- Email: [email protected]
Support for the general and specific bifactor model factors of anxiety sensitivity
Personality and Individual Differences 74 (2015) 78–83
Contents lists available at ScienceDirect
Personality and Individual Differences journal homepage: www.elsevier.com/locate/paid
Support for the general and specific bifactor model factors of anxiety sensitivity Nicholas P. Allan ⇑, Brian J. Albanese, Nicole A. Short, Amanda M. Raines, Norman B. Schmidt ⇑ Department of Psychology, Florida State University, Tallahassee, FL 32306-4301, USA
a r t i c l e
i n f o
Article history: Received 16 June 2014 Received in revised form 29 September 2014 Accepted 4 October 2014
Keywords: Anxiety sensitivity Bifactor Emotional distress disorders
a b s t r a c t Anxiety sensitivity (AS), has been conceptualized as a hierarchical construct, comprising three lowerorder dimensions. Recent findings suggest that AS may be better conceptualized as a general dimension and unrelated physical, cognitive, and social concerns dimensions (a bifactor model). The current study was designed to examine whether a bifactor model best represented AS in a sample of 878 college-age participants (Mage = 19.01, SD = 1.45). Further, given that specific relations between lower-order AS dimensions and emotional distress have been found (i.e., physical concerns and fear-based emotional distress, cognitive concerns and distress-based emotional distress), specificity between AS factors and negative affect (NA), worry, depression, social anxiety, and panic attacks was examined. The bifactor model fit the data best. Further, of all the AS factors, the general factor was most associated with NA. Accounting for general AS, cognitive concerns was related to worry and depression and social concerns was related to worry, depression, and social anxiety. Physical concerns was not related to emotional distress. These findings indicate that AS consists of a general facet, associated with emotional distress generally, and several facets more specifically associated with components of emotional distress. Ó 2014 Elsevier Ltd. All rights reserved.
1. Introduction Anxiety Sensitivity (AS) is a well-established trait-like construct reflecting a propensity to fear sensations associated with anxious arousal (Reiss & McNally, 1985). AS was initially studied as a risk factor for panic and agoraphobia. However, heightened levels of AS have now been linked to other anxiety disorders and depression (Naragon-Gainey, 2010; Olatunji & Wolitzky-Taylor, 2009), substance abuse (Schmidt, Buckner, & Keough, 2007), and increased suicidality (Capron et al., 2012). As such, AS may function as a transdiagnostic risk factor for a broad range of psychopathology. To best utilize AS as a transdiagnostic risk factor, it is important to fully understand the underlying structure of this construct. AS is typically conceptualized as a hierarchical construct comprising three distinct but related lower-order dimensions: physical concerns (fears of physiological arousal), cognitive concerns (fears of mental capacitation), and social concerns (fears of publically observable symptoms of anxiety). Although historically, AS was most commonly measured with the Anxiety Sensitivity Index (ASI; Reiss, Peterson, Gursky, & McNally, 1986), factor analytic studies ⇑ Corresponding authors at: Department of Psychology, Florida State University, P.O. Box 3064301, Tallahassee, FL 32306-4301, USA (N.B. Schmidt). E-mail addresses: [email protected] (N.P. Allan), [email protected] (N.B. Schmidt). http://dx.doi.org/10.1016/j.paid.2014.10.003 0191-8869/Ó 2014 Elsevier Ltd. All rights reserved.
have demonstrated that the recently developed Anxiety Sensitivity Index-3 (ASI-3; Taylor et al., 2007) appears to best capture the lower-order dimensions of AS (Allan et al., in press; Taylor et al., 2007; Wheaton, Deacon, McGrath, Berman, & Abramowitz, 2012). However, researchers have recently questioned whether this hierarchical model best represents the structure of AS (e.g., Ebesutani, McLeish, Luberto, Young, & Maack, 2014; Osman et al., 2010). It has recently been argued that AS may be best represented as a bifactor model. Bifactor models suggest the presence of a single general factor reflecting the common variance among all manifest variables (i.e., items), as well as orthogonal factors reflecting the variance among clusters of items (Reise, 2012). The general factor represents the broad construct being measured (e.g., AS), and group factors represent more narrow constructs (e.g., physical, cognitive, and social concerns). This is in contrast to hierarchical models that conceptualizes the general factor as a higher-order factor and the group factors as lower-order oblique factors. Several recent studies, have found that a bifactor model fits the structure of AS better than a hierarchical model (Ebesutani et al., 2014; Osman et al., 2010). However, given the relative paucity of studies examining a bifactor model of AS, it is important to replicate these findings, particularly since there is still a need to provide validity for the distinct AS factors. Whereas Ebesutani et al. (2014) found that a bifactor model of AS best fit the data, they argued that the AS should be classified as
N.P. Allan et al. / Personality and Individual Differences 74 (2015) 78–83
a unidimensional construct. Ebesutani et al. (2014) argued for AS as unidimensional because only the general AS construct was related to an external measure of anxiety, trait anxiety (as measured by the trait scale of the State-Trait Anxiety Inventory [STAI]). However, given that trait anxiety, especially as measured by the STAI, has shown non-specific relations with emotional distress disorders (i.e., mood and anxiety; e.g., Kennedy, Schwab, Morris, & Beldia, 2001), this may not be the most appropriate measure to provide external validation for AS physical, cognitive, and social concerns dimensions. In contrast to Ebesutani et al. (2014) and Osman et al. (2010) argued that the AS could be used as a unidimensional measure of AS, but also that there was support for the specific factors as well as these specific factors were correlated with external mood and anxiety measures. Emotional distress disorder symptoms provide utility for determining external validity of the physical, cognitive, and social concerns dimensions. Distress disorders are those that are characterized by pervasive sadness and worry, such as generalized anxiety disorder (GAD) and major depressive disorder (MDD). Fear disorders are those characterized by phobic avoidance of external threats, such as panic disorder (PD) and specific phobia (Clark & Watson, 2006; Krueger, 1999; Sellbom, Ben-Porath, & Bagby, 2008). Based on evidence that AS physical and cognitive concerns dimensions relate to symptoms of disorders classified as fear and distress emotional distress disorders, respectively, Allan et al. (in press) examined the relations between correlated lower-order AS factors and fear and distress disorders in a community sample of smokers. Results indicated AS physical and cognitive concerns uniquely predicted fear and distress factors, respectively. Further, AS social concerns significantly predicted both the fear and distress factors, suggesting that AS social concerns may confer a more general risk for emotional distress disorders. However, in another study examining the relations between the lower-order dimensions of AS and anxiety and depression, AS social concerns was the only unique predictor of social anxiety (Allan, Capron, Raines, & Schmidt, 2014), suggesting at least some degree of specificity to social anxiety for the AS social concerns dimension. 1.1. Current study The current study was designed to replicate the findings that AS is better represented as a bifactor model including a general AS component and distinct AS physical concerns, AS cognitive concerns, and AS social concerns factors than as a correlated factors model (Ebesutani et al., 2014; Osman et al., 2010). It was hypothesized that a bifactor model would provide the best fit of the ASI-3. The present study was also designed to provide validity for the specific AS factors by examining the relations between the AS factors and general NA, and several symptom measures that have been linked to the specific factors of AS (i.e., worry, social anxiety, depression, and panic attacks; e.g., Allan et al., 2014, in press; Olatunji & Wolitzky-Taylor, 2009). It was hypothesized that NA would be related to the common AS factor only. It was also hypothesized that the common AS factor would be associated with all measures of psychopathology. It was further hypothesized that specific AS factors would be uniquely associated with specific fear and distress constructs. Although variance common to all the items is hypothesized to be best accounted for by a general factor, variance unique to the specific AS domains should still reflect more circumscribed risk not fully captured by the common AS factor (i.e., items specific to AS cognitive concerns might reflect cognitive biases). Given that AS cognitive concerns appears to be associated with and worry and depression have been implicated as distress facets of emotional distress disorders (e.g., Allan et al., in press), it was expected that significant relations would be found for AS
79
