Positive Affect as a Buffer between Chronic Stress and Symptom Severity of Emotional Disorders

Abstract

Research has demonstrated that stressors play a critical role in the development of generalized anxiety disorder (GAD), social anxiety disorder (SAD), and major depressive disorder (MDD). Separately, deficits in positive affect (PA) have been identified in GAD, SAD, and MDD. While previous research has linked the buffering effects of PA in chronic illness, such effects have yet to be investigated for chronic stressors and emotional disorder-related symptom severity. The purpose of the present study was to examine PA as a moderator of chronic interpersonal and non-interpersonal stress on GAD, SAD, and MDD symptom severity. Using a multilevel statistical approach with a sample of adolescents and young adults (N=463), PA was found to significantly moderate the relationship between chronic interpersonal stress and symptom severity for MDD and SAD. Findings suggest that in times of chronic interpersonal stress, higher PA may serve as a buffer from development of SAD and MDD symptoms.

Introduction

Stress has long been implicated as a factor in the development of mental health disorders and is predictive of poorer psychological well-being (McGonagle & Kessler, 1990; Monroe & Simons, 1991; Pynoos, Steinberg, & Piacentini, 1999). Based on its time course, stress can be categorized into either acute (episodic) or chronic (ongoing) stress (Hammen, Kim, Eberhart, & Brennan, 2009). While acute stressful life events possess relatively distinct onset and offset, chronic stress captures the “enduring aspects of the social and/or physical environment which involve deprivation or disadvantage and create a continuous stream of threats and challenges for the individual” (Compas 1987, p. 276) and often slowly develops as a continuing, problematic condition (Wheaton, 1997). Chronic stress is typically categorized as interpersonal or non-interpersonal in nature (Connolly, Eberhart, Hammen, & Brennan, 2010; Hammen, 1991; Hammen et al., 1987). Chronic interpersonal stress arises from sustained relationship discord – wherein a given relationship may be conflictual, untrustworthy, or nonreciprocal – or, alternatively, from the absence of close, confiding relationships (Shih, Eberhart, Hammen, & Brennan, 2006). In contrast, stressors not generated from relationship dynamics are considered non-interpersonal stressors (Shih, Eberhart, Hammen, & Brennan, 2006). Chronic non-interpersonal stressors take a variety of forms, including enduring financial strain, academic or job-related difficulties, neighborhood crime, and poor health (Baum, Garofalo, & Yali, 1999; Dohrenwend et al., 1992; Rudolph et al., 2000; Sheets & Craighead, 2014).

Both acute and chronic stress have been implicated in the development of anxiety and depressive disorders (Arborelius, Owens, Plotsky, & Nemeroff, 1999; Baran, Armstrong, Niren, Hanna, & Conrad, 2009; Faravelli et al., 2012; Hammen, 1991; Heim, Newport, Mletzko, Miller, & Nemeroff, 2008; McWilliams, Cox, & Enns, 2003; Phillips, Hammen, Brennan, Najman, & Bor, 2005; Vyas, Pillai, & Chattarji, 2004). Disorder-specific research has shown that both acute and chronic stress contribute to the onset of generalized anxiety disorder (GAD; Blazer, Hughes, & George, 1987; Kendler, Hettema, Butera, Gardner, & Prescott, 2003; Kessler et al., 2008; Moffitt et al., 2007). For example, in adults (ages 18-44 years), experiencing one or more stressful life events in the past year was found to increase the risk threefold of developing GAD (Blazer, Hughes, & George, 1987). These stressful life events were both interpersonal and non-interpersonal (e.g.., personal health, legal matters involving self or family, change in residence, work or marital status), and were coded as “stressful life events” if they were unexpected, negative, and very important. A second study examining adults (ages 18-55 years) showed that experiencing more acute stressors, particularly threat-related (e.g., potential future loss, including potential trauma or threatening/dire outcomes) or loss-related stressors (i.e., reduced sense of connectedness/well-being potentially covering every aspect of life, such as loss of person, material possessions, or health), significantly predicted GAD onset (Kendler, et al., 2003). Furthermore, stressful life events during childhood have also been shown to predict GAD onset (Beesdo, Pine, Lieb, & Wittchen, 2010; Kessler et al., 2008; Moffitt et al., 2007). For example, Kessler, et al. (2008) found that childhood adversity – measured as maltreatment (neglect, physical abuse, sexual abuse) and loss (parental death, parental divorce, other long-term separation) – significantly predicted GAD (age 15-24; Kessler, Davis, and Kendler, 1997). In total, the research seems to suggest that both acute and chronic stress may predict GAD.

Furthermore, the literature suggests that chronic interpersonal stress is associated with social anxiety disorder (SAD; Davila & Beck, 2002; Wittchen, Stein, & Kessler, 1999), though the data is scarcer than that for GAD. For example, Davila and Beck (2002) found that chronic interpersonal stress in the previous six months with close relationships (i.e., friendships, social life, romantic relationships, family relationships, and independence from one’s family) was significantly associated with greater SAD symptomatology for young adults (Mage =18.72). Similarly, Wittchen, et al. (1999) found that parental separation during childhood or early adolescence was an interpersonal risk factor associated with SAD (ages 14-24 years). These findings are complemented by an experimental animal study investigating chronic interpersonal versus non-interpersonal stress on subsequent anxiety (Barsy, Leveleki, Zelena, & Haller, 2010). The results showed that rodents who were administered chronic psychosocial stressors (i.e., repeated conflict with other rats) but not non-social stressors (i.e., physical restraint) exhibited increased anxiety in a subsequent social interaction test, suggesting that chronic interpersonal stress may increase social anxiety. Thus, in total, it seems that chronic interpersonal stress may be a risk factor for developing SAD.

