Abstract
The current study examined the convergent validity and factor structure of the State-Trait Inventory of Cognitive and Somatic Anxiety in a sample of African Americans and European Americans. Validity analyses revealed similar associations; however, the factor analysis failed to support the original factor structure and factorial variance was consistently demonstrated.
According to the National Comorbidity Survey (Kessler, Chiu, Demler, & Walters, 2005), anxiety disorders are the most prevalent class of psychological disorders for adults in the United States, affecting approximately 30% of individuals. Importantly, prevalence rates for these disorders tend to vary based on ethnicity, with African Americans less likely to meet criteria for panic attacks, panic disorder, social anxiety disorder, and generalized anxiety disorder (see Lewis-Fernandez et al., 2009, for a recent review). Possible reasons for these variations include true prevalence differences, the failure of instruments to properly capture the range of symptomatic experiences, the failure of the diagnostic manual to include culturally specific presentations, or the influence of variables related to ethnicity such as education (Lewis-Fernandez et al., 2009; Ritsher, Struening, Hellman, & Guardino, 2002). One particular area of recent focus has been on the ability to transport instruments across cultural and ethnic groups. As noted by Ritsher et al., transporting measures across groups can lead to inaccurate prevalence rates as well as patterns of symptoms. Thus it is essential to determine if psychometric properties are consistent for various ethnic groups.
Given the significant prevalence rate of these disorders, continued attention has focused on the accurate assessment of the range of manifestations of the symptoms of anxiety. Particular focus has been given to creating diagnostic tools with the ability to distinguish between various forms of anxiety and the symptoms associated with their presentation (for a complete review, see Sakakibara, Miller, Orenczuk, & Wolfe, 2009). The State-Trait Anxiety Inventory (STAI; Spielberger, 1983) was thought to have fulfilled such a demand, presumably being able to identify anxiety on both a state level and a trait level. More recent literature, however, has raised questions about the validity of the STAI, predominantly its ability to distinguish between symptoms of anxiety and depression (as measured by Depression Anxiety Stress Scales; Gros, Antony, Simms, & McCabe, 2007).
Ree, MacLeod, and Locke (2008) aimed to improve on the STAI by mirroring the state-trait model but modifying the items and factor structure used to capture the domain of anxiety. In developing the State-Trait Inventory for Cognitive and Somatic Anxiety (STICSA), they began with a pool of 62 items that they administered to 576 participants. Confirmatory factor analysis was used to select items that loaded purely on domains of cognitive or somatic symptoms. The final measure included a cognitive subscale (10 items) that included items such as “I think worse things will happen” and “I picture some future misfortune,” whereas the somatic subscale (11 items) contained items such as “I feel dizzy” and “My heart beats fast.” The final measure includes two versions: a state version and a trait version, which are presented as separate instruments. In an elegant series of studies, Ree et al. (2008) validated the STICSA by demonstrating the ability of the trait scales to predict anxiety in response to both cognitive (end of term examinations) and somatic (CO2 inhalation) stressors. The development study also reported acceptable test–retest reliability (tested 7 weeks apart) for the trait version of the measure (above .65 for both subscales) with lower reliabilities for the state version (.49 for cognitive aspects and .31 for somatic), suggesting the measure is responsive to current psychological state.
Gros et al. (2007) validated the proposed factor structure of the STICSA using a clinical sample receiving outpatient care for symptoms of anxiety. In addition to supporting the factor structure of STICSA, this study revealed a strong internal consistency for both state and trait symptoms, both cognitive and somatic (all Cronbach’s alphas > .90). This study adds significant support for the use of the STICSA as the validity was demonstrated in a clinically anxious sample.
