Journal of Anxiety Disorders 26 (2012) 359–367

Contents lists available at SciVerse ScienceDirect

Journal of Anxiety Disorders

Psychometric properties of the Repetitive Thinking Questionnaire in a clinical sample Alison E.J. Mahoney a,∗ , Peter M. McEvoy b,c , Michelle L. Moulds d a Clinical Research Unit for Anxiety and Depression, University of New South Wales at St Vincent’s Hospital, Level 4 O’Brien Centre, 394-404 Victoria Street, Darlinghurst, Sydney, New South Wales, 2010, Australia b Centre for Clinical Interventions, 223 James Street, Northbridge, Perth, Western Australia, 6003, Australia c School of Psychology, University of Western Australia, 35 Stirling Hwy, Crawley, Western Australia, 6009, Australia d School of Psychology, University of New South Wales, Sydney, New South Wales, 2052, Australia

a r t i c l e

i n f o

Article history: Received 9 May 2011 Received in revised form 14 September 2011 Accepted 3 December 2011 Keywords: Repetitive negative thinking Transdiagnostic Psychometrics Rumination Worry

a b s t r a c t Repetitive negative thinking (RNT) is thought to contribute to the maintenance of many emotional disorders. Although several measures of RNT are available, the items of most of these instruments index RNT that is specific to particular diagnostic groups (e.g., RNT about depression symptoms). This has limited our ability to examine the relevance of RNT across diagnostic groups and advance our understanding of RNT as a transdiagnostic process. This study evaluated the psychometric properties of the Repetitive Thinking Questionnaire (RTQ), a transdiagnostic measure of RNT. In a clinical sample of individuals with anxiety and depressive disorders (N = 186), the RTQ demonstrated good internal consistency, convergent, and divergent validity. Supporting the transdiagnostic nature of the measure, the Repetitive Negative Thinking subscale of the RTQ was associated with a variety of negative emotions and metacognitive beliefs, and significantly predicted symptoms of multiple disorders when controlling for neuroticism. Our findings support the use of the RTQ as a transdiagnostic, trans-emotional measure of maladaptive repetitive thought following distressing events, with scope to increase efficiency and reduce burden on patients by assessing RNT in clinical settings with one short measure. Experimental and longitudinal research identifying mechanisms driving RNT using the RTQ would be informative for theory and treatment developments. © 2011 Elsevier Ltd. All rights reserved.

1. Introduction Repetitive negative thinking (RNT) involves attentive, perseverative, frequent, and relatively uncontrolled cognitive activity that is focused on negative aspects of the self and the world (Ehring & Watkins, 2008; Segerstorm, Stanton, Alden, & Shortridge, 2003). RNT has been conceptualized as a transdiagnostic process, and as such, is relevant to our understanding of the development and maintenance of multiple emotional disorders (Ehring & Watkins, 2008; Harvey, Watkins, Mansell, & Shafran, 2004). The construct of RNT encompasses many processes including depressive rumination (Nolen-Hoeksema & Morrow, 1991), worry (Borkovec, Ray, & Stober, 1998), post-event processing in social phobia (Clark & Wells, 1995; Rapee & Heimberg, 1997), perseverative cognition (Brosschot, Gerin, & Thayer, 2006), habitual negative self-thinking

∗ Corresponding author. Tel.: +61 2 8382 1407; fax: +61 2 8382 1402. E-mail addresses: [email protected] (A.E.J. Mahoney), [email protected] (P.M. McEvoy), [email protected] (M.L. Moulds). 0887-6185/$ – see front matter © 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.janxdis.2011.12.003

(Verplanken, Friborg, Wang, Trafirmow, & Woolf, 2007), and stressreactive rumination (Robinson & Alloy, 2003). Elevated levels of RNT have been observed across a large range of disorders including generalized anxiety disorder (GAD), depressive disorders, posttraumatic stress disorder (PTSD), social phobia, obsessive complusive disorder (OCD), eating disorders, pain disorder, insomnia, and hypochondriasis (Ehring & Watkins, 2008; Harvey et al., 2004). Importantly, experimental studies have found causal relationships between RNT and symptoms of emotional disorders. For example, experimentally induced rumination has been shown to exacerbate depressed mood (Nolen-Hoeksema & Morrow, 1993; Park, Goodyer, & Teasdale, 2004) and PTSD symptoms (Ehring, Fuchs, & Klasserner, 2009; Zetsche, Ehring, & Ehlers, 2009), as well as maintain symptoms of social phobia and anxiety (Wong & Moulds, 2009) and other negative mood states (e.g., anger; Blagden & Craske, 1996). Also, inducing worry has been shown to increase anxiety, depression, and negative affect, while reducing positive affect (McLaughlin, Borkovec, & Sibrava, 2007). Previous research has sought to understand differences across the various forms of RNT. First, temporal orientation may vary across forms of RNT. Depressive rumination appears to be more