Finally, both chronic and acute interpersonal and non-interpersonal stress are well-established risk factors for major depressive disorder (MDD; e.g., Hammen, 1991, 2003; Hammen, et al., 2004, 2009; Kessler, 1997; Kessler, et al., 2008; Mazure, 1998; Shih, et al., 2006). For example, Shih, et al. (2006) found that both male and female 15-year-olds experienced increased depressive severity as a result of acute and chronic interpersonal and non-interpersonal stress. In adults, Hammen, et al. (2009) investigated the relationship between chronic stress, acute stress, and major depressive episode onset in women (Mage = 41). Results from this study revealed that chronic stress was positively associated with the occurrence of acute stressors and that major depressive episode onset was positively associated with both chronic and acute stressors. These results were collapsed across both interpersonal and non-interpersonal stress. Thus, it seems that depression is associated with interpersonal, non-interpersonal, acute, and chronic stress.

Positive affect may facilitate resilience in the presence of stressors and reduce vulnerability to mental health disorders. Resilience is defined as the ability of individuals to adapt when faced with acute stress, trauma, or chronic adversity and maintain or rapidly regain psychological well-being (Carver, 1998; Charney, 2004; Fredrickson, 1998; Lazarus, 1993; Tugade & Fredrickson, 2007). Positive affect refers to one’s pleasurable engagement with the environment and includes positive emotional states, such as happiness, joy, interest, excitement, confidence, and alertness (Watson & Clark, 1991). Positive affect has been shown to promote flexibility in thinking and problem solving (Billings, Folkman, Acree, & Moskowitz, 2000; Folkman & Moskowitz, 2000b; Fredrickson & Branigan, 2005; Isen, Daubman, & Nowicki, 1987), reduce the physiological effects of negative emotions (Fredrickson, 2004; Fredrickson, Mancuso, Branigan, & Tugade, 2000; Fredrickson & Levenson, 1998; Ong, Bergeman, Bisconti, & Wallace, 2006), build enduring social resources (Keltner & Bonanno, 1997), promote effective coping strategies (Affleck & Tennen, 1996; Folkman & Moskowitz, 2000a, 2000b; Folkman, Moskowitz, Ozer, & Park, 1997; Ong, Bergeman, & Bisconti, 2004; Ong et al., 2006), and create upward spirals of improved emotional well-being (Fredrickson & Joiner, 2002).

The “dynamic model of affect” (Zautra, Smith, Affleck, & Tennen, 2001) predicts that positive and negative affect are relatively independent under ordinary circumstances but may become inversely related during stress. In support, high (but not low) positive affect significantly lowered the effects of chronic health-related stress on negative affect (Zautra et al., 2001). High positive affect also reduced the deleterious effects of pain on negative affect significantly more during times of high versus low pain (Zautra et al., 2001). Similarly, Zautra, Johnson, and Davis (2005) found that, in a sample of females experiencing chronic health-related stress and pain from osteoarthritis and/or fibromyalgia, higher weekly levels of pain and stress predicted higher negative affect. However, higher weekly levels of positive affect and average overall positive affect resulted in less negative affect, and this association was stronger during weeks of high pain. In addition, individuals with greater positive affect were less likely to demonstrate higher negative affect during weeks of interpersonal conflict when compared to those with low positive affect – though these interpersonal stressors were more acute in nature (e.g., argued with spouse/partner). Similar patterns were reported by Strand et al. (2006), where, during times of pain, individuals high in positive affect experienced less negative affect. Moreover, Tugade and Fredrickson (2007) found that cultivating positive affect promoted resilience through automatic activation of positive emotions during stress. Taken together, these findings suggest that positive affect buffers chronic interpersonal and non-interpersonal stress through reductions of negative affect.

The role of positive affect is especially relevant given evidence that low levels of positive affect are characteristic of depression (Watson & Clark, 1991) and anxiety disorders (Amies, Gelder, & Shaw, 1983; Brown, Chorpita, & Barlow, 1998; Chorpita, Plummer, & Moffitt, 2000; Kashdan, 2007; Watson, Clark, & Carey, 1988). Significant portions of factorial variance in symptoms of anxiety as well as depression are explained by positive affect (Kashdan, 2007; Prenoveau, Zinbarg, Craske, Mineka, Griffith & Epstein, 2010). Effect sizes for cross-sectional and longitudinal relationships between positive affect and anxiety are significant and indistinguishable from corresponding effect sizes for positive affect and depression (Kotov, Gamez, Schmidt & Watson, 2010; Khazanov & Ruscio, 2016). Of the anxiety disorders, the evidence for deficits in positive affect is most conclusive for SAD. Using structural equation modeling, Brown, Chorpita, and Barlow (1998) found no evidence for differential influence of positive affect while examining the paths of positive affect on SAD (−.28) and of positive affect on depression (−.29) after covarying for negative affect. Converging with these findings, a meta-analysis examining 19 studies which reported either correlations or group difference tests of positive affect and social anxiety also found a stable, moderate, inverse relationship (r = −.36; 95% CI: −.31 to −.40; Kashdan, 2007). Notably, this relationship remained consistent after covarying for depressive symptoms and disorders, which are commonly attributed to low positive affect and often comorbid with SAD (r = −.21; 95% CI: −.16 to −.26). Individuals with GAD also report low positive affect (Chambers, Power, & Durham, 2004; Power & Tarsia, 2007). In addition, an “anxious-misery” factor comprised of positive affect and anhedonic depression items from the Mood and Anxiety Symptom Questionnaire (MASQ; Keogh & Reidy, 2000; Watson & Clark, 1991) and Inventory to Diagnose Depression (IDD; Zimmerman & Coryell, 1987) was found to be associated with GAD, SAD, and MDD severity in a late adolescence to emerging adulthood sample (Prenoveau et al., 2010).

