Journal of Psychopathology and Behavioral Assessment

, Volume 36, Issue 2, pp 318–328

The Validity and Specificity of the Short-Form of the Obsessive Beliefs Questionnaire (OBQ)

Authors

    • Department of Psychology and NeuroscienceBaylor University
  • Cheryl N. Carmin
    • Department of PsychiatryUniversity of Illinois at Chicago
Article

DOI: 10.1007/s10862-013-9398-6

Cite this article as:
Fergus, T.A. & Carmin, C.N. J Psychopathol Behav Assess (2014) 36: 318. doi:10.1007/s10862-013-9398-6

Abstract

Moulding et al. (Assessment 18:357–374, 2011) created a 20-item short-form of the Obsessive Beliefs Questionnaire (OBQ; Obsessive Compulsive Cognitions Working Group in Behaviour Research and Therapy 43:1527–1542, 2005) labeled the OBQ-20. To date, the psychometric properties of the OBQ-20 have yet to be examined among clinical respondents. It is also unknown whether the OBQ-20 addresses the limited specificity of prior OBQ versions to obsessive-compulsive symptoms. In Study 1, using a small clinical sample (N = 48), each OBQ-20 scale evidenced good internal consistency and shared a nearly identical correlation in magnitude with obsessive-compulsive symptoms relative to a corresponding full-length OBQ scale. In Study 2, using a large community sample (N = 507), the OBQ-20 scales generally correlated equivalently with obsessive-compulsive symptoms relative to both depression and generalized anxiety symptoms. The OBQ-20 demonstrated strong psychometric properties, but, like prior OBQ versions, did not emerge as specific to obsessive-compulsive symptoms. Given its brevity and potential transdiagnostic importance, the OBQ-20 will likely be a useful assessment tool in both research and clinical settings.

Keywords

Dysfunctional beliefsObsessive-compulsive disorder (OCD)Obsessive beliefs questionnaire (OBQ)Short form

Cognitive-behavioral models of obsessive-compulsive disorder (OCD) purport that dysfunctional beliefs are central to the phenomenology of the disorder (Clark 2004). The Obsessive Compulsive Cognitions Working Group (OCCWG) developed an 87-item self-report measure, known as the Obsessive Beliefs Questionnaire-87 (OBQ-87; OCCWG 2001), to assess six beliefs central to OCD. These beliefs are inflated responsibility, overestimation of threat, overimportance of thoughts, need to control thoughts, intolerance of uncertainty, and perfectionism. A subsequent factor analysis of the OBQ-87 by the OCCWG (2005) led to a refined conceptualization of these beliefs and a shortened, 44-item version of the measure (i.e., OBQ-44). Three beliefs are assessed by the OBQ-44, which include responsibility/threat, importance/control of thoughts, and perfectionism/certainty.

Despite its widespread use, the available literature highlights limitations of the OBQ that have precluded advancements in our understanding of the beliefs assessed by this self-report measure. One limitation relates to the factorial validity of the measure, as previous studies have failed to replicate the OCCWG’s (2001, 2005) OBQ factor structure (Myers et al. 2008; Wu and Carter 2008). Moulding et al. (2011) sought to improve the factorial validity of the OBQ by identifying a replicable factor structure of this measure (using the OBQ-44). Using a series of exploratory and confirmatory analyses with nonclinical samples, Moulding et al. identified a modified 38-item, 4-factor version of the OBQ-44 that provided a replicable solution across their samples. This 4-factor solution provided more favorable goodness-of-fit indices relative to alternative OBQ solutions (Myers et al. 2008; OCCWG 2005; Wu and Carter 2008). Moulding et al. labeled their 4-factors as threat (e.g., “Even when I am careful, I often think bad things will happen”), responsibility (e.g., “Even if harm is very unlikely, I should try to prevent it at any cost”), importance of thoughts (also assesses need to control thoughts; e.g., “Having intrusive thoughts means I’m out of control”), and perfectionism (also assesses intolerance of uncertainty; e.g., “For me, things are not right if they are not perfect”). A notable difference between Moulding et al.’s 4-factor solution and the OCCWG’s (2005) 3-factor solution is the presence of separate threat and responsibility factors in Moulding et al.’s solution. Based on the factors represented in their solution, Moulding et al. labeled their scale OBQ-TRIP.

Although Moulding et al.’s (2011) OBQ-TRIP seems to improve the factorial validity of the OBQ, a second limitation of the OBQ relates to its length. The OBQ was reduced to 38 items in the OBQ-TRIP, but Moulding et al. still opined that “some researchers may find our final 38-item solution to be too burdensome to routinely use in their studies” (p. 369). Because of this concern, Moulding et al. developed a 20-item OBQ short-form, labeled the OBQ-20. Moulding et al. identified 20 items as being a desirable length for a 4-factor measure and sought to retain five items from each OBQ-TRIP scale for inclusion in the OBQ-20. Moulding et al. selected the OBQ-20 items by maximizing internal consistency, as well as the correlation between each OBQ-20 scale and the corresponding OBQ-TRIP scale. Moulding et al. used this method to help ensure that each OBQ-20 scale adequately captured the variance of the corresponding OBQ-TRIP scale. The OBQ-20 has the same four factors as the OBQ-TRIP. Moulding et al. found that each OBQ-20 scale showed adequate internal consistency given its brevity (Cronbach’s αs  > .76), shared a correlation exceeding .93 with its OBQ-TRIP scale counterpart, and demonstrated a nearly identical correlation with obsessive-compulsive symptoms relative to the corresponding OBQ-TRIP scale (difference in magnitude of rs was  < .03). Based on these findings, the OBQ-20 warrants further study as a potential substitute for prior, longer OBQ versions.