Additionally, Gros et al. (2007) tested the convergent validity of the STICSA with other measures of anxiety and depression. As expected, the STICSA demonstrated a stronger relationship with a measure of anxiety (Depression Anxiety Stress Scales–Anxiety Subscale; Lovibond & Lovibond, 1995) than with a measure of depression (Depression Anxiety Stress Scales–Depression Subscale). In contrast, the STAI was more strongly correlated with the measure of depression than the measure of anxiety. These findings suggest that STICSA may be a more pure measure of anxiety. Gros, Simms, and Antony (2010) once again examined the factor structure of STICSA in a sample of undergraduate students while also exploring the convergent validity of the cognitive and somatic subscales. Findings confirmed the strength of STICSA’s factor structure and demonstrated convergent validity of the subscales with a number of associated measures, including the Albany Panic and Phobia Questionnaire, Social Phobia Inventory, Social Thoughts and Beliefs Scale, and Thought Control Inventory (rs > .49).
Despite the evidence in support of STICSA as a valid measure of anxiety, to date there has been little investigation of the validity of the STICSA with diverse populations. Ree et al. (2008) noted apprehensions about STICSA’s validity in diverse populations as the original sample was quite homogeneous compared to the general population. A primary limitation of the research to date is the lack of ethnic diversity in the previous validation samples. For example, Gros et al.’s (2007) participant pool was 92% Caucasian. Their follow-up study in 2010 had a more diverse representation (with 50% Caucasian, 28.4% Asian, 10.3% African American), but it failed to meaningfully compare various ethnic groups.
A number of studies have shown lower mean levels of anxiety in African American participants than European American participants (Carter, Miller, Sbrocco, Suchday, & Lewis, 1999; Chapman, Williams, Mast, & Woodruff-Borden, 2009). These results are consistent with the prevalence rates described above. One possible cause for this difference is anxiety may be experienced, or reported, differently by various ethnic groups. In particular, research examining ethnic differences in anxiety has focused on the way African Americans experience and express cognitive and physical aspects of anxiety (Heurtin-Roberts, Snowden, & Miller, 1997). Heurtin-Roberts et al. (1997) revealed a significant variation in the expression of anxiety in the African American population in comparison to European Americans. Results of their study demonstrated that African Americans are much more likely to report somatic symptoms of anxiety (e.g., shaking, nausea, and tension) rather than psychological symptoms (e.g., fear, anger, and mental fatigue). After reporting a different factor structure for African American children on the Anxiety Sensitivity Index; Lambert, Cooley, Campbell, Benoit, and Stansbury (2004) noted, “Somatic symptoms may be particularly salient for ethnic minority youth, as demonstrated with ethnic minority adults” (p. 256). Evidence suggests that this heightened somatic expression is due to a tendency for African Americans to view physical symptoms more seriously than cognitive ones as they seek assistance with these manifestations from medical versus psychiatric personnel (Neal & Turner, 1991).
This variation in cognitive-somatic presentation may have significant implications for the STICSA as a measure of anxiety, specifically in relation to the previously supported factor structure. Chapman et al. (2009) reported such variation in the factor structure of the Beck Anxiety Inventory (BAI) when samples consisted of African American populations. Within this study, the original two-factor model of the BAI, consisting of somatic and subjective anxiety groupings, did not load as well for the African American cohort as it did for the European American population.
Kingery, Ginsburg, and Burstein (2009) similarly evaluated the validity of a popular measure of anxiety, the Multidimensional Anxiety Scale for Children (MASC), in an African American sample. Their study demonstrated strong convergent validity of the MASC within a sample of 118 African American adolescents when compared to other measures of anxiety (e.g., the screen for child anxiety-related emotional disorders). Despite its consistent convergent validity, their findings also suggest that the factor structure associated with the MASC, consisting of four factors—physical symptoms, social anxiety, separation anxiety, and harm avoidance—did not load similarly for their African American sample. Kingery et al. hypothesized that the MASC may be identifying the presence of anxiety within this population but not capturing how anxiety is experienced by African Americans. This may result from ignoring additional somatic symptoms associated with the manifestation of the disorder within African Americans.
