive nature of the disorder (Barrett and Healy, 2003), the lack of insight in children, and the high rates of comorbidity associated with this disorder during childhood and adolescence (Farrell et al., 2012). Comorbidity is the norm rather than the exception in samples of youth with OCD, with up to 80 percent of children affected having at least one comorbid diagnosis (Geller et al., 1996; Lewin and Piacentini, 2010; Storch et al., 2009; Swedo et al., 1989), and as many as 50 to 60 percent of youth experiencing two or more other mental disorders during their lifetime (Rasmussen and Eisen, 1990). Some of the most commonly co-occurring psychiatric conditions associated with paediatric OCD include other anxiety disorders (affecting 26–70 percent of children with OCD), depression (10–73 percent), tics and Tourette's Syndrome (17–59 percent), attention deficit disorders (10–50 percent), and disruptive behavioural disorders (10–57 percent) (Flament et al., 1990; Geller et al., 1996, 2001a, 2001b; Swedo et al., 1993). The presence of these comorbid disorders not only adds to the child's level of overall impairment and dysfunction, but may also lead to the misdiagnosis or absence of a diagnosis of OCD. In fact, one study has suggested that almost half of all OCD patients are misdiagnosed with a depressive or anxiety disorder, while their OCD is overlooked (Hollander et al., 1996). Given that OCD is among the most common of all psychiatric disorders (Hollander et al., 1996), yet is frequently overlooked by health care specialists, validated screening tools for OCD during childhood and youth are sorely needed. Moreover, assessment tools that are not costly, time consuming or require trained interviewers are necessary for use in educational settings and primary health care to increase the screening and early detection of OCD. There are currently limited screening tools for childhood OCD that do not require clinically trained interviewers. The 20 item self-report version of the Leyton Obsessional Inventory (LOI-CV) (Berg et al., 1988) is one popular self-report assessment of childhood OCD, which is relatively brief and has demonstrated good psychometric properties (Berg et al., 1988). The LOI-CV was derived from the Leyton Obsessional Inventory Short Form (Cooper, 1970) and includes 20 items for use in children and adolescents up to 18 years of age. Studies have found the LOI-CV to have good reliability (Cronbach's α¼ .91), sensitivity (75 percent) and specificity (84 percent), although it has a relatively low predictive value for OCD (18 percent) (Berg et al., 1988; Flament et al., 1988). In regards to the factor structure of the LOI-CV, findings across studies are mixed, with evidence to support both a multidimensional view of OCD with a three-factor model, as well as a homogenous approach with a single factor structure (Moore et al., 2010). Bamber et al. (2002) identified a three factor structure in a sample of 253 British youth, including obsessions/incompleteness, cleanliness, and compulsions, which accounted for 48 percent of the overall variance. Similarly, Berg et al. (1988) found a threefactor structure with 4551 youth in their original principal components analysis (PCA), and described their factors as—general obsessive, dirt-contamination, and numbers-luck. These two

studies were largely consistent with item loadings apart from one item (item 1), which loaded on the general obsessive factor in Bamber et al. (2002), whereas in Berg et al. (1988), item 1 loaded on the compulsions factor. A confirmatory factor analysis study has recently been conducted by Moore et al. (2010) with 517 adolescent twins, which also found support for Bamber et al. (2002) three-factor model. In contrast, Rueda-Jaimes et al. (2007) reported a PCA in 581 Columbian youth and found a one-factor solution, explaining 75 percent of variance. Therefore, there remains some inconsistency in the findings to date regarding the factor structure of the LOI-CV. To date, there are no research studies which have examined the utility of an OCD self-report measure in a sample of Chinese youth of secondary school age, despite the fact that Western studies have found that this is the common age of onset (Fogel, 2003). The scarcity of research with this population may be due to the lack of a validated assessment tool. This study examines the utility of the self-report version of the LOI-CV (Berg et al., 1988) in a large sample of Chinese youth. The aims of this study were: (a) to explore the level of obsessive–compulsive (OC) symptoms in a normative sample of Chinese youth; (b) to examine the level of various self-reported OC symptom dimensions (i.e., checking, washing, counting), across age and gender groups; (c) to explore the psychometric properties of the LOI-CV (i.e., internal consistency); and (d) to examine the factor structure of the LOI-CV.

2. Method There were two stages of the study. First stage of the study was a cross sectional epidemiological study investigating the structure, reliability and validity of the LOI-CV scale in the self-reported responses of Chinese secondary school adolescents. In the second stage of the study, a clinical interview using DSM-IV Axis I disorders, patient edition (SCID-I/P, Version 2.0) (American Psychiatric Association, 1994) to diagnose adolescents who had score of 15 or more in using LOI-CV was conducted to diagnose OCD symptoms at clinical level. 2.1. Participants A mixture of stratified cluster, random sampling and stratification was used to identify adolescents from four secondary schools from two of Beijing's administrative districts. These districts were chosen as representative of schools with different academic ranking. Prior to data collection, the researchers [Delete for peer review, Chinese Academy of Sciences Institute of Psychology] met with all school principals and project officers to explain the significance of the investigation and discuss the research strategy. They, in turn, informed the students and their parents about the study and consents were obtained from both parents and students. The research project has been approved by Research Board of the Research Institute [Delete for peer review, Institute of Psychology, Chinese Academy of Sciences]. Within two districts, four schools which expressed interests in the study were chosen. Among four schools, 72 classes were all invited to participate in the study. The total sample comprised 3221 eligible secondary school students, aged between 12 and 18 years of old within four schools. 2.2. Materials and procedure The Chinese translation of the LOI-CV was used. This version was translated and back-translated by academic staff at the research institute [Delete for peer review, Institute of Psychology, Chinese Academy of Sciences]. The LOI-CV includes 20 items, each of which is scored on a dichotomous scale (1 ¼yes, 0 ¼no). The yes