The development of short-form self-report measures has garnered increased attention in the empirical literature. Peters et al. (2012) outlined a number of reasons why the use of short forms should be favored in both research and clinical settings. In research settings, Peters et al. noted that there generally exists interest in assessing a broad range of constructs, but it is also necessary to reduce the likelihood of respondent burden. Short forms allow researchers to achieve a balance of these two interests. Peters et al. further noted that clinicians benefit from the use of economical methods of assessment, stating that economical methods of assessment allow clinicians to more regularly monitor and screen for constructs of interest. These benefits would appear to extend to the OBQ, as the beliefs assessed by this measure are given central importance within existing cognitive-behavioral conceptual models and treatments of OCD (Clark 2004).

In the service of providing researchers and clinicians with a cost-effective assessment of the beliefs assessed by the OBQ, it is important to further examine the viability of Moulding et al.’s (2011) OBQ-20. One unaddressed empirical question surrounding the OBQ-20 relates to its psychometric properties among clinical respondents. Although Moulding et al.’s use of nonclinical samples when developing the OBQ-20 is appropriate given the dimensional nature of obsessive-compulsive symptoms (Olatunji et al. 2008), it is currently unknown whether the OBQ-20 and the OBQ-TRIP function similarly among individuals who consistently endorse high scores on the measures. Evidence of the comparability of these two versions of the OBQ in a clinical sample would further strengthen the notion that the OBQ-20 is a viable substitute for prior, longer versions of the OBQ.

If the OBQ-20 is found to evidence adequate psychometric properties among clinical respondents, it is possible that the use of the OBQ-20 might also address a third limitation of the OBQ evidenced in the existing literature. This limitation relates to the specificity of the beliefs assessed by the OBQ to obsessive-compulsive symptoms. Following from the cognitive-content specificity hypothesis, the OBQ beliefs are supposed to be of particular importance to only obsessive-compulsive symptoms (Clark 2004). Whereas a number of published studies have examined whether there exist significant correlations between the OBQ beliefs and various symptom types, only a few known published studies have completed statistical tests comparing the strength of correlations between the OBQ beliefs and symptom types. In these studies, Fergus and Wu (2010, 2011) found that the OBQ beliefs shared correlations with obsessive-compulsive symptoms that were statistically equivalent in magnitude relative to their correlations with mood and anxiety symptoms. However, these prior studies used the OBQ-44 and the original 3-factor solution put forth by the OCCWG (2005) when examining symptom-level specificity. Using this factor structure is potentially problematic because, as discussed, Moulding et al. (2011) found that the OBQ actually is best operationalized as a 4-factor construct consisting of separate threat and responsibility factors. It seems particularly important to target the threat belief in isolation from the responsibility belief when examining the specificity of the OBQ to obsessive-compulsive symptoms, as Moulding et al. found that the threat belief shares an especially strong relation with obsessive-compulsive symptoms. Because the OBQ-20 operationalizes the OBQ beliefs differently than does the OBQ-44, it is possible that use of the OBQ-20 can help lead to a better understanding of the specificity of the OBQ beliefs to obsessive-compulsive symptoms.

Prior studies interested in the symptom-level specificity of the OBQ beliefs used the total scale score when assessing obsessive-compulsive symptoms rather than separate symptom dimension scores (Fergus and Wu 2010, 2011). However, when using a recently developed self-report measure that assesses for the severity of four empirically-derived obsessive-compulsive symptom dimensions (i.e., Dimensional Obsessive Compulsive Scale [DOCS]; Abramowitz et al. 2010), researchers have found results consistent with the OBQ beliefs having symptom-level specificity with obsessive-compulsive symptoms. The four symptom dimensions assessed by the DOCS are contamination, responsibility for harm, unacceptable thoughts, and symmetry. Wheaton et al. (2010) found that the OBQ beliefs (using the OCCWG’s 2005, 3-factor OBQ-44 solution) tended to cluster strongest with certain obsessive-compulsive symptom dimensions. Specifically, Wheaton et al. found that the contamination and responsibility for harm symptom dimensions shared particularly robust relations with responsibility/threat, the unacceptable thoughts symptom dimension shared a particularly robust relation with the importance/control of thoughts, and the symmetry symptom dimension shared a particularly robust relation with perfectionism/certainty. These particular OBQ-symptom dimension correlations tended to be larger in magnitude than the correlations between the OBQ beliefs and depression symptoms. Unfortunately, tests directly comparing the magnitude of these correlations were not reported, which precludes firm conclusions as to the presence of symptom-level specificity. Nonetheless, Wheaton et al.’s results highlight the possibility that the symptom-level specificity of the OBQ beliefs might only be evidenced when operationalizing obsessive-compulsive symptoms using symptom dimension scores. With this possibility in mind, we examined whether the correlations between the OBQ-20 scales and the symptom dimensions of the DOCS were significantly stronger in magnitude than the correlations between the OBQ-20 scales and depression symptoms.

We also assessed for generalized anxiety symptoms when examining the specificity of the OBQ beliefs. Assessing for generalized anxiety symptoms is important because the OBQ beliefs share particularly strong correlations with this symptom type (Fergus and Wu 2010, 2011). However, prior studies examining relations between the OBQ beliefs and generalized anxiety symptoms only used a measure of worry. Assessing for generalized anxiety symptoms only via an index of worry does not allow for a comprehensive assessment of this symptom type (Gentes and Ruscio 2011). Gentes and Ruscio further noted that parallels between worry and obsessions might bias the magnitude of correlations between the OBQ beliefs and generalized anxiety symptoms. To date, no known published study has addressed this limitation via examining whether the observed relevance of the OBQ beliefs to generalized anxiety symptoms is replicated when using a comprehensive measure of generalized anxiety symptoms.

