In psychopathology, avoidance was first studied in the anxiety disorders, in particular in the motor behaviors of avoidance of phobic stimuli. Authors such as Barlow (Reference Barlow2002) see avoidance as an essential characteristic for describing, diagnosing and treating these disorders. With respect to depressive disorders, Ferster (Reference Ferster1973) also acknowledged the central role that avoidance plays in his depression theory, a role that was later reiterated in Behavioral Activation (Jacobson, Martell, & Dimidjian, Reference Jacobson, Martell and Dimidjian2001). The relevance of avoidance study is also expanding into other areas of psychopathology, such as psychotic disorders (Vilardaga, Hayes, Atkins, Bresee, & Kambiz, Reference Vilardaga, Hayes, Atkins, Bresee and Kambiz2013) or posttraumatic stress disorder (Bordieri, Tull, McDermott, & Gratz, Reference Bordieri, Tull, McDermott and Gratz2014).
According to Blalock and Joiner (Reference Blalock and Joiner2000) longitudinal studies suggest that coping strategies characterized by avoiding stressful situations and using passive cognitive responses to life´s negative events lead to more depressed mood. Conversely, the use of strategies characterized by proactive efforts to manage stressful events can result in lower levels of depression. Also, the relationship between avoidance coping and depression has been studied by Ottenbreit and Dobson (Reference Ottenbreit and Dobson2004) using their Cognitive-Behavioral Avoidance Scale (CBAS), which not only examines the dimensions of cognitive and behavioral avoidance, but also of social/nonsocial and active/passive avoidance. These authors suggest that passivity is an implicit characteristic of avoidance coping.
Ottenbreit and Dobson (Reference Ottenbreit and Dobson2004) suggest that the functional analysis of depression, as first proposed by Ferster (Reference Ferster1973), focuses on avoidance as a coping strategy. Thus, depressed individuals tend to isolate themselves, avoid stressful situations, and take little action when faced with a stressful event. Coping avoidance thoughts and behaviors include avoiding thinking about goals or solutions, putting off decisions, turning down opportunities, isolating oneself, and giving up on tasks before they are completed. The functional perspective of depression proposes that inactivity, avoidance and isolation interfere with the possibilities of exposing oneself to success or reinforcement, because they prevent the individual from searching for and testing solutions to problems.
Moreover, Kashdan, Barrios, Forsyth, and Steger (Reference Kashdan, Barrios, Forsyth and Steger2006) have documented the effects of avoidance behavior not only on depression but on psychopathology in general. Although most of these studies were conducted on non-clinical samples, some results point to the relevance of avoidance in the clinical population. For example, Barajas (Reference Barajas2015) has found a significantly higher level of avoidance in patients with anxiety and depression than in the general population when using an experimental task. Likewise, Barlow (Reference Barlow2002) proposed that many emotional disorders are related to efforts to avoid unexpected or excessive emotional experiences. Such experiences occur frequently, not only in the clinical population but in the general population as well. Barlow found that 68% of a sample of 300 students had experienced at least one unexpected emotional experience in the last three months. He further found that the distress associated with such experiences triggered efforts to manage them using suppression or other forms of avoidance, resulting in the consequences described above.
Several avoidance types can be described. Behavioral avoidance includes putting off decisions, turning down opportunities, isolating oneself, and giving up on tasks before they are completed. Cognitive avoidance refers to a wide variety of coping strategies such as rumination, intentionally trying to suppress thoughts, and attempting to dissociate from negative experiences. Hayes, Wilson, Gifford, Follette, and Strosahl (Reference Hayes, Wilson, Gifford, Follette and Strosahl1996) took a different approach, described experiential avoidance as the phenomenon that occurs when a person is unwilling to remain in contact with particular private experiences and takes steps to alter the form or frequency of these events and the contexts that occasion them. Along these lines, Kashdan et al. (Reference Kashdan, Barrios, Forsyth and Steger2006) noted that everyone experiences moments of pain and suffering, which are a part of the total spectrum of human emotions. These include potentially intense moments of distress, such as panic attacks, and a range of evaluative thoughts, such as doubts about oneself and about one’s ability to handle a given situation. The form and content of these events are part of being human and living in the present moment, and are not in themselves problematic or dysfunctional. Moreover, working toward valuable goals requires confronting a wide range of emotional experiences, many of them painful. According to Hayes et al. (Reference Hayes, Wilson, Gifford, Follette and Strosahl1996), such events become pathological when avoidance persists even when it is useless or life distorting. In this sense, it has been hypothesized that mindfulness implies a contact with external and internal events just as they are and increases willingness to tolerate uncomfortable emotions and sensations (Levitt, Brown, Orsillo, & Barlow, Reference Levitt, Brown, Orsillo and Barlow2004). On the contrary, strategies as avoidance can hinder the extinction of emotional responses. Some studies have found a direct correlation between experiential avoidance and symptoms associated with most psychological disorders, as well as an inverse correlation between experiential avoidance and health and quality of life measurements (Hayes, Luoma, Bond, Masuda, & Lillis, Reference Hayes, Luoma, Bond, Masuda and Lillis2006). In addition, on using the CBAS it has been found an inverse correlation between cognitive-behavioral avoidance and mindfulness, both in clinical population as in general population (Barajas & Garra, Reference Barajas and Garra2014).