cognitive concerns and worry and depression. Although AS social concerns appears to generalize to most emotional distress disorders symptoms, it was hypothesized that AS social concerns would be significantly associated with social anxiety only as it is thought that the general links between AS social concerns and emotional distress disorders is better explained by the AS common variance. Finally, given that AS physical concerns appears to be associated with and social anxiety and panic attacks have been implicated as fear facets of emotional distress disorders (e.g., Allan et al., in press), it was hypothesized that AS physical concerns would be significantly associated with social anxiety and panic attacks. 2. Methods 2.1. Participants The sample included 878 participants recruited from a large southern university. Participants were primarily female (65.3% female) with ages ranging from 17 to 33 (M = 19.01, SD = 1.45). The racial composition of the sample was distributed as such: 81.5% Caucasian, 7.7% African American, 3.1% Asian, .1% American Indian, 6.3% other (e.g., bi-racial) and 1.3% declined to respond. Regarding ethnicity, 83.7% of the sample identified as nonhispanic. 2.2. Measures 2.2.1. Anxiety sensitivity Anxiety sensitivity was assessed using the ASI-3, an 18-item self-report questionnaire measuring feared consequences of anxious arousal (Taylor et al., 2007). The ASI-3 is composed of three subscales, physical concerns, cognitive concerns, and social concerns. Previous research has demonstrated that the ASI-3 is a psychometrically sound and valid measure of AS (Taylor et al., 2007). Within the current investigation, the ASI-3 and the physical, social, and cognitive concerns subscales demonstrated good to excellent internal consistency (a’s = .91, .82, .80, and .88, respectively). 2.2.2. Depression Depression was measured using the Beck Depression Inventory2 (BDI-2). The BDI-2 is a 21-items self-report questionnaire assessing various symptoms of depression experienced over the past two weeks (Beck, Steer, & Carbin, 1988). The BDI-2 has strong psychometric properties, which include high internal consistency and good test–retest reliability (Beck et al., 1988). The BDI-2 demonstrated excellent internal consistency (a = .90) in the present investigation. 2.2.3. Negative affect Negative affect was measured using ten items from the Negative Affect (NA) scale of the Positive and Negative Affect Schedule-Expanded Form (PANAS-X; Watson & Clark, 1999). Previous research has demonstrated acceptable internal consistency, temporal reliability, and convergent validity for the PANAS-X (Watson, 1999). Within the current investigation, the 10-item NA subscale demonstrated good internal consistency (a = .90). This measure was administered to only 66% of the sample, although there were no differences across levels of the other measures or demographics for those who received the NA subscale versus those who did not. 2.2.4. Panic Panic symptoms were assessed using the Panic Disorder Severity Scale (PDSS). The PDSS is a 7-item self-report questionnaire assessing various panic related variables such as frequency of panic attacks, fear and avoidance, and impairment in occupational and
80
N.P. Allan et al. / Personality and Individual Differences 74 (2015) 78–83
social domains (Shear et al., 1997). The PDSS has been demonstrated to be a reliable and valid measure of panic symptoms (Houck, Spiegel, Shear, & Rucci, 2002). In the current study, only item one of the PDSS was used to measure panic attack frequency. A single-item self-report measure of having a panic attack has demonstrated high agreement with a structured clinical interview diagnosis of panic attacks (j = .64; Schmidt, Lerew, & Jackson, 1997). This item was only available for 66% of the sample. Similar to NA, no differences in the other measures or on demographics were observed for participants who did and did not receive this item. 2.2.5. Social anxiety Symptoms of social anxiety were measured using the Social Interaction Anxiety Scale (SIAS). The SIAS is a 20-item self-report measure designed to assess fears in numerous social interaction situations (Mattick & Clarke, 1998). Prior research has demonstrated that the SIAS has good internal consistency and test–retest reliability (Mattick & Clarke, 1998). The SIAS demonstrated excellent internal consistency in the present investigation (a = .93). 2.2.6. Worry The Penn State Worry Questionnaire (PSWQ) is a 16-item scale measuring symptoms related to generalized anxiety disorder (Meyer, Miller, Metzger, & Borkovec, 1990). The PSWQ has excellent internal consistency and excellent test–retest reliability over an 8–10 week period (Meyer et al., 1990). The PSWQ demonstrated excellent internal consistency (a = .93) in the present study. 2.3. Procedure Participants were recruited from the undergraduate psychology research pool. Upon arrival to the laboratory, participants were consented and then completed a battery of self-report questionnaires. The study took approximately 60 min to complete. Upon completion, participants were debriefed and dismissed. Participants received course credit as compensation for their participation. All study procedures were approved by the university’s Institutional Review Board. 2.4. Data analytic plan ASI-3 models were examined in Mplus version 5.2 (Muthén & Muthén, 1996–2008), treating the items as categorical with the robust weighted least squares estimator (WLSMV). Missing values are treated as pairwise missing by convention with this estimator. A series of nested models were compared using the DIFFTEST option in Mplus. The one-factor model (all items on a single factor) was nested in the three-factor model (covarying AS physical, cognitive, and social concerns factors), which was nested in the bifactor model (orthogonal general AS and AS physical, AS cognitive, and AS social concerns factors).1 Overall model fit was examined using the v2 statistic. A nonsignificant value indicates good model fit. However, the v2 statistic can be too restrictive, especially with larger sample sizes and many items per factor (Hu & Bentler, 1999; Moshagen, 2012; Mulaik, 2007). Therefore, the comparative fit index (CFI), and the root mean square error of approximation (RMSEA) were also used to provide additional model fit information. CFI values greater than .90 and RMSEA values less than .08 indicate adequate fit and CFI values greater than .95 and RMSEA values less than .06 indicate good fit (Brown, 2006; Browne & Cudeck, 1992; Yu, 2002). 1 The correlated factors model can be obtained from the bifactor model by freeing the orthogonality constraint on the AS physical, cognitive, and social concerns factors and fixing the loadings of the items on the general AS factor to 0 (see Reise, 2012).
Once the best-fitting model was determined, separate structural equation models (SEMs) were conducted to examine (1) the relations between the AS factors and NA, (2) the relations between the AS factors and PSWQ Worry, SIAS Social Anxiety, and BDI-2 Depression, and (3) the presence of panic attacks, treated as a dichotomous variable (0 = no panic attacks, 1 = one or more panic attacks). Models for NA and panic attacks were conducted separately because these measures were only available for a subsample of the full sample. 3. Results 3.1. Descriptive statistics, correlations, and data examination All continuous outcome variables (i.e., NA, PSWQ Worry, SIAS Social Anxiety, and BDI-2 Depression) were examined for violations of normality according to procedures recommended by Tabachnick and Fidell (2007). All continuous variables were first examined for outliers, classified as values ±3 standard deviations (SDs) from the mean. There were several outliers at the upper boundary, including 6 each for NA and SIAS Social Anxiety and 12 for BDI-2 Depression. These values were bounded at +3 SD from their respective mean. Following this, outcome variables were examined for skew and kurtosis. All variables demonstrated significant skew (i.e., skew divided by standard error of skew greater than 2; Tabachnick & Fidell, 2007) and the appropriate transformations were conducted. NA was log transformed and PSWQ Worry, SIAS Social Anxiety, and BDI-2 Depression were square-root transformed. Means for the ASI-3 lower-order scales, the untransformed NA, PSWQ Worry, SIAS Social Anxiety, and BDI-2 Depression scales are provided in Table 1 as well as correlations between all variables, including the dichotomous Panic Attacks variable (25.2% presented as having at least mild or limited symptom attacks less than once a week, up to extreme panic attacks more than once a day). No participants were missing item-level data on the ASI-3. For the outcome variables, missing data was present for one, one, and two participants, respectively on PSWQ Worry, SIAS Social Anxiety, and BDI-2 Depression. These participants were included in the analyses as the WLSMV estimator is robust to missing data. For PANAS NA and panic attacks, 295 and 291 participants were missing data because they did not receive the measure. Data for these participants were not included in these analyses. Therefore, the analysis including PANAS NA was based on 583 participants and the analysis including panic attacks was based on 587 participants. 3.2. Confirmatory factor analysis of ASI-3 CFAs of the one-factor, the three-factor, and the bifactor model of the ASI-3 were compared sequentially (see Table 2). The bifactor model was the best-fitting model. This model demonstrated significantly improved model fit as compared to the three-factor model (Dv2 = 80.71, df = 13, p < .001). In addition, the model demonstrated adequate overall model fit (v2 = 338.15, p < .05, CFI = .95, RMSEA = .07). All items loaded significantly on both the global AS factor and their specific factors (i.e., AS physical concerns, AS cognitive concerns, AS social concerns; see Fig. 1 for loadings [loadings are from SEM model, but do not differ substantively from loadings in the bifactor model]). 3.3. Structural equation models examining relations between ASI-3 factors and NA and psychopathology A SEM model including the associations between the general AS factor as well as the specific factors and NA demonstrated ade-
81
N.P. Allan et al. / Personality and Individual Differences 74 (2015) 78–83 Table 1 Descriptive statistics for ASI-3 subscales, NA, and mood and anxiety scales.
