Factor Structure and Validation of the Obsessive Compulsive Inventory-Revised (OCI-R)
Factor structure and validation of the Obsessive Compulsive Inventory-Revised (OCI-R) in a Greek non-clinical sample
Ioannis Angelakis1-2* MariaPanagioti3& Jennifer L. Austin2
1Panteion University, Athens, Greece;
2University of South Wales, School of Psychology, Pontypridd, Wales, UK;
3Institute of Population Health, University of Manchester, UK
RUNNING HEAD: FACTOR STRUCTURE AND VALIDATION OF A HELLENIC VERSION OF OCI-R
*Correspondence: Dr Ioannis Angelakis, University of South Wales, School of Psychology, Pontypridd, Wales, CF37 1DL, e-mail:
Abstract
Thisstudy assessed and validated a Greek version of the Obsessive-Compulsive Inventory-Revised (OCI-R) in a large community sample. A total of 1379 Greek adults, including university students and individuals from the community, participated. Overall, the Greek OCI-R supported the six-factor solution, namely checking, ordering, obsessing, hoarding, washing and neutralizing, which was suggested by the original scale development and was later confirmed in other cultural settings. OCI-R was found to have very good to excellent psychometric properties as demonstrated by the application of traditional and alternative validating methods. Further, a five-factor structure that excluded the hoarding scale provided a slightly better conceptual fit of the data. In light of new recommendations, the Greek version of OCI-R provides compelling evidence of its efficacy to clearly differentiated between high- and low self-reports of OCD symptoms. We discuss that scores from the Greek community sample were commensurate with those observed in clinical samples.
Factor structure and validation of the Obsessive Compulsive Inventory-Revised (OCI-R) in a Greek non-clinical sample
Obsessive-Compulsive Disorder (OCD) is a disabling mental health illness characterized by recurrent intrusive thoughts and repetitive covert or overt acts, namely obsessions and compulsions, which elicit a considerable degree of discomfort to the sufferer (American Psychiatric Association, 2000, 2013). OCD has been described as a highly heterogeneous disorder, which makes its clinical presentation and subsequent treatment difficult to understand and manage (Lochner & Stein, 2006; Mataix-Cols, Rosario-Campos, & Leckman, 2005). Recently, it was removed from the anxiety disorders chapter to compose a new clinical entity of obsessive-compulsive and related disorders according to DSM-5 (American Psychiatric Association, 2013). OCD shares some common characteristics with a number of other clinical syndromes, such as Body Dysmorphic disorder (BDD) and Hoarding disorder (HD), which now are reflected in the diagnostic criteria for these disorders. However, the unique features of each of these conditions still allow their classification as separate mental disorders. Although changes to the criteria and methodological limitations (e.g., Angst et al., 2004; Gibbs, 1996) render an estimate of its prevalence difficult, epidemiological studies suggest that the lifetime OCD prevalence varies from 1.5% to 3.5% (Angst et al., 2004; Crino, Slade, & Andrews, 2005; Subramaniam, Abdin, Vaingankar, & Chong, 2012), whereas 28.2% of the general population report OCD symptoms at least once in their lifetime (Ruscio, Stein, Chiu, & Kessler, 2010).
Numerous studies highlight the distress and apprehension that obsessions and compulsions cause to people with OCD (e.g., Veale & Roberts, 2014). Further, OCD is associated with great reductions in quality of life and severe impairments in social and occupational functioning (Bobes, Gonzalez, Bascaran, Arango, Saiz, & Bousono, 2001; Subramaniam, Soh, Vaingankar, Picco, & Chong, 2013). Consequently, a number of self-report measures have been developed to diagnoseOCD and to assess the frequency and severity of OCD symptoms. The most well-known of these measures is the Obsessive-Compulsive Inventory - Revised (OCI-R), which was developed to assess the full dimensions of OCD symptoms and their severity, as well as the overall distress caused from these experiences (Foa, et al., 2002).
OCI-R is a shortened version of the Obsessive-Compulsive Inventory (OCI) (Foa, Kozak, Salkovskis, Coles, & Amir, 1998). The main advantage of OCI-R is that it requires less time to be completed than the original OCI. In comparison to the 42-item structure of OCI, OCI-R consists of only 18 items, which measure symptom severity on a 5-point Likert-type scale and form six sub-scales of the most common symptom categories encountered in OCD patients, including a washing, checking, ordering, obsessing, hoarding and neutralizing category. The OCI-R has been found to have excellent psychometric properties (Foa, et al., 2002; Hajcak, Huppert, Simons, & Foa, 2004). In particular, OCI-R has shown very good to excellent internal consistency, test-retest reliability and convergent validity both in clinical populations and in normal controls.