Given the relationship between a) positive affect and SAD, GAD, and MDD and b) positive affect and chronic stress, it is plausible that higher levels of positive affect may mitigate the effects of chronic stress upon these disorders. The goals of the present study were to investigate the role of chronic interpersonal and non-interpersonal stress in the development of SAD, GAD, and MDD symptom distress and impairment (i.e., symptom severity) and to examine whether positive affect moderated this relationship. The present study uses data from the Youth Emotion Project (e.g., Zinbarg et al., 2010), a longitudinal study aimed at assessing the predictors of the development of mood and anxiety disorders (e.g., neuroticism) from adolescence into young adulthood.

We hypothesized that chronic interpersonal stress would predict increased distress (e.g., relationship discord) and impairment relating to SAD and MDD symptom severity. Furthermore, we hypothesized that chronic non-interpersonal stress (e.g., financial, health, school/work, etc.) would predict increased distress and impairment relating to GAD and MDD symptom severity. Positive affect is associated with greater resilience during stress (Tugade & Fredrickson, 2007), and SAD, GAD, and MDD are associated with lower positive affect (e.g., Chambers, Power, & Durham, 2004; Kashdan, 2007). Thus, we hypothesized an interaction between positive affect and chronic interpersonal stress such that higher positive affect would moderate the effects of stress to predict decreased severity of SAD and MDD. Similarly, we hypothesized an interaction between positive affect and chronic non-interpersonal stress such that higher positive affect would moderate the effects of stress to predict decreased severity of GAD and MDD. Lastly, to examine specificity of the effects of positive affect, we analyzed the effects of negative affect in moderating the relationship between a) chronic interpersonal and non-interpersonal stress and b) GAD, SAD, and MDD symptom severity.

Methods

Participants

Participants enrolled in a larger two-site, 8- to 10-year longitudinal study (total N = 627) who completed a chronic life stress interview, a structured interview for psychiatric disorders, and a self-report measurement of positive affect at successive time points during the study were eligible for inclusion in the analyses. Our multilevel statistical approach allowed us to include multiple observations nested within individuals (see Statistical Approach). The larger study examined risk factors for psychopathology in late adolescence and early adulthood (Youth Emotion Project, YEP; Zinbarg et al., 2010) in three consecutive cohorts of high school juniors from two demographically diverse public high schools in suburban Chicago and suburban Los Angeles. To increase the incidence of psychopathology onset during the longitudinal phase, at-risk students exhibiting high levels of neuroticism (i.e., top tertile) as measured by the Neuroticism subscale of the revised 23-item Eysenck Personality Questionnaire Neuroticism Scale (EPQ–R–N; Eysenck & Eysenck, 1975) were oversampled (Clark, Watson, & Mineka, 1994; Hayward, Killen, Kraemer, & Taylor, 2000). See Zinbarg et al. (2010) for details regarding sampling procedures. Given that risk for mania is associated with abnormal elevations in positive emotion (e.g., Gruber, Johnson, Oveis, & Keltner, 2008), participants who endorsed symptoms of Bipolar I, Bipolar II, or Cyclothymia during the structured interview for psychiatric disorders at any timepoint during the study were excluded from the present analyses (n=15). Using these criteria, 1898 observations nested within 463 individuals were eligible for inclusion in the analyses (Mage = 16.94). The resulting sample was 68.7% female and 49.0% White, 15.1% Hispanic/Latino, 13.8% African-American, 4.5% Asian, 0.9% Native American/Pacific Islander/Alaskan Native, 12.1% Multiracial, and 4.5% who identified with another unlisted racial/ethnic group. Based on Hollingshead socioeconomic status scores, the sample on average was upper middle class (M = 48.96, SD = 12.37; Hollingshead, 1975).

Measures

Chronic interpersonal and non-interpersonal stress.