1 2 3 4 5 6 7 8
ASI-3 physical ASI-3 cognitive ASI-3 social PANAS NA PSWQ Worry SIAS Social Anxiety BDI-2 CYPAS panic attacks
Mean (% panic attacks) Standard deviation
1
2
3
– .66* .56* .35* .43* .35* .36* .32*
– .51* .39* .41* .38* .43* .31*
– .36* .44* .54* .39* .31*
4.14 4.10
2.89 4.06
7.37 4.89
4
5
6
7
8
– .43* .46* .53* .31* 18.66 6.86
– .43* .48* .41* 46.01 14.00
– .48* .28* 22.79 14.28
– .38* 8.33 7.31
– 25.2%
Note: ASI-3 = Anxiety Sensitivity Index-3. PANAS NA = Positive and Negative Affect Schedule Negative Affect Scale. APPQ Agoraphobia = Albany Panic and Phobia Questionnaire Agoraphobia subscale. PSWQ = Penn State Worry Questionnaire. SIAS Social = Social Interaction Anxiety Scale. BDI-2 = Beck Depression Inventory-2. Panic Attacks were coded dichotomous (0 = no panic attacks, 1 = 1 or more panic attacks). * p < .05.
Table 2 Confirmatory factor analyses of one-factor, three-factor, and bifactor models of ASI-3. Models
v2
Dv2
Ddf
CFI
RMSEA
One-factor Three-factor Bifactor
992.74 385.34 338.15
– 292.54*** 80.71***
– 3 13
.82 .94 .95
.13 .07 .07
Note: CFI = Comparative fit index. WRMR = Weighted root mean square residual. RMSEA = Root mean square error of approximation. Nested models were compared sequentially and are listed in the order that they were compared. Best fitting model in italics. *** p < .001.
quate fit to the data (v2 = 234.54, p < .05, CFI = .96, RMSEA = .07). In this model, significant relations were found for NA and AS (b = .43,
p < .001), AS cognitive concerns (b = .13, p < .05), and AS social concerns (b = .12, p < .01), but not AS physical concerns (b = .01, p > .05). The AS factors accounted for 22% of the variance in NA (p < .001). A SEM model including the associations between the AS factors and PSWQ Worry, SIAS Social Anxiety, and BDI-2 Depression also demonstrated adequate model fit (v2 = 393.37, p < .05, CFI = .93, RMSEA = .06). Parameters for this model are provided in Fig. 1. In this model, significant relations were found for PSWQ Worry and AS (b = .43, p < .001), AS cognitive concerns (b = .13, p < .05), and AS social concerns (b = .19, p < .001). Significant relations were also found for SIAS Social Anxiety and AS (b = .47, p < .001) and AS social concerns (b = .40, p < .001). Finally, significant relations were found for BDI-2 Depression and AS (b = .40, p < .001), AS cognitive concerns (b = .26, p < .001), and AS social concerns (b = .20, p < .001).
Fig. 1. Structural equation model of the relations between anxiety sensitivity factors and measures of mood and anxiety. Residuals are not displayed for clarity.
82
N.P. Allan et al. / Personality and Individual Differences 74 (2015) 78–83
The AS factors accounted for 29% of the variance in PSWQ Worry, 38% of the variance in SIAS Social Anxiety, and 27% of the variance in BDI-2 Depression. Finally, A SEM model including the relations between the AS factors and the odds of presenting with a panic attack provided adequate fit to the data (v2 = 258.10, p < .05, CFI = .95, RMSEA = .07).2 The odds of presenting with a panic attack was significant for the AS factor (OR = 1.66, 95% Confidence Interval CI [1.47, 1.85]). The odds of presenting with a panic attack were not significant for the AS physical concerns (OR = .99, 95% CI [.78, 1.20]), AS cognitive concerns (OR = .93, 95% CI [.71, 1.15]), or AS social concerns factors (OR = 1.14, 95% CI [.97, 1.31]). The AS factors accounted for 28% of the variance in panic attacks (p < .001). 4. Discussion Results of the current study indicate that the structure of AS, as measured by the ASI-3, is best conceptualized as a bifactor model, with orthogonal general and specific factors. These findings are consistent with two prior studies of AS that also compared a bifactor model to the more common hierarchical model of AS (e.g., Ebesutani et al., 2014; Osman et al., 2010). Together, these findings suggest that when the general AS factor is accounted for, AS physical, cognitive, and social concerns are best conceptualized as orthogonal constructs. Ebesutani et al. (2014) speculate that this may indicate that the three AS subdomains may be linked to differential neuro-biological processes involved in processing information and fear events in addition to a more general neurobiological system implicated in the global fear of anxious arousal. Bifactor modeling approaches, combined with recent advances in identification of AS-relevant biomarkers (e.g., Sehlmeyer et al., 2010) allow for researchers to address whether distinct neurobiological domains for the distinct AS subdomains can be identified. The general AS factor was associated with NA, worry, depression, social anxiety, and panic attacks. These findings mirror past findings implicating AS as a risk factor for most emotional distress disorders (Naragon-Gainey, 2010; Olatunji & Wolitzky-Taylor, 2009). In addition, these findings replicate the findings of Ebesutani et al. (2014), who reported that general AS was associated with state anxiety. Together, these findings provide support for the role of AS as a transdiagnostic risk factor for emotional distress disorders. Regarding the more specific AS dimensions and the fear and distress emotional distress symptoms, our hypotheses were only partially supported. Accounting for the general AS factor, unique relations emerged for AS cognitive concerns and worry and depression symptoms. In addition, AS social concerns was uniquely associated with worry, depression, and social anxiety, although the relation between AS social concerns and social anxiety appeared to be the most robust. These findings are similar to prior studies by Allan et al. (2014, in press) in which AS cognitive concerns were related to a distress disorder factor as well as specific distress disorders and AS social concerns showed more general relations to mood and anxiety disorders, but, was significantly more associated with social anxiety symptoms than with all other measures of psychopathology, with the exception of depression symptoms (e.g., Allan et al., in press). However, AS physical concerns was not significantly associated with panic attacks in the current study. One possible explanation for this is that, because AS physical concerns is a foundational aspect of AS, the general AS factor best captures 2 In this model ASI-3 item 2 was only allowed to load on the AS factor (and removed from the AS cognitive concerns factor) to account for a linear dependency when this item was allowed to load on both factors. There were no substantive differences in the analysis when this item was omitted from the AS cognitive concerns factor.