360

A.E.J. Mahoney et al. / Journal of Anxiety Disorders 26 (2012) 359–367

past-focused whereas worry is more future-focused (Papageorgiou & Wells, 1999; Watkins, Moulds, & Mackintosh, 2005). Second, the content of various forms of RNT also varies because it often reflects diagnosis-specific concerns and symptoms (Ehring & Watkins, 2008). For example, depressive rumination may involve dwelling on difficulties associated with low mood, poor concentration, and anhedonia (Nolen-Hoeksema & Morrow, 1991). In contrast, the post-event rumination that is characteristic of individuals with social phobia or PTSD is more likely to involve thoughts about distressing social experiences or thoughts about the meaning and consequences of traumatic events, respectively (Abbott & Rapee, 2004; Ehlers & Clark, 2000). Interestingly, when differences in temporal orientation are controlled for, and diagnosis-specific information is removed from measures of RNT, items on measures of worry, rumination, and post-event processing load on one generic RNT factor (McEvoy, Mahoney, & Moulds, 2010). Most of the work regarding the similarities between different forms of RNT has been limited to depressive rumination and worry. Responses on measures of worry and rumination correlate with one another and with measures of anxiety and depression (de Jong-Meyer, Beck, & Riede, 2009; Segerstorm, Tsao, Alden, & Craske, 2000; Watkins, 2004). Worry and rumination also share many characteristics (e.g., frequency, duration, intrusiveness, and controllability) and are associated with similar emotions (Watkins et al., 2005). Individuals also tend to appraise and respond to rumination and worry in similar ways (e.g., with distraction or reassurance seeking, Papageorgiou & Wells, 1999; Watkins, 2004; Watkins et al., 2005). Moreover, worry and rumination are both associated with constructs that have been shown to perpetuate emotional disorders, such as intolerance of uncertainty (de JongMeyer et al., 2009), metacognitive beliefs (Papageorgiou & Wells, 2001; Watkins & Moulds, 2005; Wells, 1995), thought control strategies (Watkins & Moulds, 2009), and avoidance (Borkovec, Alcaine, & Behaar, 2004; Moulds, Kandris, Starr, & Wong, 2007). In sum, there appear to be more similarities than differences in the various forms of RNT. However, the lack of generic or transdiagnostic measures of RNT has limited researchers’ capacity to examine the relationships between RNT and a broad array of emotions and symptoms of psychopathology. Measures of RNT typically contain disorderspecific content or reference specific forms of RNT that are pertinent to particular disorders (e.g., every item in the Penn State Worry Questionnaire refers to worry, the core diagnostic feature of GAD; Meyer, Miller, Metzger, & Borkovec, 1990). Generic RNT measures are needed to enable research that will progress our understanding of the core similarities and differences in RNT across disorders, as well as clarify the interplay of RNT and other important constructs that maintain psychopathology. The development of such instruments will in turn have direct implications for transdiagnostic models and treatments. With these issues in mind, two transdiagnostic measures of RNT have recently been developed (Ehring et al., 2011; McEvoy et al., 2010). This study sought to extend the limited body of research in this area by evaluating the psychometric properties of one such generic RNT measure, the Repetitive Thinking Questionnaire (McEvoy et al., 2010), in a clinical sample. The Repetitive Thinking Questionnaire (RTQ) was developed by combining items from commonly used measures of worry, rumination, and post-event processing, and then modifying the items to remove diagnosis-specific content. In order to standardize temporal orientation, respondents complete the items in reference to a recent distressing event. In a student sample, exploratory factor analysis was used to derive the 31-item RTQ which consisted of two subscales labeled Repetitive Negative Thinking (RNT) and Absence of Repetitive Thinking (ART). The RNT subscale demonstrated high internal consistency and was associated with symptoms of anxiety