An annual evaluation of ongoing, typical chronic life stress was conducted using the UCLA Life Stress Interview (LSI; Hammen, 1991; Hammen et al., 1987), a semi-structured interview of ongoing, typical conditions in ten life domains. Each domain of the LSI was rated by trained interviewers on objective circumstances, indicating severity of chronic stress in the given role domain on a 5-point scale ranging from 1 to 5, using half-point increments (1 = minimal stress, 5 = very stressful circumstances). Per interview protocol, raters were trained to consider scores of 1 and 5 to be rare and relatively extreme cases and used general probes to elicit relevant, objective information from participants. To assess chronic interpersonal and non-interpersonal stress, role domains of 1) close friendship, 2) social life, 3) intimate (romantic) relationships, and 4) relationships with family members were categorized as interpersonal, while the domains of 5) neighborhood environment, 6) academic performance, 7) work environment, 8) financial status, 9) personal health, and 10) health of the individual’s family were considered non-interpersonal. For example, when evaluating the ‘close friendship’ role domain raters assessed the degree to which participants felt trust, stability, closeness, reciprocality, conflict within their closest friendship. For this domain, a rating of 1 would indicate a close friendship posing little chronic stress and that the relationship is close, confiding, mutually satisfying, and stable with good conflict resolution, whereas a rating of 5 would indicate an absence of a close, confiding friendship or a highly conflictual relationship (Shih, Eberhart, Hammen, & Brennan, 2006). Similar probes and criteria were used to evaluate other domains. Composite scores for chronic interpersonal and non-interpersonal stress used for analyses were calculated by averaging category-relevant domain scores. To determine baseline reliability of chronic life stress in the present study, intraclass correlation coefficients (ICCs) were calculated using 76 intersite and intrasite rated audiotaped interviews. The ICCs ranged from 0.57 to 0.91 for each domain, and averaged 0.73 across interpersonal and non-interpersonal domains (Doane et al., 2013).

Positive and negative affect.

Positive affect was measured using the Mood and Anxiety Symptom Questionnaire (MASQ; Watson & Clark, 1991), a self-report measure in which participants are asked rate the strength of emotional experiences over the past week using a 5-point Likert scale ranging from 1 (not at all) to 5 (extremely). Using factor analysis, high positive affect items were found to possess higher loadings (b = 0.73-0.78) on the general positive affect factor than low positive affect items (b = 0.42-0.46) (see Kendall et al., 2016). This may be due to unintended contamination of negative affect into items constructed to assess low positive affect on this measure (e.g., “Thought of death or suicide”) (Kendall et al., 2016). In addition, high positive affect items are considered to be more robust markers of the underlying positive affect factor (see Watson et al., 1988). As a result, we chose the mean of the 14 high positive affect items (e.g., “Felt optimistic”, “Felt like I had a lot of interesting things to do”) from the Anhedonic Depression subscale of the MASQ (MASQ-AD) to represent positive affect. Using the current sample of adolescents, the stability of the MASQ-AD was investigated over three years through a series of trait-state-occasion (TSO) models (see Kendall et al., 2016). A trait-only model was found to best represent the data, suggesting there was a significant trait component in positive affect as measured by the MASQ-AD. Furthermore, more than three quarters of the variance (77%) of the MASQ-AD, on average, was found to be stable and attributable to the trait component. As a result, while this measure was designed to assess time-variant experiences, we felt it was appropriate for this measure to represent a more stable measure of positive affect. Negative affect was measured by using the mean score of 15 disorder nonspecific items from the General Distress Mixed subscale (MASQ-GDM) which capture general negative affectivity (e.g. “Felt irritable”) also featured on the MASQ (Watson & Clark, 1991). We included negative affect in our models to 1) appropriately account for negative affect’s influence on disorder-related symptom severity (e.g., Clark, Watson, & Mineka, 1994; Watson & Clark, 1984) and 2) to demonstrate the unique moderating effects of positive affect for chronic interpersonal and non-interpersonal stress on disorder-related symptom severity.

Social anxiety, generalized anxiety, and major depressive disorder symptom severity.

The Structured Clinical Interview for the DSM-IV, non-patient edition (SCID-I/NP; Spitzer, Robert, Gibbon, & Williams, 2002) was administered to assess for Axis I DSM-IV disorders. Participants were administered the SCID-I/NP on an annual basis to evaluate diagnostic symptomatology since the prior interview (see Zinbarg et al., 2010, for details). After administering screening questions to determine which diagnostic sections required further inquiry, interviewers administered endorsed diagnostic sections and rated the clinical severity of each current diagnosis using the 0 to 8 Clinical Severity Rating (CSR) Scale (Di Nardo & Barlow, 1988). A CSR score of 4 or higher for a diagnosis is indicative of clinically significant, distressing or impairing symptoms. All participants received a CSR rating for all possible Axis I diagnoses. Thus, CSR was used as an index for symptom severity for SAD, GAD, and MDD. Inter-rater reliability (Cohen’s D; Cohen, 1960) for the present study was found to be acceptable to good for SAD (κ = .65), GAD (κ = .85), and MDD (κ = .83) diagnoses (see Uliaszek et al., 2009 for details). As aforementioned, participants endorsing symptoms of Bipolar I, Bipolar II, or Mania on the SCID-I/NP at any time point were excluded from analyses.

Statistical Approach

Hierarchical multilevel modeling was used to examine the effects of positive affect, chronic interpersonal stress, and chronic non-interpersonal stress on SAD, GAD, and MDD symptom severity. Analyses were performed in HLM 7.0 (Raudenbush, Bryk, Cheong, Congdon, & Du Toit, 2011). Multilevel modeling accounts for nonindependence of repeated measures within individuals and is a superior statistical approach for data with nested sources of variability. Thus, utilizing a multilevel approach allowed us to include multiple observations nested within individuals while remaining robust against violations of homoschedasticity and sphericity (Quené & Van den Bergh, 2004).