Melka, Lancaster, Adams, Howarth, and Rodriquez (2010) evaluated the validity of the factor structures of the Fear of Negative Evaluation and the Social Avoidance and Distress Scales, two widely used evaluations of social anxiety, for both African American and European American populations. Consistent with previous literature, results indicated significant discrepancies in the structural model between the two ethnic samples. Melka et al. hypothesized that these variations may indicate that such scales operate differently for each of the two samples, thus making it difficult to compare results if such confounds are not corrected. Given the previous findings demonstrating problematic variations in factor structure for various anxiety measures, examining the structure of the STICSA in a diverse population seems prudent.
The current study evaluates the convergent validity and factor structure of STICSA as a measure of anxiety. Expanding on previous research (Gros et al., 2010; Ree et al., 2008), this study will compare results of European American participants with that of African American participants, in an effort to validate its use in both populations. Testing structural consistency is key as “evaluating structural equivalence of the model across groups represents one step in testing the cross-cultural validity of an instrument” (Ritsher et al., 2002, p. 200). Given previous research that indicates strong convergent validity across various demographics but inconsistent factor structures for other measures of anxiety when tested on diverse samples (Kingery et al., 2009), it can be hypothesized that the STICSA will demonstrate consistent results for convergent validity across both European American and African American samples but varying results in relation to the previously established factor structure of the STICSA within these samples. More specifically, it is predicted that similar strengths of relationships will emerge for both samples in relation to convergent findings. Given previously identified measurement discrepancies in questionnaires measuring anxiety symptomatology (Chapman et al., 2009; Melka et al., 2010), it is hypothesized that conflicting findings will emerge for the two ethnic samples for confirmatory factor analysis, as the original factor structure should demonstrate a poor fit for the African American sample.
Method
Participants
Participants were drawn from a larger study of 514 undergraduates (169 African Americans, 269 European Americans, and 76 from other ethnic groups not included in the current study) from a public rural Midwestern university, who received course-extra credit for their participation. The survey measures used in the current study were part of a larger project examining symptoms of stress (including how they coped with stressful events) and anxiety in college undergraduates. Surveys were administered in groups of 5 to 20 students and were completed in standard paper-and-pencil formatting. Due to the aims of the larger project, no counterbalancing (or alternative orders) were used in the current project. The order of measures in the current project was the Trait version of the STICSA, the Center for Epidemiologic Studies Depression Scale (CES-D), the Social Interaction Anxiety Scale, the State version of the STICSA, the Posttraumatic Stress Disorder (PTSD) Checklist–Specific Event Version (PCL-S), and finally the Positive and Negative Affect Schedule (see below for descriptions of all measures). Four African American participants were removed from the final sample due to missing data. This resulted in a final sample of 165 African American participants. Since fit index values in factor analysis are often affected by sample size, it is recommended that comparison of measurement characteristics across samples use an identical number of participants (Brown, 2006). Because the total sample for the current study consisted of more European American than African American participants, the SPSS random sampling function was employed to select an equivalent number of 165 European American participants that were blocked on gender so that an equivalent number of men and women were in each sample. Thus, both samples were 35% male (n =58). The mean age of the final sample was 19.32 years (SD = 3.49). In terms of relationship status, the sample was consistent with that of many college populations with the majority of the sample being single (80.9%, n = 267), a substantial minority partnered but not cohabitating (15.2%, n = 50), and a small number married (2.7%, n = 9) or divorced (0.6%, n = 2). The vast majority of the sample described themselves as middle class (81.8%, n = 270), 11.2% (n = 37) as upper class, and 7% (n = 23) as lower class.