J. Sun et al. / Journal of Affective Disorders 164 (2014) 19–27

answers were further assessed using Likert scale based on interference level (0 is no interference, and 1 to 3 indicate increasing level of interference). A total LOI-CV yes or interference score is calculated by summing all individual item scores. A high score represents a higher level of OCD symptoms. A clinical cut-off ‘yes' score of 15 or more or ‘interference' score of 25 or more are typically used, in which yes scores of less than 15 or interference score of less than 25 are categorised as non-clinical, with yes scores of 15 or above or interference scores of 25 or above representing a clinical level of OCD symptoms, at least two standard deviation above the normal population mean (Flament et al., 1988). All surveys were sent to secondary schools and distributed to students in grades 7 to 12, who were asked to fill in the survey in the classroom. Of the 3221 eligible participants, 3185 responded, yielding a response rate of 98.9 percent. A questionnaire was considered invalid if answers were missing for more than 50 percent of questions. Surveys with complete student data were analysed. Ninety-nine (3.1 percent) cases were found to have 50 percent data incomplete and therefore excluded from further analysis. Other missing items across variables were imputed using the multiple imputation method with five imputations to assimilate the missing data to the closest value in the dataset. 2.3. Statistical analysis The dataset for 3185 students were split into two sub-samples. For the first subsample, data analysis was conducted using SPSS version 18.0 for exploratory factor analysis (EFA) to identify the latent factor structure of the LOI-CV. For the second subsample, Mplus version 6.0 was used for confirmatory factor analysis (CFA) to confirm the factor structure extracted from the first subsample. Statistical analyses were conducted on weighted data. A maximum likelihood analysis was conducted to investigate the subscale structure of the scales. Direct oblimin rotation was considered the appropriate method that minimizes the number of variables that have high loadings on each factor for the current study. EFA was used to extract factor structure of the scale. Reliability was assessed from analyses of internal consistency using Cronbach’s α. Structural equation modelling was used to examine the replicability of the factor structure extracted from the first subsample, using model fit and residuals indices. We report several fit statistics to facilitate evaluation of how well the hypothesized model is replicated in the second subsample. The sample size used in these analyses precludes reliance on the traditional χ2 discrepancy test as a sole measure of model fit because the large sample size increases the likelihood that the χ2 test will be significant, even if the hypothesized model represents just a small deviation from the sample data (Hu and Bentler, 1995). Accordingly, additional fit measures are presented. First, the relative χ2 index should be close to 1, with values below 3 considered indicative of a close fit between the hypothetical model and the sample data. Indices of fit (e.g., Comparative Fit Index, TFI), vary between 0 and 1; values greater than .90 are indicative of a good fit. RMSEA compares the model optimal parameter values with the population covariance matrix. Values less than .05 indicate good fit, and values between .05 and .08 indicate reasonable fit (Hu and Bentler, 1995). A multi-step analysis of invariance (Bollen, 1989) was employed to examine whether the structure of the LOI-CV scale is similar across groups relating the presence or absence of OCD, age and gender. Four multi-step models were used. Model A, the ‘unconstrained’ model, assessed whether the structure of the scale (the number of factors and the items per factor) was acceptable for all groups. It is the less restrictive model and implies that the measurement model holds across OCD, gender and age groups. Model B, the ‘constrained’ model assessed if factor loadings of the

21

items were equal across groups to evaluate the extent to which items are answered similarly by different groups without bias. In Model C, the error variance was constrained to be equal across groups. It tested whether error variance of the variables were equal across groups, while the factor loadings were still constrained. In Model D the factor variances were kept equal across groups, while the factor loadings and error variances were still constrained. Since the Chi-square difference test is highly sensitive to sample size, fit indices are recommended to assess whether the fits of the unrestricted and restrictive models are acceptable. These fit indices are the CFI, TLI, SRMR, RMSEA. CFI and TLI values should be more than 0.90, and RMSEA and SRMR values less than 0.08 (Hu and Bentler, 1995). Differences between the OCD group and non-OCD group, between males and females, and between age groups on the four factors (compulsive thoughts, contamination and cleanliness, number counting, repeating and checking, and total scores) were compared using the Univariate Analysis of Variance (ANOVA) test when the variables were normally distributed. Log transform methods were used to ensure the normality of the data for any of the four factors when they were not normally distributed.

3. Results 3.1. Participant demographics The mean age for the 3185 participants was 14.68 years (SD ¼1.75). Among the respondents, 1664 (52.4 percent) were males, 1510 (47.6 percent) were females. A total of 546 (17.2 percent) students were in Year 7, 561 (17.6 percent) were in Year 8, 558 (17.5 percent) were in Year 9, 548 (17.2 percent) were in Year 10, 522 (16.4 percent) were in Year 11, and 448 (14.1 percent) were in Year 12. There were 2957 (93.1) students who were Han Chinese and 218 (6.9 percent) were other minority nationalities. The mean time to complete the questionnaire was 10 min. 3.2. Prevalence of OCD and interference score distribution in gender group Using cut off score of either total yes score of 15 (Flament et al., 1988), 13.6 percent (n¼ 434) of adolescents had OCD symptoms. Fifteen percent of male adolescents (n ¼250) and 12.2 percent of female adolescents (n ¼ 184) had OCD symptoms. In the second stage of the study, a clinical interview using the DSM-IV was used to further diagnose adolescent who had LOI-CV score of 15. Of these 288 adolescents had clinical symptoms of OCD, representing a 9.04 percent prevalence of OCD in the study population. Based on this result, ROC sensitivity and specificity analysis further confirms the 66 percent of the screening ability of LOI-CV. 3.3. Equivalence of responders and non-responders Individuals were excluded when they failed to complete 4 50 percent of the questionnaire items. There was no significant difference in age groups (χ2 ¼2.79, p ¼ 0.95) or gender (χ2 ¼0.72, p¼ 0.40) between the respondents and non-respondents. 3.4. Exploratory factor analysis The results of the maximum likelihood analysis with direct oblimin rotation for the secondary school students using interference score, LOI-CV are detailed in Tables 1 and 2. A total of four factors were returned with eigenvalues greater than one. The Scree test (Cattell, 1966) also suggested the existence of four latent factors.