In sum, in this two-part study, we examined the psychometric properties of the OBQ-20 in a clinical sample in Study 1 and the specificity of the OBQ-20 to obsessive-compulsive symptoms in a community sample in Study 2. Given the noted dimensional nature of obsessive-compulsive symptoms, in which differences are considered quantitative rather than qualitative in nature (Olatunji et al. 2008), we expected that the pattern of results obtained using Moulding et al.’s (2011) nonclinical samples would be similar to the results obtained in our clinical sample in Study 1. Our clinical sample consisted of individuals diagnosed with OCD, as well as individuals diagnosed with mood and anxiety disorders. The decision to include respondents with a mood or anxiety disorder was informed by prior findings that individuals with mood or anxiety disorders often endorse high levels of the beliefs assessed by the OBQ (Belloch et al. 2010). Our Study 1 data analytic approach followed the approach used by Moulding et al. (2011) when comparing the OBQ-20 and the OBQ-TRIP. We predicted that each OBQ-20 scale would (a) demonstrate adequate internal consistency (Cronbach’s αs  > .75), (b) share a correlation exceeding  > .90 with its corresponding OBQ-TRIP scale, and (c) share a nearly identical correlation with obsessive-compulsive symptoms relative to its corresponding OBQ-TRIP scale (difference in magnitude of rs of  < .03). Finally, although the size of our Study 1 sample (N = 48) precluded an examination of the OBQ-20 factor structure, we examined correlations among the OBQ-20 scales to investigate the distinctiveness of these scales in a clinical sample. Finding relatively moderate intercorrelations among the OBQ-20 scales would offer indirect support for the structural distinctiveness of the OBQ-20 scales. Moulding et al. did not present correlations among the OBQ-20 scales, but we expected that the OBQ-20 scale intercorrelations would be consistent with those found by Moulding et al. among the OBQ-TRIP scales (rs ranging from .49 to 64).

In Study 2, we compared the strength of correlations between the OBQ-20 scales and symptom types. Following Fergus and Wu’s (2010, 2011) findings, we predicted the OBQ-20 scales would appear common to the targeted symptom types when obsessive-compulsive symptoms were operationalized using a total scale score. Following Wheaton et al.’s (2010) findings, when obsessive-compulsive symptoms were operationalized as symptom dimensions, we predicted that the OBQ-20 importance of thoughts scale would cluster with the unacceptable thoughts symptom dimension, the OBQ-20 perfectionism scale would cluster with the symmetry symptom dimension, and both the OBQ-20 threat and responsibility scales would cluster with the contamination and responsibility for harm symptom dimensions.

Study 1

Method

Participants

Participants were 48 patients presenting for treatment in an outpatient specialty clinic in stress and anxiety disorders under the direction of the second author. The primary diagnosis for these patients was OCD, a mood disorder, or an anxiety disorder, and it was unaccompanied by diagnoses of psychotic disorder or an active (untreated) substance use disorder. Diagnoses were based upon the Anxiety Disorders Interview Schedule for DSM-IV (ADIS-IV; Brown et al. 1994). The ADIS-IV was administered by master or doctoral level clinicians who had prior experience conducting the interview.

Twenty-six (54.2 %) participants received only one diagnosis. These diagnoses were OCD (n = 9, 34.6 %); social phobia (n = 8, 30.8 %); generalized anxiety disorder (n = 4, 15.4 %); panic disorder (n = 3, 11.5 %); and dysthymia (n = 2, 7.7 %). Twenty-two (45.8 %) participants received more than one diagnosis. These participants had two or more of the following diagnoses: social phobia (n = 15, 68.2 %); OCD (n = 11, 50.0 %); specific phobia (n = 10, 45.5 %); panic disorder (n = 8, 36.4 %); dysthymia (n = 8, 36.4 %); major depression (n = 4, 18.2 %); generalized anxiety disorder (n = 3, 13.6 %); and posttraumatic stress disorder (n = 3, 13.6 %). The mean age of the sample was 36.2 (SD = 13.7) years and the sample consisted of an equal gender representation. The majority of participants self-identified as Caucasian (n = 41, 85.4 %) and reported either working at least part-time (n = 24, 50.0 %) or being a student (n = 11, 22.9 %).

Measures

Obsessive Beliefs Questionnaire-TRIP (OBQ-TRIP; Moulding et al. 2011)

As described, the OBQ-TRIP is a 38-item measure that assesses the four beliefs already introduced: threat (9 items), responsibility (6 items), importance of thoughts (12 items), and perfectionism (11 items). Items are rated using a 7-point scale ranging from 1 (disagree very much) to 7(agree very much). Moulding et al.’s OBQ-20 consists of four separate 5-item scales that assess for each of the four beliefs using the same ordered-category scale. Moulding et al. found that the OBQ-TRIP and the OBQ-20 both shared moderate correlations with obsessive-compulsive symptoms (OBQ-TRIP: rs ranging from .30 to .55; OBQ-20: rs ranging from .30 to .54).