AS physical concerns, and including a distinct factor to assess AS physical concerns is unnecessary. There are several implications for the current findings. From a research perspective, the bifactor model appears to be a viable alternative to the correlated factors model if researchers were interested in examining the unique relations between the specific AS factors and external covariates (e.g.,Chen, Hayes, Carver, Laurenceau, & Zhang, 2012). For clinicians interested in using AS as a risk factor or correlate of anxiety and other disorders, the strong associations between the general AS factor and the mood and anxiety symptoms as well as the generally higher factor loadings on the AS factor as compared to the factor loadings on the more specific factors suggests that using the summed ASI-3 total score should prove an adequate measure of risk conferred by this construct. There are several limitations to consider. First, this and the prior two studies in adult populations have used convenience samples. There is a need to examine whether the same factor structure emerges in samples containing participants with clinically elevated levels of mood and anxiety disorders. This would be expected given that the tested factor structure of the ASI-3 (i.e., the hierarchical and correlated factors models) is similar in clinical and community samples (e.g., Taylor et al., 2007). Second, only a subset of fear and distress disorders were available for inclusion in the present sample. Finally, the current study was conducted with cross-sectional data. Examining the role of the AS factors prospectively would provide more support for a bifactor modeling approach for this construct. There are several strengths of the current study. Support for a bifactor model of AS was found in the current study. In addition, external validity of the factors was provided. The general AS factor was associated with general NA as well as more specific mood and anxiety disorders. Specificity was also found for the more narrowband AS cognitive concerns and social concerns dimensions, but not for the AS physical concerns dimension. This suggests that whereas there is something domain specific being captured by the AS cognitive and social concerns dimensions above and beyond general AS this is not the case for AS physical concerns. References Allan, N. P., Capron, D. W., Raines, A. M., & Schmidt, N. B. (2014). Unique relations among anxiety sensitivity factors and anxiety, depression, and suicidal ideation. Journal of Anxiety Disorders, 28, 266–275. Allan, N. P., Norr, A. M., Capron, D. W., Raines, A. M., Zvolensky, M. J., & Schmidt, N. B. (in press). Specific associations between anxiety sensitivity dimensions and fear and distress dimensions of emotional distress disorders. Journal of Psychopathology and Behavior Assessment. Beck, A. T., Steer, R. A., & Carbin, M. G. (1988). Psychometric properties of the Beck Depression Inventory: Twenty-five years of evaluation. Clinical Psychology Review, 8, 77–100. Brown, T. A. (2006). Confirmatory factor analysis for applied research. New York, NY: The Guilford Press. Browne, M. W., & Cudeck, R. (1992). Alternative ways of assessing model fit. Sociological Methods and Research, 21, 230–258. Capron, D. W., Fitch, K., Medley, A., Blagg, C., Mallott, M., & Joiner, T. (2012). Role of anxiety sensitivity subfactors in suicidal ideation and suicide attempt history. Depression and Anxiety, 29, 195–201. Chen, F. F., Hayes, A., Carver, C. S., Laurenceau, J. P., & Zhang, Z. (2012). Modeling general and specific variance in multifaceted constructs: A comparison of the bifactor model to other approaches. Journal of Personality, 80, 219–251. Clark, L. A., & Watson, D. (2006). Distress and fear disorders: An alternative empirically based taxonomy of the ‘‘mood’’ and ‘‘anxiety’’ disorders. The British Journal of Psychiatry: The Journal of Mental Science, 189, 481–483. Ebesutani, C., McLeish, A. C., Luberto, C. M., Young, J., & Maack, D. J. (2014). A bifactor model of anxiety sensitivity: Analysis of the Anxiety Sensitivity Index3. Journal of Psychopathology and Behavioral Assessment, 36, 452–464. Houck, P. R., Spiegel, D. A., Shear, M. K., & Rucci, P. (2002). Reliability of the selfreport version of the panic disorder severity scale. Depression and Anxiety, 15, 183–185. Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6, 1–55.
N.P. Allan et al. / Personality and Individual Differences 74 (2015) 78–83 Kennedy, B. L., Schwab, J. J., Morris, R. L., & Beldia, G. (2001). Assessment of state and trait anxiety in subjects with anxiety and depressive disorders. The Psychiatric Quarterly, 72, 263–276. Krueger, R. F. (1999). The structure of common mental disorders. Archives of General Psychiatry, 56, 921–926. Mattick, R. P., & Clarke, J. C. (1998). Development and validation of measures of social phobia scrutiny fear and social interaction anxiety. Behaviour Research and Therapy, 36, 455–470. Meyer, T. J., Miller, M. L., Metzger, R. L., & Borkovec, T. D. (1990). Development and validation of the Penn State Worry Questionnaire. Behaviour Research and Therapy, 28, 487–495. Moshagen, M. (2012). The model size effect in SEM: Inflated goodness-of-fit statistics are due to the size of the covariance matrix. Structural Equation Modeling: A Multidisciplinary Journal, 19, 86–98. Mulaik, S. (2007). There is a place for approximate fit in structural equation modelling. Personality and Individual Differences, 42, 883–891. Muthén, B. O., & Muthén, L. K. (1996–2008). Mplus (Version 5.2). Los Angeles, CA. Naragon-Gainey, K. (2010). Meta-analysis of the relations of anxiety sensitivity to the depressive and anxiety disorders. Psychological Bulletin, 136, 128–150. Olatunji, B. O., & Wolitzky-Taylor, K. B. (2009). Anxiety sensitivity and the anxiety disorders: A meta-analytic review and synthesis. Psychological Bulletin, 135, 974–999. Osman, A., Gutierrez, P. M., Smith, K., Fang, Q., Lozano, G., & Devine, A. (2010). The Anxiety Sensitivity Index-3: Analyses of dimensions, reliability estimates, and correlates in nonclinical samples. Journal of Personality Assessment, 92, 45–52. Reise, S. P. (2012). The rediscovery of bifactor measurement models. Multivariate Behavioral Research, 47, 667–696. Reiss, S., & McNally, R. (1985). Expectancy model of fear. In S. Reiss & R. Bootzin (Eds.), Theoretical Issues in Behavior Therapy (pp. 107–121). San Diego, CA: Academic. Reiss, S., Peterson, R. A., Gursky, D. M., & McNally, R. J. (1986). Anxiety sensitivity, anxiety frequency and the prediction of fearfulness. Behaviour Research and Therapy, 24, 1–8.
83
Schmidt, N. B., Buckner, J. D., & Keough, M. E. (2007). Anxiety sensitivity as a prospective predictor of alcohol use disorders. Behavior Modification, 31, 202–219. Schmidt, N. B., Lerew, D. R., & Jackson, R. J. (1997). The role of anxiety sensitivity in the pathogenesis of panic: Prospective evaluation of spontaneous panic attacks during acute stress. Journal of Abnormal Psychology, 106, 355–364. Sehlmeyer, C., Konrad, C., Zwitserlood, P., Arolt, V., Falkenstein, M., & Beste, C. (2010). ERP indices for response inhibition are related to anxiety-related personality traits. Neuropsychologia, 48, 2488–2495. Sellbom, M., Ben-Porath, Y. S., & Bagby, R. M. (2008). On the hierarchical structure of mood and anxiety disorders: Confirmatory evidence and elaboration of a model of temperament markers. Journal of Abnormal Psychology, 117, 576–590. Shear, M. K., Brown, T. A., Barlow, D. H., Money, R., Sholomskas, D. E., Woods, S. W., et al. (1997). Multicenter collaborative panic disorder severity scale. The American Journal of Psychiatry, 154, 1571–1575. Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics. Boston, MA: Pearson Education Inc.. Taylor, S., Zvolensky, M. J., Cox, B. J., Deacon, B., Heimberg, R. G., Ledley, D. R., et al. (2007). Robust dimensions of anxiety sensitivity: Development and initial validation of the Anxiety Sensitivity Index-3. Psychological Assessment, 19, 176–188. Watson, D., & Clark, L. A. (1999). The PANAS-X: Manual for the positive and negative affect schedule-expanded form. Ames: University of Iowa. Wheaton, M. G., Deacon, B. J., McGrath, P. B., Berman, N. C., & Abramowitz, J. S. (2012). Dimensions of anxiety sensitivity in the anxiety disorders: Evaluation of the ASI-3. Journal of Anxiety Disorders, 26, 401–408. Yu, C. Y. (2002). Evaluating cutoff criteria of model fit indices for latent variable models with binary and continuous outcomes (Doctoral dissertation). Los Angeles: University of California.