22

J. Sun et al. / Journal of Affective Disorders 164 (2014) 19–27

Table 1 Results of exploratory factor analysis for Leyton Obsessional Inventory—Child Version. Item

Factor I

II

III

IV

2. Thoughts or words kept going over and over in my mind even though I didn’t want them to 0.454 16. I often felt guilty because I had done something even though no one else thought it was bad 0.374 0.349 1. I felt I had to do certain things even though I knew I didn’t really have to (like always having to count the steps as I went up them). 0.366 I felt something bad would happen if I didn’t 15. I had a special number that I liked to count up to or I had to do things just that number of times 0.362 13. I had to count in a special way several times or go through numbers in my mind 0.332  0.302 17. I worried a lot if I did something not exactly the way I liked 0.331 12. I had to do things over and over again before they seemed quite right 0.300 6. It was hard for me to make up my mind 0.586 5. I felt that if someone used or touched something it was spoilt for me 0.419 7. I worried about being clean enough 0.410 8. I was fussy about keeping my hands clean 0.387 4. I hated dirt and dirty things 0.378 10. I got angry if other people messed up my things at school 0.358  0.300 9. When I put things away at night they had to be put away just right (i.e., in a special order or a special way) 0.322 20. I had special numbers or words that I said because I hoped they kept bad luck or bad things away  0.713 19. I moved or talked in a special way to avoid bad luck  0.561 14. I had trouble finishing my schoolwork or other jobs because I had to do something over and over again  0.392 18. I kept on thinking about things that I had done because I wasn’t sure that they were the right things to do 0.352  0.566 11. I spent a lot of extra time checking my homework to make sure it was just right  0.482 3. I had to check things several times (e.g., that switches were turned off or windows closed)  0.315 Cronbach alpha 0.68 0.71 0.72 0.52 Note. Factor loadings o 0.3 are not displayed.

Table 2 Model fit indices from confirmatory factor analysis.

20-Item models Current study Berg et al. (1988) Rueda-Jaimes et al. (2007) 11-Item models Bamber et al. (2002) Moore et al. (2010)

Items

Factors

χ2/df

CFI

TFI

RMSEA

20 20 20

4 4 1

1.64 1.87 4.31

0.97 0.92 0.69

0.95 0.91 0.65

0.02 0.02 0.05

11 11

3 3

5.11 2.02

0.81 0.95

0.75 0.94

0.05 0.02

For the four factors extracted, Items 1, 2, 12, 13, 15, 16, and 17 were loaded onto an obsessive thoughts factor; items 4, 5, 6, 7, 8, 9, and 10 loaded on a cleanliness and tidiness factor; Items 14, 19 and 20 loaded on a lucky number dimension; and Items 3, 11, and 18 loaded on a repetitiveness and repeated checking factor. The variance explained by each of the four factors was 23.35 percent, 8.13 percent, 6.31 percent, and 5.54 percent respectively, yielding a total variance explained for the four factors of 43.34 percent. 3.5. Reliability (internal consistency) The reliability level for the total scale based on interference score was at reasonable level, with a total Cronbach’s α of .82, and 0.68, 0.71, 0.72, 0.52 respectively for the four factors: obsessive thoughts, cleanliness and tidiness, lucky number; and repetitiveness and repeated checking. 3.6. Confirmatory factor analysis Confirmatory Factor Analysis (CFA) was used to examine the model fit for the four factor structure. The four- factor model derived in the exploratory factor analysis (see Table 2, Fig. 1) was examined for model fit using the Comparative Fit Index (CFI) (Bentler, 1980), Tucker Fit Index (TFI) (Hu and Bentler, 1995), and Root Mean Squared Error of Approximation (RMSEA) (Hu and Bentler, 1995). The factor structure of the LOI-CV was tested using the Mplus restrictive factor analysis approach with maximum

likelihood estimations. The goodness-of-fit indicators are summarized in Table 3. The model derived in this study was compared against other models previously proposed by Berg et al. (1988), Bamber et al. (2002), Rueda-Jaimes et al. (2007) and Moore et al. (2010). It was found that the four-factor model represented a good fit to the data in the second subsample (CFI¼ 0.97), and showed similar or superior fit to the four other previously reported 11 item (Bamber et al., 2002; Moore et al., 2010) and 20 item (Berg et al., 1988; Rueda-Jaimes et al., 2007) factor structures. Additional analyses showed that the items converged relevantly and significantly on the respective factors hypothesized in this model. On the basis of model fit results, the four factor structure was determined to be a superior fit to the data in subsample 2, compared to the structure proposed previously by us. The four factor structure was found to be a parsimonious model in terms of Chi Square, CFI, and RMSEA results. The fit indices results of the multi-step invariance analysis (see Table 3) across the OCD and non-OCD groups demonstrate that the configural model (Model A), factor loading model (Model B), and factor error variance (Model C) supported adequate to excellent fit to the data. This suggests an equal form or identical factor structure for the LOI-CV scale in both OCD and non-OCD groups. The test for factor variances (Model 4) resulted in a statistically significant decrement in model fit, in which all model fit indices are at an unacceptable level. The estimate of the variances of factor 1 (1.45), factor 2 (1.57), factor 3 (1.48) and factor 4 (0.96) in the OCD group are higher than among the non-OCD group participants in the estimates of the variances of the four factors, which are 0.12, 0.10, 0.16, and 0.46 respectively. Invariance across gender (see Table 3) was acceptable in the configural model (Model A), factor loading model (Model B), and factor invariance model (Model C). This suggests an equal form or identical factor structure for the LOI-CV scale in both the male and female groups. In the factor variance model (Model D), the measurement errors with RMSEA were increased 0.097 which was at an unacceptable level, suggesting the model does not have a good fit when more constraints were added to the model. Results of the multi-step invariance analysis across age groups (see Table 3) found the configural model (Model A), factor loading