Measures
State-Trait Inventory for Cognitive and Somatic Anxiety
As reviewed above, Ree et al. (2008) developed the STICSA to separately and accurately measure somatic and cognitive domains of anxiety. In addition to distinguishing between cognitive and somatic aspects, the measure was developed to distinguish between trait and state aspects of anxiety. As noted above, the cognitive domain included items such as “I think worse things will happen” and “I picture some future misfortune,” whereas the somatic domain included items such as “I feel dizzy” and “My heart beats fast.” Participants rated each item on a Likert-type scale with response options ranging from 1 (almost never) to 4 (almost always). Participants were asked to complete both the state and trait versions of the measure. In African American participants, Cronbach’s alphas in this study were .847 for State-Cognitive, .909 for State-Somatic, .828 for Trait-Cognitive, and .851 for Trait-Somatic. In European American participants, Cronbach’s alphas in this study were .908 for State-Cognitive, .901 for State-Somatic, .896 for Trait-Cognitive, and .877 for Trait-Somatic.
Social Interaction Anxiety Scale
The Social Interaction Anxiety Scale (SIAS; Mattick & Clarke, 1998) is a frequently used measure of social anxiety that contains 20 items describing affective, behavioral, and cognitive reactions to potentially distressing social situations. Participants are asked to indicate how characteristic these statements are to them with responses ranging from 0 (not at all) to 4 (extremely). This measure has shown excellent test–retest reliability (.92 over 4 weeks in a clinical sample; Mattick & Clarke, 1998) and validity (Cronbach’s α = .88 in a college sample; Mattick & Clarke, 1998). Heimberg, Mueller, Holt, Hope, and Liebowitz (1992) reported a similarly strong test–retest correlation in a student sample (.86 over 2 weeks). Mattick and Clarke reported that the SIAS was able to discriminate those with social anxiety disorder. Cronbach’s alphas in this sample were .874 for European Americans and .868 for African Americans.
Center for Epidemiologic Studies Depression Scale
The CES-D (Radloff, 1977) was developed to assess a wide range of depression related experiences. Participants were asked to indicate how often they experienced relevant feelings and behaviors within the past week on a scale that ranges from 0 (rarely or none of the time) to 3 (most or all of the time). This scale has demonstrated strong internal reliability (.85; Radloff, 1987). Test–retest reliability is somewhat lower (range = .51-.57; Radloff, 1977); however, this is to be expected as the measure inquires about “current” functioning and thus is designed to vary with changes in perceived mood. Cronbach’s alphas in this sample were .768 for European Americans and .774 for African Americans.
Positive and Negative Affect Schedule
The Positive and Negative Affect Schedule (Watson, Clark, & Tellegen, 1988) was developed to assess levels of general positive and negative affect. This measure asked participants to assess the degree to which they have experienced 10 positive and 10 negative feelings over “the past few weeks.” The scale uses a Likert-type scale ranging from 1 (very slightly or not at all) to 5 (extremely). Scores are computed separately for Positive Affect (PA) and Negative Affect (NA) consistent with the orthogonal nature of these constructs (Clark & Watson, 1991). Crawford and Henry (2004) examined the psychometric properties of the Positive and Negative Affect Schedule in a large general adult population and reported the measure demonstrated both validity (a confirmatory factor analysis supported the proposed factor structure and the subscales demonstrated convergent validity) and reliability (internal consistencies of .89 for PA scale and .85 for NA scale) of the measure. Cronbach’s alphas in this sample for African Americans are .916 (PA) and .877 (NA) and for European Americans are .921 (PA) and .805 (NA).
Posttraumatic Stress Disorder Checklist–Specific Event Version
The PCL-S (Weathers, Litz, Herman, Huska, & Keane, 1993) is a 17-item scale developed to measure PTSD symptoms associated with a specific event on three domains; reexperiencing, avoidance/numbing, and physical symptoms. The scale uses a Likert-type scale ranging from 1 (not at all) to 5 (extremely). The measure has demonstrated satisfactory convergent validity with the Clinician-Administered PTSD Scale (r = .93; Blanchard, Jones-Alexander, Buckley, & Forneris, 1996), as well as excellent internal consistency (α = .86) and test–retest reliability (Ventureyra, Yao, Cottraux, Note, & De Mey-Guillard, 2002). Cronbach’s alphas in this sample are .906 for European Americans and .910 for African Americans.
