A review of the literature establishes that anxiety and depression are common in adolescents with a 12-month prevalence rate of 6.9% for anxiety and 1.7% for depression (Roberts et al. 2006). These disorders can develop during childhood or later in adolescence and can potentially create a significant burden for the adolescent, family, and society. Most research has focused on treatment, but attention also needs to be directed towards the prevention of these disorders. Identifying reliable, valid, and modifiable risk factors is one way of achieving this goal. Anxiety sensitivity is one specific cognitive risk factor that has generated interest towards this end; unfortunately, as a unified variable it demonstrates adequate convergent validity, but has poor discriminate validity. To date, researchers have only examined anxiety sensitivity as a unified variable, as it relates to adolescent anxiety and depression, and not as component parts as they individually relate to specific anxiety disorders and depression. The purpose of this study is to address this gap in the literature, by examining the component parts of anxiety sensitivity to specific anxiety disorders and depression in an effort to demonstrate improved discriminate validity.

Prevention and Cognitive Risk Factors

Anxiety and depression are common mental health conditions in the general population (Costello et al. 2003; Kessler and Walters 1998) and often first develop in childhood or adolescence (Costello et al. 2006). Anxiety and depression create a heavy burden on both society and the individuals who suffer from these disorders (e.g., decreased quality of life, decreased academic achievement). A study of adults in the United States during the 1990s found the annual cost for treatment of anxiety disorders to be $42.3 billion, and for depression, $43.7 billion (Greenberg et al. 1993, 1999).

While numerous evidence-based treatments for anxiety and depression exist (e.g., cognitive behavioral therapy, interpersonal psychotherapy, medications), attention needs to be paid to the prevention and early intervention of these disorders as recommended by the subcommittee on Children and Families created by an executive order of the President’s New Freedom Commission on Mental Health Care (Huang et al. 2005). Weisz et al. (2005) in their meta-analytic review of the literature on the effectiveness of prevention programs for adolescents found that universal prevention (i.e., prevention programs available to anyone in the eligible population) demonstrated a range of small to large effect (mean effect ranging from .24 to .93). These results suggest prevention, as anticipated, is an important and effective goal to pursue.

There are many types of prevention programs, however, the one that is most appropriate for preventing a mental health condition is an “indicated prevention intervention” (Weisz et al. 2005). An indicated preventive intervention is one that is “directed at individuals who are already showing signs or symptoms that foreshadow a mental or behavioral disorder” (Huang et al. 2005, pp. 262–263). In order to develop an indicated prevention intervention, there needs to be reliable, valid, and modifiable risk factors identifying people who are vulnerable to the development of these disorders.

Three potential cognitive risk factors of anxiety and depression will be examined in this study: anxiety sensitivity, negative affectivity and positive affectivity. Anxiety sensitivity is defined as the fear of bodily sensations (Reiss and McNally 1985). Negative affectivity is defined as a general factor of emotional distress (e.g., worry, disgust, fear, sadness, anger, and guilt) and is viewed as a risk factor to the development of both anxiety and depression (Clark and Watson 1991). Positive affectivity is defined as interacting positively with the world and as a “zest for life” (Clark and Watson 1991, p. 321). The research on anxiety sensitivity will be reviewed first followed by research on positive and negative affectivity.

Anxiety Sensitivity

Several studies have found associations between anxiety sensitivity and panic attacks, anxious symptoms, and anxiety disorders after controlling for trait anxiety and depression (e.g., Calamari et al. 2001; Joiner et al. 2002; Muris et al. 2001). Research prospectively found anxiety sensitivity to increase the risk of panic attacks in a group of normal adolescents, especially when the adolescent had stable and high levels of anxiety sensitivity (Hayward et al. 2000; Weems et al. 2002).

An important consideration, when identifying risk factors to be used in an indicated preventive intervention, is whether or not the risk factor is modifiable. Hazen et al. (1996), in a study with adults, found that anxiety sensitivity could be lowered using cognitive behavioral methods such as providing psychoeducation about anxiety, cognitive restructuring, and introceptive exposures and in-vivo exposures. In addition, several studies found parental behaviors such as parent hostility, threats, rejecting behaviors, and parental reinforcement of sick-role behavior in their children played a contributory role in the development of anxiety sensitivity (Cohan et al. 2004; Scher and Stein 2003; Watt et al. 1998). This information is not to be used to blame parents, but suggests potential avenues for intervention with parents exhibiting these behaviors, while co-jointly treating the child.