Padua Inventory-Washington State University Revision (PI-WSUR; Burns et al. 1996)

The PI-WSUR is a 39-item revision of the original measure that assesses the severity of obsessive-compulsive symptoms. The symptoms assessed by the PI-WSUR include checking (e.g., “I tend to keep on checking things more often than necessary”), contamination (e.g., “I feel my handy are dirty when I touch money”), grooming (e.g., “I feel obliged to follow a particular order in dressing, undressing, and washing myself”), harm thoughts (e.g., “I think or worry at length about having hurt someone without knowing it”), and harm impulses (e.g., “When I see a train approaching, I sometimes think I could throw myself under its wheels”). Items are rated using a 5-point scale ranging from 0 (not at all) to 4 (very much). The PI-WSUR has shown a moderate convergent correlation with other obsessive-compulsive symptom measures (e.g., r = .56; Anholt et al. 2009). The PI-WSUR was specifically chosen to assess obsessive-compulsive symptoms in the present research because this measure was used by Moulding et al. (2011) in their initial investigation of the OBQ-20. We operationalized the PI-WSUR as a total scale score because of limitations assessing the four empirically established obsessive-compulsive symptom dimensions when using this measure (Abramowitz et al. 2010).

Procedure

Study 1 was approved by the institutional review board at the second author’s affiliated university. Participants were recruited from an outpatient specialty clinic. Participants completed pencil-and-paper questionnaires at the time of their initial assessments, as well as the ADIS-IV to determine their eligibility for participation. As described, eligibility for participation included participants receiving a primary diagnosis of OCD, a mood disorder, or an anxiety disorder. Moreover, participants could not meet criteria for a psychotic disorder or an active (untreated) substance use disorder. Participants were informed that their responses were confidential and they were free to withdraw from the study without penalty.

Results

Descriptive Statistics and Zero-Order Correlations

Descriptive statistics and correlations among the study variables are presented in Table 1. As predicted, each OBQ-20 scale demonstrated good internal consistency (all αs  > .84) and shared a correlation exceeding .90 with its OBQ-TRIP counterpart (rs ranged from .93 to .99, ps  < .01). Each OBQ-20 scale shared a significant moderate to strong correlation with the PI-WSUR. Importantly, the magnitude of correlations between each OBQ-20 scale and the PI-WSUR was nearly identical to the correlation between the corresponding OBQ-TRIP scale and the PI-WSUR (difference in magnitude of rs ranged from .00 to .04). Of note, tests of dependent correlations (Meng et al. 1992) revealed that each OBQ-20 scale shared a statistically equivalent correlation with obsessive-compulsive symptoms relative to the corresponding OBQ-TRIP scale (magnitude of z-statistics ranged from 0.00 to 0.89, ns). However, given our small sample size, these tests might have been overly conservative. Finally, as expected, the OBQ-20 scales shared moderate to strong intercorrelations (rs ranging from .49 to 75, ps  < .01).
Table 1

Correlations among study 1 variables

Variable

Mean

SD

1

2

3

4

5

6

7

8

9

1. OBQ-TRIP-Threat

28.47

14.28

(.91)

        

2. OBQ-TRIP-Responsibility

20.77

10.27

.73

(.88)

       

3. OBQ-TRIP-Importance of Thoughts

30.40

16.68

.72

.69

(.92)

      

4. OBQ-TRIP-Perfectionism

44.66

19.23

.77

.59

.60

(.93)

     

5. OBQ-20-Threat

15.06

8.24

.96

.74

.76

.73

(.86)

    

6. OBQ-20-Responsibility

16.40

8.74

.74

.99

.71

.57

.75

(.86)

   

7. OBQ-20-Importance of Thoughts

14.80

8.40

.69

.68

.93

.58

.71

.68

(.86)

  

8. OBQ-20-Perfectionism

19.02

8.49

.73

.51

.57

.97

.69

.49

.54

(.85)

 

9. PI-WSUR

25.24

22.60

.74

.62

.60

.49

.74

.63

.56

.48

(.94)

N = 48. All rs significant at p < .01 (two-tailed). Cronbach’s α values listed in parentheses along the diagonal

OBQ obsessive beliefs questionnaire, PI-WSUR Padua inventory-Washington State University Revision

Study 1 Discussion

There is a need for short forms of commonly assessed constructs. Moulding et al. (2011) developed an OBQ short-form (i.e., OBQ-20) that assesses for beliefs purportedly central to OCD. However, the comparability of Moulding et al.’s OBQ-20 to prior, longer versions of the OBQ had yet to be investigated in a sample of respondents who consistently evidence high scores on the measures. Addressing this gap in the literature, we examined the convergence of Moulding et al.’s OBQ-20 to their OBQ-TRIP using a clinical sample. Our results paralleled those obtained by Moulding et al., as each OBQ-20 scale evidenced (a) good internal consistency, (b) a robust correlation with the full-length counterpart, and (c) a nearly identical zero-order correlation in magnitude with obsessive-compulsive symptoms relative to its full-length counterpart. Although our small sample size precluded an examination of the OBQ-20 factor structure, our findings that the OBQ-20 scales shared intercorrelations that were comparable in magnitude to the intercorrelations among the OBQ-TRIP scales found by Moulding et al. using a nonclinical sample provide indirect support for the structural distinctiveness of the OBQ-20 scales among clinical respondents. Overall, and taken with results from Moulding et al., Study 1 results indicate that Moulding et al.’s OBQ-20 appears to be a viable substitute for prior, longer versions of the OBQ.

Study 2

Given that the OBQ-20 was found to evidence adequate psychometric properties among clinical respondents in Study 1, it is possible that the use of the OBQ-20 might address the lack of specificity of prior versions of the OBQ to obsessive-compulsive symptoms. More precisely, symptom-level specificity of the OBQ beliefs might be found when using the unique 4-factor solution of Moulding et al.’s (2011) OBQ-20. In Study 2, we examined whether the OBQ-20 scales correlated significantly more strongly with obsessive-compulsive symptoms than either depression or generalized anxiety symptoms.