Although experiential avoidance and cognitive-behavioral avoidance seem to be related and could be different dimensions of the same problem, there are no studies analyzing such relation because those concepts come from different theoretical approaches. Moulds, Kandris, Starr, and Wong (Reference Moulds, Kandris, Starr and Wong2007) proposed that subsequent studies would benefit by using experiential avoidance measures such as AAQ to explore the relationship between experiential avoidance and cognitive-behavioral avoidance. They also pointed out the need to validate the CBAS with clinical samples. We opted to explore both ideas in our study.
The present paper describes several aims: to adapt the CBAS in a Spanish sample made up of clinical patients, general population and undergraduate students, to test differences in avoidance between the three samples, and to assess the factorial structure and psychometric properties of the Spanish version. The scale was hypothesized to show positive correlations with the AAQ-II as well as the depression and anxiety measures, and to show negative correlation with the mindfulness scale. Lastly, it was hypothesized that the clinical patients would show more cognitive-behavioral avoidance than the general population and the undergraduate students.
Method
Participants
Clinical group: 100 patients meeting the Diagnostic and Statistical Manual of Mental Disorders (4th Ed., Text Rev.; American Psychiatric Association, 2000) criteria for an anxiety disorder (n = 50), mood disorder (n = 39), or borderline personality disorder (BPD; n = 11). The first subgroup included panic disorder with agoraphobia, agoraphobia without history of panic disorder, social phobia, obsessive-compulsive disorder, and generalized anxiety disorder. The second group included major depressive disorder and dysthymia. The BPD subgroup, although not part of the previous categories, is included because the anxious-depressive symptomatology is markedly present.
Control group: 100 subjects with no prior history of mental disorders, who were not taking psychoactive drugs, and who had not seen a psychologist or psychiatrist for any reason in the last two years. These subjects were recruited through advertising and through the “snowball” method by the researchers and other participants.
Undergraduate group: 54 undergraduate students across different academic years and areas of study (Nursing, Speech Therapy and Occupational Therapy).
Participants’ age and gender is as follows: control group (M = 36.21, SD = 11.95; 71 females, 29 males), undergraduate group (M = 20.43, SD = 2.1; 42 females, 12 males), clinical group (M = 35.39, SD = 10.74; 58 females, 42 males). All participants were aged between 18 and 60.
Procedure
Authorization for the study was requested and subsequently granted by the Clinical Research Ethics Committee of the University Hospital of Guadalajara (Spain). All subjects signed an informed consent release before participating in the study. The SCID-I (First, Spitzer, Gibbon, & Williams, Reference First, Spitzer, Gibbon and Williams1999) was conducted to confirm the appropriate diagnosis for the clinical sample and to screen for mental disorders in the control group. All instruments were applied to clinical and control groups by first author in his office at cited Hospital. Additionally, instruments were applied to undergraduate group by the second author in a lecture room at University of Castilla-La Mancha. The evaluation instruments were administered according to the random order established.
The CBAS was translated into Spanish using the back translation technique. For this technique, one translation team translates the scale into the target language, and then the second team translates it back into the original language. The fidelity of the translation is judged by how closely it matches the original version. As Muñiz and Hambleton (Reference Muñiz and Hambleton1996) point out, this popular method is not without its flaws. They suggest that the translators not only know both languages, but also be familiar with both cultures. These recommendations were taken into account in the selection of translators for the study. The translators worked independently of one another, and no relevant differences were found in the way the items were expressed. The authors subsequently reached a consensus on both versions with the translators. Lastly, an English studies professor and other psychologists tweaked several items in order to make them more understandable to the general population. Efforts were made to ensure that the length of the items was equivalent to the original scale.