Anxiety sensitivity is a broad construct and, therefore, does not discriminate well among the various anxiety disorders and depression (e.g., Kearney et al. 1997; Mannuzza et al. 2002). Reiss (1991) initially viewed anxiety sensitivity as one of the fundamental fears within the expectancy theory of fear. Further research has demonstrated that rather than being a unitary construct, anxiety sensitivity is more likely composed of three (physiological, mental, and social concerns) or four factors (mental incapacitation concerns, disease concerns, unsteady concerns, and social concerns) (e.g., Muris 2002; Muris et al. 2001; Silverman et al. 1999, 2003).

Silverman et al. (2003) reported strong evidence that anxiety sensitivity is represented by four factors. According to these researchers, the four factors of anxiety sensitivity are (1) mental incapacitation concerns, worry that indicates something is mentally wrong with them; (2) disease concerns, worry that physical sensations represent a disease or sickness; (3) unsteady concerns, worry that symptoms will make them unsteady (e.g., faint); and (4) social concerns, worry that others will find out that they are anxious. The next step, in order to pursue this further, is to examine the relationship of these four factors as outlined by Silverman et al. as they relate to specific anxiety disorders and depression.

There were no studies found that examined how the four factors of anxiety sensitivity relate to specific anxiety disorders and depression in adolescents, therefore, we consulted the literature regarding adult populations. Rector et al. (2004) found mental incapacitation and social concerns were positively related to generalized anxiety disorder and physiological concerns (i.e., disease concerns and unsteady concerns) were positively related to panic. Last and Perrin (1993) found physiological concerns were related to separation anxiety. Zinbarg et al. (1997) found social concerns were positively related to social anxiety disorder. Joiner et al. (2002) found mental incapacitation concerns were positively related to depression. No research was found that examined the factors of anxiety sensitivity to OCD in adults. Therefore, an exploratory model with all four factors of anxiety sensitivity will be used.

Positive and Negative Affectivity

Research supports the role of negative affectivity being positively related to anxiety and depression and low positive affectivity (i.e., anhedonia) being related to depression (e.g., Austin and Chorpita 2004; Chorpita 2002; Chorpita and Daleiden 2002; Chorpita et al. 2000a; Lambert et al. 2004). For example, Jacques and Mash (2004) examined the relationships between positive and negative affectivity and anxiety and depression in a sample of elementary and high school aged boys and girls (n = 472) and found general support using structural equation modeling (SEM). As predicted, negative affectivity was highly correlated with both depression and anxiety and positive affectivity was negatively correlated with depression. Lonigan et al. (2003) evaluated the stability of negative and positive affectivity and its relation to anxiety and depression. Consistent with the hypothesis, negative and positive affectivity were related to anxiety and depression; additionally, the results indicated the stability of positive and negative affectivity across time. However, there was not any research found to determine if positive and negative affectivity were modifiable.

In summary, anxiety sensitivity and negative affectivity have been identified as risk factors to the development of anxiety, while negative affectivity, low positive affectivity, and a component part of anxiety sensitivity, mental concerns, have been linked to depression. The specific aims of the study are: (1) to examine how the four factors of anxiety sensitivity as defined by Silverman et al. (2003) relate to specific anxiety disorders and depression in an effort to improve discriminate validity, and (2) to examine the role of positive and negative affectivity in anxiety and depressive disorders. It is hypothesized that when using the specific factors of anxiety sensitivity as related to specific anxiety disorders, the model will demonstrate good convergent and discriminate validity will be demonstrated. Additionally, it is hypothesized that low positive affectivity will be related to depression and negative affectivity will be positively related to both anxiety and depression. Finally, it is hypothesized that negative affectivity and mental incapacitation concerns will be related to OCD.