In these analyses, we expected to replicate prior studies (Fergus and Wu 2010, 2011) and find that the OBQ-20 scales correlated statistically equivalently with the symptom types when obsessive-compulsive symptoms were operationalized as a total scale score. Following Wheaton et al. (2010), we further predicted that evidence of specificity would emerge when obsessive-compulsive symptoms were operationalized as separate symptom dimensions. Based on Wheaton et al.’s results, we predicted that the OBQ-20 importance of thoughts scale would correlate significantly stronger with the unacceptable thoughts symptom dimension and the OBQ-20 perfectionism scale would correlate significantly stronger with the symmetry symptom dimension relative to both depression and generalized anxiety symptoms. We further predicted that the OBQ-20 threat and responsibility scales would correlate significantly more strongly with the contamination and responsibility for harm symptom dimensions relative to both depression and generalized anxiety symptoms. Exploratory analyses were then completed to compare the magnitude of correlations between the OBQ-20 scales and each obsessive-compulsive symptom dimension in comparison to the other three obsessive-compulsive symptom dimensions.

Method

Participants

The sample consisted of 507 adults recruited through the Internet. The mean age of the sample was 33.1 years (SD = 11.3; ranging from 18 to 71) and respondents predominantly self-identified as female (54.6 %). In terms of racial/ethnic identification, 79.1 % of the sample self-identified as Caucasian, 6.5 % as African American, 6.3 % as Asian, 3.7 % as Latino, 3.6 % as bi-or multi-racial, and 0.8 % as “Other.” In terms of educational background, the majority of the sample reported receiving a two-year college degree or higher (57.3 %), being currently employed at least part-time (68.1 %), and as currently non-married (65.7 %). In addition to completing the OBQ-20 (Moulding et al. 2011), participants completed the following measures.

Measures

The Dimensional Obsessive-Compulsive Scale (DOCS; Abramowitz et al. 2010)

The DOCS is a 20-item measure that assesses the severity of four obsessive-compulsive symptom dimensions over the past month. The four DOCS scales are contamination, responsibility for harm, unacceptable thoughts, and symmetry. Items are rated using a 5-point scale ranging from 0 to 4, with the response options differing based on the item. Each DOCS scale assesses for the time spent (ranging from none at all to 8 h or more each day), avoidance (ranging from none at all to extreme avoidance of nearly all things), distress (ranging from not at all distressed/anxious to extremely distressed/anxious), interference (ranging from no disruption at all to my life is completely disrupted and I cannot function at all), and attempts of control (not at all difficult to extremely difficult) surrounding the respective symptom dimension. Each DOCS scale shares moderate to strong correlations (rs ranging from .39 to .88) with measures assessing the corresponding symptom dimension (Abramowitz et al. 2010).

Depression, Anxiety, and Stress Scale-21-item Version (DASS-21; Lovibond and Lovibond 1995)

The DASS-21 is a 21-item measure that assesses depression, anxiety, and stress symptoms over the past week using three, 7-item scales. Responses are provided on a 4-point scale ranging from 0 (did not apply to me at all) to 3 (applied to me very much, or most of the time). The depression scale (e.g., “I couldn’t seem to experience any positive feeling at all”; I “felt down-hearted and blue”) – the DASS-21 scale of interest in the present research – shares a strong convergent correlation (r = .79) with another measure of depression symptoms (Antony et al. 1998).

Generalized Anxiety Disorder-7 Item scale (GAD-7; Spitzer et al. 2006)

The GAD-7 is 7-item measure that assesses generalized anxiety symptoms (e.g., “Feeling nervous, anxious or on edge”; “Worrying too much about different things”) over the past 2 weeks using a 4-point scale ranging from 0 (not at all) to 3 (nearly every day). The GAD-7 shares a strong convergent correlation (r = .64) with another measure of generalized anxiety symptoms (Kertz et al. 2013).

Procedure

Participants were recruited using Amazon’s Mechanical Turk (MTurk), an Internet-based platform that allows individuals to request the completion of jobs (e.g., survey completion) for monetary compensation. Respondents completing surveys through MTurk have been found to produce high quality data (Buhrmester et al. 2011). The use of an Internet sample in the present research is supported by the equivalency of the OBQ scales across paper-and-pencil versus Internet administration (Coles et al. 2006). Study 2 was approved by the institutional review board affiliated with the first author’s university. Recruitment was limited to MTurk workers over 18 years of age and located in the United States. Participants were required to provide electronic consent and there was no penalty for withdrawing from the study. Upon completion of the study, participants were debriefed and paid in full. Compensation was $1, an amount consistent with the compensation given to MTurk workers completing prior studies of similar length (Buhrmester et al. 2011).

Results

Planned Comparisons of Symptom-Level Specificity

Correlations among the study variables are presented in Table 2. Tests of dependent correlations (Meng et al. 1992) were used to test whether the OBQ-20 scales correlated significantly strongest with certain symptom types (obsessive-compulsive versus either depression or generalized anxiety). The increased possibility of Type I error associated with completing multiple planned tests was addressed using the False Discovery Rate (FDR; Benjamini and Hochberg 1995). Results from the FDR indicated that a familywise alpha level of p < .015 (two-tailed) should be used when completing the following reported comparisons.
Table 2

Correlations among study 2 variables

Variable

Mean

SD

1

2

3

4

5

6

7

8

9

10

11

1. OBQ-20-Threat

16.21

7.07

(.86)

          

2. OBQ-20-Responsibility

20.34

7.07

.53

(.85)

         

3. OBQ-20-Importance of Thoughts

12.96

6.79

.55

.44

(.87)

        

4. OBQ-20-Perfectionism

18.53

7.24

.71

.53

.54

(.83)

       

5. DOCS-Contamination

2.69

3.24

.31a

.20a

.30

.28

(.86)

      

