Elsevier

Journal of Anxiety Disorders

Volume 38, March 2016, Pages 79-87
Journal of Anxiety Disorders

The structure of obsessive compulsive symptoms and beliefs: A correspondence and biplot analysis

https://doi.org/10.1016/j.janxdis.2016.01.003Get rights and content

Highlights

  • Correspondence analysis with biplot was applied to examine the association between obsessive-compulsive symptoms and obsessive-compulsive beliefs.

  • Samples of individuals with OCD (N = 398), other anxiety disorders (N = 104), and undergraduates (N = 285) were collected.

  • The results across three subsamples indicate that inflated responsibility was significantly associated with OC symptoms, for mild symptoms.

  • Importance and control of thoughts were associated with OC symptoms across all levels of symptom severity.

Abstract

Research has suggested that obsessive-compulsive (OC) beliefs are differentially predictive of OC symptom dimensions. One additional way in which beliefs and symptoms may be related is by severity; that is, the role of beliefs may vary as a function of symptom severity. In order to evaluate this possibility, correspondence analysis with biplot was applied to evaluate the association between OC beliefs and OC symptom severity across three subsamples, individuals with obsessive-compulsive disorder (OCD; N = 398), individuals with anxiety disorders (N = 104), and undergraduate students (N = 285). To do so, we generated five row categories of symptom severity and six columns based on the Obsessive Compulsive Beliefs Questionnaire (OBQ) for three subsamples. Unlike factor analyses of inter-variable correlations (or covariances), the CA-biplot paradigm calibrates simultaneously row and column information and estimates dimensional coordinates (analogous to factor loadings) separately for rows and columns. We used the first two dimensions from each subsample because they accounted for most variance (on average 89%) so as to construct a hypothetical plane with them. Then, we visually inspected associations among five severity categories (rows) and six OBQ subscales (columns) in the plane and also calculated their correlations. The visual configurations and numerical correlations were consistent across three subsamples, indicating that inflated responsibility was significantly associated with OC symptoms, but only for those with the least severe symptoms. Importance and control of thoughts were associated with OC symptoms across all levels of symptom severity. The implications of these findings for the cognitive model of OCD are considered.

Introduction

Obsessive-compulsive disorder (OCD) is characterized by obsessions and compulsions (American Psychiatric Association, 2013). Contemporary cognitive-behavioral models of OCD (Clark, 2004, Frost and Steketee, 2002, Salkovskis, 1996) emphasize the role of dysfunctional beliefs in giving rise to obsessions and compulsions (Obsessive Compulsive Cognitions Workgroup [OCCWG], 2001). Although these belief domains have been associated with obsessive-compulsive (OC) symptoms, it is unclear whether the belief domains have specificity to OCD (McKay et al., 2014) or whether it is necessary to target each belief domain in treatment (McKay et al., 2015; Wilhelm, Berman, Keshaviah, Schwartz, & Steketee, 2015).

The OCCWG (1997, 2001) identified, based on a review of the research literature, six types of beliefs that are associated with OC symptoms: (1) inflated personal responsibility (e.g., “If I foresee harm, then I am responsible for preventing harm”), (2) the tendency to overestimate threat (e.g., “Dangers are lurking everywhere”), (3) perfectionism (e.g., “It is essential that I perform perfectly at my job”), (4) intolerance of uncertainty (e.g., “I can’t stand being uncertain about the possibility of danger”), (5) overimportance of the significance of one's thoughts (e.g., “Thinking bad thoughts is morally equivalent to committing bad deeds”), and (6) belief in the necessity to control one’s thoughts (e.g., “It is essential to control my unwanted thoughts”). The most well validated and widely used measure of these dysfunctional beliefs is the Obsessive Beliefs Questionnaire (OBQ; OCCWG, 2003, 2005). In the original development of the scale, it was recognized that at least some belief domains assessed with the scale were not uniquely predictive of OCD (i.e., perfectionism). However, other belief domains were considered particularly salient to OCD (i.e., inflated responsibility).

Since the development of the cognitive model of OCD, and the associated belief domains, researchers have been attempting to address three broad elements related to the defined beliefs (Tolin, Worhunsky, & Maltby, 2006). First, to what extent do these beliefs have generality to the various symptoms associated with OCD? That is, do all individuals with OCD have obsessive-compulsive (OC) beliefs? Second, to what extent do OC beliefs have specificity to OCD? That is, are these beliefs predominantly present in individuals with OCD, or are these belief domains also present in other conditions? And finally, do OC beliefs have congruence, in that they are related to different symptoms in meaningful ways? To illustrate, Viar, Bilsky, Armstrong, and Olatunji (2011) found that OC symptoms are associated with some form of OC belief (generality), that different OC beliefs related to different OC symptoms in a meaningful way (congruence), and that individuals with OCD did not endorse OC beliefs more strongly than individuals with generalized anxiety disorder (specificity).