Methods

Sample, Sample Size, and Participants

The initial sampling frame was drawn at a large, private mental health practice in the Baltimore–Washington region. The practice included psychiatrists, psychologists, and clinical social workers, and had contracts with most health insurance providers (public and private), the school system, and child protective services. The sampling frame contained the names of 426 adolescents with an anxiety and/or depressive disorder. The sample list was reviewed for redundant entries and sibling pairs. There were 61 redundant entries and 21 sibling entries that were removed. Examining surnames for siblings reduced, but may not have eliminated all possibility of having siblings’ complete the questionnaires. Thus, study instructions explained that if a family received two measures, only one child per family was to complete the measures. The final sample size was 344. The goal was to achieve at least a 50% response rate for a sample of 150 to have an ample sample size for up to ten predictors (Cohen et al. 2003). A simple random sample of 315 potential participants was selected using Microsoft Office Excel 2003 random number generator, = RAND ().

Data were collected on 185 adolescents. Their average age was 15.09 years (SD = 1.9) with 108 (58.4%) being male and 77 (41.6%) being female. The median family income for the participants was $90,000 (SD = $36,489). The adolescents’ ethnicities were Caucasian (n = 150, 81.1%), Hispanic or Latino (n = 20, 10.8%), African-American (n = 7, 3.8%), Asian (n = 5, 2.7%), and American Indian or Alaska Native (n = 1, .5%). Parents’ marital status was reported as: married or remarried (n = 135, 73%), divorced (n = 34, 18%), separated (n = 8, 4.3%), widowed (n = 2, 1.1%), other (n = 2, 1.1%).

Materials and Procedure

This study employed four-timed mailings sent to a random sample of adolescents who were currently in mental health treatment for an anxiety and/or depressive disorder. The study was approved by the university IRB. The mailings involved an introductory letter, followed by a packet that contained the scales and consent and assent forms, a follow-up post-card, and finally, a new packet sent to those who had not responded. This procedure yielded a 61.1% response rate, which is considered an adequate response rate (Meyers et al. 2006) and there was very little missing data.

Instrumentation

The adolescents completed the 18-item Childhood Anxiety Sensitivity Index by indicating the extent to which the statement is like them on a 3-point scale: 1 (none), 2 (some), or 3 (a lot) with higher scores reflecting a greater degree of anxiety sensitivity. (CASI; Silverman et al. 1991, 1999, 2003). Internal consistency was very good (α = .87) and test–retest reliability (2 week interval) was .76. There was also strong evidence of construct validity, including across ethnic groups (e.g., Lambert et al. 2004). The CASI was used to operationalize the four factors (e.g., predictor variables) of anxiety sensitivity: mental incapacitation concerns, disease concerns, unsteady concerns, and social concerns.

The PANAS-C has 30 different feelings or emotions designed to measure positive (e.g., positive feelings) and negative affectivity (e.g., emotional distress) (Laurent et al. 1999). Adolescents circle how much they felt each feeling/emotion during the past week—very slightly or not at all, a little, moderately, quite a bit, or extremely. The scale was developed using a non-clinical sample of children (n = 349) in grades 4 through 8 followed by a similarly composed replication sample (n = 358). Internal consistency was also good for both the positive affectivity subscale (α = .87–.89) and for the negative affectivity subscale (α = .92–.94) (Laurent et al. 1999). There is evidence that the scale has good convergent and discriminant validity (Chorpita and Daleiden 2002; Laurent et al. 1999).

The Revised Child Anxiety and Depression Scale (RCADS; Chorpita et al. 2000b) is a scale designed to measure DSM-IV criteria for anxiety and depression in children and adolescents 6–18 years old and was used to operationalize the anxiety disorders (i.e., generalized anxiety, panic, separation anxiety, social phobia, and obsessive compulsive disorder) and depression. Reliability was good: separation anxiety = .81, panic disorder = .85, generalized anxiety = .80, major depression = .76, social anxiety = .78, and obsessive-compulsive disorder = .71. One-week retest reliability was also good; the test–retest correlations ranged from .63 to .87. There was evidence of both convergent and discriminant validity for the RCADS subscales with the Children’s Depression Inventory and the Revised Children’s Manifest Anxiety Scale (Chorpita et al. 2000b).