Data Analysis
To examine differences in mean levels of functioning between groups, independent sample t tests were calculated to compare all measures in African American and European American samples (see Table 1 for results). To examine differences between cognitive and somatic anxiety within each ethnic group, paired sample t tests were conducted and the results are presented in Table 1. To examine convergent validity, separate correlation analyses were conducted with the African American and European American samples. Results of these analyses can be found in Table 2. To compare the strength of the correlations between the ethnic groups, we used Fisher’s r to z transformations and compared z scores as outlined in Cohen, Cohen, West, and Aiken (2003).
Means and Standard Deviations by Ethnic Group.
Note. PCL = PTSD (Posttraumatic Stress Disorder) Checklist; PA = Positive Affect from the PANAS (Positive and Negative Affect Schedule); NA = Negative Affect from the PANAS; CES-D = Center for Epidemiologic Studies Depression Scale; SIAS = Social Interaction Anxiety Scale; STICSAT = State-Trait Inventory for Cognitive and Somatic Anxiety Trait; STICSAS = State-Trait Inventory for Cognitive and Somatic Anxiety State; C = Cognitive subscale; S = Somatic subscale. Boldface indicates significant difference between African American and European Americans participants at p < .05.
Indicates that Cognitive scale is significantly different from Somatic scale within ethnic group at p < .05.
Correlations of All Variables Split by Ethnicity.
Note. STICSAT = State-Trait Inventory for Cognitive and Somatic Anxiety Trait; STICSAS = State-Trait Inventory for Cognitive and Somatic Anxiety State; C = Cognitive subscale; S = Somatic subscale; SIAS = Social Interaction Anxiety Scale; PCL = PTSD Checklist; CES-D = Center for Epidemiologic Studies Depression Scale; NA = Negative Affect from the PANAS (Positive and Negative Affect Schedule); PA = Positive Affect from the PANAS. All correlations significant at p < .05 except those marked by †. Correlations for African Americans are below the midline of the table while correlations for European Americans are above the midline. Boldface indicates that correlation was significantly stronger in European Americans than African Americans as measured using Fisher’s r to z comparisons.
Confirmatory factor and measurement invariance analyses were conducted with AMOS 4.0 (Arbuckle & Wothke, 1997). Measurement invariance was examined, comparing African American and European American participants on both state and trait domains of the STICSA. Invariance analysis is thoroughly reviewed by Brown (2006). The procedure involves constraining factor loadings, variable means, and indicator residuals based on values obtained after structural analysis using the whole sample. If significant chi-square change is observed following constraint, measurement variance is present. Invariance tests of configuration, factor loading, indicator intercept, and indicator error variance were completed as part of the present investigation. The aforementioned analyses consisted of 2 latent variables and 21 observed variables for each form of the measure (state vs. trait).
It is suggested that investigators employ several fit indices when evaluating overall model fitness (Jaccard & Wan, 1996; Kline, 2004; Weston & Gore, 2006). The STICSA’s factor structure was examined using chi-square, the Comparative Fit Index (CFI; Bentler, 1990), Root Mean Square Error of Approximation (RMSEA; Steiger, 1990), and Tucker–Lewis Index (TLI; Tucker & Lewis, 1973). These indices are widely accepted measures of model fit for confirmatory factor analysis (Bentler, 1990; Bollen 1990; Gerbing & Anderson, 1993; Hu & Bentler, 1995; Marsh, Hau, & Wen, 2004; Weston, Gore, Chan, & Catalano, 2008). When evaluating model fit, a number of criteria values are suggested (Hu & Bentler, 1995). TLI and CFI values of .95 or above and RMSEA values below .06 are thought optimal. Nonetheless, TLI and CFI values of .90 may be adequate (Kline, 2004). Similarly, Brown and Cudeck (1993) suggested that RMSEA values of .08 are acceptable. Chi-square is not recommended as an indicator of poor fit given its sensitivity to larger sample size (Schumacker & Lomax, 2004). Despite the concern, chi-square is used in the current study to examine the STICSA’s measurement qualities. Significant increase of chi-square, following constraint of factor loading, indicator intercept, and indicator error, indicated measurement variance between African American and European American participants.
