مقیاس نگرانی ارزیابی اجتماعی: اعتبار سنجی روان با یک نمونه بالینی از بیماران مبتلا به اختلال اضطراب اجتماعی

Abstract The Appraisal of Social Concerns (ASC) Scale was created by Telch et al. (2004) to improve upon existing self-report measures of social anxiety–related cognition. In a largely nonclinical sample, the ASC was found to possess three factors and was psychometrically sound. In a smaller clinical sample, the ASC demonstrated sensitivity to the effects of cognitive behavioral therapy. In the present study, the psychometric properties of the ASC were examined in a larger sample of patients with social anxiety disorder. In this sample, the ASC exhibited a 2-factor structure; the nature of the factors was similar to the primary factors originally reported by Telch et al. The ASC also demonstrated strong validity, internal consistency, and sensitivity to treatment effects. It is concluded that the ASC may be useful in the assessment of cognition and cognitive change in patients with social anxiety disorder. Individuals with social anxiety disorder experience persistent and often intense fears related to social interaction or performance situations in which they anticipate scrutiny by others (American Psychiatric Association [APA], 2000). At some point in their lives, more than 12% of the population meet diagnostic criteria for this disorder (Kessler, Berglund, et al., 2005), although estimates of the prevalence of clinically significant social anxiety disorder are closer to 4% (Narrow, Rae, Robins, & Regier, 2002). Once considered the neglected anxiety disorder (Liebowitz, Gorman, Fyer, & Klein, 1985), increasing knowledge of the prevalence and impact of social anxiety disorder has led researchers to examine the mechanisms of its maintenance, and several theorists have suggested that cognitive factors are likely to be especially important (Amir and Foa, 2001, Clark and Wells, 1995 and Rapee and Heimberg, 1997). A central fear in social anxiety disorder is of negative evaluation, and this fear may be maintained in several ways. For instance, compared to controls, individuals with social anxiety disorder overestimate the likelihood that negative social events will occur, underestimate the likelihood that positive social events will occur, and appraise negative social events as substantially more costly (Foa et al., 1996, McManus et al., 2000 and Poulton and Andrews, 1994). Furthermore, as delineated by recent cognitive behavioral models of social anxiety (Clark and Wells, 1995 and Rapee and Heimberg, 1997), individuals with social anxiety disorder are likely to enter a social situation with negative predictions concerning the situation’s outcome, which may be related to biases in the allocation of attention toward threat and in the tendency to interpret ambiguous situations as negative. This process acts to maintain anxiety (see reviews of attentional and interpretational biases in social anxiety by Hirsch and Clark, 2004 and Ledley and Heimberg, in press). Therefore, it is likely that individuals experiencing social anxiety assume that threat is always present in social situations (Heimberg & Becker, 2002). Persons with social anxiety disorder commonly engage in avoidance behaviors and isolate themselves from social interaction (e.g., as reviewed by Rapee & Heimberg, 1997), making it unlikely that they will receive disconfirmatory feedback, which might otherwise allow for emotional processing of feared stimuli (Foa & Kozak, 1986). Furthermore, it has been demonstrated that, with successful treatment, socially anxious individuals rate the probability of negative social events as less likely (Foa et al., 1996 and McManus et al., 2000) and the cost of such events as less catastrophic (Foa et al., 1996, Hofmann, 2004 and McManus et al., 2000). Foa et al. and Hofmann have concluded that changes in negative cognitions may mediate anxiety reduction in social anxiety disorder. These findings underscore the central role of cognition in social anxiety disorder as well as the importance of valid and reliable assessment of these cognitive features. Such assessment may better elucidate the nature of social anxiety morbidity, maintenance, and symptom reduction. Telch et al. (2004) introduced the Appraisal of Social Concerns (ASC) scale as a tool to measure distress