Data Analyses

Predictors of specific anxiety disorders and depression will be examined utilizing multiple regression analyses. Predictors were entered using hierarchal entry. The adolescent’s sex and age were entered first in order to statistically control for these variables, followed by negative affectivity, which is predictive of both anxiety and depression, and finally, the individual factors of anxiety sensitivity depending on the disorder. Only the specific predictors that are supported by prior research (borrowed from the adult research) will be included in each model. There will be one exploratory model evaluated for OCD, as there was no prior research found. Thus, all factors of anxiety sensitivity will be used in this model. The first model examines mental incapacitation concerns, social concerns, and negative affectivity in relation to generalized anxiety disorder. The second model examines social concerns and negative affectivity as related to social anxiety disorder. The third and forth models examine how physiological concerns, as defined by the two physiological based subscales of the CASI—unsteady concerns and disease concerns, related to both separation anxiety and panic. The fifth model examines mental incapacitation concerns and negative affectivity related to OCD. The sixth model involves mental incapacitation concerns, negative affectivity and low positive affectivity related to depression.

All data were analyzed using SPSS 14.0 for Windows. Descriptive statistics (e.g., frequencies and percentages) were used to describe the sample. Predictors of specific anxiety disorders and depression were examined in six separate regression equations. Age and gender were controlled for and entered on the first step for all equations. The second step involved simultaneous entry of specific predictor variables for generalized anxiety, separation anxiety, panic, social anxiety, obsessive compulsive disorder, and depression. Assumptions were adequately met for all six regression analyses. Means, standard deviations, and coefficient alphas for all scales are presented in Table 1.

Table 1 Means, standard deviations, coefficient alphas (n = 185)
Full size table

Results

Predictors of Generalized Anxiety Disorder

The model examining predictors of generalized anxiety scores was statistically significant (F(5, 179) = 31.75, p < .005) and accounted for 47% of the variance (adjusted R 2 = 45.5) (see Table 2 for regression coefficients and confidence intervals). Higher levels of reported negative affectivity (p < .005) and mental incapacitation concerns (p = .025) were related to generalized anxiety.

Table 2 Predictors of anxiety disorders and depression scores (n = 185)
Full size table

Predictors of Social Anxiety Disorder

The model examining predictors of social anxiety scores was statistically significant (F(4, 184) = 22.29, p < .005) and accounted for 33.1% of the variance (adjusted R 2 = 31.6) (see Table 2 for regression coefficients and confidence intervals). Higher levels of reported negative affectivity (p < .005) and social concerns (p < .005) were related to social anxiety.

Predictors of Panic Disorder

The model examining predictors of panic disorder scores was statistically significant (F(5, 179) = 50.30, p < .005) and accounted for 58.4% of the variance (adjusted R 2 = 57.3) (see Table 2 for regression coefficients and confidence intervals). Higher levels of reported negative affectivity (p < .005), unsteady concerns (p = .004), and disease concerns (p < .005) were related to panic disorder.

Predictors of Separation Anxiety Disorder

The model examining predictors of separation anxiety disorder scores was statistically significant (F(5, 179) = 18.78, p < .005) and accounted for 34.4% of the variance (adjusted R 2 = 32.6) (see Table 2 for regression coefficients and confidence intervals). Younger age (p = .004) and higher levels of reported negative affectivity (p < .005) and disease concerns (p = .001) were related to separation anxiety disorder.

Predictors of Obsessive Compulsive Disorder (OCD)

The model examining predictors of OCD scores was statistically significant (F(7, 177) = 14.17, p < .005) and accounted for 36% of the variance (adjusted R 2 = 33) (see Table 2 for regression coefficients and confidence intervals). Higher levels of reported negative affectivity (p < .005), higher levels of mental incapacitations concerns (p = .017), and higher levels of disease concerns (p < .005) were related to OCD scores.

Predictors of Depression

The model examining predictors of depression scores was statistically significant (F(5, 179) = 35.48, p < .005) and accounted for 49.8% of the variance (adjusted R 2 = 48.4) (see Table 2 for regression coefficients and confidence intervals). Higher levels of reported negative affectivity (p < .005) and lower levels of positive affectivity (p = .002) were related to depression.

Discussion

This is the first study that we are aware of that examines predictors of specific anxiety disorders and depression in a clinical group of adolescents using the specific individual factors of anxiety sensitivity. Prior research indicates adolescents experience high rates of depression and anxiety disorders. The identification of factors that predict anxiety and depressive disorders is a critical issue for the development of prevention and early intervention programs with adolescents at risk. This research adds support to the current literature on adults. There is general support for the first hypothesis that using the specific factors of anxiety sensitivity to specific anxiety disorders demonstrates good discriminate and convergent validity. There is also support for the second hypothesis that low positive affectivity is related to depression and negative affectivity is related to both anxiety and depression. There is also general support for negative affectivity and mental incapacitation concerns being related to OCD.