J. Sun et al. / Journal of Affective Disorders 164 (2014) 19–27

23 err err err err err err err

err err err err err err err

err err err

err err err

Fig. 1. Confirmatory factor analysis item loadings on each of the four latent factors.

Table 3 Analysis of invariance of the Leyton Obsession Inventory across OCD group, gender and age groups. χ2

df

p

CFI

TFI

RMSEA

SRMR

Δχ2

Invariant by OCD group Model A 505 Model B 511 Model C 1028 Model D 1672

356 344 364 373

0.16 0.09 o 0.001

0.91 0.97 0.89 0.79

0.90 0.97 0.88 0.78

0.055 0.04 0.058 0.124

0.05 0.05 0.06 2.19

1.41 1.48 2.82 4.48

Invariant by gender Model A Model B Model C Model D

508 511 1231 3288

344 343 360 392

0.51 0.11 o 0.001

0.97 0.97 0.97 0.89

0.96 0.96 0.96 0.89

0.04 0.05 0.05 0.097

0.03 0.03 0.03 0.09

1.47 1.48 3.41 8.38

Invariant by age Model A Model B Model C Model D

521 529 1240 4376

371 346 368 1176

0.19 0.08 o 0.001

0.95 0.95 0.95 0.87

0.95 0.95 0.95 0.88

0.06 0.06 0.06 0.10

0.05 0.04 0.05 0.08

1.40 1.53 3.36 3.72

Δχ2 is adjusted chi-square derived from χ2/df. Model A: is configure model that the equal factor loading across groups are constrained. Model B is factor loading model and assumed equal factor loading across groups; Model C is factor loading and error variance were constrained; Model D is factor loading, error variance, and factor variance were constrained.

(Model B) and error variance (Model C) model were acceptable as indicated by the good to excellent model fit indices. In factor variance model (Model 4), the model fit indices of CFI and TFI were

reduced from 0.95 and 0.95 to 0.87 and 0.88, the measurement error of RMSEA increased from 0.065 to 0.10, and SRMR increased from 0.05 to 0.08. This suggests the unequal form or factor

24

J. Sun et al. / Journal of Affective Disorders 164 (2014) 19–27

Table 4 Comparison between OCD and non-OCD groups in four factors and total score. OCD group (N ¼ 223)

Factor I: Compulsive thoughts Factor II: Concerns of cleanliness Factor III Lucky number Factor IV: Repeating and checking Total Score

Non-OCD group (N ¼ 1369)

M

SD

M

SD

8.85 8.52 2.50 2.87 38.54

3.65 3.85 2.65 2.17 31.29

3.12 3.07 0.38 0.65 7.74

2.60 2.78 0.97 1.05 6.23

F

P

1314.87 1074.79 21.53 21.02 2131.54

o 0.001 o 0.001 o 0.001 o 0.001 o 0.001

F value indicates ANOVA test as parametric test was used when data were normally distributed. Factor 3 number, Factor 4 checking, and total score are not normally distributed from in OCD and non-OCD groups, the data were log transformed before the ANOVA analysis was conducted.

Table 5 Comparison between male and female students (based on interference score) for Leyton Obsession Inventory. Male (N ¼837)

Factor I: Compulsive thoughts Factor II: Concerns of cleanliness Factor III Lucky number Factor IV: Repeating and checking Total Score

Female (N¼ 747)

M

SD

M

SD

3.90 3.81 0.72 1.00 12.55

3.33 3.44 1.53 1.50 17.07

3.63 3.78 0.53 0.78 11.33

3.22 1.71 1.31 1.28 16.28

F

P

F ¼5.28 F ¼4.51 Z ¼3.03 Z ¼4.05 Z ¼3.44

0.02 0.03 0.002 o 0.001 0.001

F value indicates ANOVA test as parametric test was used when data were normally distributed. Factor 3 number, Factor 4 checking, and total score are not normally distributed from in both male and female groups, the data were log transformed before the ANOVA analysis was conducted.

Table 6 Comparison between male and female students (based on interference score) for Leyton Obsession Inventory. LOV-CV

13 Years (n ¼299)

Factor I: Compulsion Factor II: Cleanliness Factor III: Number Factor IV: Checking Total scores