Our model included measurements of positive and negative affect at time point_t, chronic interpersonal and non-interpersonal stress at time point_{t + 1}, and disorder symptom severity at time point_{t + 2}. Each of these time points was separated by approximately one year. This was done to establish temporal precedence and to assess direction of effects. Establishing temporal precedence made certain that 1) any effects of positive or negative affect occurred on subsequent chronic stress and 2) chronic stress was experienced before the occurrence of disorder symptoms.

Covarying for previous symptom severity time point (i.e., lagged symptom severity) was not included in these present analyses due to threats of violation of the assumption of independence of random effects and covariates in mixed models. The hierarchical linear models in the current study include a random intercept which represents the combined effect on y of all unobserved variables that do not change over time. Because the model applies to all time points, the random intercept effect would likely include effects of baseline differences on the outcome that would covary with an observed lagged symptom severity variable. This violation of this independence can bias both the coefficient for the lagged dependent variable and the coefficients for other variables. The present hierarchical linear models control for person-level effects, which in part controls for previous time point symptom severity. Given that the majority of individuals had a disorder symptom severity of ‘0’, or no disorder symptom severity, at any given time point our outcome variables of interest possessed zero-inflated distributions. To appropriately account for this distribution, multilevel Poisson regressions were performed for our moderator analyses. Models were examined separately for interpersonal chronic stress and non-interpersonal stress as predictors due to avoid issues relating to multicollinearity of the two variables. SAD, GAD, and MDD symptom severity were examined separately as outcomes using the following model:

• Level-1 Model

  • $$\begin{gathered} \mathrm{E}({\text{Symptom Severity}}_{t+2(i)}\mid {\pi }_i)={\lambda }_{\mathit{ti}} \\ \log [{\lambda }_{\mathit{ti}}]={\eta }_{\mathit{ti}} \\ {\eta }_{\mathit{ti}}={\beta }_{0i}+{\beta }_{1i}({\text{Positive Affect}}_{t(\mathit{ij})})+ \\ {\beta }_{2i}({\text{Negative Affect}}_{t(\mathit{ij})})+ \\ {\beta }_{3i}({\text{Chronic Stress}}_{t+1(\mathit{ij})})+ \\ {\beta }_{4i}({\text{Positive Affect}}_{t(\mathit{ij})}\ast {\text{Negative Affect}}_{t(\mathit{ij})})+ \\ {B}_{5i}({\text{Positive Affect}}_{t(\mathit{ij})}\ast {\text{Chronic Stress}}_{t+1(\mathit{ij})})+ \\ {B}_{6i}({\text{(Negative Affect}}_{t(\mathit{ij})}\ast {\text{Chronic Stress}}_{t+1(\mathit{ij})}) \end{gathered}$$
• Level-2 Model

  • $$\begin{gathered} {\beta }_{0j}={\gamma }_{00}+u_{0j} \\ {\beta }_{1j}={\gamma }_{10}+u_{1j} \\ {\beta }_{2j}={\gamma }_{20}+u_{2j} \\ {\beta }_{3j}={\gamma }_{30}+u_{3j} \\ {\beta }_{4j}={\gamma }_{40}+u_{4j} \\ {\beta }_{5j}={\gamma }_{50}+u_{5j} \\ {\beta }_{6j}={\gamma }_{60}+u_{6j} \end{gathered}$$

Results

Descriptive Statistics

Descriptive statistics for all variables of interest are presented in Table 1. The overall mean of chronic interpersonal stress across all time points indicated participants on average experienced mild to moderate levels of chronic interpersonal (M = 2.29, SD = 0.42) and non-interpersonal stress (M = 2.23, SD = 0.33). Mean scores on the modified MASQ-GDM (M =1.90, SD = 0.67) and MASQ-AD (M =2.89, SD = 0.83) were consistent with previously collected undergraduate samples (e.g., Watson et al., 1995). Rates of clinically significant GAD (1.8%; M = 0.14, SD = 0.72) , SAD (3.3%; M = 0.30, SD = 0.95), and MDD (2.8%; M = 0.16, SD = 0.82) – indicated by a CSR of 4 or higher – corresponded with prevalence rates found in large-scale adolescent and young adult samples (see Beesdo, Knappe, & Pine, 2009).

Table 1.

Means and Standard Deviations of Positive Affect, Chronic Interpersonal Stress, and Disorder Symptom Severity

Measure	M	SD	Min	Max	N
Positive affecta	2.89	0.83	1.00	5.00	2626
Negative affecta	1.90	0.67	1.00	4.93	2626
Chronic interpersonal stressb	2.29	0.42	1.25	4.50	2626
Chronic non-interpersonal stressb	2.23	0.33	1.42	3.83	2621
Social anxiety disorder symptom severityc	0.30	0.95	0	7.00	1898
Generalized anxiety symptom severityc	0.14	0.72	0	6.00	1897
Major depressive disorder symptom severityc	0.16	0.82	0	7.00	1898

Note.