Although Hernández-Guzmán et al. (Reference Hernández-Guzmán, Dobson, Caso-Niebla, Gónzalez-Montesinos, Epp, Arratíbel-Siles and Wierzbicka-Szymczak2009) conducted a study with a Spanish version of the CBAS on a sample of Mexican undergraduate students, the published paper included only a summary of the items, rather than the complete scale used in the study. As it was impossible to access the complete version, we opted to translate and adapt the original English-language scale (see Appendix) developed by Ottenbreit and Dobson (Reference Ottenbreit and Dobson2004), which was kindly provided to us by second author.
Measures
Structured Clinical Interview for DSM-IV Axis I Disorders (SCID-I; First et al., Reference First, Spitzer, Gibbon and Williams1999)
The objective of this semi-structured interview is to make the major DSM-IV-TR axis I diagnoses. The clinician version (SCID-CV), a later adaptation that was fine-tuned for use in clinical settings, was used for this study. The decision to use the SCID-CV is based on its widespread use in those publications that seek to add rigor when establishing diagnoses.
Cognitive-Behavioral Avoidance Scale (CBAS; Ottenbreit & Dobson, Reference Ottenbreit and Dobson2004)
The CBAS is a multidimensional measure of avoidance. It consists of 31 items which are evaluated using a 5-point Likert scale. The higher the score on the scale, the higher the level of avoidance. The total score ranges from 31 to 155. The factor analysis of the CBAS carried out by the original authors found four factors that accounted for 44.95% of the variance: Behavioral Social (BS) (e.g., “I avoid attending social activities”), made up of eight items and accounting for 27.54%; Cognitive Nonsocial (CN) (e.g., “I try not to think about my future and what I will do with my life”), made up of ten items and accounting for 7.87%; Cognitive Social (CS) (e.g., “I try not to think about problems in my personal relationships”), made up of seven items and accounting for 5.30%; and Behavioral Nonsocial (BN) (e.g., “I quit activities that challenge me too much”), made up of six items and accounting for 4.24%. Ottenbreit and Dobson (Reference Ottenbreit and Dobson2004) initially proposed several dimensions as important in the conceptualization of the avoidance construct. These included cognitive/behavioral, active/passive and social/nonsocial avoidance. Nevertheless, the authors found, contrary to their hypothesis, that the active versus passive dimension was not important in the empirical analysis of the CBAS. These authors did not suggest cut-offs. Ottenbreit and Dobson (Reference Ottenbreit and Dobson2004) conducted their study of the CBAS construction with 391 undergraduate students. They found correlations between the four subscales ranging from .39 to .57, suggesting that they were measuring related constructs. They also found elevated correlations between the scores of the subscales and the total score: .78–.80. The Cronbach’s alpha for the total scale was .91, indicating a high degree of internal consistency. For the four subscales, the Cronbach’s alpha ranged from .75 to .86.
Acceptance and Action Questionnaire-II (AAQ-II; Bond et al., Reference Bond, Hayes, Baer, Carpenter, Guenole, Orcutt and Zettle2011)
The AAQ-II evaluates the constructs of acceptance, experiential avoidance, and psychological inflexibility. A 7-point Likert scale is used to measure responses. Higher scores indicate greater experiential avoidance. There are two versions of the AAQ-II, one consisting of 7 items and the other of 10; we used the second version. Bond et al. (Reference Bond, Hayes, Baer, Carpenter, Guenole, Orcutt and Zettle2011) affirm that the translations and empirical studies conducted using the 10-item version are valid, given that they found a correlation of .96 between the versions of 7 and 10 items. Ruiz, Langer Herrera, Luciano, Cangas, and Beltrán (Reference Ruiz, Langer Herrera, Luciano, Cangas and Beltrán2013) recently published the Spanish version of the AAQ-II, and concurred with Bond et al. (Reference Bond, Hayes, Baer, Carpenter, Guenole, Orcutt and Zettle2011) that the 10-item version is not significantly weaker than the 7-item version. In our study, the mean score was 44.65 (SD = 10.84) for the clinical group and 24.08 (SD = 6.74) for the control group.