3.51 4.74 0.55 1.05 15.58

(3.09)a (3.71)a (1.36) (1.56)a (22.44)a

14 Years (n ¼284) 3.40 3.92 0.60 0.82 12.16

(3.20)a (3.23)b (1.36) (1.34) (17.28)b

15 years (n¼ 258) 3.74 4.08 0.56 0.81 11.95

(3.25)a (3.62)c (1.29) (1.31) (15.96)b

16 years (n ¼277) 3.85 3.25 0.74 0.97 10.28

(3.32)a (3.28)d (1.58) (1.55) (12.70)b

17 years (n¼249) 3.67 2.77 0.53 0.76 9.30

(3.09)a (2.74)d (1.34) (1.25)b (12.37)b

18 years (n¼ 223) 4.59 3.26 0.78 0.97 12.38

(3.79)b (3.28)d (1.68) (1.40) (16.55)b

F 7.60nnn 23.46nnn 2.76n 3.35nn 9.10nnn

Factor 3 number, Factor 4 checking, and total score are not normally distributed from 13 years old to 18 years old group, the data were log transformed before the ANOVA analysis was conducted. a to d indicate the significant post-hoc comparisons between age groups. n

po 0.05. p o0.01. nnn p o0.001. nn

structure for the LOI-CV scale across age groups when different age groups require different content in the scale. In summary, three of the four multi-step invariance tests exhibited acceptable configural, factor loading and factor error variance across OCD vs non-OCD, and age and gender groups. The fourth model (factor variance test) in relation to OCD vs non-OCD, gender and age groups exhibited an unacceptable level of model fits. 3.7. Discriminant validity of LOI-CV test Table 4 shows that there were significant differences in four factors and total interference scores between OCD and non-OCD group and between male and female group (see Table 5). There were also significant differences in four factors and total scores between age groups (see Table 6). 4. Discussion This study aimed to validate a short Leyton questionnaire for symptoms of OCD for use with adolescents in China, with

the results identifying four factors: (1) compulsive thoughts; (2) cleanliness and tidiness; (3) lucky number and; (4) repeating and checking. Using an interference score (0 to 3 answers) in the exploratory factor analysis, the four factors explained 23.35 percent, 8.13 percent, 6.31 percent, and 5.54 percent of the variance respectively, yielding a total variance of 43.34 percent. This level of variance is similar to findings of previous research (Berg et al., 1988; Sans et al., 2012), suggesting that factors specified in LOI-CV are adequate to explain the dimensions of OCD in the Chinese adolescent population. It should also be noted that the reliability level for the total scale was acceptable with a Cronbach’s α of 0.82 for the overall scale, and 0.68, 0.71, 0.72, and 0.52, respectively, for the four factors. This is consistent with previous clinical and epidemiological studies using this scale. Using a high cut off yes score of 15 or more (Flament et al., 1988), 13.6 percent of Chinese adolescents demonstrated OCD symptoms. Additional clinical interviews, using the DSM-IV, confirmed that 66 percent of this group of adolescents had clinical symptoms of OCD. The clinical interview results confirmed the good screening ability of the LOI-CI instrument to identify the condition. Based on these results, 9.0 percent of the total cohort population had OCD symptoms, a much higher prevalence rate

J. Sun et al. / Journal of Affective Disorders 164 (2014) 19–27

than the 1.9 percent found in the American population (Flament et al., 1988), 0.25 percent of England adolescents (Heyman et al., 2003), 8.0 percent of Spanish adolescents (Sans et al., 2012), but lower than the 11.8 percent of a Colombian non-clinical sample (Rueda-Jaimes et al., 2007). The high percentage of clinically diagnosed OCD in the Chinese sample may be explained by the highly competitive environment and study pressure imposed on Chinese adolescents, factors that have been shown to contribute to the onset of the condition (Real et al., 2011). Over 90 percent of participants were single children reared in a one-child family environment, increasing the pressure on them as the ‘only child’ to perform well at school (Liu et al., 1999). The cultural and family expectations of male students may also explain the gender difference whereby males have higher scores on the four factors. Among the four factors extracted from the exploratory factor analysis in the current study, the first factor was characterized by compulsive thoughts and behaviour, and accounted for the most variance (23.35 percent). Compulsive thoughts (e.g., thoughts or words kept going over and over in mind) and behaviour (doing things again as again and counts number several times) have consistently been identified as the area of OCD symptoms that have been associated with early onset of OCD in adolescents (Bamber et al., 2002; Berg et al., 1988). In both the Berg et al. (1988) and Bamber et al. (2002) studies, compulsion was found to be the most important factor explaining OCD symptoms in an American adolescent sample. Three items, “I moved or talked in a special way to avoid bad luck”, and “I had special numbers or words that I said because I hoped they kept bad luck or bad things away” and “I had to count in a special way several times or go through numbers in my mind”, were not loaded on the compulsion factor in the Chinese sample, a finding that differed from the either American (Bamber et al., 2002) or Spanish sample (Sans et al., 2012). These differences may be because the lucky number and number counting is represented as a separate factor in the Chinese sample. The second factor, described as contamination and cleanliness, accounted for 8.13 percent of the symptom variance in the Chinese sample. Contamination and cleaning symptoms load together on one factor, a finding that is consistent with previous studies (Bamber et al., 2002; Berg et al., 1988). Seven items relating to contamination concerns and somatic obsession (due to fear of illness resulting from contamination) are consistent with factor loadings in previous studies. These studies have reported dirt phobia and contamination concerns as the most frequent obsessive–compulsive phenomena in children and adolescents (Berg et al., 1988; Sans et al., 2012). This was also reflected in current Chinese adolescent sample and reflects the stability of this factor across cultures. The third factor accounted for 6.31 percent of the symptom variance and was characterized by superstition/mental compulsion in lucky number and number counting. Superstitious behaviour is often reported in previous studies as an important aspect of OCD symptoms in adolescents (Berg et al., 1988; Stewart et al., 2007). This item loading is consistent with some previously published studies, but not others (Bamber et al., 2002; Sans et al., 2012). The differences may be due to chance fluctuation of the symptom categories across age groups e.g. low frequency in the younger age group and more frequent occurring in older age group in the Chinese sample. Future work is needed to examine whether there is systematic change in this factor across the lifespan to examine whether this factor can be understood as a stable characteristic of adolescents with OCD. The fourth factor accounted for 5.54 percent of the total variance and was characterised as repeating and checking behaviour. Repeating and checking symptoms load together on one