Other self-report measures for OCD also have been devised, but deemed insufficient for a variety of reasons. For instance, Maudsley Obsessive-Compulsive Inventory (MOCI) (Hodgson & Rachman, 1977) has received major criticism because it has poor internal consistency and also fails to fully capture all the symptoms of OCD (Thordarson, Radomsky, Rachman, Shafran, Sawchuk, & Hakstianet, 2004). Other self-report measures, such as the Padua Inventory of Obsessive-Compulsive Symptoms (Sanavio, 1988) or the Vancouver Obsessional Compulsive Inventory (Thordarson, et al., 2004), also have been criticized due to their large size and the prolonged time required for their completion. Fairly recently, Abramowitz et al. (2010) developed the Dimensional Obsessive-Compulsive Scale (DOCS) to capture and assess the heterogeneous nature of the OCD symptomatology. Although DOCS has been found to be a sound measure of OCD, it also can betime consuming to administer. Further, its properties have not yet fully been re-examined across diverse cultural settings.
On the contrary, several studies have translated and validated OCI-R in different cultural settings, including Spain, Germany, France, Korea, China, Turkey and Iceland, verifying that the scale has very good or excellent psychometric properties (Aydin, Boysan, Kalafat, Selvi, Beşiroglu, & Kagan, 2014; Fullana, Tortella-Feliu, Caseras, Andion, Torrubia, & Mataix-Cols, 2005; Gonner, Leonhart, & Ecker, 2008; Smári, Ólason, Eypórsdóttir, & Frölunde, 2007; Tang, Yu, He, Wang, Chasson, 2015; Zermatten, van der Linden, Jermann, & Ceschi, 2006; Woo, Kwon, Lim, & Shin, 2010).Of note, nearly half of the translated versions of OCI-R have been validated among non-clinical populations (Fullana et al., 2005; Smari et al., 2007; Zermatten et al., 2006). Validating OCI-R in community samples is important because the frequency of OCD-related behaviors is considerably high in the general population. For instance, Fullana et al. (2010) found a lifetime prevalence of 13% in any of the symptom categories that characterize OCD in the general population. Research evidence also suggests that non-clinical participants, who self-report high scores in obsessive-compulsive inventories, tend to produce stable obsessive-compulsive behaviors across time, which closely correspond to those observed in clinical patients diagnosed with OCD (Burns, Formea, Keortge, & Sternberger, 1995). Moreover, the use of a community sample facilitates the investigation of the factor structure and the evaluation of the psychometric properties of OCI-R (e.g., Sica et al., 2009).
Among the limitations ofOCI-R’s scaleis thatits discriminant validity is only adequate to good,likelybecause OCD usually co-exists with other disorders, with major depression possessinga significant place among Axis I comorbid conditionscommonly diagnosed with OCD (e.g., Abramowitz & Deacon, 2006; Foa et al., 2002; Tükel, Polat, Özdemir, Aksüt, & Türksoy, 2002).Further, order effects have been observed when it is administered after another measure of OCD (Hajcak et al., 2004; Woo et al., 2010). Gender effects also have been found in some cultural settings. For instance,Spanish men have been found tohavemorehoarding and checking complaints than Spanish women (Fullana et al., 2005), whereas Italian men score higher on washing, checking and obsessing sub-scales than women of the same culture (Sica et al., 2009). In Iceland, these findings are reversed with women reporting more complaints on the checking and orderingsub-scales compared tomen (Smári et al., 2007).
A Greek version of the OCI-R does not exist. This is an important gap in the literature that needs to be addressed for both research and clinical reasons. In terms of research, a Greek version of the OCI-R would facilitate the research process and the direct comparison of studies focusing on OCD in the Greek settings with studies conducted in other European or US settings. Clinically, there is evidence that the frequency of OCD-related behaviors is considerably higher in Greek samples compared to samples from other cultures. For instance, the mean score of the Hellenic version of Spence Children’s Anxiety Scale (SCAS-GR; Mellon & Moutavelis, 2007) measuring compulsive behavior in children aged from 9 to 12 years old was 82% higher compared to the mean of the original scale (Spence, 1997), which was comprised of Australian children of the same age.