Mindful Attention Awareness Scale (MAAS; Brown & Ryan, Reference Brown and Ryan2003)
This scale focusses on the presence or absence of attention and awareness of what is occurring in the present moment. It consists of 15 items which are evaluated using a 6-point Likert scale. Higher scores indicate greater mindfulness. The MAAS has a single-factor structure and can be applied to both clinical and general population. Researchers have reported satisfactory psychometric properties to date. Our study used the Spanish version of MAAS adapted by Barajas and Garra (Reference Barajas and Garra2014), which yielded a high degree of internal consistency (α = .88). The mean score of the clinical group was 4.64 (SD = 0.61) and the mean score of the control group was 3.57 (SD = 0.77).
Beck Depression Inventory-II (BDI-II; Beck, Steer, & Brown, Reference Beck, Steer and Brown1996)
The BDI-II is a highly regarded self-report inventory used to detect and measure the severity of depression. It was included in this study because it is widely used, and also because of its outstanding psychometric properties. The Sanz, Navarro, and Vázquez (Reference Sanz, Navarro and Vázquez2003) version of the instrument was used in this study. The mean score of the clinical group was 21.37 (SD = 10.38) and the mean score of the control group was 4.50 (SD = 3.9).
Beck Anxiety Inventory (BAI; Beck & Steer, Reference Beck and Steer1993)
The BAI is one of the most widely used self-report inventories for evaluating anxiety symptomatology. This study used the BAI version developed by Sanz and Navarro (Reference Sanz and Navarro2003). In the present study, the mean score for the clinical group was 21.41 (SD = 11.37), while the mean score for the control group was 4.19 (SD = 3.7).
Data analysis
The demographic characteristics of the groups differed significantly, which could have resulted in potential confounding factors if they were also associated with the CBAS score. It was necessary to assess the potential confounding factors using the following analyses (see Table 1): Spearman´s Rho for age, the Mann-Whitney U and Wilcoxon W tests for gender, and the Chi-squared and Kruskal-Wallis tests for marital status, education, socioeconomic status and occupation. Adjusted p-values for age, gender, education, socioeconomic status and occupation have been provided. The distribution of the CBAS variable was found to be highly asymmetric. The Kolmogorov-Smirnov test yielded a value of p < .001 for both the total scale and the four subscales, (Total-CBAS: z = 2.41; CBAS-BS: z = 2.75; CBAS-BN: z = 2.12; CBAS-CS: z = 2.06; CBAS-CN: z = 2.51), which indicates that these variables were not normally distributed. In particular, the distribution was asymmetric to the right, with the lowest scores appearing most frequently, and the higher scores less frequently (gamma distribution). Specifically, the Total-CBAS variable presented values that ranged from 32 to 146 (M = 62.72, SD = 22.9; N = 254), with some positive skewness (1.07; S.E. = 0.153). The quartiles were: 45; 55.5 and 75.25. As a result, ANOVA could not be performed. Generalized linear models (GLM) with log link function were used instead. The statistical analyses were carried out with SPSS version 17.
Potential confounding factors analyses
Notes: BS (behavioral social), BN (behavioral nonsocial), CS (cognitive social), CN (cognitive nonsocial)
** p < .01 *p < .05
Additionally, to examine the validity of CBAS, Confirmatory Factor Analysis (CFA) was also performed. CFA was conducted using IBM-SPSS AMOS-19. In CFA, maximum likelihood was used to estimate all model parameters. As multivariate normality assumption was not met, bootstrapping technique was used as it does not rely on assumption of multivariate normality (Byrne, Reference Byrne2010). The number of bootstrap samples for this study was set at 2000 samples. In order to evaluate the model fit the comparative fit index (CFI; Hu & Bentler, Reference Hu and Bentler1999), a root-mean-square error of approximation (RMSEA; Browne & Cudeck, Reference Browne, Cudek, Bollen and Long1993) and a standardized root mean square residual (SRMR) were used. According to Bentler (Reference Bentler1992), CFI values greater than .90 are indicative of an acceptable fit. As regards RMSEA, values that are less than .05 represent a good fit and values up to .08 represent a reasonable fit (Byrne & Campbell, Reference Byrne and Campbell1999). For SRMR, Byrne (Reference Byrne2010) suggests a cut-off point of < .08. Regarding χ2, as the significance of the χ2 statistic is influenced by the sample size, some authors suggest the use of χ2/df ratio as a better measure of the goodness-of-fit of the overall model (Byrne, Reference Byrne2010). Tabachnick and Fidell (Reference Tabachnick and Fidell2001) recommend a χ2/df ratio < 2 for well- fitting models. Potential model improvements were assessed by inspecting standardized residuals, modification indices and factor loadings. Finally, ROC (receiver operating characteristic) analysis was used to evaluate the screening characteristics of the CBAS.