25

factor, a finding that is consistent with previous studies (Berg et al., 1988; Rueda-Jaimes et al., 2007). In Bamber et al. (2002) study, repeating and checking is loaded on the compulsive thought factor, and in Sans et al. (2012) study, repeating and checking is loaded across different factors. This symptom may occur in the context of different types of obsessions, or may reflect its diversity across thinking concerns and behaviour manifestations. In our study of adolescents, repeating and checking tends to occur in the context of repeating symptoms and ordering of doing things and arranging objects. This could relate to a more developmentally specific response to concerns regarding ordering in Chinese adolescents. The confirmatory and exploratory factor analyses confirm offers support for the multi-dimensional nature of the scale, consistent with Moore et al. (2010) study. As a result, the present findings imply that the use of an overall score is justifiable of one big concept of overall OCD status because the different dimensions of the Leyton scale define a single higher-order construct that may represent the concept of overall OCD status. An examination of Table 2 showed that the compulsive and obsessive factors are separate factors. While this finding is consistent with the data reported by Moore et al. (2010), Bamber et al. (2002) and Sans et al. (2012), the factor loadings on each dimension varied across studies. Although the observed discrepancies across studies might be attributable to subject and cultural differences, re-analyses of the data in the current study in terms of both EFA and CFA confirm the new structure of the instrument for Chinese adolescents so that more definitive conclusions can be drawn. The design of the study permitted a test of measurement invariance in relation to the variables of OCD, gender and age. The analysis supported configural invariance, factor loading invariance, factor error variance invariance, and factor variance in the OCD, gender and age groups. This supports previous studies, which found that the LOI-CV has good internal consistency (Cronbach’s alpha ¼ 0.81), particularly in a large non-clinical population (Berg et al., 1988; Flament et al., 1988). Notably, there was a lack of good model fit in factor variance model in the OCD group invariance analysis. This may be because the symptoms become more complex and severe in OCD group of adolescents which requires the use of a more sensitive and clinical tool in explaining the OCD symptoms in OCD group. In the non-OCD group, the variances explained by the four factors of the LOI-CV were significantly lower than those in the OCD group. This suggests a low content validity of the LOI-CV in the non-OCD group when non-OCD adolescents do not have OCD symptoms. Our results suggest that the content of LOI-CV is inadequate to diagnose OCD symptoms when symptoms are severe and complex. This supports Storch et al. (2011) contention that the LOI-CV has inadequate psychometric properties in clinical populations. For example, he found that the LOI-CV had low internal consistency level across four factors (general obsessive (Cronbach alpha ¼0.53), dirt-contamination (Cronbach¼ 0.49), numbersluck (Cronbach¼0.66), and school (Cronbach¼ 0.56)) in children with OCD. The sensitivity of the LOI-CV was poor, in that only 6 percent of the clinical OCD adolescents could be diagnosed by the LOI-CV. In the factor variance model (Model 4) of the invariance test in gender group, the variance explained by each item and four factors of the LOI-CV were similar across male and female groups. The increased measurement errors in the factor variance test in the gender group may be due to the large sample effect and additional constraint added to the analysis, a finding that has been previously reported using an invariance test (Krank et al., 2011). In the factor variance model (Model 4) in relation to the age variable, the variance explained by factor 3 (number counting) and factor 4 (repeating and checking behaviour) in the 17–18 age group was

26

J. Sun et al. / Journal of Affective Disorders 164 (2014) 19–27

significantly less than other age groups. The increased measurement errors as indicated by the RMSEA (0.10 and SRMR (0.08) may be due to the different responses from different age groups. In the 14–16 age group, the content of the LOI-CV is appropriate. However, it may require more age appropriate items for the 17–18 age group, a time when adolescents’ cognitive and social abilities are changing. This result supports the findings of King et al. (1995), whereby the test–retest reliability of the LOI-CV was found to be acceptable up to the 14–16 age group. Due to the scarcity of published literature in this area, our study was the first to report that the LOI-CV is inadequate to assess adolescents aged 17–18 years with OCD and non-OCD symptoms. The present data demonstrate the importance of using confirmatory factor analysis in the examination and confirmation of the structure of the Leyton scale. This is important because there have been few studies that have used the confirmatory factor analysis approach and higher-order factor analyses in uncovering the structure of the Leyton scale in mainland Chinese adolescents in secondary schools. The invariance relating to the OCD, gender and age groups suggests that the LOI-CV scale has good application in measuring OCD in a Chinese non-clinical adolescent sample. 4.1. Limitations There are a number of limitations of this study. First, the dataset was a non-clinical sample of high school students, rather than a sample of individuals with OCD. Comparison of the latent structure of OCD symptoms in clinical and non-clinical populations is only meaningful if it is assumed that the latent structure is similar in these two groups. While all four studies against which we compared our results (Bamber et al., 2002; Berg et al., 1988; Moore et al., 2010; Rueda-Jaimes et al., 2007) used non-clinical samples, we believe that the lack of investigation of OCD symptoms in a clinical population remains a limitation of our method, and that of previous research. Notwithstanding these limitations, our results suggest that further research should be conducted to re-examine the factor structure of the LOI-CV in different populations and ethnic groups. Such research would assist in clarifying the reasons for the differences observed in our results compared to those of previous authors, and elucidate more fully the psychometric properties of the LOI-CV when used in different populations. Examination of the factor structure within a clinical population would also provide information about whether the latent structure of symptoms is different in clinical and non-clinical populations. Examination of the similarities and differences in latent structure of OCD symptoms in different cultures may help shed light on the cultural/societal influences on OCD symptoms. This may, in turn, enhance our overall understanding of the phenomenology and etiology of the condition. If future research was to demonstrate that the latent structure of OCD symptoms was different in various cultural and ethnic groups, this would provide evidence for the impact of cultural and societal factors in determining the final manifestation of the illness. As mentioned earlier, we believe that investigating the latent structure of OCD symptoms using only non-clinical participants is a limitation of the current research, as well as the four previously published factor analytic studies of the LOI. Future research may examine symptom structure in clinical populations to assess whether this structure fits into both clinical and community population. 5. Conclusions The Leyton-20 was shown to be reliable and valid in Chinese adolescent. The clinical implication is that the scale could be

an important instrument to assess OCD in adolescents at both population and clinical level. Using this instrument may provide a foundation for interventions such as training programs and school-family collaborative programs to foster mental health in adolescents in Chinese secondary schools.