^aMood and Anxiety Symptom Questionnaire (MASQ; Watson & Clark, 1991) ranges 1 (not at all) to 5 (extremely);

^bLife Stress Inventory (LSI; Hammen, 1991; Hammen et al., 1987) ranges from 1 (minimal stress) to 5 (very stressful circumstances);

^eClinical Severity Rating (CSR; Di Nardo & Barlow, 1988) ranges from 0 (no distress/interference) to 8 extreme distress/interference).

Bivariate between-person correlations are presented in Table 2. Chronic interpersonal stress was significantly correlated with SAD (r = .092, p < .01), GAD (r = .101, p < .01) and MDD symptom severity (r = .104, p < .01). Similarly, chronic non-interpersonal stress was significantly correlated with SAD (r = .046, p < .05), GAD (r = .122, p < .01) and MDD symptom severity (r = .107, p < .01). Positive affect was significantly negatively correlated with SAD (r = −.074, p < .01) and MDD symptom severity (r = −.060, p < .01) but not with GAD symptom severity (r = −.041, p = .071). Furthermore, SAD, GAD, and MDD symptom severity were all positively significantly positively correlated with each other. SAD and GAD (r = .205, p < .01), SAD and MDD (r = .144, p < .01) GAD and MDD (r = .211, p < .01). Finally, negative affect was significantly positively correlated with SAD (r = .118, p < .01), GAD (r = .127, p < .01), and MDD symptom severity (r = .133, p <.01).

Table 2.

Bivariate Between Group Correlations Among Positive Affect, Negative Affect, Chronic Interpersonal Stress, and Disorder Symptom Severity

Measure	1	2	3	4	5	6	7
1. Positive affect	—
2. Negative affect	−.259**	—
3.. Chronic interpersonal stress	−.216**	.218**	—
4. Chronic non-interpersonal stress	−.168**	.143**	.470**	—
5. Social anxiety disorder symptom severity	−.074**	.118**	.092**	.046*	—
6. Generalized anxiety symptom severity	−.041	.127**	.101**	.122**	.205**	—
7. Major depressive disorder symptom severity	−.060**	.133**	.104**	.107**	144**	.211**	—

Note.

^**p < .01,

^*p <.05, Pairwise deletion

Symptom Severity, Chronic Interpersonal Stress, and Positive Affect

Main effects of chronic interpersonal and noninterpersonal stress on symptom severity.

To establish whether chronic interpersonal and non-interpersonal stress were predictors of future symptom severity of SAD, GAD, and MDD, we first analyzed the main effects of chronic interpersonal and non-interpersonal stress on disorder-related symptom severity ratings using separate Poisson regression models. Chronic interpersonal stress was a significant predictor of future SAD (b = .516, SE = .123, p < .001), GAD (b = .479, SE = .161, p = .003), and MDD symptom severity (b = .344, SE = .153, p = .029). Chronic non-interpersonal stress was a significant predictor of future GAD (b = 1.161, SE = .346, p < .001) and MDD symptom severity (b = .719, SE = .189, p = <.001), but not SAD symptom severity (b = −.004, SE = .239, p = .985).

Moderating effects of positive affect for chronic interpersonal stress on symptom severity.

To examine whether positive affect dampened the effect of chronic interpersonal stress on future SAD, GAD, and MDD symptom severity, we analyzed the main effects of positive affect, negative affect, chronic interpersonal stress, and their two-way interactions using separate multilevel Poisson regression models (Table 3). Supporting our hypothesis, positive affect significantly moderated the effects of chronic interpersonal stress on future SAD symptom severity (b = −.148, SE = .058, p = .011) and MDD symptom severity (b = −.657, SE = .242, p = .007), but not GAD symptom severity (b = −.080 SE = .081, p = .324). Given the chronological nature of our analyses – in which positive affect preceded chronic stress, which preceded disorder symptom severity – this suggests that increased positive affect reduced the effect of chronic interpersonal stress on SAD and MDD symptom severity, resulting in lower SAD¹ and MDD severity. Simple slopes tests were subsequently conducted to determine at which level positive affect moderates the effects of chronic interpersonal stress on SAD and MDD symptom severity (see Aiken &West, 1991). Poisson analyses at one standard deviation above (i.e., high positive affect) and below (i.e., low positive affect) the mean were executed by mean-shifting the data. For both SAD and MDD symptom severity, the effect of chronic interpersonal stress on symptom severity was significant when positive affect was low (SAD b = .38, SE = .20, p = .04; MDD b = .32, SE = .12, p = .01) but insignificant when positive affect was high (SAD b = −.10, SE = .14, p = .46; MDD b = −.33, SE = .25, p = .19). Thus, it appears positive affect possesses buffering capabilities beginning at high levels for SAD and MDD symptom severity. There were no significant results of negative affect moderating the effects of chronic interpersonal stress on SAD, GAD, or MDD (ps > .170). Thus, overall, positive affect buffered the effects of chronic interpersonal stress on symptom severity of SAD and MDD but not GAD; additionally, negative affect did not significantly moderate with the effects of chronic interpersonal stress in relation to symptom severity with SAD, GAD, and MDD.

Table 3.