Results
Confirmatory factor analysis
As bootstrapping technique was used, Bollen-Stine p-value was calculated to assess model fit, where p-values of .05 and more indicate a good model fit (Bollen & Stine, Reference Bollen and Stine1992). In the CFA, the original four factors structure reported by Ottenbreit and Dobson (Reference Ottenbreit and Dobson2004) was tested. This model did not show a good fit to the observed data (CFI = .810, RMSEA = .083, SRMR = .181, χ2(429) = 1184.07, p < .001 and a χ2/df = 2.76). Additionally, Bollen-Stine p was .004, so the model fit was not adequate. An exhaustive examination of the items indicated that items 3, 10 and 22 showed high standardized residuals (> 2.58). So these items were deleted in order to obtain a better-fitting model. This revised model shows a good fit to the observed data (CFI = .929, RMSEA = .057, SRMR = .051, χ2(333) = 603.28, p < .001 and a χ2/df = 1.81). Bollen-Stine p was .050, so the model fit was also adequate. Standardized parameters estimates for this revised model were obtained using bootstrapping technique. As can be seen in Table 2, factor loadings were between .52 and .85 with p < .05 in all cases. Bias corrected percentile method was used. Effect sizes were calculated using Cohen’s f 2 (Cohen, Reference Cohen1992). Cohen’s f 2 values between .15 and .34 indicate a medium effect size and values higher than 0.35 indicate a large effect size. Cohen’s f 2 were between 0.30 and 2.57 in all cases, and only one item (30) showed value < 0.35 (see Table 2).
Factor loadings obtained in the CFA
Notes: Items 3, 10 and 22 were deleted. BS (behavioral social), BN (behavioral nonsocial), CS (cognitive social), CN (cognitive nonsocial).
Reliability
The internal consistency of the revised CBAS consisting of 28 items was calculated using the Cronbach’s alpha for the total sample (N = 254) made up of patients, controls and undergraduates. The degree of internal consistency of the total scale was very high, (α = .95), and was also satisfactory for each of the four subscales: CBAS-BS (α = .93), CBAS-BN (α = .80), CBAS-CS (α = .74) and CBAS-CN (α = .87). The correlations between subscales and between the subscales and total scale were significant at the p < .01 level, except for one combination, which was at the p < .05. This result was repeated in the clinical, control and undergraduate samples (see Table 3). The subscale correlations ranged from .29 to .74. With respect to the correlations between each subscale and the total scale, the values obtained were high and statistically significant, ranging from .59 to .86.
Correlation coefficients between Total-CBAS and its subscales
Notes: BS (behavioral social), BN (behavioral nonsocial), CS (cognitive social),
CN (cognitive nonsocial). Items 3, 10 and 22 were deleted. **p < .01; *p < .05
Validity
We also tested the validity of the new version of CBAS formed by 28 items. Respect to construct validity, significant differences between the mean scores were found for the clinical and control groups (GLM, p < .001), and for the clinical and undergraduate groups (GLM, p < .001), both for the total scores and for the four subscale scores (see Table 4). The clinical group showed higher scores all around. No significant differences were found between the control group and undergraduate group. No significant gender differences were found in the Total-CBAS scores of the three groups (GLM, Patients: men – M = 72.05, SD = 19.13 – vs. women – M = 75.81, SD = 21.95 –, p = .234; Controls: men – M = 47.52, SD = 10.38 – vs. women – M = 42.46, SD = 10.76 –, p = .142; Undergraduates: men – M = 51.08, SD = 13.67 – vs. women – M = 44.38, SD = 12.42 –, p = .190), so there is no evidence to suggest that one gender exhibits greater cognitive-behavioral avoidance than the other. We did not find significant age differences in the Total-CBAS scores in any of the groups (Patients: rho = .139, p = .167; Controls: rho = .170, p = .090; Undergraduates: rho = .017, p = .902).