Contents lists available at ScienceDirect
Personality and Individual Differences journal homepage: www.elsevier.com/locate/paid
Support for the general and specific bifactor model factors of anxiety sensitivity Nicholas P. Allan ⇑, Brian J. Albanese, Nicole A. Short, Amanda M. Raines, Norman B. Schmidt ⇑ Department of Psychology, Florida State University, Tallahassee, FL 32306-4301, USA
a r t i c l e
i n f o
Article history: Received 16 June 2014 Received in revised form 29 September 2014 Accepted 4 October 2014
Keywords: Anxiety sensitivity Bifactor Emotional distress disorders
a b s t r a c t Anxiety sensitivity (AS), has been conceptualized as a hierarchical construct, comprising three lowerorder dimensions. Recent findings suggest that AS may be better conceptualized as a general dimension and unrelated physical, cognitive, and social concerns dimensions (a bifactor model). The current study was designed to examine whether a bifactor model best represented AS in a sample of 878 college-age participants (Mage = 19.01, SD = 1.45). Further, given that specific relations between lower-order AS dimensions and emotional distress have been found (i.e., physical concerns and fear-based emotional distress, cognitive concerns and distress-based emotional distress), specificity between AS factors and negative affect (NA), worry, depression, social anxiety, and panic attacks was examined. The bifactor model fit the data best. Further, of all the AS factors, the general factor was most associated with NA. Accounting for general AS, cognitive concerns was related to worry and depression and social concerns was related to worry, depression, and social anxiety. Physical concerns was not related to emotional distress. These findings indicate that AS consists of a general facet, associated with emotional distress generally, and several facets more specifically associated with components of emotional distress. Ó 2014 Elsevier Ltd. All rights reserved.
1. Introduction Anxiety Sensitivity (AS) is a well-established trait-like construct reflecting a propensity to fear sensations associated with anxious arousal (Reiss & McNally, 1985). AS was initially studied as a risk factor for panic and agoraphobia. However, heightened levels of AS have now been linked to other anxiety disorders and depression (Naragon-Gainey, 2010; Olatunji & Wolitzky-Taylor, 2009), substance abuse (Schmidt, Buckner, & Keough, 2007), and increased suicidality (Capron et al., 2012). As such, AS may function as a transdiagnostic risk factor for a broad range of psychopathology. To best utilize AS as a transdiagnostic risk factor, it is important to fully understand the underlying structure of this construct. AS is typically conceptualized as a hierarchical construct comprising three distinct but related lower-order dimensions: physical concerns (fears of physiological arousal), cognitive concerns (fears of mental capacitation), and social concerns (fears of publically observable symptoms of anxiety). Although historically, AS was most commonly measured with the Anxiety Sensitivity Index (ASI; Reiss, Peterson, Gursky, & McNally, 1986), factor analytic studies ⇑ Corresponding authors at: Department of Psychology, Florida State University, P.O. Box 3064301, Tallahassee, FL 32306-4301, USA (N.B. Schmidt). E-mail addresses: [email protected] (N.P. Allan), [email protected] (N.B. Schmidt). http://dx.doi.org/10.1016/j.paid.2014.10.003 0191-8869/Ó 2014 Elsevier Ltd. All rights reserved.
have demonstrated that the recently developed Anxiety Sensitivity Index-3 (ASI-3; Taylor et al., 2007) appears to best capture the lower-order dimensions of AS (Allan et al., in press; Taylor et al., 2007; Wheaton, Deacon, McGrath, Berman, & Abramowitz, 2012). However, researchers have recently questioned whether this hierarchical model best represents the structure of AS (e.g., Ebesutani, McLeish, Luberto, Young, & Maack, 2014; Osman et al., 2010). It has recently been argued that AS may be best represented as a bifactor model. Bifactor models suggest the presence of a single general factor reflecting the common variance among all manifest variables (i.e., items), as well as orthogonal factors reflecting the variance among clusters of items (Reise, 2012). The general factor represents the broad construct being measured (e.g., AS), and group factors represent more narrow constructs (e.g., physical, cognitive, and social concerns). This is in contrast to hierarchical models that conceptualizes the general factor as a higher-order factor and the group factors as lower-order oblique factors. Several recent studies, have found that a bifactor model fits the structure of AS better than a hierarchical model (Ebesutani et al., 2014; Osman et al., 2010). However, given the relative paucity of studies examining a bifactor model of AS, it is important to replicate these findings, particularly since there is still a need to provide validity for the distinct AS factors. Whereas Ebesutani et al. (2014) found that a bifactor model of AS best fit the data, they argued that the AS should be classified as
N.P. Allan et al. / Personality and Individual Differences 74 (2015) 78–83
a unidimensional construct. Ebesutani et al. (2014) argued for AS as unidimensional because only the general AS construct was related to an external measure of anxiety, trait anxiety (as measured by the trait scale of the State-Trait Anxiety Inventory [STAI]). However, given that trait anxiety, especially as measured by the STAI, has shown non-specific relations with emotional distress disorders (i.e., mood and anxiety; e.g., Kennedy, Schwab, Morris, & Beldia, 2001), this may not be the most appropriate measure to provide external validation for AS physical, cognitive, and social concerns dimensions. In contrast to Ebesutani et al. (2014) and Osman et al. (2010) argued that the AS could be used as a unidimensional measure of AS, but also that there was support for the specific factors as well as these specific factors were correlated with external mood and anxiety measures. Emotional distress disorder symptoms provide utility for determining external validity of the physical, cognitive, and social concerns dimensions. Distress disorders are those that are characterized by pervasive sadness and worry, such as generalized anxiety disorder (GAD) and major depressive disorder (MDD). Fear disorders are those characterized by phobic avoidance of external threats, such as panic disorder (PD) and specific phobia (Clark & Watson, 2006; Krueger, 1999; Sellbom, Ben-Porath, & Bagby, 2008). Based on evidence that AS physical and cognitive concerns dimensions relate to symptoms of disorders classified as fear and distress emotional distress disorders, respectively, Allan et al. (in press) examined the relations between correlated lower-order AS factors and fear and distress disorders in a community sample of smokers. Results indicated AS physical and cognitive concerns uniquely predicted fear and distress factors, respectively. Further, AS social concerns significantly predicted both the fear and distress factors, suggesting that AS social concerns may confer a more general risk for emotional distress disorders. However, in another study examining the relations between the lower-order dimensions of AS and anxiety and depression, AS social concerns was the only unique predictor of social anxiety (Allan, Capron, Raines, & Schmidt, 2014), suggesting at least some degree of specificity to social anxiety for the AS social concerns dimension. 1.1. Current study The current study was designed to replicate the findings that AS is better represented as a bifactor model including a general AS component and distinct AS physical concerns, AS cognitive concerns, and AS social concerns factors than as a correlated factors model (Ebesutani et al., 2014; Osman et al., 2010). It was hypothesized that a bifactor model would provide the best fit of the ASI-3. The present study was also designed to provide validity for the specific AS factors by examining the relations between the AS factors and general NA, and several symptom measures that have been linked to the specific factors of AS (i.e., worry, social anxiety, depression, and panic attacks; e.g., Allan et al., 2014, in press; Olatunji & Wolitzky-Taylor, 2009). It was hypothesized that NA would be related to the common AS factor only. It was also hypothesized that the common AS factor would be associated with all measures of psychopathology. It was further hypothesized that specific AS factors would be uniquely associated with specific fear and distress constructs. Although variance common to all the items is hypothesized to be best accounted for by a general factor, variance unique to the specific AS domains should still reflect more circumscribed risk not fully captured by the common AS factor (i.e., items specific to AS cognitive concerns might reflect cognitive biases). Given that AS cognitive concerns appears to be associated with and worry and depression have been implicated as distress facets of emotional distress disorders (e.g., Allan et al., in press), it was expected that significant relations would be found for AS
79