related to negative outcomes in challenging social situations and to improve upon the noted shortcomings of existing cognitive assessment scales (also see Hofmann & DiBartolo, 2000) such as the Irrational Beliefs Test (Jones, 1969), the Social Interaction Self-Statement Test (Glass, Merluzzi, Biever, & Larsen, 1982), the Fear of Negative Evaluation Scale (Watson & Friend, 1969), and unstructured thought-listing protocols. Telch et al. contend that the ASC offers the benefit of asking participants to rate the degree to which they feel concerned about social threat, as opposed to rating the frequency and cost of past negative social outcomes, of which they may not be completely aware. Therefore, the instructions of the ASC direct respondents to think about themselves in social situations and rate their concerns with the listed negative social negative outcomes, as opposed to directing them to indicate the frequency of or distress related to these outcomes. The ASC directs the respondent to rate the degree to which he or she would be concerned by the particular outcomes if they occurred in challenging social situations. It consists of three subscales based on factor analyses of the responses of 550 college undergraduates: Negative Evaluation, Observable Symptoms, and Social Helplessness. According to Telch et al. (2004), the Negative Evaluation subscale measures the degree to which the respondent is concerned with the negative judgments of others in social situations (e.g., “People ridiculing you”; “Appearing weird”). The Observable Symptoms subscale assesses the degree to which the respondent is concerned with experiencing (ostensibly) visible physical symptoms (e.g., “Twitching”; “Blushing”) in social situations. The Social Helplessness subscale presumably measures the respondent’s concern with experiencing social outcomes beyond his or her control (e.g., “Losing control [screaming, running out, etc.]”). The psychometric properties of the ASC were initially established in this sample as well as a smaller sample of individuals with social anxiety disorder (N = 86; Telch et al., 2004). The authors concluded that the ASC has strong psychometric properties, including good internal consistency and test-retest reliability, good convergent and discriminant validity, as well as sensitivity to treatment effects. Telch et al. further suggest that the ASC may aid the clinician in focusing on specific targets for intervention. It is both logical and imperative to further examine the psychometric characteristics of the ASC in a large, clinical sample of socially anxious patients. In the current study, data from a sample of 204 patients with social anxiety disorder were analyzed. We examined the factor structure to determine if the data from this clinical sample would fit Telch et al.’s three-factor solution, which was derived from their undergraduate sample. We also examined the internal consistency of the ASC as well as its convergence with measures of social anxiety and divergence with measures of other constructs. Effect sizes were also calculated to determine the ASC’s sensitivity to the effects of cognitive behavioral group therapy (Heimberg & Becker, 2002); the monamine oxidase inhibitor phenelzine sulfate, which has also been shown to be efficacious in the treatment of social anxiety (Heimberg et al., 1998); and pill placebo.

Results Preliminary analyses Demographic characteristics of patients from the three collaborative sites (Albany, Philadelphia, and New York) are presented in Table 1. Comparisons across sites revealed no differences in age, gender, marital status, or education. However, there were site differences in ethnicity, as the New York sample was more racially diverse than the Albany and Philadelphia samples, which were both predominantly Caucasian. Furthermore, ethnicity was related to full-scale ASC scores, as Caucasian participants scored 6.06 points lower than non-Caucasians, t(1, 200) = 2.04, p = .043. This significant difference among ethnic groups persisted when measures of social anxiety (i.e., SIAS, SPS, and BFNE) were controlled, F(1, 191) = 4.87, p = .03. Therefore, it appears that the ethnic groups in the present sample may differ specifically in terms of the appraisal of social threat. Table 1. Descriptive characteristics of the total