To date, there is no evidence with regard to the frequency of occurrence of OCD symptoms in a Greek adult community sample, most probably due to the absence of a validated self-report measure of OCD in Greece. Given the necessity of the existence of such an instrument, which would reliably assess the frequency of OCD symptoms, an effort has been undertaken to translate and validate OCI-R in the Greek context.The overarching aim of the present study was (a) to evaluate the factor structure and the psychometric properties of the Greek-version of OCI-R in a large community sample, (b) to overcome its previous limitations (e.g., discriminant validity) by applying traditional and alternative validating methods, and (c) totest different structural models according to the new research findings. Last, in view of the inconsistent findings that have been produced with regard to the existence of both gender and order effects in other cultural settings (Fullana et al., 2005; Hajcak et al., 2004; Sica et al., 2009, Smári et al., 2007; Woo et al., 2010), thepotential influence of gender differences and order effects on the administration of the Greek version of OCI-R also was examined.
Method
Participants
In total, 1379 participants from the general community(31% males and 69% females) aged from 18 to 83 years old (M = 27.54, SD = 10.76) completed the OCI-R scale. Potential participants had to satisfy the following criteria to be included in the study: (a) speak Greek fluently, and (b) be Greek residents/citizens. Participation was voluntarily for all participants and informed consent was obtained prior to scale completion. Approximately less than half of the participants (44%) were university students, whereas 33.2% of the remaining participants had received university education (e.g., first degree or master’s level). Only 5.6% of the study sample was unemployed and only 2% had retired. With regard to the socio-economic status (SES), 24.8% of the study sample declared an average household income of600 euros per month (low SES), 46.6% declareda monthly income ranging from 1200 to 2100 (middle SES), whereas the rest 28.6% of the sample declared makingmore than 2100 euros per month (high SES). Fewer participants completed the total study measures as detailed in Tables 6 and 7. No payments or course credits were offered for their participation.
Measures
Obsessive-compulsive inventory-revised (OCI-R) (Foa et al., 2002)
OCI-R consists of 18 items that measure distress caused by OCD symptoms. Participants’ scores are rated in a 5-point scale ranging from 0 = never to 4 = very much. OCI-R also consists of six sub-scales, including washing, checking, orderliness, hoarding, neutralization and obsessions. In the current study a Greek version of the OCI-R was used. The scale was translated from English to Greek and back to English by six independent raters (three raters translated the scale from Greek to English and the other three translated it from English to Greek). The accuracy of these translations was reviewed by the authors of the study (IA, MP). In this study, the alpha coefficient was 0.89 for the overall scale (see,Table 2). The final version of the Greek version of the scale is presented in Appendix.
Maudsley obsessive-compulsive inventory (MOCI) (Hodgson & Rachman, 1977)
MOCI consists of 30 true-false statements that assess the frequency of OCD symptoms. It covers four areas of complaints, that is, checking, cleanliness, slowness and doubting. MOCI has good internal consistency and good to very good test-retest reliability and convergent validity (Emmelkamp, Kraaijkamp, & Van den Hout, 1999; Stoylen, Larsen, & Kvale, 2000). The alpha coefficient for this scale was found to be very good, Cronbach’s α =0.81, whereas the 30-day test-retest reliability was excellent (0.83) in this study, suggesting proper psychometric qualities (McCrae, Kurtz, Yamagata, & Terracciano, 2011).
Launay-Slade hallucinations scale (LSHS) (Launay & Slade, 1981)
LSHS is a 16-item self-report questionnaire that measures healthy individuals’ predisposition towards hallucinations. Responses are recorded in a 4-point scale, which range from 1 = never to 4 =very frequently. It incorporates three types of hallucinations, namely, intrusive mental events, auditory religious and visual religious hallucinations. In this dataset, LSHS was found to have excellent internal consistency, Cronbach’s a= 0.88, and an excellent 30-day test-retest reliability (0.82), suggesting appropriate psychometric properties (McCrae, Kurtz, Yamagata, & Terracciano, 2011).
Centre for epidemiological studies depression scale (CES-D) (Radloff, 1977)
CES-D consists of 20 items that measure the severity of depressive symptoms in the general population. Participants self-report their answers in a 4-point scale by marking the number that better describe their mood or feelings during the past week. CES-D scores range from 0 = rarely to 3 = most of the times. It has been found to have very good to excellentpsychometric properties (Fountoulakis et al., 2001). In the current study, the alpha coefficient was high,Cronbach’s a = 0.86, and the 1-month test-retest reliability was excellent (0.83).