cognitive concerns and worry and depression. Although AS social concerns appears to generalize to most emotional distress disorders symptoms, it was hypothesized that AS social concerns would be significantly associated with social anxiety only as it is thought that the general links between AS social concerns and emotional distress disorders is better explained by the AS common variance. Finally, given that AS physical concerns appears to be associated with and social anxiety and panic attacks have been implicated as fear facets of emotional distress disorders (e.g., Allan et al., in press), it was hypothesized that AS physical concerns would be significantly associated with social anxiety and panic attacks. 2. Methods 2.1. Participants The sample included 878 participants recruited from a large southern university. Participants were primarily female (65.3% female) with ages ranging from 17 to 33 (M = 19.01, SD = 1.45). The racial composition of the sample was distributed as such: 81.5% Caucasian, 7.7% African American, 3.1% Asian, .1% American Indian, 6.3% other (e.g., bi-racial) and 1.3% declined to respond. Regarding ethnicity, 83.7% of the sample identified as nonhispanic. 2.2. Measures 2.2.1. Anxiety sensitivity Anxiety sensitivity was assessed using the ASI-3, an 18-item self-report questionnaire measuring feared consequences of anxious arousal (Taylor et al., 2007). The ASI-3 is composed of three subscales, physical concerns, cognitive concerns, and social concerns. Previous research has demonstrated that the ASI-3 is a psychometrically sound and valid measure of AS (Taylor et al., 2007). Within the current investigation, the ASI-3 and the physical, social, and cognitive concerns subscales demonstrated good to excellent internal consistency (a’s = .91, .82, .80, and .88, respectively). 2.2.2. Depression Depression was measured using the Beck Depression Inventory2 (BDI-2). The BDI-2 is a 21-items self-report questionnaire assessing various symptoms of depression experienced over the past two weeks (Beck, Steer, & Carbin, 1988). The BDI-2 has strong psychometric properties, which include high internal consistency and good test–retest reliability (Beck et al., 1988). The BDI-2 demonstrated excellent internal consistency (a = .90) in the present investigation. 2.2.3. Negative affect Negative affect was measured using ten items from the Negative Affect (NA) scale of the Positive and Negative Affect Schedule-Expanded Form (PANAS-X; Watson & Clark, 1999). Previous research has demonstrated acceptable internal consistency, temporal reliability, and convergent validity for the PANAS-X (Watson, 1999). Within the current investigation, the 10-item NA subscale demonstrated good internal consistency (a = .90). This measure was administered to only 66% of the sample, although there were no differences across levels of the other measures or demographics for those who received the NA subscale versus those who did not. 2.2.4. Panic Panic symptoms were assessed using the Panic Disorder Severity Scale (PDSS). The PDSS is a 7-item self-report questionnaire assessing various panic related variables such as frequency of panic attacks, fear and avoidance, and impairment in occupational and
80
N.P. Allan et al. / Personality and Individual Differences 74 (2015) 78–83
social domains (Shear et al., 1997). The PDSS has been demonstrated to be a reliable and valid measure of panic symptoms (Houck, Spiegel, Shear, & Rucci, 2002). In the current study, only item one of the PDSS was used to measure panic attack frequency. A single-item self-report measure of having a panic attack has demonstrated high agreement with a structured clinical interview diagnosis of panic attacks (j = .64; Schmidt, Lerew, & Jackson, 1997). This item was only available for 66% of the sample. Similar to NA, no differences in the other measures or on demographics were observed for participants who did and did not receive this item. 2.2.5. Social anxiety Symptoms of social anxiety were measured using the Social Interaction Anxiety Scale (SIAS). The SIAS is a 20-item self-report measure designed to assess fears in numerous social interaction situations (Mattick & Clarke, 1998). Prior research has demonstrated that the SIAS has good internal consistency and test–retest reliability (Mattick & Clarke, 1998). The SIAS demonstrated excellent internal consistency in the present investigation (a = .93). 2.2.6. Worry The Penn State Worry Questionnaire (PSWQ) is a 16-item scale measuring symptoms related to generalized anxiety disorder (Meyer, Miller, Metzger, & Borkovec, 1990). The PSWQ has excellent internal consistency and excellent test–retest reliability over an 8–10 week period (Meyer et al., 1990). The PSWQ demonstrated excellent internal consistency (a = .93) in the present study. 2.3. Procedure Participants were recruited from the undergraduate psychology research pool. Upon arrival to the laboratory, participants were consented and then completed a battery of self-report questionnaires. The study took approximately 60 min to complete. Upon completion, participants were debriefed and dismissed. Participants received course credit as compensation for their participation. All study procedures were approved by the university’s Institutional Review Board. 2.4. Data analytic plan ASI-3 models were examined in Mplus version 5.2 (Muthén & Muthén, 1996–2008), treating the items as categorical with the robust weighted least squares estimator (WLSMV). Missing values are treated as pairwise missing by convention with this estimator. A series of nested models were compared using the DIFFTEST option in Mplus. The one-factor model (all items on a single factor) was nested in the three-factor model (covarying AS physical, cognitive, and social concerns factors), which was nested in the bifactor model (orthogonal general AS and AS physical, AS cognitive, and AS social concerns factors).1 Overall model fit was examined using the v2 statistic. A nonsignificant value indicates good model fit. However, the v2 statistic can be too restrictive, especially with larger sample sizes and many items per factor (Hu & Bentler, 1999; Moshagen, 2012; Mulaik, 2007). Therefore, the comparative fit index (CFI), and the root mean square error of approximation (RMSEA) were also used to provide additional model fit information. CFI values greater than .90 and RMSEA values less than .08 indicate adequate fit and CFI values greater than .95 and RMSEA values less than .06 indicate good fit (Brown, 2006; Browne & Cudeck, 1992; Yu, 2002). 1 The correlated factors model can be obtained from the bifactor model by freeing the orthogonality constraint on the AS physical, cognitive, and social concerns factors and fixing the loadings of the items on the general AS factor to 0 (see Reise, 2012).
Once the best-fitting model was determined, separate structural equation models (SEMs) were conducted to examine (1) the relations between the AS factors and NA, (2) the relations between the AS factors and PSWQ Worry, SIAS Social Anxiety, and BDI-2 Depression, and (3) the presence of panic attacks, treated as a dichotomous variable (0 = no panic attacks, 1 = one or more panic attacks). Models for NA and panic attacks were conducted separately because these measures were only available for a subsample of the full sample. 3. Results 3.1. Descriptive statistics, correlations, and data examination All continuous outcome variables (i.e., NA, PSWQ Worry, SIAS Social Anxiety, and BDI-2 Depression) were examined for violations of normality according to procedures recommended by Tabachnick and Fidell (2007). All continuous variables were first examined for outliers, classified as values ±3 standard deviations (SDs) from the mean. There were several outliers at the upper boundary, including 6 each for NA and SIAS Social Anxiety and 12 for BDI-2 Depression. These values were bounded at +3 SD from their respective mean. Following this, outcome variables were examined for skew and kurtosis. All variables demonstrated significant skew (i.e., skew divided by standard error of skew greater than 2; Tabachnick & Fidell, 2007) and the appropriate transformations were conducted. NA was log transformed and PSWQ Worry, SIAS Social Anxiety, and BDI-2 Depression were square-root transformed. Means for the ASI-3 lower-order scales, the untransformed NA, PSWQ Worry, SIAS Social Anxiety, and BDI-2 Depression scales are provided in Table 1 as well as correlations between all variables, including the dichotomous Panic Attacks variable (25.2% presented as having at least mild or limited symptom attacks less than once a week, up to extreme panic attacks more than once a day). No participants were missing item-level data on the ASI-3. For the outcome variables, missing data was present for one, one, and two participants, respectively on PSWQ Worry, SIAS Social Anxiety, and BDI-2 Depression. These participants were included in the analyses as the WLSMV estimator is robust to missing data. For PANAS NA and panic attacks, 295 and 291 participants were missing data because they did not receive the measure. Data for these participants were not included in these analyses. Therefore, the analysis including PANAS NA was based on 583 participants and the analysis including panic attacks was based on 587 participants. 3.2. Confirmatory factor analysis of ASI-3 CFAs of the one-factor, the three-factor, and the bifactor model of the ASI-3 were compared sequentially (see Table 2). The bifactor model was the best-fitting model. This model demonstrated significantly improved model fit as compared to the three-factor model (Dv2 = 80.71, df = 13, p < .001). In addition, the model demonstrated adequate overall model fit (v2 = 338.15, p < .05, CFI = .95, RMSEA = .07). All items loaded significantly on both the global AS factor and their specific factors (i.e., AS physical concerns, AS cognitive concerns, AS social concerns; see Fig. 1 for loadings [loadings are from SEM model, but do not differ substantively from loadings in the bifactor model]). 3.3. Structural equation models examining relations between ASI-3 factors and NA and psychopathology A SEM model including the associations between the general AS factor as well as the specific factors and NA demonstrated ade-
81
N.P. Allan et al. / Personality and Individual Differences 74 (2015) 78–83 Table 1 Descriptive statistics for ASI-3 subscales, NA, and mood and anxiety scales.