sample and patients from the three collaborative sites Total sample (N = 204) Albany (n = 24) Philadelphia (n = 123) New York (n = 57) χ2 n % n % n % n % Gender 0.85 Female 83 41.3 8 33.3 52 43.3 23 40.4 Male 118 58.7 16 66.7 68 56.7 34 59.6 Martial status 2.13 Single, never married 140 70.0 16 66.7 82 67.2 42 77.8 Currently or previously married 60 30.0 8 33.3 40 32.8 12 22.2 Race 31.01** Caucasian 135 66.8 22 91.7 91 75.2 22 38.6 Other 67 33.2 2 8.3 30 24.8 35 61.4 Education 6.4 Some college or less 102 50.5 10 41.6 57 46.3 35 63.6 College graduate 57 28.2 7 29.2 40 32.5 10 18.2 Postgraduate 43 21.3 7 29.2 26 21.1 10 18.2 Age F 0.17 Mean 33.32 32.42 33.26 33.84 SD 10.02 9.73 10.35 9.55 Range 19–65 20–46 19–65 20–61 Note. Variations in sample size are the result of missing data. Percentages may not sum to 100% due to errors in rounding. * p < .05; ** p < .01. Table options There were no site differences on the mean scores of any measures used in the current study (see Table 2). On average, clients endorsed clinically significant levels of fear in 6.62 social situations (SD = 2.13) in response to the social phobia module of the ADIS-IV-L. A minority (46.5%) of clients endorsed 4 or more panic-related symptoms in social situations, 1 and the mean number of clinically significant symptoms reported was 3.61 (SD = 2.56). The number of panic symptoms experienced in social situations correlated .39 (p < .05) with the ASI’s Social Concerns subscale. Table 2. Severity scores of the total sample and patients from the three collaborative sites Total sample (N = 204) Albany (n = 24) Philadelphia (n = 123) New York (n = 57) M SD M SD M SD M SD F Appraisal of Social Concerns 52.96 20.15 52.52 19.63 51.53 19.02 56.05 22.48 0.99 ASC Subscale 1 59.34 22.96 58.68 23.51 58.62 21.50 61.17 25.92 0.25 ASC Subscale 2 40.13 22.10 38.89 18.51 37.71 22.12 45.88 22.74 2.75 Brief Fear of Negative Evaluation Scalea 30.01 7.05 30.73 7.25 30.21 6.61 29.30 7.93 0.44 Social Phobia Scale 31.10 15.94 32.67 11.72 29.77 15.81 33.34 17.62 1.10 Social Interaction Anxiety Scale 49.23 15.21 50.08 12.23 48.64 15.94 50.16 14.88 0.23 Beck Depression Inventory 12.87 8.62 10.04 5.41 13.01 8.96 13.76 8.85 2.09b Anxiety Sensitivity Index–Full Scale 24.89 12.73 26.17 10.39 23.71 12.65 26.93 16.66 1.38 ASI–Physical Concerns 11.43 6.74 11.46 6.12 10.97 6.90 12.45 6.64 0.925 ASI–Mental Incapacitation Concerns 5.19 3.99 6.42 3.74 4.72 3.92 5.71 4.13 2.53 ASI–Social Concerns 9.68 3.52 10.29 2.68 9.42 3.48 9.98 3.91 0.915 No differences significant at p < .05. a BFNE scores calculated on the 8 straightforwardly-worded items only. b Brown Forsyth F* tests conducted to account for heterogeneity of variances (as measured by Levene’s test) among collaborative sites. Table options Descriptive statistics for the ASC in patients with social anxiety disorder For the full sample, the ASC total score was normally distributed, with kurtosis of −0.25 and skewness of −0.28. The mean score was 51.92 (SD = 20.11). The ASC demonstrated strong internal consistency (α = .92), comparable to the internal consistency reported for Telch et al.’s (2004) nonclinical sample (α = .94). Confirmatory factor analysis A confirmatory factor analysis was conducted to assess the fit of the three-factor structure obtained by Telch et al. (2004) to the ASC responses of individuals with social anxiety disorder (with the use of AMOS 5, a structural equation modeling software program; Arbuckle, 2002). Correlations between the latent variables in the model were allowed, and a maximum likelihood method of covariance analysis was utilized. Several indices suggest a poor match of the data to the original structure, as they failed to meet proposed cutoff scores of adequate fit (Hu & Bentler, 1999). Examined fit indices included the chi-square test of covariance equivalence, χ2(168) = 614.96, p < .01; the Tucker Lewis Index (TLI; Tucker and Lewis, 1973) = 0.72; the Comparative Fit Index (CFI; Bentler, 1990) = 0.77; and the root mean square error of approximation (RMSEA) = 0.12, 90% CI = 0.11, 0.13. The assumption of multivariate normality was not satisfied, however. Hu, Bentler, and Kano (1992) suggest that the violation of this assumption may bias tests of fit in a confirmatory factor analysis, such that too many models are rejected. Dropping the