Distribution of CBAS scores as a function of the sample group
Notes: GLM (Generalized Linear Models, log function). Items 3, 10 and 22 were deleted.
P vs. C = Patients vs. Controls; C vs. S = Controls vs. Students; P vs. S = Patients vs. Students
BS (behavioral social), BN (behavioral nonsocial), CS (cognitive social), CN (cognitive nonsocial)
The convergent validity was evaluated through correlations between the CBAS and its subscales and the AAQ-II. As can be seen in Table 5, we encountered significant correlations between the Total-CBAS and the AAQ-II in both the clinical and control groups. The correlations between the four CBAS subscales and the AAQ-II were also significant (ranging from .24 to .42), with exception of the correlation between CBAS-BS and AAQ-II for the control group (rho = .14, p = .15). The discriminant validity was examined through correlations between the CBAS and MAAS. As expected, we found statistically significant negative correlations with the MAAS in both the clinical and control groups (ranging from –.23 to –.47). We have also found that of all the instruments studied, the highest and most consistent correlations were obtained with the MAAS, except in the case of the Behavioral Nonsocial (BN) subscale, which showed a slighter better correlation with the AAQ-II.
Correlations (Spearman´s rho) between CBAS and its subscales with AAQ-II, MAAS, BDI-II and BAI
Notes: A = clinical group (N = 100). B = control group (N = 100). F-Z-t = Fisher´s Z-test for the difference between two independent correlation coefficients (z-score), all not significant except: CBAS-BS x BDI-II. BS (behavioral social), BN (behavioral nonsocial), CS (cognitive social), CN (cognitive nonsocial). *p < .05; **p < .01
Correlations with criterion measures were conducted to determinate if cognitive-behavioral avoidance was related to depression and anxiety. The correlation between CBAS and BDI-II was significant (see Table 5), both for the total (rho = .36, p < .01) and for the four subscales in the clinical group (ranging from .28 to .32). This was not the case for the control group, where we only found a significant correlation with CBAS-CS (rho = .22, p < .05). Finally, and contrary to what we expected, we did not find significant correlations between CBAS and BAI in any of the two groups.
Screening characteristics of the CBAS
To evaluate sensitivity and specificity of the CBAS at a various cut-off points, we used ROC analysis. In order to construct a ROC curve, sensitivity and specificity were calculated for several scores of the CBAS. The sample used in this analysis was formed by the control and clinical groups. Sensitivity, specificity, positive likelihood ratio (LR+) and negative likelihood ratio (LR-) are shown in Table 6. The results suggested a score of 53 as the optimum cut-off point. At this point, sensitivity and specificity are both 86%.
Diagnostic performance of the CBAS at different cut-off scores for clinical populations (N = 200)
Notes: Items 3, 10 and 22 were deleted. LR+ = positive likelihood ratio; LR-= negative likelihood ratio.
Prevalence of clinical diagnosis was 50%.
Figure 1 shows the ROC curve for the CBAS. The area under the curve (AUC) calculated with ROC analysis was 0.91 (95% CI: 0.87 – 0.95; p < .001). As can be seen, the curve is substantially above the random ROC (AUC = 0.5 represents sensitivity and specificity of random guessing whether an individual is a case or not). The fact that the AUC is significantly greater than 0.5 means that the null hypothesis (i.e., the CBAS provides no useful information) can be rejected.
ROC curve of the CBAS for clinical populations.
Discussion
The main objectives of this study were, on one hand, to assess the factorial structure and psychometric properties of the Spanish version of CBAS, and on the other hand, to examine how cognitive-behavioral avoidance differs between clinical patients, general population and undergraduate students.