1 2 3 4 5 6 7 8
ASI-3 physical ASI-3 cognitive ASI-3 social PANAS NA PSWQ Worry SIAS Social Anxiety BDI-2 CYPAS panic attacks
Mean (% panic attacks) Standard deviation
1
2
3
– .66* .56* .35* .43* .35* .36* .32*
– .51* .39* .41* .38* .43* .31*
– .36* .44* .54* .39* .31*
4.14 4.10
2.89 4.06
7.37 4.89
4
5
6
7
8
– .43* .46* .53* .31* 18.66 6.86
– .43* .48* .41* 46.01 14.00
– .48* .28* 22.79 14.28
– .38* 8.33 7.31
– 25.2%
Note: ASI-3 = Anxiety Sensitivity Index-3. PANAS NA = Positive and Negative Affect Schedule Negative Affect Scale. APPQ Agoraphobia = Albany Panic and Phobia Questionnaire Agoraphobia subscale. PSWQ = Penn State Worry Questionnaire. SIAS Social = Social Interaction Anxiety Scale. BDI-2 = Beck Depression Inventory-2. Panic Attacks were coded dichotomous (0 = no panic attacks, 1 = 1 or more panic attacks). * p < .05.
Table 2 Confirmatory factor analyses of one-factor, three-factor, and bifactor models of ASI-3. Models
v2
Dv2
Ddf
CFI
RMSEA
One-factor Three-factor Bifactor
992.74 385.34 338.15
– 292.54*** 80.71***
– 3 13
.82 .94 .95
.13 .07 .07
Note: CFI = Comparative fit index. WRMR = Weighted root mean square residual. RMSEA = Root mean square error of approximation. Nested models were compared sequentially and are listed in the order that they were compared. Best fitting model in italics. *** p < .001.
quate fit to the data (v2 = 234.54, p < .05, CFI = .96, RMSEA = .07). In this model, significant relations were found for NA and AS (b = .43,
p < .001), AS cognitive concerns (b = .13, p < .05), and AS social concerns (b = .12, p < .01), but not AS physical concerns (b = .01, p > .05). The AS factors accounted for 22% of the variance in NA (p < .001). A SEM model including the associations between the AS factors and PSWQ Worry, SIAS Social Anxiety, and BDI-2 Depression also demonstrated adequate model fit (v2 = 393.37, p < .05, CFI = .93, RMSEA = .06). Parameters for this model are provided in Fig. 1. In this model, significant relations were found for PSWQ Worry and AS (b = .43, p < .001), AS cognitive concerns (b = .13, p < .05), and AS social concerns (b = .19, p < .001). Significant relations were also found for SIAS Social Anxiety and AS (b = .47, p < .001) and AS social concerns (b = .40, p < .001). Finally, significant relations were found for BDI-2 Depression and AS (b = .40, p < .001), AS cognitive concerns (b = .26, p < .001), and AS social concerns (b = .20, p < .001).
Fig. 1. Structural equation model of the relations between anxiety sensitivity factors and measures of mood and anxiety. Residuals are not displayed for clarity.
82
N.P. Allan et al. / Personality and Individual Differences 74 (2015) 78–83
The AS factors accounted for 29% of the variance in PSWQ Worry, 38% of the variance in SIAS Social Anxiety, and 27% of the variance in BDI-2 Depression. Finally, A SEM model including the relations between the AS factors and the odds of presenting with a panic attack provided adequate fit to the data (v2 = 258.10, p < .05, CFI = .95, RMSEA = .07).2 The odds of presenting with a panic attack was significant for the AS factor (OR = 1.66, 95% Confidence Interval CI [1.47, 1.85]). The odds of presenting with a panic attack were not significant for the AS physical concerns (OR = .99, 95% CI [.78, 1.20]), AS cognitive concerns (OR = .93, 95% CI [.71, 1.15]), or AS social concerns factors (OR = 1.14, 95% CI [.97, 1.31]). The AS factors accounted for 28% of the variance in panic attacks (p < .001). 4. Discussion Results of the current study indicate that the structure of AS, as measured by the ASI-3, is best conceptualized as a bifactor model, with orthogonal general and specific factors. These findings are consistent with two prior studies of AS that also compared a bifactor model to the more common hierarchical model of AS (e.g., Ebesutani et al., 2014; Osman et al., 2010). Together, these findings suggest that when the general AS factor is accounted for, AS physical, cognitive, and social concerns are best conceptualized as orthogonal constructs. Ebesutani et al. (2014) speculate that this may indicate that the three AS subdomains may be linked to differential neuro-biological processes involved in processing information and fear events in addition to a more general neurobiological system implicated in the global fear of anxious arousal. Bifactor modeling approaches, combined with recent advances in identification of AS-relevant biomarkers (e.g., Sehlmeyer et al., 2010) allow for researchers to address whether distinct neurobiological domains for the distinct AS subdomains can be identified. The general AS factor was associated with NA, worry, depression, social anxiety, and panic attacks. These findings mirror past findings implicating AS as a risk factor for most emotional distress disorders (Naragon-Gainey, 2010; Olatunji & Wolitzky-Taylor, 2009). In addition, these findings replicate the findings of Ebesutani et al. (2014), who reported that general AS was associated with state anxiety. Together, these findings provide support for the role of AS as a transdiagnostic risk factor for emotional distress disorders. Regarding the more specific AS dimensions and the fear and distress emotional distress symptoms, our hypotheses were only partially supported. Accounting for the general AS factor, unique relations emerged for AS cognitive concerns and worry and depression symptoms. In addition, AS social concerns was uniquely associated with worry, depression, and social anxiety, although the relation between AS social concerns and social anxiety appeared to be the most robust. These findings are similar to prior studies by Allan et al. (2014, in press) in which AS cognitive concerns were related to a distress disorder factor as well as specific distress disorders and AS social concerns showed more general relations to mood and anxiety disorders, but, was significantly more associated with social anxiety symptoms than with all other measures of psychopathology, with the exception of depression symptoms (e.g., Allan et al., in press). However, AS physical concerns was not significantly associated with panic attacks in the current study. One possible explanation for this is that, because AS physical concerns is a foundational aspect of AS, the general AS factor best captures 2 In this model ASI-3 item 2 was only allowed to load on the AS factor (and removed from the AS cognitive concerns factor) to account for a linear dependency when this item was allowed to load on both factors. There were no substantive differences in the analysis when this item was omitted from the AS cognitive concerns factor.