items that were not univariate normal did not produce multivariate normality. To satisfy the condition of multivariate normality, the items on each of Telch et al.’s three factors were randomly paired and summed, with the following constraints: (a) items that were not univariate normal (e.g., skewness greater than 1.0) were added to items that were univariate normal; (b) factors with odd numbers of items were left with one item that was not paired with another (two items total; for a discussion of the appropriateness of this technique see Gorsuch, 1997 and West et al., 1995). This process of pairing items yielded multivariate normality in the sample. Subsequently, a second confirmatory factor analysis was conducted on the summed, multivariate normal scores and fit indices were again examined, which further confirmed a poor fit of the proposed factor structure to the clinical sample, χ2(42) = 261.28, p < .01; TLI = 0.77; CFI = 0.83; RMSEA = 0.17, 90% CI = 0.15, 0.19. Thus, the poor fit of the model is unlikely to be due to lack of multivariate normality alone. Exploratory factor analysis Given the poor fit of the Telch et al. (2004) factor structure to the current sample, an exploratory factor analysis was conducted. A principal components analysis with an oblique promax rotation2 (as recommended by Gorsuch, 1997 and Floyd and Widaman, 1995) was utilized to examine the structure of the ASC’s 20 items. A scree plot of eigenvalues (Cattell, 1966) was examined, and a two-factor solution was deemed most appropriate. As suggested by Floyd and Widaman (1995), a maximum likelihood factor analysis was then conducted to examine the relationships of observed variables to latent variables. Items with factor loadings greater than 0.40 were retained, and any item with less than a 0.10 difference in loadings between the factors was rejected. Thus, 19 items were retained and 1 item, “Concerns about blushing,” was rejected. Promax rotated factor loadings are presented in Table 3. Subscale scores were calculated by averaging the individual scores of items loading on each factor. Table 3. Rotated factor loadings for common factor analysis of the Appraisal of Social Concerns (ASC) Item Subscale 1 Subscale 2 Subscale 1: Concerns With Negative Evaluation and Rejection Concerns about appearing stupid (ASC 2) 0.696 0.111 Concerns about people laughing at you (ASC 3) 0.786 0.003 Concerns about people ignoring you (ASC 5) 0.657 − 0.074 Concerns about people staring at you (ASC 6) 0.450 0.310 Concerns about appearing incompetent (ASC 9) 0.743 0.077 Concerns about being incoherent (not making sense) (ASC 10) 0.534 0.025 Concerns about not performing adequately (ASC 12) 0.499 0.265 Concerns about appearing weird (ASC 14) 0.584 0.189 Concerns about people ridiculing you (ASC 15) 0.839 0.000 Concerns about not being able to think (mind going blank) (ASC 16) 0.453 0.283 Concerns about appearing ugly (ASC 17) 0.757 − 0.243 Concerns about appearing weak (ASC 18) 0.494 0.121 Concerns about people rejecting you (ASC 19) 0.980 − 0.295 Subscale 2: Concerns With Physical Symptoms Concerns about trembling (ASC 1) − 0.274 0.869 Concerns about twitching (ASC 7) − 0.120 0.675 Concerns about voice quality (cracking, stuttering, squeaking, etc.) (ASC 8) 0.254 0.460 Concerns about losing control (screaming, running out, etc.) (ASC 11) 0.022 0.509 Concerns about being tense (ASC 13) 0.097 0.494 Concerns about sweating (ASC 20) 0.013 0.510 Rejected Item Concerns about blushing (ASC 4) 0.161 0.394 Table options Both subscales evidenced a normal distribution (Subscale 1: skewness = −0.458, kurtosis = −0.320; Subscale 2: skewness = 0.464, kurtosis = −0.401). Subscale 1’s mean score was 59.34 (SD = 22.96), and its internal consistency was excellent (α = .92). Subscale 2’s mean score was 40.13 (SD = 22.10), and it demonstrated strong internal consistency as well (α = .76). Subscales 1 and 2 demonstrated a correlation of .52 (p < .001). Subscale 1 was labeled Concerns With Negative Evaluation and Rejection because the items loading on this factor assess an individual's concern with negative social outcomes, such as appearing inadequate, performing poorly, and being judged or rejected. Subscale 1 consisted of 13 items and accounted for 40.78% of the variance. With the exception of “Concerns about not being