6. DOCS-Responsibility for Harm

3.44

3.74

.46a

.31a

.35

.38

.50

(.91)

     

7. DOCS-Unacceptable Thoughts

3.54

4.05

.43

.20

.41a

.36

.38

.50

(.91)

    

8. DOCS-Symmetry

3.06

3.86

.37

.24

.41

.41a

.44

.53

.40

(.93)

   

9. DOCS-Total

12.73

11.47

.52a

.31a

.48a

.47a

.73

.82

.76

.77

(.93)

  

10. DASS-21-Depression

6.01

5.51

.55b

.29b

.33b

.51b

.22

.40

.46

.32

.46

(.92)

 

11. Generalized Anxiety Disorder-7

7.01

5.33

.57b

.31b

.35b

.52b

.27

.46

.48

.40

.53

.69

(.88)

N = 507. All rs significant at p < .01 (two-tailed). Cronbach’s αs listed in parentheses along the diagonal. Column counterparts with different superscripts compared via planned comparisons

OBQ obsessive beliefs questionnaire, DOCS dimensional obsessive-compulsive scale, DASS-21 depression, anxiety, stress scale-21-item version

Our predictions were largely supported when operationalizing obsessive-compulsive symptoms as a total scale score. Specifically, the threat, responsibility, and perfection scales of the OBQ-20 all shared statistically equivalent correlations with the DOCS total scale score relative to either the DASS-21-Depression or the GAD-7 (magnitude of z-statistics ranged from 0.00 to 1.46, ns). Contrary to expectations, the importance of thoughts scale of the OBQ-20 correlated significantly more strongly with the DOCS total scale score relative to either the DASS-21-Depression (magnitude of z-statistic = 3.65, p < .015) or the GAD-7 (magnitude of z-statistic = 3.40, p < .015).

Contrary to expectations, specificity was not evidenced when operationalizing obsessive-compulsive symptoms as separate symptom dimensions. In fact, the threat scale of the OBQ-20 correlated significantly more strongly with both the DASS-21-Depression and the GAD-7 relative to DOCS-Contamination (magnitude of z-statistics = 5.04 and 5.68, ps  < .015). The threat scale of the OBQ-20 also correlated significantly more strongly with the GAD-7 (magnitude of z-statistic = 2.92, p < .015), but not the DASS-21-Depression (magnitude of z-statistic = 2.26, ns), relative to DOCS-Responsibility for Harm. Results further revealed that the responsibility scale of the OBQ-20 correlated significantly equivalently with both the DASS-21-Depression and the GAD-7 relative to DOCS-Contamination (magnitude of z-statistics = 1.69 and 2.14, ns) and DOCS-Responsibility for Harm (magnitude of z-statistics = 0.44 and 0.00, ns). The importance of thoughts belief scale of the OBQ-20 correlated statistically equivalently with both the DASS-21-Depression (magnitude of z-statistic = 1.90, ns) and the GAD-7 (magnitude of z-statistic = 1.46, ns) relative to DOCS-Unacceptable Thoughts. Finally, the perfectionism scale of the OBQ-20 correlated significantly more strongly with the GAD-7 (magnitude of z-statistic = 2.66, p < .015), but not the DASS-21-Depression (magnitude of z-statistic = 2.28, ns), relative to DOCS-Symmetry.

Exploratory Comparisons of Symptom-Level Specificity

Exploratory comparisons were completed to examine whether any of the OBQ-20 scales correlated more strongly with certain obsessive-compulsive symptom dimensions than others. Because of their exploratory nature, an alpha level of p < .05 (two-tailed) was used. These exploratory comparisons are presented in Table 3. The threat scale of the OBQ-20 generally correlated equivalently with the scales of the DOCS with the following exceptions. The threat scale of the OBQ-20 correlated significantly more strongly with DOCS-Responsibility for Harm relative to DOCS-Contamination (magnitude of z-statistic = 3.74, p < .05) and DOCS-Symmetry (magnitude of z-statistic = 2.35, p < .05). The threat scale of the OBQ-20 also correlated significantly more strongly with DOCS-Unacceptable Thoughts relative to DOCS-Contamination (magnitude of z-statistic = 2.67, p < .05). The remaining comparisons among the correlations between the threat scale of the OBQ-20 and the DOCS scales were statistically equivalent in magnitude (magnitude of z-statistic ranged from 0.78 to 1.38, ns).
Table 3

Exploratory comparisons of specificity between OBQ-20 scales and obsessive-compulsive symptom dimensions

 

OBQ-20 Scale

Variable

T

R

I

P

DOCS-Contamination

.31c

.20b

.30b

.28b

DOCS-Responsibility for Harm

.46a

.31a

.35a,b

.38a

DOCS-Unacceptable Thoughts

.43a,b

.20b

.41a

.36a,b

DOCS-Symmetry

.37b,c

.24a,b

.41a

.41a

N = 507

Column counterparts with different superscripts significantly different (p < .05) via exploratory comparisons

OBQ obsessive beliefs questionnaire, T threat, R responsibility, I importance of thoughts, P perfectionism, DOCS dimensional obsessive-compulsive scale

The responsibility scale of the OBQ-20 generally correlated equivalently with the scales of the DOCS with the following exceptions. The responsibility scale of the OBQ-20 correlated significantly more strongly with DOCS-Responsibility for Harm relative to DOCS-Contamination (magnitude of z-statistic = 2.58, p < .05) and DOCS-Unacceptable Thoughts (magnitude of z-statistic = 2.58, p < .05). The remaining comparisons among the correlations between the responsibility scale of the OBQ-20 and the DOCS scales were statistically equivalent in magnitude (magnitude of z-statistic ranged from 0.00 to 1.70, ns).