This result suggests that adolescents with high levels of anxiety sensitivity may be at greater risk for future anxiety disorders, and that negative affectivity may function as a general risk factor to the development of anxiety and depression, and low positive affectivity is identified as a risk factor for depression. Results from this study also provide support for using the component parts of anxiety sensitivity: disease concerns, mental incapacitation concerns, unsteady concerns, and social concerns, as predictors of specific anxiety disorders.

Consistent with prior research, these results demonstrate negative affectivity and mental concerns are related to generalized anxiety (Chorpita 2002; Rector et al. 2004); negative affectivity and social concerns are related to social anxiety (Chorpita 2002; Zinbarg et al. 1997); negative affectivity, disease concerns, and unsteady concerns are related to panic scores (Chorpita 2002; Rector et al. 2004); negative affectivity and disease concerns are related to separation anxiety (Chorpita 2002; Last and Perrin 1993); and negative affectivity and low positive affectivity are related to depression (Chorpita 2002). Also consistent with prior research, negative affectivity is shown to be related to obsessive compulsive disorder scores (Chorpita 2002).

Negative affectivity is a general risk factor for the development of anxiety and depression. The individual components of anxiety sensitivity are risk factors for specific anxiety disorders and therefore are important constructs to understand. More research is needed to examine the role of negative affectivity and the individual components of anxiety sensitivity in the development of anxiety and depression. Two important questions are: (1) can the level of anxiety sensitivity and negative affectivity be modified in children and adolescents and (2) will the decrease in the level of anxiety sensitivity and negative affectivity decrease the likelihood of an anxiety or depressive disorder developing. Future research should elaborate on predictor models to include variables such as how parenting styles (e.g., degree of reinforcement in sick role) play a role in the development of these disorders.

Interestingly, OCD and GAD share two of the same factors—negative affectivity and mental incapacitation concerns. This suggests that these two disorders share more similarities than originally thought or these disorders are on a continuum of psychopathology. From a topographical point of view, the two disorders are similar in that they both have intrusive and difficult to control worried-based-thoughts. Thus, this suggests the possibility of GAD being conceptualized as a form of pure obsessional OCD; that is, the obsessive worries without overt compulsions. Previous research suggests that OCD tends to be comorbid with GAD (Relative Risk = 2.3).

Two important practice implications can be drawn from this study. First, social workers can use the components of anxiety sensitivity to help distinguish between different anxiety disorders and depression when trying to render a diagnosis. For example, if an adolescent is struggling with anxiety, and it appears to be social anxiety disorder, an elevated score on the social concerns subscale could provide support for the diagnosis of social anxiety disorder. In addition, social workers can use the positive affectivity subscale to help distinguish between anxiety and depression, with low scores of positive affectivity indicating depression. Finally, information from the anxiety sensitivity subscales can assist a social worker during treatment. For example, when a social worker is treating an adolescent with an anxiety disorder, knowing what components of anxiety sensitivity are elevated can be helpful in determining treatment interventions. For instance, if an adolescent with panic disorder has elevated scores on the disease concerns subscales, cognitive restructuring techniques could be used to address these fear concerns.

As with all studies, there are limitations here as well. Two limitations are the cross-sectional nature of the data and the mono-method effect. In cross-sectional data the theory specifies the direction of the relationship, but cannot be used to draw causal conclusions. Operationalizing a concept with only one measure is dangerous because the whole concept may not be covered adequately. An additional limitation is that the median income of the sample was higher than the community from which the sample was drawn. However, the ethnic/racial variation in this sample did match the community from which it was drawn.

Results from this study and from prior research suggest that we may be able to identify cognitive risk factors involved in the development of various anxiety and depressive disorders. Understandably more research is needed as this is the first study to examine the specific individual risk factors of anxiety sensitivity in a sample of adolescents. However, if we are able to identify cognitive risk factors and thus screening tools, prevention and early intervention programs can be developed aimed specifically at adolescents who would be at risk. This is an exciting time as we can now consider developing a knowledge base of risk factors that may support early intervention or prevention strategies for anxiety and depressive disorders, rather than only treating them.