Results
As noted in Table 1, European Americans reported higher levels of cognitive and somatic anxiety in both the state and trait versions of the STICSA. To examine differences between cognitive and somatic anxiety, scale totals were divided by the number of items as the number of items differs on the STICSA scales. Inconsistent with our hypotheses, African American participants reported higher levels of state cognitive anxiety relative to somatic, t(164) = 5.96, p < .001, as well as higher levels of trait cognitive anxiety relative to somatic, t(164) = 7.63, p < .001. Results were similar for European American participants with significantly higher levels of cognitive anxiety in the trait version, t(164) = 4.53, p < .001, as well as higher levels of trait cognitive anxiety relative to somatic, t(164) = 8.811, p < .001.
As presented in Table 2, correlations between domains of the STICSA and relevant domains were relatively consistent. That is, African American and European American scores on the STICSA were generally similarly related to measures of other psychological variables. One exception to this finding is that the cognitive domain of the trait version of the STICSA was more strongly related to scores on a measure of depression than were scores for African American participants. These findings suggest that the cognitive domain of the STICSA may be a more pure measure of anxiety for African Americans than it is for European Americans. Also of note, and counter to hypotheses, the STICSA subscales were more highly correlated with a measure of depression than with a measure of social anxiety across both ethnic groups, which calls into question the discriminant validity of the measure.
Results of confirmatory and measurement invariance analyses of both state and trait measures for African American and European American participants are summarized in Tables 3 and 4. Findings are indicative of relatively poor fit for African Americans and European Americans. Additional analyses were then conducted to identify individual items that may have consistently loaded poorly on their respective subscales. Only one item (“I think that others won’t approve of me”) showed a relatively low standardized regression weight in two estimations; for the state (0.396) and trait (0.402) analyses for African American participants. No other items showed a low standardized value for more than one subscale. Full results of these analyses are available from the second author.
Measurement Invariance Results for State Full Form.
Note. df = degrees of freedom; AIC = Akaike information criterion; CFI = comparative fit index; TLI = Tucker–Lewis index; RMSEA = root mean square error of approximation.
p < .05. **p < .01. ***p < .001.
Measurement Invariance Results for Trait Full Form.
Note. df = degrees of freedom; AIC = Akaike Information Criterion; CFI = comparative fit index; TLI = Tucker–Lewis index; RMSEA = root mean square error of approximation.
p < .05. **p < .01. ***p < .001.
Additional analyses revealed that measurement variance was observed when examining the instruments. Significant chi-square change was present following each set of constraints, suggesting variance across ethnic groups with regard to factor loadings, indicator intercept, and indicator error variance. Additional tests of structure and measurement invariance were conducted, isolating each latent variable and associated observed variables. These analyses were included in an effort to ascertain if the measurement variance observed was present only for a single subscale. Measurement variance was observed for both the somatic and cognitive subscales of the state and trait instruments, indicating that measurement qualities were discrepant for all subscales. Full results of the subscale analyses are available from the second author.
Discussion
In the current study, we evaluated the psychometric properties of the STICSA through an examination of the convergent validity and the factor structure in African American and European American samples. In doing so, we sought to validate the STICSA for both demographic groups, potentially expanding the empirical support for the scale. Based on previous research, we hypothesized that similar convergent validity would be demonstrated between both samples, but the factor structure would differ significantly between groups.
As hypothesized, results demonstrated strong convergent validity. Consistent with the findings of Ree et al. (2008), who found the STICSA to correlate strongly with the STAI, we found similar patterns, with the STICSA converging with other domains of anxiety including social anxiety and PTSD symptoms. However, the subscales of the STICSA generally failed to show discriminant validity. Of note, the STICSA was more highly correlated with a measure of depression than social anxiety across all subscales. This finding is consistent with earlier work that demonstrated very high relationships between measures of anxiety and depression. For example, Williams, Peeters, and Zautra (2004) reported correlations between the STAI and CESD of .76 and .63 for African Americans and European Americans, respectively. This preliminary evidence suggests the measure performs similarly for both demographic groups. Thus while the STICSA shows strong convergent validity, it shows the same difficulties discriminating anxiety from depression as has been reported for other measures of anxiety.