and depression, and a range of negative emotions. Consistent with cognitive models that seek to explain the maintenance of RNT (e.g., Self-Regulatory Executive Function model, Wells & Matthews, 1996; the avoidance theory of worry, Borkovec et al., 2004), the RNT subscale was significantly associated with a number of constructs thought to perpetuate rumination and worry, namely metacognitive beliefs, cognitive avoidance strategies, and maladaptive thought control strategies. An abbreviated 10-item version of the RNT subscale demonstrated similar psychometric properties to the full-scale version (McEvoy et al., 2010). In contrast, the Absence of Repetitive Thinking subscale did not correlate well with symptoms of anxiety and depression, negative emotions, or constructs thought to perpetuate RNT. While these initial findings suggest that the RTQ is a promising tool with which to measure RNT, these findings now need to be replicated and extended in a clinical sample. The 15-item Perseverative Thinking Questionnaire (PTQ; Ehring et al., 2011) is another measure of repetitive negative thinking. Like the RTQ, the items of the PTQ exclude diagnosis-specific content. However, its construction is based on a set of proposed core characteristics and features of RNT rather than being derived from existing measures of RNT like the RTQ. The PTQ’s factor structure, internal consistency, temporal stability, convergent validity, and predictive validity have been demonstrated in clinical and nonclinical samples (Ehring et al., 2011). The measure is trait-like in nature, whereas the RTQ assesses RNT in reference to a recent distressing event. This places the RTQ in the unique position to assess RNT following a naturally occurring event, and has facilitated the exploration of the relationships between RNT and a range of transient negative emotions as well as persistent symptoms of anxiety and depression. Further research is required to examine the relationships between RNT and a broader array of negative emotions, diagnoses, and symptoms of psychopathology in order to elucidate the trans-emotional and transdiagnostic nature of RNT in clinical populations. To this end, we sought to evaluate the validity and reliability of the RTQ in a mixed clinical sample. We hypothesized that the RTQ would be internally consistent and that RNT mean scores would be greater than those reported in a student sample (McEvoy et al., 2010). We also hypothesized that the RNT scale would demonstrate acceptable convergent validity as indicated by positive relationships with measures of: (a) negative emotion, (b) symptoms of various anxiety and depressive disorders, (c) neuroticism, and (d) metacognitive beliefs. We also sought to evaluate the utility of specific diagnoses (GAD, social phobia, OCD, panic disorder with or without agoraphobia, and depression) in predicting RNT. Consistent with the transdiagnostic nature of the RTQ, we predicted that the RNT scales would be associated with multiple diagnoses as well as the degree of comorbidity in the sample. Moreover, we predicted that RNT would demonstrate good discriminate validity as shown by a lack of significant relationships with measures of alcohol use and extraversion. Our last prediction was that the 10-item RNT scale would demonstrate comparable psychometric properties to the full scale version. 2. Material and methods 2.1. Participants Participants (N = 186, 51.6% women) were recruited from a specialist anxiety disorders treatment service. Participants had a mean age of 35.54 years (SD = 11.66, range = 17–68) and 78.5% had completed high school. Regarding relationship status, 34.6% reported that they were married or in de facto relationships, 55.1% were never married, 9.7% separated or divorced, and .5% widowed. Prior to treatment, participants completed the Anxiety Disorders Interview Schedule for DSM-IV (Brown, DiNardo, & Barlow, 1994).