able to think (mind going blank)” and “Concerns about appearing incoherent (not making sense),” all items loading on Subscale 1 represent a combination of Telch et al.’s (2004) factors labeled Concerns With Negative Evaluation and Concerns With Social Helplessness (p. 221). Subscale 2 was labeled Concerns With Physical Symptoms, as the items loading on the factor assess an individual’s concern about experiencing physical symptoms, such as trembling and twitching. Subscale 2 consisted of 6 items and accounted for 10.44% of the variance. All items but “Concerns about losing control (screaming, running out, etc.)” from Subscale 2 loaded on Telch el al.’s Observable Symptoms Scale (p. 221). Thus given the similarity between Telch et al.’s factors and those of the present study, coefficients of congruence (Gorsuch, 1983) were calculated (see Table 4). Telch et al.’s Negative Evaluation subscale and Subscale 1 of the present study were strongly related (coefficient of congruence = .90). Similarly, Telch et al.’s Observable Symptoms subscale was strongly related to Subscale 2 in the present study (coefficient of congruence = .90). Therefore, despite the poor fit of the current study’s clinical data to Telch et al.’s factor structure in the confirmatory factor analysis, the factors appear to be quite similar. Table 4. Coefficients of congruence between Appraisal of Social Concerns factors from the present study and Telch et al. (2004) Telch et al. Present study I II I .90 .26 II .15 .90 III .45 .15 Note. Coefficients of congruence (Gorsuch, 1983) were derived using factor loadings from factor pattern matrices. Factor labels assigned by Telch et al. (2004): Factor I, Concerns With Negative Evaluation; Factor II, Concerns With Observable Physical Symptoms; Factor III, Concerns With Social Helplessness. Factor labels assigned in the present study: Factor I, Concerns With Negative Evaluation and Rejection; Factor II, Concern With Physical Symptoms. Table options Convergent validity of the ASC A Bonferroni correction was applied to control for the number of comparisons between the ASC (and each of its subscales) and the four measures of social anxiety (.05/4 = .0125). The ASC total score was significantly correlated with all measures of social anxiety. Subscale 1, reflecting concerns with negative evaluation and rejection, was also significantly correlated with all measures of social anxiety. Subscale 2, reflecting concerns with physical symptoms, was significantly correlated only with the BFNE, the SPS, and the Social Concerns subscale of the ASI (see Table 5). Table 5. Zero-order correlations of the Appraisal of Social Concerns (ASC) and its factors with measures of social anxiety ASC Subscale 1 Subscale 2 Brief Fear of Negative Evaluation Scale .591* .665* .236* Social Interaction Anxiety Scale .444* .522* .122 Social Phobia Scale .594* .524* .513* Anxiety Sensitivity Index–Social Concerns .451* .283* .650* Note.Ns vary from 198 to 204. Bonferroni correction based on the number of comparisons per column. BFNE scores calculated on straightforward-worded items only. *p < .0125 (.05/4). Table options In both the Telch et al. (2004) study and the current study, the ASC was found to measure different facets of social concerns (e.g., concern with negative evaluation and with physical anxiety symptoms). Therefore, it was expected that the ASC’s subscales would demonstrate measurement specificity and each would correlate more strongly with questionnaires that measure presumably similar constructs. Therefore, tests of correlated correlation coefficients, as described by Meng, Rosenthal, and Rubin (1992), were conducted to determine if the ASC and its subscales were related more strongly to specific convergent measures than to others. A Bonferroni correction was applied to control the error rate in multiple comparisons of the ASC’s (and each of its subscales’) correlations with the four measures used to assess convergent validity in this study (.05/4 = .0125). The ASC’s total score was highly correlated with all convergent measures; however, its correlation with the BFNE was significantly stronger than its correlation with the SIAS, z = 2.90, p = .004, and its correlation with the SPS was stronger than its correlations with both the SIAS and the Social Concerns subscale of the