Role of funding source Funding organisation has provided funding support. It has not influenced the design of the study, data collection, data analysis and publication work. This publication work represents authors’ opinion.

Conflict of interest Authors declare that there is no conflict ofinterest.

Acknowledgement The authors wish to thank Institute of Psychology, Chinese Academy of Science for data collection work support.

References Allsopp, M., Verduyn, C., 1990. Adolescents with obsessive–compulsive disorder: a case note review of consecutive patients referred to a provincial regional adolescent psychiatry unit. J. Adolesc. 13, 157–169. American Psychiatric Association, 1994. Diagnostic and Statistical Manual of Mental Disorders (DSM-IV)American Psychiatric Association, Washington, DC. Amir, N., Freshman, M., Foa, E.B., 2000. Family distress and involvement in relatives of obsessive–compulsive disorder patients. J. Anxiety Disord. 14, 209–217. Bamber, D., Tamplin, A., Park, R.J., Kyte, Z.A., Goodyer, I.M., 2002. Development of a short Leyton Obsessional Inventory for children and adolescents. J. Am. Acad. Child Adolesc. Psychiatry 41, 1246–1252. Barrett, P.M., Healy, L.J., 2003. An examination of the cognitive processes involved in childhood obsessive–compulsive disorder. Behav. Res. Ther. 41, 15 (15). Barrett, P.M., Rasmussen, P., Healy, L., 2001. The effects of childhood obsessive– compulsive disorder on sibling relationships in late childhood and early adolescents: preliminary findings. Australian Educ. Dev. Psychol. 17, 82–102. Bentler, P.M., 1980. Multivariate analysis with latent variables: causal modeling. Annu. Rev. Psychol. 31, 419–456. Berg, C.Z., Whitaker, A., Davies, M., Flament, M.F., Rapoport, J.L., 1988. The survey form of the Leyton Obsessional Inventory—Child Version: norms from an epidemiological study. J. Am. Acad. Child Adolesc. Psychiatry 27, 759–763. Bollen, K.A., 1989. Structural Equations with Latent Variables. John Wiley, New York. Cattell, R.B., 1966. The Scree test tor the number of factors. Multivariate Behav. Res. 1, 245–276. Chia, B.H., 1981. A study of obsessive compulsive neurosis in private practice. Ann. Acad. Med. Singapore 10, 140–144. Cooper, J., 1970. The Leyton Obssessional Inventory. Psychol. Med. 1, 48–64. Cooper, J., 1996. Obsessive–compulsive disorder: effects on family members. Am. J. Orthopsychiatry 66, 296–304. Douglass, H.M., Moffitt, T.E., Dar, R., McGee, R., Silva, P., 1995. Obsessive–compulsive disorder in a birth cohort of 18-year-olds: prevalence and predictors. J. Am. Acad. Child Adolesc. Psychiatry 34, 1424–1431. Farrell, L., Waters, A., Milliner, E., Ollendick, T., 2012. Comorbidity and treatment response in pediatric obsessive–compulsive disorder: a pilot study of group cognitive-behavioral treatment. Psychiatry Res. 199, 115–123. Flament, M.F., Koby, E., Rapoport, J.L., Berg, C.J., Zahn, T., Cox, C., Denckla, M., Lenane, M., 1990. Childhood obsessive–compulsive disorder: a prospective follow-up study. J. Child Psychol. Psychiatry 31, 363–380. Flament, M.F., Whitaker, A., Rapoport, J.L., Davies, M., Berg, C.Z., Kalikow, K., Sceery, W., Shaffer, D., 1988. Obsessive compulsive disorder in adolescence: an epidemiological study. J. Am. Acad. Child Adolesc. Psychiatry 27, 764–771. Fogel, J., 2003. An epidemiological perspective of obsessive–compulsive disorder in children and adolescents. Can. Child Adolesc Psychiatry Rev. 12, 33–36. Geller, D.A., Biederman, J., Faraone, S., Agranat, A., Cradock, K., Hagermoser, L., Kim, G., Frazier, J., Coffey, B.J., 2001a. Developmental aspects of obsessive compulsive disorder: findings in children, adolescents, and adults. J. Nerv. Ment. Dis. 189, 471–477. Geller, D.A., Biederman, J., Faraone, S.V., Bellordre, C.A., Kim, G.S., Hagermoser, L., Cradock, K., Frazier, J., Coffey, B.J., 2001b. Disentangling chronological age from age of onset in children and adolescents with obsessive–compulsive disorder. Int. J. Neuropsychopharmacol. 4, 169–178. Geller, D.A., Biederman, J., Griffin, S., Jones, J., Lefkowitz, T.R., 1996. Comorbidity of juvenile obsessive–compulsive disorder with disruptive behavior disorders. J. Am. Acad. Child Adolesc. Psychiatry 35, 1637–1646. Heyman, I., Fombonne, E., Simmons, H., Ford, T., Meltzer, H., Goodman, R., 2003. Prevalence of obsessive–compulsive disorder in the British nationwide survey of child mental health. Int. Rev. Psychiatry 15, 178–184.