A.E.J. Mahoney et al. / Journal of Anxiety Disorders 26 (2012) 359–367

The principal diagnoses of the sample included social phobia (42.9%), generalized anxiety disorder (19.0%), panic disorder with or without agoraphobia (22.3%), obsessive compulsive disorder (6.5%), major depressive disorder (4.3%), dysthymic disorder (1.1%), specific phobia (1.6%), posttraumatic stress disorder (1.6%), and somatization disorder (.5%). Comorbidity was common in the sample; 33.9% met diagnostic criteria for two disorders, 28.0% reported three disorders, 10.8% reported four disorders, and 5.9% reported symptoms consistent with five or more disorders. Comorbid diagnoses included GAD (28.0%), social phobia (19.4%), OCD (4.3%), panic disorder with or without agoraphobia (9.1%), depressive disorder (i.e., major depressive disorder and/or dysthymia, 40.3%), specific phobia (16.7%), alcohol use disorder (11.3%), and drug use disorder (2.7%). 2.2. Measures 2.2.1. Anxiety Disorders Interview Schedule for DSM-IV (ADIS-IV) The ADIS-IV (Brown et al., 1994) is a structured diagnostic interview for the anxiety, mood, somatoform, and substance use disorders according to criteria in the Diagnostic and Statistical Manual (DSM-IV; American Psychiatric Association, 1994). Brown, Di Nardo, Lehman, and Campbell (2001) provided evidence of acceptable inter-rater reliability for the anxiety disorders investigated in the present study ( = .59–.79). Inter-rater reliability ( = .63) for the combined depressive disorders group (major depressive disorder and dysthymia) was also acceptable. Evidence of construct validity, including discriminant and convergent validity, has been demonstrated (Brown, Chorpita, & Barlow, 1998). Diagnosticians in this study were four clinical psychologists and six psychiatric registrars. Training involved: (a) thorough reading of the ADIS-IV protocol, (b) observation of an experienced interviewer conducting an ADIS-IV, and (c) administration of an ADIS-IV while being observed by an experienced interviewer. After the training interviews, diagnosticians compared and reviewed diagnoses. All clinicians had extensive experience in the assessment and treatment of internalizing disorders. Principal diagnoses were determined collaboratively by asking clinicians and participants to choose which disorder was the most distressing and life-interfering disorder at the time of interview. 2.2.2. Repetitive Thinking Questionnaire The RTQ (McEvoy et al., 2010) is a 31-item self-report measure of transdiagnostic perseverative thinking. Items were drawn from the Penn State Worry Questionnaire (Meyer et al., 1990), the Ruminative Responses Scale of the Response Styles Questionnaire (Nolen-Hoeksema & Morrow, 1991) and the Post-Event Processing Questionnaire-Revised (McEvoy & Kingsep, 2006), and were modified to remove diagnosis-specific content. The questionnaire comprises two subscales; Repetitive Negative Thinking (27 items) and Absence of Repetitive Thinking (4 items). Items are rated along a 5-point scale: Not at all true (1), Somewhat true (3), or Very true (5). As described previously, we have demonstrated the factor structure, internal consistency (˛ = .72–.93), convergent validity, and predictive utility of the RTQ in a student sample (McEvoy et al., 2010). As in the McEvoy et al. (2010) study, temporal orientation was standardized by asking participants to complete the RTQ items in reference to their experience following a recently occurring distressing event. Participants specified this event at the start of the questionnaire. The RNT-10 is a short version of the RNT scale comprised the 10 highest loading items on the RNT factor. The short scale was highly correlated with the full scale (r = .95, p < .001) and demonstrated almost identical psychometric properties (McEvoy et al., 2010).

361

2.2.3. Positive and Negative Affect Scale The Negative Affect subscale (10 items) of the Positive and Negative Affect Scale (PANAS, Watson, Clarke, & Tellegen, 1998) was used to index the range and strength of negative emotions experienced by participants during distressing situations that were referenced in the RTQ. The Positive Affect subscale was not administered. The negative emotions include: distressed, upset, guilty, scared, hostile, irritable, ashamed, nervous, jittery, and afraid, and are rated on a 5-point response scale; Very slightly or not at all (1), A little (2), Moderately (3), Quite a bit (4) or Extremely (5). Crawford and Henry (2004) have provided evidence of the tool’s internal consistency (˛ = .85–.89) and construct validity (including convergent and divergent validity). In order to be consistent with McEvoy et al. (2010), we examined associations between the RTQ and PANAS total score, as well as four subscales derived from a factor analysis in the original study: Anxiety (˛ = .84) included the items nervous, afraid, scared, and jittery. Anger (˛ = .79) included the items irritable and hostile. Shame (˛ = .73) included the items guilty and ashamed. General distress (˛ = .68) included the items distressed and upset. Average scale scores were calculated for each of these factors.