The importance of thoughts scale of the OBQ-20 generally correlated equivalently with the scales of the DOCS with the following exceptions. The importance of thoughts scale of the OBQ-20 correlated significantly more strongly with DOCS-Unacceptable Thoughts relative to DOCS-Contamination (magnitude of z-statistic = 2.43, p < .05). The importance of thoughts scale of the OBQ-20 correlated significantly more strongly with DOCS-Symmetry relative to DOCS-Contamination (magnitude of z-statistic = 2.55, p < .05). The remaining comparisons among the correlations between the importance of thoughts scale of the OBQ-20 and the DOCS scales were statistically equivalent in magnitude (magnitude of z-statistic ranged from 0.00 to 1.53, ns).

The perfectionism scale of the OBQ-20 generally correlated equivalently with the scales of the DOCS with the following exceptions. The perfectionism scale of the OBQ-20 correlated significantly more strongly with DOCS-Responsibility for Harm relative to DOCS-Contamination (magnitude of z-statistic = 2.42, p < .05). The perfectionism scale of the OBQ-20 correlated significantly more strongly with DOCS-Symmetry relative to DOCS-Contamination (magnitude of z-statistic = 3.00, p < .05). The remaining comparisons among the correlations between the perfectionism scale of the OBQ-20 and the DOCS scales were statistically equivalent in magnitude (magnitude of z-statistic ranged from 0.49 to 1.73, ns).

Study 2 Discussion

Replicating prior studies (Fergus and Wu 2010, 2011), the OBQ-20 scales generally appeared common to all symptom types when operationalizing obsessive-compulsive symptoms as a total scale score. The specificity of the importance of thoughts OBQ belief to obsessive-compulsive symptoms (total score) found in Study 2 has been inconsistently identified in the extant literature (Fergus and Wu 2010, 2011). Measurement choice could be one reason for these inconsistent findings, as prior studies used a measure that asks about the occurrence of specific types of obsessions and compulsions. We used Abramowitz et al.’s (2010) DOCS to assess obsessive-compulsive symptoms in Study 2. The DOCS provides an assessment of the severity of obsessive-compulsive symptoms that is independent from the occurrence of specific types of obsessions and compulsions. Based on these considerations, we believe that the present results are more likely to generalize across studies than the results of prior studies using alternative measures of obsessive-compulsive symptoms.

Contrary to expectations, symptom-level specificity of the OBQ-20 scales was not evidenced when examining the separate obsessive-compulsive symptom dimensions. Other studies have found that certain OBQ beliefs cluster strongest with certain obsessive-compulsive symptom dimensions (Wheaton et al. 2010). However, these studies used previous versions of the OBQ and did not statistically compare the strength correlations. Consistent with findings operationalizing obsessive-compulsive symptoms as a total scale score, the OBQ-20 scales tended to correlate equivalently in magnitude with the obsessive-compulsive symptom dimensions relative to either depression or generalized anxiety symptoms. Moreover, there was not a clear pattern of results when examining whether the OBQ-20 scales correlated strongest with certain obsessive-compulsive symptom dimensions relative to other obsessive-compulsive symptom dimensions. In particular, each OBQ-20 scale appeared broadly relevant to each obsessive-compulsive symptom dimension.

General Discussion

The OBQ-20 demonstrated strong psychometric properties and appears to be a viable substitute for prior, longer OBQ versions. Nonetheless, the lack of symptom-level specificity of the OBQ beliefs to obsessive-compulsive symptoms appears to be a relatively robust finding that spans across versions of the OBQ and different measures of obsessive-compulsive, depression, and generalized anxiety symptoms. There are a number of potential explanations for the lack of symptom-level specificity of the OBQ beliefs. One potential explanation is that the targeted beliefs are not adequately operationalized by the OBQ items. Although possible, it should be noted that prior studies have found that the OBQ beliefs (using the OBQ-44) prospectively predict obsessive-compulsive symptoms (Abramowitz et al. 2006). This finding provides strong empirical support that the beliefs as assessed by the OBQ items are important to the phenomenology of OCD. Another potential explanation for the lack of specificity of the OBQ beliefs was put forth by Tolin et al. (2006). These researchers asserted that the OBQ beliefs could be related to psychopathology in general rather than OCD per se. Tolin et al. based this conclusion on findings that the OBQ beliefs did not differentiate between an OCD patient group and anxious controls after controlling for the depression and non-specific anxiety symptoms. The commonality of nearly all of the OBQ-20 scales to obsessive-compulsive, depression, and generalized anxiety symptoms in the present research provides further evidence for such a possibility.

There is growing interest in identifying variables that span across psychopathology (i.e., transdiagnostic variables; Mansell et al. 2008). Nolen-Hoeksema and Watkins (2011) developed a heuristic for understanding how transdiagnostic variables might serve as both general risk factors for multiple symptom types (i.e., multifinality) and as specific risk factors for certain symptom types (i.e., divergent trajectories). In particular, Nolen-Hoeksema and Watkins purport that there exist moderating variables that determine whether a transdiagnostic variable leads to certain symptom types. For example, coupled with dysfunctions in the basal ganglia, Nolen-Hoeksema and Watkins suggested that a transdiagnostic variable might lead to obsessive-compulsive symptoms. Alternatively, coupled with experiences of loss, rejection, or failure, the same transdiagnostic variable might lead to depressive symptoms. If an individual has dysfunctions in the basal ganglia and experiences loss, rejection, or failure, the transdiagnostic variable could in turn lead to both obsessive-compulsive and depressive symptoms. Although these examples are overly simplistic explanations for symptom development, they highlight the importance in shifting the empirical focus from examining the main effects of the OBQ beliefs to examining interactive effects. Adopting such an empirical focus can elucidate under what conditions the OBQ beliefs relate to obsessive-compulsive symptoms versus other symptom types (e.g., depression, generalized anxiety).