Procedure
The first author, who had a first degree in Psychology and a postgraduate professional qualification in Psychology, together with undergraduate Psychology students from Panteion University, approached prospective participants. The majority of the student population was comprised of participants mainly from Panteion University or National & Kapodistian University of Athens. The non-student population was recruited from local coffee shops, cinemas and other merchants. The first author and research assistants visited these establishments and recruited participants by asking customers if they would like to complete the survey.Those that agreed were given a written consent form that informed them of the voluntary nature of the study. If they signed the form, they were then taken to a quiet area of the establishment to complete the survey.After the completion of the survey, participantswere verbally debriefed regarding the aims of the study using a script, so that debriefing statements were consistent across researchers. Data on a number of key demographic characteristics (i.e., age, gender, occupation, socio-economic status) also were recorded. Completion of these measures lasted approximately 20 min. To examine test-retest reliability, the OCI-R was distributed again to 108 participants (56.5%of the sample was comprised of undergraduate students) 30 days after its initial administration.
Data analyses
All statistical analyses were conducted using the IBM SPSS® (version 23.0) or the AMOS®(version 23.0) statistical software.Study’s variables were tested for normality by assessing the measure of skewness for every item, which revealed no deviation from a normal distribution.In the current analysisboth anexplanatory factor analysis (EFA) and a confirmatory factor analysis (CFA) were utilized. This decision was based on three reasons: (a) the symptoms that describe a number of mental health disorders are possible to differ at least slightly among different cultural contexts,(b) this is the first attempt to adapt the OCI-R in a Greek setting and therefore it is critical to explore the structure of the Greek-version of the OCI-R (e.g., Gerbing & Hamilton, 1996), and (c)the mean scores of the overall scale and the sub-scales of OCI-R were found to be exceptionally high,nearly approaching the mean scores of the original study for the patient population sample. EFAalso is more appropriate for appraising the factor structure of an instrument, which has been in an initial stage of development or validation within a particular cultural setting, and reducing multiple observed variables into fewer components that summarize their variance (see, Byrne, 2001). To this end, anEFA with promax rotation was employed, because the factors of the scale correlatedsignificantly with one another, together with a CFA. For the CFA, the maximum likelihood estimation method was used.We tested the fit of the overall scale, a five-factor model by omitting the hoarding sub-scale in accordance with the new recommendations and a six-factor model that have been proposed by the vast majority of the studies validating OCI-R in either clinical or community samples across different social settings. Test-retest reliability was assessed by producing Pearson product-moment correlation coefficients from Time 1 (initial administration) to Time 2 (30 days later) administrations of the scale. Pearson’s r coefficients also were produced to examine the inter-factor correlations of the OCI-R and to assess the convergent and discriminant validity of the scale. In addition, we explored convergent validity by computing Average Variance Extracted (AVE) scores, discriminant validity by producing square roots for AVE, and the internal consistency by calculating Composite Reliability (CR) scores for each of the suggested factors of the model.Independent-samples t-tests were conducted to explore potential gender differences on the scores of OCI-R for the total, non-collegeand the college study’s samples and to examine the possibility of an order effect.
Missing values
Initial screening revealed 76 cases with missing values in the OCI-R and in additional study measures. A non-significant Little’s MCAR test, x2(220) = 208.09,p = 0.71, suggested data missing completely at random (Little, 1988). To this end, single imputation methods using maximization expectation algorithm were performed (Scheffer, 2002). The imputed scores were used for both the EFA and CFA analyses. Then, we repeated all factor analyses to ensure that our findings remained unchanged and unbiased. With regard to subsequent analyses (e.g., correlations) and in line with recommendations (e.g., Rubin, 1987), we performed multiple imputations analyses. These scores entered the final analyses. Again, in order to ensure that our interpretation was not biased, we performed all the analyses with and without the imputed values.
Results
Descriptive statistics
The means and standard deviations of the OCI-R across the male and female participantsof the total, non-collegeand college study samples are detailed in Table 1. In general, the overall mean of the OCI-R of the current study (M = 26.82, SD = 12.79) closely resembled the overall mean reported for the patient population (M = 28.01, SD = 13.52), but not for the non-anxious controls (M = 18.82, SD = 11.10) of the original study (Foa et al., 2002). A number of additional differences also were found. For example, higher means in the ordering (M = 5.96, SD = 3.24) and hoarding sub-scales (M = 4.75, SD = 2.96), but lower means in the obsessing (M = 5.13, SD = 3.24), neutralizing (M = 2.51, SD = 2.63) and washing sub-scales (M = 3.72, SD = 3.11) were reported in this sample compared to the patient one of the original OCI-R version. The mean of the checking sub-scale (M = 4.73, SD = 3.32) was found to be similar to the one reported by Foa et al. (2002).