Multilevel Poisson Models for Positive Affect, Negative Affect, Chronic Interpersonal Stress, Social Anxiety, Generalized Anxiety, and Major Depressive Disorder Symptom Severity

	Social anxiety disorder symptom severity					Generalized anxiety disorder symptom severity					Major depressive disorder symptom severity
Measure	β	SE	t-ratio	~ d.f.	p	β	SE	t-ratio	~ d.f.	p	β	SE	t-ratio	~ d.f.	p
Positive affect	−.194	.078	−2.458	462	.014	−.290	.114	−2.545	462	.011	.106	.182	.580	462	.562
Negative affect	.097	.110	.886	462	.376	−.197	.145	−1.365	462	.173	.901	.284	3.175	462	.002
Chronic interpersonal stress	.194	.181	1.072	462	.284	.316	.251	1.262	462	.208	.465	.336	1.383	462	.167
Positive affect × Negative affect	−.115	.060	−1.932	462	.054	−.027	.077	−2.212	462	.027	.126	.252	.500	462	.617
Chronic interpersonal stress × Positive affect	−.148	.058	−2.548	462	.011	−.080	.081	.988	462	.324	−.657	.242	−2.714	462	.007
Chronic interpersonal stress × Negative affect	.031	.063	.492	462	.623	−.103	.077	−1.374	462	.170	−.110	.217	−.509	462	.611

Moderating effects of positive affect for chronic non-interpersonal stress on symptom severity.

To examine whether positive affect dampened the effects of chronic non-interpersonal stress on future SAD, GAD, and MDD symptom severity, we paralleled the analyses conducted above (Table 4). Positive affect did not moderate the effects of chronic non-interpersonal stress with SAD, GAD, or MDD (ps > .231). However, negative affect significantly and positively moderated the effects of chronic non-interpersonal stress on future MDD symptom severity (b = .778, SE = .242, p = .001). As negative affect increased, the effect of chronic non-interpersonal stress on distress and impairment of MDD symptoms was amplified. Thus, overall, positive affect did not significantly moderate the effects of non-interpersonal stress on symptom severity of SAD, GAD, and MDD. However, negative affect enhanced the effects of non-interpersonal stress on MDD symptom severity, whereas no effects were observed with SAD and GAD symptom severity.

Table 4.

Multilevel Poisson Models for Positive Affect, Negative Affect, Chronic Non-Interpersonal Stress, Social Anxiety, Generalized Anxiety, and Major Depressive Disorder Symptom Severity

	Social anxiety disorder symptom severity					Generalized anxiety disorder symptom severity					Major depressive disorder symptom severity
Measure	β	SE	t-ratio	~ d.f.	p	β	SE	t-ratio	~ d.f.	p	β	SE	t-ratio	~ d.f.	p
Positive affect	.162	.082	1.955	462	.051	−.154	.115	−1.339	462	.181	−.044	.166	−.264	462	.792
Negative affect	.104	.107	.977	462	.329	−.243	.135	−1.798	462	.073	.305	.154	1.978	462	.049
Chronic non-interpersonal stress	.337	.212	1.591	462	.112	1.287	.302	4.625	462	<.001	−.277	.392	−.706	462	.481
Positive affect × Negative affect	−.158	.059	−2.657	462	.008	−.231	.082	−2.821	462	.005	−.310	.119	−2.607	462	.009
Chronic non-interpersonal stress × Positive affect	.068	.059	1.149	462	.251	−.097	.074	−1.199	462	.231	.233	.154	1.978	462	.481
Chronic non-interpersonal stress × Negative affect	−.084	.054	−1.551	462	.122	.037	.082	.501	462	.617	.778	.242	3.208	462	.001

Discussion

Research has reliably demonstrated that stress plays a critical role in the development of GAD (e.g., Blazer et al., 1987), SAD (e.g., Rosellini, Rutter, Bourgeois, Emmert-Aronson, & Brown, 2013), and MDD (e.g., Hammen, 1991). In a separate line of research, deficits in positive affect have been identified in individuals with GAD (e.g., Power & Tarsia, 2007), SAD (e.g., Brown et al., 1998), and MDD (Brown et al., 1998). The present study serves to both bridge and extend these two bodies of research by examining chronic interpersonal and non-interpersonal stress, largely unstudied dimensions of stress within the context of anxiety disorders, as predictors of SAD and GAD symptom severity and – in addition – MDD symptom severity. Furthermore, we examined the potential moderating role of positive affect on the effects of chronic interpersonal and non-interpersonal stress on SAD, GAD, and MDD symptom severity. Whereas previous research has linked the buffering effects of positive affect on negative affect in the context of chronic illness (e.g., Zautra, Johnson, & Davis, 2005), these findings have yet to be investigated for chronic stress and symptom severity of anxiety and depressive disorders. Thus, the present study was an initial investigation into the dynamics of positive affect and chronic stress for anxiety and depression symptom severity.

The present study had several main findings. First, chronic interpersonal stress had a significant main effect in predicting SAD, GAD, or MDD symptom severity. Chronic non-interpersonal stress had a significant main effect in predicting GAD and MDD – but not SAD – symptom severity. Most notably, higher positive affect significantly reduced the effects of chronic interpersonal stress on SAD and MDD – but not GAD – symptom severity. Negative affect did not significantly moderate the effects of chronic interpersonal stress on SAD, GAD, and MDD symptom severity, suggesting specificity of the effects of positive affect. In contrast to our chronic interpersonal stress findings, positive affect did not significantly moderate the effects of chronic non-interpersonal stress on SAD, GAD, and MDD symptom severity. Higher negative affect, however, enhanced the effects of chronic non-interpersonal stress on MDD symptom severity, but had no effects on SAD or GAD symptom severity.