Regarding the first objective, the confirmatory factor analysis shows that the four factors structure provide a good fit to the data of our sample. This supports the factor structure proposed by Ottenbreit and Dobson (Reference Ottenbreit and Dobson2004), comprising Behavioral Social, Cognitive Nonsocial, Cognitive Social and Behavioral Nonsocial factors. Our analysis also shows the existence of three problematic items in the study sample. Items 3 (from Behavioral Nonsocial factor), 10 and 22 (both from Cognitive Social factor) were deleted from the scale to obtain a better fit to the factor model. Perhaps, the exclusion of these items is related to the type of sample used. In the original study, the sample was composed of 391 undergraduate students, whereas in our study, the sample was made up of general population, clinical population and undergraduate students. As a result, replication of the factorial structure must be achieved in different samples and future studies are needed to determine the optimum scale structure. It should be noted that until now most studies have only applied the CBAS to undergraduate students (Carvalho & Hopko, Reference Carvalho and Hopko2011; Hernández-Guzmán et al., Reference Hernández-Guzmán, Dobson, Caso-Niebla, Gónzalez-Montesinos, Epp, Arratíbel-Siles and Wierzbicka-Szymczak2009; Moulds et al., Reference Moulds, Kandris, Starr and Wong2007; Ottenbreit & Dobson, Reference Ottenbreit and Dobson2004), whereas our study explores new ground by also applying it to a clinical sample and general population. In this sense, we would like to point out that if one of the basic objectives of this scale is its use on subjects suffering from psychological disorders, such studies should be carried out using equivalent samples. In any case, the good reliability scores obtained after eliminating these three items allowed us to obtain an improved version of the CBAS for Spanish people. Thus, we found similar internal consistency results (α values) to those reported by Ottenbreit and Dobson (Reference Ottenbreit and Dobson2004): Total-CBAS (.91), CBAS-BS (.86), CBAS-BN (.75), CBAS-CS (.78) and CBAS-CN (.80). And, similar to those reported by Röthlin et al. (Reference Röthlin, Holtforth, Bergomi, Berking, Ottenbreit and Caspar2010) for the German version: Total-CBAS (.92), CBAS-BS (.86), CBAS-BN (.70), CBAS-CS (.80) and CBAS-CN (.87). Therefore, we can conclude that the CBAS is a reliable instrument for use in different types of population.
The four subscale correlations are significant and of average intensity, which would support Ottenbreit and Dobson’s (Reference Ottenbreit and Dobson2004) view that they measure distinct, albeit related, constructs. The subscale correlations we obtained are similar to those obtained by Ottenbreit and Dobson (Reference Ottenbreit and Dobson2004; ranging from .39 to .57) and by Moulds et al. (Reference Moulds, Kandris, Starr and Wong2007; ranging from .31 to .56.). Finally, our correlations between each subscale and the total scale were also similar to those found by the authors of the original scale, in the .78–.80 range, as well as those found by Moulds et al. (Reference Moulds, Kandris, Starr and Wong2007), who reported significant correlations ranging from .73 to .80.
The CBAS has shown to be an instrument capable of distinguishing between a normal and clinical population (Total-CBAS score of 43.93 vs. 74.23). Along the same lines, Ottenbreit, Dobson, and Quigley (Reference Ottenbreit, Dobson and Quigley2014) have found significantly higher levels of Total-CBAS avoidance in two clinical samples, one made of 60 women with depression and another one made of 30 women with social anxiety disorder, than in a non-clinical group. Although, as these authors have pointed out, they employed only women, and thus, the results cannot be generalized to men. Our results of the ROC analysis show that the CBAS is sensitive, specific and has good likelihood ratios (both positive and negative), taking the total score of 53 (excluding items 3, 10 and 22) as the cut-off point to detect people with anxiety or mood disorders. To date it has not been informed any cut-off point by other authors. There are no significant gender differences in the CBAS scores. Although Ottenbreit and Dobson (Reference Ottenbreit and Dobson2004) reported higher scores for the males for the both total scale and subscales, except for the BN subscale where the females scored higher, neither Moulds et al. (Reference Moulds, Kandris, Starr and Wong2007) nor Carvalho and Hopko (Reference Carvalho and Hopko2011) reported gender differences. Avoidance is not related to age, at least as measured by the CBAS, so we cannot affirm that younger or older people display greater or lesser avoidance.
The results of our study indicate that there is an inverse relationship between cognitive-behavioral avoidance and mindfulness. With regard to the relationship between cognitive-behavioral avoidance and experiential avoidance, our results indicate that subjects with higher cognitive-behavioral avoidance levels also present higher levels of experiential avoidance, this result agree with the correlation found by Röthlin et al. (Reference Röthlin, Holtforth, Bergomi, Berking, Ottenbreit and Caspar2010) between Total-CBAS and AAQ in a sample of 71 students. Therefore, as Ottenbreit and Dobson (Reference Ottenbreit and Dobson2004) pointed, the concurrent and discriminant validity of the scale seems well established.