J. Sun et al. / Journal of Affective Disorders 164 (2014) 19–27

Hollander, E., Kwon, J.H., Stein, D.J., Broatch, J., Rowland, C.T., Himelein, C.A., 1996. Obsessive–compulsive and spectrum disorders: overview and quality of life issues. J. Clin. Psychiatry 57 (Suppl. 8), 3–6. Hu, L.T., Bentler, P.M., 1995. Evaluating model fit. In: HOYLE, R.H. (Ed.), Structural Equation Modeling: Concepts, Issues, and Applications. Sage, Thousand Oaks, CA. King, N., Inglis, S., Jenkins, M., Myerson, N., Ollendick, T., 1995. Test–retest reliability of the survey form of the Leyton Obsessional Inventory—Child Version. Percept. Mot. Skills 80, 1200–1202. Krank, M., Stewart, S.H., O’Connor, R., Woicik, P.B., Wall, A.M., J.C.P., 2011. Structural, concurrent, and predictive validity of the Substance Use Risk Profile Scale in early adolescence. Addict. Behav. 36 (37-46). Lewin, A.B., Piacentini, J., 2010. Evidence-based assessment of child obsessive compulsive disorder: recommendations for clinical practice and treatment research. Child Youth Care Forum 39, 73–89. Liu, X., Kurita, H., Sun, Z., Wang, F., 1999. Risk factors for psychopathology among Chinese children. Psychiatry Clin. Neurosci. 53, 497–503. Lo, W.H., 1967. A follow-up study of obsessional neurotics in Hong Kong Chinese. Br. J. Psychiatry 113, 823–832. Moore, J., Smith, G.W., Shevlin, M., O’Neill, F.A., 2010. Alternative factor models and heritability of the Short Leyton Obsessional Inventory—Children’s Version. J. Abnorm. Child Psychol. 38, 921–934. Rasmussen, S.A., Eisen, J.L., 1990. Epidemiology of obsessive compulsive disorder. J. Clin. Psychiatry, 51 (Suppl., 10–14). Real, E., Labad, J., Alonso, P., Segalàs, C., Jiménez-Murcia, S., Bueno, B., Subirà, M., Vallejo, J., Menchón, J.M., 2011. Stressful life events at onset of obsessive– compulsive disorder are associated with a distinct clinical pattern. Depress. Anxiety 28, 367–376. Rueda-Jaimes, G.E., Díaz-Martínez, L.A., Escobar-Sánchez, M., Franco-López, J.A., Navarro-Mancilla, A.A., Cadena-Afanador, L.P., 2007. Validation of the short version of the Leyton Obsessional Inventory for children and adolescents in Bucaramanga, Colombia. Aten. Primaria 39, 75–80.

27

Sans, J.C., Hernández-Martínez, C., Muñoz, S.C., García, L.L., Trallero, J.T., 2012. The Leyton Obsessional Inventory—Child Version: validity and reliability in Spanish non-clinical population. Int. J. Clin. Health Psychol. 12, 81–96. Shaffer, D., Fisher, P., Dulcan, M.K., Davies, M., Piacentini, J., Schwab-Stone, M.E., Lahey, B.B., Bourdon, K., Jensen, P.S., Bird, H.R., Canino, G., Regier, D.A., 1996. The NIMH diagnostic interview schedule for children version 2.3 (DISC-2.3): description, acceptability, prevalence rates, and performance in the MECA study. Methods for the epidemiology of child and adolescent mental disorders study. J. Am. Acad. Child Adolesc. Psychiatry 35, 865–877. Stewart, S.E., Rosario, M.C., Brown, T.A., Carter, A.S., Leckman, J.F., Sukhodolsky, D., Katsovitch, L., King, R., Geller, D., Pauls, D.L., 2007. Principal components analysis of obsessive–compulsive disorder symptoms in children and adolescents. Biol. Psychiatry 61, 7 (7). Storch, E.A., Khanna, M., Merlo, L.J., Loew, B.A., Franklin, M., Reid, J.M., Goodman, W.K., Murphy, T.K., 2009. Children’s Florida obsessive compulsive inventory: psychometric properties and feasibility of a self-report measure of obsessive–compulsive symptoms in youth. Child Psychiatry Hum. Dev. 40, 467–483. Storch, E.A., Park, J.M., Lewin, A.B., Morgan, J.R., Jones, A.M., Murphy, T.K., 2011. The Leyton Obsessional Inventory—Child Version Survey Form does not demonstrate adequate psychometric properties in American youth with pediatric obsessive–compulsive disorder. J. Anxiety Disord. 25, 574–578. Swedo, S.E., Leonard, H.L., Schapiro, M.B., Casey, B.J., Mannheim, G.B., Lenane, M.C., Rettew, D.C., 1993. Sydenham’s chorea: physical and psychological symptoms of St Vitus dance. Pediatrics 91, 706–713. Swedo, S.E., Rapoport, J.L., Cheslow, D.L., Leonard, H.L., Ayoub, E.M., Hosier, D.M., Wald, E.R., 1989. High prevalence of obsessive–compulsive symptoms in patients with Sydenham’s chorea. Am. J. Psychiatry 146, 246–249. Valleni-Basile, L.A., Garrison, C.Z., Jackson, K.L., Waller, J.L., McKeown, R.E., Addy, C.L., Cuffe, S.P., 1994. Frequency of obsessive–compulsive disorder in a community sample of young adolescents. J. Am. Acad. Child Adolesc. Psychiatry 33, 782–791. Zohar, A.H., 1999. The epidemiology of obsessive–compulsive disorder in children and adolescents. Child Adolesc. Psychiatry Clin. N. Am. 8, 445–460.