AS physical concerns, and including a distinct factor to assess AS physical concerns is unnecessary. There are several implications for the current findings. From a research perspective, the bifactor model appears to be a viable alternative to the correlated factors model if researchers were interested in examining the unique relations between the specific AS factors and external covariates (e.g.,Chen, Hayes, Carver, Laurenceau, & Zhang, 2012). For clinicians interested in using AS as a risk factor or correlate of anxiety and other disorders, the strong associations between the general AS factor and the mood and anxiety symptoms as well as the generally higher factor loadings on the AS factor as compared to the factor loadings on the more specific factors suggests that using the summed ASI-3 total score should prove an adequate measure of risk conferred by this construct. There are several limitations to consider. First, this and the prior two studies in adult populations have used convenience samples. There is a need to examine whether the same factor structure emerges in samples containing participants with clinically elevated levels of mood and anxiety disorders. This would be expected given that the tested factor structure of the ASI-3 (i.e., the hierarchical and correlated factors models) is similar in clinical and community samples (e.g., Taylor et al., 2007). Second, only a subset of fear and distress disorders were available for inclusion in the present sample. Finally, the current study was conducted with cross-sectional data. Examining the role of the AS factors prospectively would provide more support for a bifactor modeling approach for this construct. There are several strengths of the current study. Support for a bifactor model of AS was found in the current study. In addition, external validity of the factors was provided. The general AS factor was associated with general NA as well as more specific mood and anxiety disorders. Specificity was also found for the more narrowband AS cognitive concerns and social concerns dimensions, but not for the AS physical concerns dimension. This suggests that whereas there is something domain specific being captured by the AS cognitive and social concerns dimensions above and beyond general AS this is not the case for AS physical concerns. References Allan, N. P., Capron, D. W., Raines, A. M., & Schmidt, N. B. (2014). Unique relations among anxiety sensitivity factors and anxiety, depression, and suicidal ideation. Journal of Anxiety Disorders, 28, 266–275. Allan, N. P., Norr, A. M., Capron, D. W., Raines, A. M., Zvolensky, M. J., & Schmidt, N. B. (in press). Specific associations between anxiety sensitivity dimensions and fear and distress dimensions of emotional distress disorders. Journal of Psychopathology and Behavior Assessment. Beck, A. T., Steer, R. A., & Carbin, M. G. (1988). Psychometric properties of the Beck Depression Inventory: Twenty-five years of evaluation. Clinical Psychology Review, 8, 77–100. Brown, T. A. (2006). Confirmatory factor analysis for applied research. New York, NY: The Guilford Press. Browne, M. W., & Cudeck, R. (1992). Alternative ways of assessing model fit. Sociological Methods and Research, 21, 230–258. Capron, D. W., Fitch, K., Medley, A., Blagg, C., Mallott, M., & Joiner, T. (2012). Role of anxiety sensitivity subfactors in suicidal ideation and suicide attempt history. Depression and Anxiety, 29, 195–201. Chen, F. F., Hayes, A., Carver, C. S., Laurenceau, J. P., & Zhang, Z. (2012). Modeling general and specific variance in multifaceted constructs: A comparison of the bifactor model to other approaches. Journal of Personality, 80, 219–251. Clark, L. A., & Watson, D. (2006). Distress and fear disorders: An alternative empirically based taxonomy of the ‘‘mood’’ and ‘‘anxiety’’ disorders. The British Journal of Psychiatry: The Journal of Mental Science, 189, 481–483. Ebesutani, C., McLeish, A. C., Luberto, C. M., Young, J., & Maack, D. J. (2014). A bifactor model of anxiety sensitivity: Analysis of the Anxiety Sensitivity Index3. Journal of Psychopathology and Behavioral Assessment, 36, 452–464. Houck, P. R., Spiegel, D. A., Shear, M. K., & Rucci, P. (2002). Reliability of the selfreport version of the panic disorder severity scale. Depression and Anxiety, 15, 183–185. Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6, 1–55.
N.P. Allan et al. / Personality and Individual Differences 74 (2015) 78–83 Kennedy, B. L., Schwab, J. J., Morris, R. L., & Beldia, G. (2001). Assessment of state and trait anxiety in subjects with anxiety and depressive disorders. The Psychiatric Quarterly, 72, 263–276. Krueger, R. F. (1999). The structure of common mental disorders. Archives of General Psychiatry, 56, 921–926. Mattick, R. P., & Clarke, J. C. (1998). Development and validation of measures of social phobia scrutiny fear and social interaction anxiety. Behaviour Research and Therapy, 36, 455–470. Meyer, T. J., Miller, M. L., Metzger, R. L., & Borkovec, T. D. (1990). Development and validation of the Penn State Worry Questionnaire. Behaviour Research and Therapy, 28, 487–495. Moshagen, M. (2012). The model size effect in SEM: Inflated goodness-of-fit statistics are due to the size of the covariance matrix. Structural Equation Modeling: A Multidisciplinary Journal, 19, 86–98. Mulaik, S. (2007). There is a place for approximate fit in structural equation modelling. Personality and Individual Differences, 42, 883–891. Muthén, B. O., & Muthén, L. K. (1996–2008). Mplus (Version 5.2). Los Angeles, CA. Naragon-Gainey, K. (2010). Meta-analysis of the relations of anxiety sensitivity to the depressive and anxiety disorders. Psychological Bulletin, 136, 128–150. Olatunji, B. O., & Wolitzky-Taylor, K. B. (2009). Anxiety sensitivity and the anxiety disorders: A meta-analytic review and synthesis. Psychological Bulletin, 135, 974–999. Osman, A., Gutierrez, P. M., Smith, K., Fang, Q., Lozano, G., & Devine, A. (2010). The Anxiety Sensitivity Index-3: Analyses of dimensions, reliability estimates, and correlates in nonclinical samples. Journal of Personality Assessment, 92, 45–52. Reise, S. P. (2012). The rediscovery of bifactor measurement models. Multivariate Behavioral Research, 47, 667–696. Reiss, S., & McNally, R. (1985). Expectancy model of fear. In S. Reiss & R. Bootzin (Eds.), Theoretical Issues in Behavior Therapy (pp. 107–121). San Diego, CA: Academic. Reiss, S., Peterson, R. A., Gursky, D. M., & McNally, R. J. (1986). Anxiety sensitivity, anxiety frequency and the prediction of fearfulness. Behaviour Research and Therapy, 24, 1–8.
83
Schmidt, N. B., Buckner, J. D., & Keough, M. E. (2007). Anxiety sensitivity as a prospective predictor of alcohol use disorders. Behavior Modification, 31, 202–219. Schmidt, N. B., Lerew, D. R., & Jackson, R. J. (1997). The role of anxiety sensitivity in the pathogenesis of panic: Prospective evaluation of spontaneous panic attacks during acute stress. Journal of Abnormal Psychology, 106, 355–364. Sehlmeyer, C., Konrad, C., Zwitserlood, P., Arolt, V., Falkenstein, M., & Beste, C. (2010). ERP indices for response inhibition are related to anxiety-related personality traits. Neuropsychologia, 48, 2488–2495. Sellbom, M., Ben-Porath, Y. S., & Bagby, R. M. (2008). On the hierarchical structure of mood and anxiety disorders: Confirmatory evidence and elaboration of a model of temperament markers. Journal of Abnormal Psychology, 117, 576–590. Shear, M. K., Brown, T. A., Barlow, D. H., Money, R., Sholomskas, D. E., Woods, S. W., et al. (1997). Multicenter collaborative panic disorder severity scale. The American Journal of Psychiatry, 154, 1571–1575. Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics. Boston, MA: Pearson Education Inc.. Taylor, S., Zvolensky, M. J., Cox, B. J., Deacon, B., Heimberg, R. G., Ledley, D. R., et al. (2007). Robust dimensions of anxiety sensitivity: Development and initial validation of the Anxiety Sensitivity Index-3. Psychological Assessment, 19, 176–188. Watson, D., & Clark, L. A. (1999). The PANAS-X: Manual for the positive and negative affect schedule-expanded form. Ames: University of Iowa. Wheaton, M. G., Deacon, B. J., McGrath, P. B., Berman, N. C., & Abramowitz, J. S. (2012). Dimensions of anxiety sensitivity in the anxiety disorders: Evaluation of the ASI-3. Journal of Anxiety Disorders, 26, 401–408. Yu, C. Y. (2002). Evaluating cutoff criteria of model fit indices for latent variable models with binary and continuous outcomes (Doctoral dissertation). Los Angeles: University of California.