Consistent with hypotheses and previous work (Carter et al., 1999; Chapman et al., 2009), significant differences were found between racial groups on the levels of reported cognitive and somatic anxiety. In particular, European Americans reported significantly higher levels of anxiety for both forms of the measure (state and trait) and both types of anxiety (cognitive and somatic). Previous work (Heurtin-Roberts et al., 1997) has suggested that African American participants are more likely to report the somatic rather than cognitive symptoms of anxiety. However, the current results not only are inconsistent with these earlier findings but also go further in that cognitive symptoms were significantly higher than somatic symptoms for African American participants. While it is unclear why these results differed from earlier work, one potential explanation involves the development of the current measure. As noted by Ree et al. (2008), the sample used in the development of the measure was not ethnically diverse or representative. Given this limitation, it is possible that somatic symptoms that may have been more relevant for African American participants may have been excluded or not fully represented in the current measure. While it would be premature to suggest changes are necessary based on one study; if replicated, these results may suggest more research is needed to examine the content validity of the STICSA for African Americans.
Another interesting and potentially important finding in the current project was that the cognitive subscale of the trait version of the STICSA was more strongly related to depression in European American participants than African American participants. These results are somewhat inconsistent with the earlier work of Carter et al. (1999) who concluded, “It may also be that the relationship between anxiety and depression for African Americans is conceptually different (perhaps more overlapping) than is traditionally conceptualized among their European American counterparts” (p. 531). One potential explanation is African Americans may be more likely to experience and report psychological distress in somatic manners. While somatization is certainly not limited to African Americans, somatic symptoms are more common (15%) in this group than in Caucasians (9%; Robins & Regier, 1991). Furthermore, African Americans with depression are more likely to exhibit somatic symptoms than European Americans (Das, Olfson, McCurtis, & Weissman, 2006; Myers et al., 2002).
It has been previously pointed out that the use of somatic metaphors and the increased reporting of somatic symptoms may complicate the accurate detection and diagnosis of psychological distress in African American patients (Ahmed & Bhugra, 2007; Das et al., 2006). This area of research was bolstered by a novel study conducted by Ayalon and Young (2003) that examined the relationship between individual symptoms and the severity of depression. They concluded that African Americans are less likely to report cognitive-affective symptoms of depression than European Americans at similar levels of the severity of depression. Furthermore, they concluded that somatic symptoms were more strongly related to the overall construct of depression in African Americans than European Americans.
In addition to the work above examining general psychological distress and depression, Kingery et al. (2009) reviewed earlier work demonstrating that African American youth have a greater tendency to report anxiety-related symptoms in somatic ways. They go on to note, “It is possible that physical symptoms serve as a more important proxy of anxiety among African American youth and may influence their scores on extant anxiety rating scales” (p. 289). Given these earlier findings, it is very possible that the stronger correlations found in the current study are a result of a stronger link between cognitive aspects of anxiety and distress in European Americans than in African Americans. That is, as African Americans are more likely to experience distress in somatic manners, measures that focus exclusively, or even heavily, on cognitive conceptualizations of psychological constructs are likely to miss essential elements of the experience of these constructs in African Americans.