Investigations with non-clinical samples have considerable relevance for understanding clinical phenomena (Abramowitz et al., 2014). In the case of OCD, symptoms are well known to occur in individuals who do not meet criteria for the diagnosis, and it is widely accepted that symptoms are dimensional in nature (Olatunji, Williams, Haslam, Abramowitz, & Tolin, 2008). Accordingly, investigations with both clinical and non-clinical samples are useful in addressing the questions of generality, specificity and congruence. Research has shown that obsessive-compulsive symptoms and beliefs are related. Taylor and Jang (2011), for example, using a sample of 307 twin pairs, showed that a causal model where beliefs predicted symptoms was the best fit for the data, but that these beliefs accounted for a comparably small amount of variance. In another large sample investigation, Taylor et al. (2010) conducted structural equation modeling to evaluate direct and indirect effects of OC beliefs on symptoms. Using a sample of N = 5015, the results showed that inflated responsibility and overestimation of threat were predictive of all six OC symptom dimensions assessed (checking, hoarding, neutralizing, obsessive, ordering, and washing). Additional investigations with non-clinical samples have suggested at least some obsessive-compulsive beliefs are related to symptoms (i.e., control over thoughts predicting obsessions, perfectionism predicting symmetry/ordering; Tolin, Woods, & Abramowitz, 2003) and other investigations have suggested that most or all obsessive-compulsive symptoms are associated with most or all obsessive-compulsive beliefs (i.e., all higher-order belief factors associated with symptoms, Taylor, McKay, & Abramowitz, 2005) all lower order belief domains associated with all symptoms, Myers, Fisher, & Wells, 2008).

Research with clinical samples has found that at least some beliefs are related to symptom expression. For example, Brakoulias et al. (2014) found that specific OC symptoms were uniquely associated with specific beliefs in a sample of 154 individuals with OCD, such that doubts and checking were associated with increased responsibility/overestimation of threat, unacceptable thoughts was associated with importance and control of thoughts, and symmetry/ordering was most strongly associated with perfectionism and intolerance of uncertainty. Fergus and Carmin (2014, study 1) found that obsessive-compulsive beliefs were associated with all obsessive-compulsive symptoms in a clinical sample of 48 individuals with OCD.

Based on the research conducted on OC beliefs and their association with OC symptoms, it appears that additional clarification is warranted since there are occasions where generality, specificity, and congruence are evident, and other occasions where they are not. Although it appears that OC beliefs are associated with OC symptoms, these relations are not uniformly supported (i.e., in some investigations, there are a subset of obsessive-beliefs that are not strongly associated with symptoms). One possibility is that symptom severity and OC beliefs are not uniformly related, but instead differentially related based on severity. That is, at higher levels of symptom severity some OC beliefs are more like to be strongly held than at less severe levels. This is possible as OC symptoms and beliefs vary by samples, such that control over thoughts had specificity to those with OCD compared to anxious and non-anxious (student) controls (McKay et al., 2014). Accordingly, this investigation was undertaken in an attempt to determine whether there are distinctive correlation patterns among OC beliefs associated with OCD based on severity level rather than on symptom subtype, and whether these correlation patterns distinguish OCD from other disorders or from control participants. Since there are no compelling models to guide prediction of specific beliefs to symptom domains, another way to evaluate the role of dysfunctional beliefs is to determine whether it serves as a proxy for severity.

For the present research, we conducted biplot along with correspondence analysis (CA; see description in Section 2) to analyze OBQ data including three subsamples—individuals with OCD, individuals with other anxiety disorders, and undergraduate students. Participants also completed measures of obsessive-compulsive symptoms. Our analyses aimed to address the following questions about the architecture of OC-related dysfunctional beliefs: (1) what are the first two dimensions from both row and column categories? (2) What are dominant categories contributing most in a plane constructed by the first two dimensions? (3) What are the correlations between severities of symptom categories (rows), dysfunctional belief categories (columns) in the plane? (4) Are these correlation patterns similar or different across different samples?

Section snippets

Participants

The sample consisted of (1) 398 people with a primary diagnosis of OCD, defined according to DSM-IV criteria (American Psychiatric Association, 2000), (2) people with an anxiety-related disorder other than OCD (anxious controls, AC; N=104), and (3) undergraduate students (UG) (N=285). The demographic features of the OCD sample were as follows: mean age of 36 years (SD=12); 55% women; 96% with at least high-school education; 94% white; 45% married or cohabiting; 25% unemployed or on disability

Results

Each group was examined separately with CA. The first two dimensions for the groups were estimated. The contributions of the first two eigenvalues (with specific eigenvalues in parentheses) in OCD were 66% (.010) and 29% (.004), respectively; those for AC were 73% (.023) and 19% (.006), respectively; and those for UG were 49% (.012) and 31% (.008). With the CA results, individual category contributions (%) were computed in each dimension and in a plane for three groups (OCD, AC, and UG). Then,

Discussion

This study evaluated, using CA and associated biplot, the hierarchical association between OC beliefs, as assessed with the OBQ, and obsessive-compulsive symptom severity. Estimation of such association is a novel application of CA given that it requires discretization of otherwise continuous data. However, this method has advantages noted earlier, specifically by its utility in evaluating the distance between scores on a measure that has assumed interval quality, and examining this visually

References (48)

  • American Psychiatric Association

    Diagnostic and statistical manual of mental disorders

    (2000)
  • American Psychiatric Association

    Diagnostic and statistical manual of mental disorders

    (2013)
  • E.J. Beh et al.