We find a significant correlation between CBAS and BDI-II, but only in those clinical subjects with anxious-depressive symptomatology. Furthermore, we should acknowledge that although these correlations are significant, they are nevertheless weak. In normal subjects, the relationship between cognitive-behavioral avoidance and depression does not appear to be significant. Taking into consideration the results of previous researchers, we would have expected to find correlations between CBAS and BDI-II in each case. In fact, it is worth noting that we found a significant correlation in the clinical group but not in the control group, given that previous researchers only used undergraduate samples (which would likely be more similar to a normal sample than a clinical sample) and they did find a correlation. For example, Ottenbreit and Dobson (Reference Ottenbreit and Dobson2004) found a correlation of .48, Moulds et al. (Reference Moulds, Kandris, Starr and Wong2007) of .62 and Carvalho and Hopko (Reference Carvalho and Hopko2011) of .75 between Total-CBAS and BDI-II. The first two research teams also reported significant correlations between the CBAS subscales and the BDI-II. In addition, our results do not support a relation between cognitive-behavioral avoidance and anxiety, at least as measured by the CBAS and the BAI. While Moulds et al. (Reference Moulds, Kandris, Starr and Wong2007) did find a significant correlation of .49 between the Total-CBAS and BAI, and Carvalho and Hopko (Reference Carvalho and Hopko2011) found a correlation of .45, also significant. We thus note that both Moulds et al. (Reference Moulds, Kandris, Starr and Wong2007) and Carvalho and Hopko (Reference Carvalho and Hopko2011) found a higher correlation between avoidance and depression than between avoidance and anxiety. Our study only found an avoidance-depression association in the clinical sample. Ottenbreit and Dobson (Reference Ottenbreit and Dobson2004) found a significant correlation between avoidance and their measurements of depression and anxiety, but in this case, the relationship was slightly higher for anxiety.
Therefore, although other authors have reported a moderate association between cognitive-behavioral avoidance and depression/anxiety, it is important to remember that those studies were done with undergraduate students, most of whom studied psychology (Carvalho & Hopko, Reference Carvalho and Hopko2011; Moulds et al., Reference Moulds, Kandris, Starr and Wong2007; Ottenbreit & Dobson, Reference Ottenbreit and Dobson2004), and their mean scores on the Beck inventories of anxiety and depression fell in general between the limits of minimal and mild symptomatology. While in the Ottenbreit and Dobson (Reference Ottenbreit and Dobson2004) study do not appear the mean scores, in the Moulds et al. (Reference Moulds, Kandris, Starr and Wong2007) study, the mean score on the BAI was 11.82, which falls into the range of “mild anxiety” and not “minimal anxiety.” The mean score on the BDI-II was 10.75, which, although falling within the “minimal depression” range, is very close to the cut-off for “mild depression” (13). And in the Carvalho and Hopko (Reference Carvalho and Hopko2011) study, the mean scores for anxiety and depression exceeded the minimal range (BAI: 10.37 and BDI-II: 13.59). These mean scores are clearly different from those in our control group, which are much lower, indicating normal or minimal levels of anxiety and depression (BAI: 4.19, BDI-II: 4.5). This calls into question the representativeness of student samples, and whether the data obtained from those studies can be applied to non-student populations, and even whether the results obtained from one type of student can be applied to a general student population. It is fair to wonder whether this difference in the sample type is the reason we did not find an association between avoidance and depression in the control group, and between avoidance and anxiety in both the control and clinical groups.
The primary limitation of this research is the cross-sectional nature of the study which does not allow to establish causality relations. An additional limitation is the use of self-report measures to assess all variables of interest because common-method variance may be affecting the relations between variables. Finally, we should note that sample size of undergraduates is small, thus the comparative results relative to this sample should be interpreted with caution. Future studies will have to further consider the relationship between cognitive-behavioral avoidance and anxiety and depression, given that, despite the fact that it seems clear in theory that a relationship would exist, the empirical results are inconsistent and tend to change depending upon the sample type used. In particular, further work will need to be done to study these relationships in clinical samples. For that work, we would recommend using the CBAS, as it has proven itself to be a useful and reliable instrument to assessing avoidance.
Appendix.
Complete items of the present Spanish CBAS.