The potential transdiagnostic importance of beliefs thought to be specific to OCD might also be associated with other important conceptual implications. Diagnostically, OCD is no longer classified as an anxiety disorder (American Psychiatric Association, APA 2013). Despite this reclassification, OCD still appears to share psychological processes (e.g., dysfunctional beliefs) with anxiety disorders. These processes appear common to depressive disorders as well. As such, symptoms of different classes of disorders appear to be marked by similar psychological processes in the form of the OBQ beliefs. Given the APA’s stated interest in identifying and assessing cross-cutting variables that span across diagnoses, future research should seek to shed further light onto the potential transdiagnostic utility of the OBQ beliefs.

Results from the present research suggest that Moulding et al.’s (2011) OBQ-20 will likely be a useful assessment tool for explicating the potential transdiagnostic importance of the OBQ beliefs. The OBQ-20 seems to function similarly to other versions of the measure (the OBQ-TRIP in the present research) and, importantly, is at least approximately half the length of prior versions of the measure. Using the OBQ-20 would thus allow researchers to include the measure as part of longer questionnaire batteries or research protocols. The relative brevity of the OBQ-20 also offers clinicians the opportunity to more regularly monitor and screen for beliefs that might span across diagnoses. The OBQ-20 also allows for the separate assessment of threat and responsibility beliefs, which, as discussed, seems to be an important practice to follow. For these reasons, the OBQ-20 might be considered a preferred version of the OBQ for use by researchers and clinicians going forward.

The OBQ-20 scales generally functioned similarly across the two studies, with a few exceptions. One difference across studies was in relation to the pattern of correlations between responsibility and obsessive-compulsive symptoms. Relative to the other OBQ-20 scales, responsibility consistently shared the lowest association with obsessive-compulsive symptoms in Study 2. Of note, this pattern of findings is consistent with those found by Moulding et al. (2011). However, in Study 1, the association between responsibility and obsessive-compulsive symptoms was relatively strong in comparison to the other OBQ-20 scales. Measurement choice is unlikely the reason for these inconsistent findings, as the DOCS was used in Study 2 and Moulding et al. used the PI-WSUR. Sample differences might help account for these findings, with responsibility appearing to show weaker relations with obsessive-compulsive symptoms in nonclinical (Moulding et al. 2011; Study 2) relative to clinical (Study 1) samples when compared to the other OBQ-20 scales.

Another difference across studies relates to the pattern of relations among the OBQ-20 scales. In Study 1, the OBQ-20 scales generally shared similar interrelations in magnitude, but, in Study 2, the threat and perfectionism scales clustered particularly strongly with one another. Unfortunately, as noted, Moulding et al. (2011) did not present interrelations among the OBQ-20 scales and so it is unclear whether this finding is consistent among nonclinical respondents. Of note, using the OBQ-44, Wu and Carter (2008) found that a number of perfectionism/certainty items loaded on the same factor as responsibility/threat items in a nonclinical sample. As discussed, one important difference in OBQ factor structures between the OBQ-44 and OBQ-20 is that the OBQ-20 has separate threat and responsibility factors. Taken with Wu and Carter’s findings, one possibility is that threat and perfectionism share especially strong relations among nonclinical respondents. Because the OBQ-20 separately assesses responsibility and threat, this version of the measure appears well-positioned to help disentangle how these two beliefs relate to obsessive-compulsive symptoms, as well as the other OBQ beliefs, in future studies.

Limitations surrounding the present research must be acknowledged. First, although use of a clinical sample in Study 1 was one of the strengths of the present study, our clinical sample was only modest in size. Although difficult data to obtain, it will be important for future studies to replicate the present findings using larger clinical samples, which would allow for an examination of the OBQ-20 factor structure among clinical respondents. However, it should be noted that the consistent pattern of findings from Study 1 and those obtained by Moulding et al. (2011) using large nonclinical samples supports the generalizability of our Study 1 results. Further, reliability and severity rating data were not available for diagnoses in Study 1. Although analyses at the diagnostic-level were not a part of the present research, it might be useful for future research to extend prior findings examining the specificity of the OBQ scales at the diagnostic-level (e.g., Tolin et al. 2006) using the OBQ-20 scales. Our Study 1 sample was also marked by diagnostic heterogeneity. Replicating the present results among only individuals diagnosed with OCD seems warranted. Fourth, across both studies, relations among our study variables were likely inflated as a result of our monomethod assessment. Fifth, although the OBQ beliefs are conceptualized as causally impacting symptoms, the cross-sectional nature of these data precludes causal conclusions. Finally, it remains important to examine the commonality of the OBQ beliefs to other symptom types related to OCD, such as symptoms of eating disorders and hoarding. This line of research will help further speak to the potential transdiagnostic importance of the OBQ beliefs.

Limitations notwithstanding, the present results further support the OBQ-20 as being a psychometrically sound version of the OBQ for use by researchers and clinicians. The lack of symptom-level specificity of the OBQ-20 scales observed in the present research suggests that this measure might be of interest to a broader range of researchers and clinicians than previously thought. Because the OBQ beliefs appear relatively common to multiple symptom types, it is recommended that researchers begin to seek to identify under what conditions these beliefs relate to certain symptom types. Future studies identifying moderators of associations between the OBQ beliefs and symptoms will further our conceptualization and treatment of symptoms with these cognitive underpinnings. The OBQ-20 appears to be an assessment tool that will aid researchers in this pursuit.

Copyright information

© Springer Science+Business Media New York 2013