    Correspondence analysis: theory practice and new strategies

    (2014)
  • V. Brakoulias et al.

    The relationship between obsessive-compulsive symptoms and cognitions in obsessive-compulsive disorder

    Psychiatric Quarterly

    (2014)
  • T.A. Brown et al.

    Anxiety disorders interview schedule for DSM-IV (ADIS-IV)

    (1994)
  • D.A. Clark

    Cognitive-behavioral therapy for OCD

    (2004)
  • L. Doey et al.

    Correspondence analysis applied to psychological research

    Tutorials in Quantitative Methods for Psychology

    (2011)
  • T.A. Fergus et al.

    The validity and specificity of the short-form of the Obsessive Beliefs Questionnaire (OBQ)

    Journal of Psychopathology and Behavioral Assessment

    (2014)
  • T.A. Fergus et al.

    Do symptoms of generalized anxiety and obsessive-compulsive disorder share cognitive processes?

    Cognitive Therapy and Research

    (2010)
  • R.O. Frost et al.

    Cognitive approaches to obsessions and compulsions: theory assessment and treatment

    (2002)
  • K.R. Gabriel

    The biplot graphic display of matrices with application to principal component analysis

    Biometrika

    (1971)
  • K.R. Gabriel et al.

    Biplots in biomedical research

    Statistics in Medicine

    (1990)
  • W.K. Goodman et al.

    The Yale–Brown Obsessive Compulsive Scale. I. Development, use, and reliability

    Archives of General Psychiatry

    (1989)
  • J. Gower et al.

    Understanding biplots

    (2011)
  • Cited by (20)

    • Age moderated–anxiety mediation for multimodal treatment outcome among children with obsessive-compulsive disorder: An evaluation with correspondence analysis

      2021, Journal of Affective Disorders
      Citation Excerpt :

      Assuming there is a hypothetical map where the horizontal axis represents the first dimension estimated by CA and the vertical axis the second dimension, each category's position on the map is determined by the first and second dimensional coordinates of the category. Using these dimensional coordinates, we estimated angles between categories, and the angles between categories are statistically equivalent to correlations for any pairs of categories (for a technical discussion, see (Egghe and Leydesdorff, 2009; Kim and Annunziato, 2018; Kim et al., 2019; Kim et al., 2016; Kim and Frisby, 2019; Kim and Grochowalski, 2019). Given that these analyses result in a large number of correlations, to control Type I error we employed the False Discovery Rate (FDR; Benjamini and Hochberg, 1995).

    • Eating disorders with over-exercise: A cross-sectional analysis of the mediational role of problematic usage of the internet in young people

      2021, Journal of Psychiatric Research
      Citation Excerpt :

      Impulsivity, compulsivity and obsessionality, when considered together they are found as prevalent behaviors in purging anorexia (Hoffman et al., 2012). Obsessional thinking is strongly positively correlated with compulsive behavior (Kim et al., 2016). Impulsivity and compulsivity exist cross-diagnostically in latent functionally impairing forms which are positively correlated (Chamberlain et al., 2018) and are both predictive of ED and PUI.

    • Assessing treatment efficacy by examining relationships between age groups of children with autism spectrum disorder and clinical anxiety symptoms: Prediction by correspondence analysis

      2020, Journal of Affective Disorders
      Citation Excerpt :

      Therefore, without controlling such linearity, examining the mean differences among these outcome measures using ANCOVA (to control the pre-treatment effect) as done in the previous studies (e.g., Storch et al., 2015; Woods et al., 2015), would be biased. On the other hand, correspondence analysis (CA) can examine the group differences in terms of score differences or magnitudes of associations, without causing any biases because of analyzing the related variables/categories (e.g., Kim, 2019; Kim and Annunziato, 2018; Kim et al., 2016). Also, CA relies on chi-square statistics where the tests are robust with sample sizes typical of treatment trials, and does not require the multivariate normality assumption.

    • Profile Analysis of Psychological Symptoms Associated With Misophonia: A Community Sample

      2018, Behavior Therapy
      Citation Excerpt :

      However, importance and control of thoughts was not found significantly correlated with the symptom profile. These cognitive domains, while initially assigned significance to OCD (i.e., OCCWG, 2005), may be more general and characterize multiple domains of psychopathology (i.e., Kim et al., 2016; McKay et al., 2014). Since these analyses do not suggest a strong relationship between obsessive-compulsive symptom dimensions and misophonia, the finding that these cognitive domains are significantly correlated with the resultant profile may better reflect elevations in dysfunctional cognitions generally among individuals with misophonia.

    View all citing articles on Scopus
    View full text