Despite the strong finding related to convergent validity, an evaluation of the factor structure and factorial invariance of the STICSA were inconsistent with hypotheses. The scale showed unexpectedly poor loadings within each group as well as consistent variance between samples. The structural analysis of the STICSA identified irregularities in the factor structure compared to the four-factor structure identified by Ree et al. (2008) and supported by further evidence (Gros et al., 2007; Gros et al., 2010). Inconsistent with expectations, which led us to hypothesize that the African American sample would demonstrate poor fit to the original model, neither African Americans nor European Americans demonstrated good fit with the four-factor structure identified for this measure. The lack of fit for the European American sample is particularly surprising as Gros et al. (2007) demonstrated strong goodness of fit for this model with a participant pool consisting of primarily Caucasian patients diagnosed with an anxiety disorder. Because previous studies employed predominately Caucasian samples, findings in support of the psychometric characteristics for this measure may reflect measurement qualities only for that group but fail to support the generalizability of these previous results for other ethnic groups. Results of the current study suggest the need for further psychometric inquiry examining the properties of the measure within diverse populations.
Furthermore, measurement variance was observed for both samples, with discrepancies seen in factor loadings, indicator intercepts, and error variance. Analyses were unable to identify a single subscale that may account for these findings as results indicate heterogeneity of the variance in each of the four subscales: Cognitive-State, Cognitive-Trait, Somatic-State, and Somatic-Trait. Given the findings noted above related to the failure of the confirmatory analyses, these results should be interpreted with caution as the participant characteristics in this sample may in some unknown way be influencing the structure of these items. These findings suggest further analyses are needed to evaluate the structural properties of the STICSA in order to validate its use as an indicator of anxiety.
As with any project, this study contained a number of strengths and limitations. A limitation of the current study was that both the European American and African American groups were sampled from one Midwest undergraduate population. The homogeneity of the sample may have accounted for some of the unexpected results. Future studies should include a more heterogeneous sample to potentially overcome this limitation and to add to the generalizability of the STICSA. Due to the aims of the broader study, it was not possible to counterbalance or change the order of the measures as administered. This may have affected the results in unknown ways and should be an area of future inquiry. Nonetheless, current results clearly highlight the importance of examining the characteristics of many widely used psychological instruments in diverse populations. Further examination focusing on psychometric qualities and discrepancies based on gender, ethnicity, age, and clinical status are recommended.
There were also a number of notable strengths of this study. First, the sample sizes for both the African American and European American groups were large, resulting in analyses that are representative of both cohorts. This factor provides depth to the understanding of the scale, as conclusions can be made based on responses of both African Americans and European Americans separately, and appropriate comparisons could be examined. Additionally, the analytic strategy used by researchers provided two different, but important, avenues for examining the validity of the STICSA.
Conclusion
The overall aim of this study was to evaluate the convergent validity and factor structure of the STICSA in African American and European American samples. Results indicate that convergent validity was generally consistent for both groups, suggesting that the measure predicts anxiety similarly for both African Americans and European Americans. These findings align with the body of literature that provides support for the use of the STICSA in identifying anxiety and expand on its utility by providing evidence for its efficacy within diverse samples. Somewhat surprisingly, the results of the factor analysis contrast with that of previous evidence (Gros et al., 2007; Gros et al., 2010; Ree et al., 2008). These findings call for additional investigation in the structural understanding of the STICSA for both African Americans and European Americans. Future research replicating the statistical methods used in this study and using diverse racial samples would provide insight into the nature of the STICSA’s factor structure and the validity of the measure.
Despite the need for additional investigation, the STICSA does provide clinically relevant and novel information needed for the assessment of anxiety. Unlike other frequently cited measures of anxiety that focus more on symptoms of social anxiety (SIAS; Mattick & Clarke, 1998) and symptoms of panic (BAI; Chapman et al., 2009), the STICSA evaluates anxiety symptoms across both cognitive and somatic domains. Recognizing variations in symptom presentation for different populations, counselors can use the cognitive and somatic frame of the STICSA to identify unique clusters of symptoms presented by individuals. With specific profiles defining how individuals experience anxiety, including cognitive and somatic symptoms and state and trait variations, clinicians can select and tailor appropriate interventions and treatment plans to influence change toward psychological well-being. Given the potential clinical utility of the measure, additional psychometric and cross-cultural studies of this measure will provide additional credibility to its continued use.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.