2.2.4. Metacognitions Questionnaire-30 The MCQ-30 (Wells & Cartwright-Hatton, 2004) measures five domains of metacognitive beliefs, metacognitive monitoring, and judgments of cognitive confidence. The five subscales are: (1) positive beliefs about worry, (2) negative beliefs about uncontrollability and danger, (3) cognitive confidence (assessing confidence in attention and memory), (4) negative beliefs concerning the consequences of not controlling thoughts, and (5) cognitive self-consciousness (the tendency to focus attention on thought processes). Wells and Cartwright-Hatton (2004) reported good internal consistency (r = .72–.92) and test–retest reliability (r = .59–.87) across the total score and subscales, as well as providing evidence of the MCQ-30’s factor structure and convergent validity. The 4-point response scale is: Do not agree (1), Agree slightly (2), Agree moderately (3), or Agree very much (4). As in the RTQ, references to worry were replaced by references to thoughts and thinking (e.g., ‘Worrying about my problems helps me cope’ became ‘Thinking about my problems helps me cope’). Cronbach’s alphas ranged from .80 to .92 in this study.

2.2.5. Positive Beliefs about Rumination Scale – Adapted (PBRS-A) This 9-item scale was adapted from the Positive Beliefs about Rumination Scale developed by Papageorgiou and Wells (2001). The PBRS-A assesses the degree to which respondents hold beliefs about the helpfulness of repetitive thinking. The PBRS-A replaces references to depression and rumination in the original PBRS with the generic descriptors of feeling and thinking. For example, ‘I need to ruminate about my problems to find answers to my depression’ became ‘I need to think about things to find answers to how I feel’. Evidence of the PBRS-A’s internal consistency (˛ = .89), convergent validity, and discriminant validity has been reported (Watkins & Moulds, 2005). The 4-point response scale is identical to the MCQ30. Internal consistency in the current study was .84.

2.2.6. Eysenck Personality Questionnaire (EPQ) The 23-item neuroticism subscale (EPQ-N) and 21-item extraversion subscale (EPQ-E) of the EPQ (Eysenck & Eysenck, 1975) were used. Internal consistency (˛ = .82 for both subscales; Loo, 1979) and test–retest reliability (r = .82 and .92 over 1 month for neuroticism and extraversion, respectively; Eysenck & Eysenck, 1975) are good, and data demonstrating construct validity are extensive (e.g., Barrett, Petrides, Eysenck, & Eysenck, 1998; Caruso, Witkiewitz, Belcourt-Dittloff, & Gottlieb, 2001; Steele & Kelly,