ASI, z = 2.62, p = .009; z = 2.64, p = .008. Subscale 1 was more highly correlated with the BFNE than with either the SIAS or the SPS, z = 3.17, p < .001; z = 2.70, p = .007. Subscale 1’s correlations with the BFNE, SPS, and SIAS were all stronger than its correlation with the Social Concerns subscale of the ASI, z = 5.60, p < .001; z = 4.24, p < .001; z = 3.25, p = .001, respectively. Subscale 2 demonstrated a stronger correlation with the SPS than with the SIAS, z = 6.07, p < .001, or the BFNE, z = 4.91, p < .001. Also, Subscale 2’s correlation with the ASI’s Social Concerns subscale was stronger than its correlations with the SIAS and BFNE, z = 6.48, p < .001; z = 5.80, p < .001, but not stronger than Subscale 2’s correlation with the SPS, z = 2.31, ns. Further, Subscale 1 demonstrated significantly stronger correlations than Subscale 2 with the BFNE, z = 7.24, p < .001, and the SIAS, z = 5.95, p < .001, but not the SPS, z = -0.14, ns (all comparisons Bonferroni corrected; .05/4 = .0125). Lastly, Subscale 2 demonstrated a stronger linear relationship with the Social Concerns subscale of the ASI than did Subscale 1, z = 6.29, p < .001. Discriminant validity of the ASC A Bonferroni correction was applied to control the error rate in multiple comparisons of the ASC’s (and each of its subscales’) correlations with the measures used to evaluate discriminant validity in this study (.05/2 = .025). The ASC total score, Subscale 1, and Subscale 2 were all positively correlated with these measures (see Table 6). Table 6. Zero-order correlations of the Appraisal of Social Concerns (ASC) and its factors with discriminant validity measures ASC Subscale 1 Subscale 2 Anxiety Sensitivity Index–Physical and mental incapacitation concerns .352* .252* .473* Beck Depression Inventory .404* .419* .231* Note.Ns vary from 199 to 204. Bonferroni correction based on the number of comparisons per column. *p < .025 (.05/2). Table options Tests of correlated correlation coefficients were again conducted to determine whether the ASC’s linear relationships and those of its subscales were stronger with measures of social anxiety than with the discriminant measures. A Bonferroni correction was applied to the comparisons of each ASC score’s correlation with convergent and discriminant measures (.05/8 = .0063). The ASC total score was more strongly related to the BFNE and the SPS than it was to the nonsocial (Mental Incapacitation and Physical Concerns) subscales of the ASI, z = 2.99, p = .003; z = 3.86, p < .001, and the BDI, z = 2.96, p = .003; z = 2.92, p = .003. Other comparisons were nonsignificant. There were also significant differences between Subscale 1’s correlations with the BFNE and its correlations with the nonsocial subscales of the ASI, z = 5.33, p < .001, and BDI, z = 4.12, p < .001. Further, Subscale 1’s correlations with the SIAS and SPS were significantly stronger than its correlation with the nonsocial subscales of the ASI, z = 3.10, p = .002; z = 3.91, p < .001. Subscale 2 evidenced a significantly stronger correlation with the ASI Social Concerns subscale than with the ASI’s nonsocial subscales, z = 4.18, p < .001. Also, Subscale 2’s correlations with the SPS and the Social Concerns subscale of the ASI were significantly stronger that its correlation with the BDI, z = 4.01, p < .001; z = 5.77, p < .001. Finally, Subscale 2 demonstrated stronger relationships with the nonsocial subscales of the ASI than with the BFNE, z = 3.01, p < .001, or the SIAS, z = 3.87, p < .001, but not the SPS, z = 0.92, ns. Sensitivity to treatment Within-group effect sizes (d; Cohen, 1988) were calculated for participants treated with CGBT (n = 43), phenelzine sulfate (n = 25), and pill placebo (n = 18). The ASC total score demonstrated effect sizes of 0.99, 1.10, and 0.89 for CBGT, phenelzine, and placebo, respectively. For Subscale 1, effect sizes for the three conditions of 0.90, 1.20, and .82 were obtained. For Subscale 2, effect sizes of 0.70, .94, and .67 were obtained. Subscale 2’s effect sizes for the CBGT and placebo conditions were the only ones to fall below Cohen’s suggested cutoff for a large effect size (0.80), as they were in the medium effect size range. It should be noted that these are uncontrolled effect sizes, which are less conservative than controlled effect sizes.