362

A.E.J. Mahoney et al. / Journal of Anxiety Disorders 26 (2012) 359–367

1976). In this study, internal consistencies were ˛ = .84 and .78 for neuroticism and extraversion, respectively. 2.2.7. Body Sensations Questionnaire (BSQ) and Agoraphobic Cognitions Questionnaire (ACQ) The 17-item BSQ and 14-item ACQ (Chambless, Caputo, Bright, & Gallagher, 1984) are established measures of panic disorder and agoraphobia symptoms. The scales measure physical sensations and thoughts respondents typically experience when they are nervous or frightened. Internal consistency is good (˛ = .80 and .87 for ACQ and BSQ respectively), and evidence of temporal stability (r = .86 and .67 for ACQ and BSQ respectively over 31 days) and construct validity (including convergent and discriminant validity) has been provided (Chambless, Beck, Gracely, & Grisham, 2000; Chambless et al., 1984; Chambless & Gracely, 1989). Internal consistencies found in this study were ˛ = .86 (ACQ) and ˛ = .94 (BSQ). 2.2.8. Beck Depression Inventory (BDI-II) The BDI-II (Beck, Steer, & Brown, 1996) is a 21-item measure of depression symptoms experienced during the previous fortnight. Internal consistency (˛ = .92) and test–retest reliability (r = .93 over 1 week) are established (Beck et al., 1996), and evidence for construct validity has been demonstrated (e.g., Dozois, Dobson, & Ahnberg, 1998; Osman, Kopper, Barrios, Gutierrez, & Bagge, 2004). Support for convergent and discriminant validity has also been reported (Osman et al., 1997; Steer, Ball, Ranieri, & Beck, 1997). In this study, internal reliability was ˛ = .94. 2.2.9. Penn State Worry Questionnaire (PSWQ) The PSWQ (Meyer et al., 1990) is a 16-item measure of worry with good internal consistency (˛ = .86–.95) and temporal stability (r = .92 over 8–10 weeks and r = .74–.93 over 4 weeks; Meyer et al., 1990; Molina & Borkovec, 1994). Evidence of construct validity, including discriminant and convergent validity, has been demonstrated in clinical and community populations (e.g., Brown, Antony, & Barlow, 1992; Meyer et al., 1990; van Rijsoort, Emmelkamp, & Vervaeke, 1999). Internal consistency was ˛ = .73 in this sample. 2.2.10. Padua Inventory – Washington State University Revision (PI) The PI (Burns, 1995) is a widely used 39-item self-report measure of OCD symptoms (e.g., ‘I feel my hands are dirty when I touch money’). Evidence for convergent and discriminant validity, as well as factor structure has been demonstrated (Burns, Keortge, Formea, & Sternberger, 1996; Jónsdóttir & Smári, 2000). Internal consistency (˛ = .92) and temporal stability (r = .61–.84 across subscales over 6–7 months) are good (Burns et al., 1996). In this study, the total score was used and internal reliability was high (˛ = .93). 2.2.11. Social Phobia Scale (SPS) and Social Interaction Anxiety Scale (SIAS) The SPS and SIAS (Mattick & Clarke, 1998) are widely used 20-item measures of performance and interaction anxiety, respectively. The SPS describes situations in which the person is the focus of attention and observed by others, such as eating and writing. The SIAS contains items reflecting cognitive, affective, and behavioral reactions to interactional situations, such as nervousness when speaking to people in authority. Mattick and Clarke (1998) provide evidence of good internal reliability (SPS ˛ = .89; SIAS ˛ = .93), test–retest reliability (r = .91 and .92 for the SPS and SIAS respectively across 3–5 weeks), convergent validity and discriminant validity. In this sample internal reliability was good (SPS ˛ = .95; SIAS ˛ = .87).

2.2.12. Alcohol Use Disorders Identification Test The 10-item AUDIT (Saunders, Aasland, Bbor, de le Fuente, & Grant, 1993) is a widely used screening measure which identifies hazardous and harmful alcohol consumption. Evidence of internal consistency (˛ range = .75–.94) and convergent and discriminant validity is extensive (Allen, Litten, Fertig, & Babor, 1997). Internal consistency was ˛ = .89 in this sample. 2.3. Procedure Prior to treatment at a specialist anxiety disorders clinic, participants completed the ADIS-IV and a series of questionnaires which formed part of the clinic’s standard intake battery. Questionnaires included the RTQ, PANAS, MCQ-30, PBSR-A, EPQ-N, EPQ-E, BSQ, ACQ, BDI-II, PSWQ, PI, SAIS, SPS, and AUDIT. Participants consented for their data to be used for research purposes and the use of the data was approved by the Hospital’s Human Research Ethics Committee. No participant refused to participate. 3. Results 3.1. Internal consistencies and descriptive statistics of the RTQ Scales Cronbach’s alpha suggested that internal consistency was excellent for the Repetitive Negative Thinking scale (RNT, ˛ = .88, average inter-item correlation = .38) but moderate for the ART scale (˛ = .62, average inter-item correlation = .29). The correlation between the RNT and ART scales was low to moderate (Pearson’s r = −.33, p < .001). Mean total scores (SDs) of the RNT and ART scales were 86.96 (23.87) and 7.14 (2.99), respectively. Average scores were 3.22 (.88) and 1.78 (.75), respectively. Note that the RNT mean was higher than that reported in a student sample in our previous study (M = 71.97, SD = 22.02), whereas the ART mean was lower than the student mean (M = 8.41, SD = 3.37; McEvoy et al., 2010). Independent samples t-tests directly comparing the original student data published in McEvoy et al. (2010) and the clinical sample used in this study demonstrated that the differences were highly statistically significant for both the RNT scale, t(458) = 6.92, p < .001, d = .65, and the ART scale, t(464) = -4.15, p < .001, d = −.63. Independent samples t-tests were conducted to compare scores for males and females in the current sample, but failed to find a significant difference for the RNT, t(179) = −.37, p = .71, d = −.06, or ART, t(181) = .01, p = .99, d =