We found that positive affect reduced the effects of chronic interpersonal stress on SAD and MDD symptom severity. This is the first study to implicate positive affect as a buffering agent between a) chronic interpersonal stress and b) SAD and MDD symptom severity. These results were not limited to those with a diagnosis of SAD or MDD, but applied to individuals with symptoms ranging from below clinical severity to extreme clinical severity. Conceivably, for SAD, relationship discord experienced during chronic interpersonal stress may give rise to increased fear of negative evaluation and rejection that may generalize to relationships broadly. These fears may then motivate SAD-related behaviors, such as avoidance of social interactions, and in turn reinforce social fears and increase symptom severity. Similarly, for MDD, relationship discord may give rise to depressed mood, anhedonia, behavioral inactivation, and various maladaptive cognitions (e.g., negative self-statements). Positive affect may protect individuals from these sorts of downstream effects and reduce SAD and MDD symptom severity. As previously noted, individuals with high positive affect have been found to build enduring social resources (Fredrickson & Branigan, 2001; Keltner & Bonanno, 1997). Thus, in times of chronic interpersonal stress, individuals with high positive affect may be more likely to cope partly due to support from their social network. From an associative and instrumental learning perspective, seeking support and feeling supported by others may provide positive social associations that inhibit negative, excitatory associations such as “if I ask for help, I will be rejected,” which are implicated in the development of SAD (see Sewart & Craske, in press). Thus, in times of chronic interpersonal stress, positive affect may provide motivation to seek out social support, which therein provides corrective inhibitory learning on socially-relevant feared outcomes. Similarly, individuals high in positive affect may engage in social forms of behavioral activation, leading to reward-based learning and experiences (e.g., “if I ask for help, people will support me; if I interact with other people, I will enjoy myself’) and reduction in MDD symptom severity.

These results of the benefits of positive affect converge with recent research evaluating the effects of targeting positive versus negative affect in the treatment of anxiety and depression (Craske, Meuret, Ritz, Treanor, Dour & Rosenfield, in submission). In this study, participants were admitted based on elevated anxiety or depression and were randomized to receive treatment focused on either increasing positive affect or decreasing negative affect (see Craske et al., in submission for details). While each treatment included cognitive-behavioral components, the focus of each treatment differed. For example, in the behavioral component, the negative affect condition engaged in behaviors to disconfirm negative outcomes (e.g., exposures), whereas the positive affect condition engaged in behaviors to increase positive outcomes (e.g., pleasant events scheduling). Similarly, in the cognitive component, the negative affect condition engaged in cognitive restructuring to reduce negative cognitions (e.g., reduce perceived likelihood and magnitude of negative outcomes), whereas the positive affect condition engaged in cognitive restructuring to enhance positive cognitions (e.g., increase anticipation/motivation for future positive effects, increase perceived ownership of positive outcomes). The results showed that the positive affect treatment condition was superior to the negative affect treatment condition in terms of symptoms of depression and anxiety, and for participants higher on pre-treatment levels of stress-related symptoms.

Our results also showed that for individuals with higher negative affect, experiencing later chronic non-interpersonal stress resulted in increases in MDD symptom severity. Given the heterogeneity in non-interpersonal stressors, we are unable to specify which of the many types of non-interpersonal stressors predicted this effect (e.g., finances, health, school/job, neighborhood crime). This suggests that if individuals already have high negative affect, experiencing later chronic non-interpersonal stress will combine with negative affect to increase MDD symptom severity.

Regarding positive affect’s buffering effect on chronic interpersonal stress, a plausible explanation for this specificity lies in a theorized function of positive emotions. According to the ‘broaden-and-build’ theory, positive emotions not only give rise to exploratory and novel thoughts and actions - thus broadening one’s behavioral range (i.e., thought-action repertoire), but also encourage the strengthening of enduring personal resources and relationships (Fredrickson 1998, 2001; Fredrickson, 2004). Given that one of the primary functions of positive affect is theorized to urge individuals to build upon interpersonal relationships to optimize well-being (Fredrickson & Branigan, 2001; Keltner & Bonanno, 1997), positive affect and associated behaviors may, as a result, also serve to directly remediate the psychological consequences chronic interpersonal stress. Although our findings did not support positive affect as a buffer for non-interpersonal chronic stress, this domain still deserves further attention. Given that certain forms of chronic non-interpersonal stress (e.g., financial problems) are likely to be higher for those with low socioeconomic status, future studies should examine if factors such as education or income may also play a predictive role in this relationship.

Furthermore, the present study found that positive and negative affect had an inverse interaction in predicting SAD, GAD, and MDD symptom severity within the chronic non-interpersonal stress model; a similar effect was observed in the chronic interpersonal stress model with GAD and trended with SAD. One interpretation of this finding is that higher positive affect lowers the effects of negative affect on later symptom severity. This is consistent with findings that positive affect reduces the physiological effects of negative emotions (e.g., Fredrickson, 2004; Fredrickson et al., 2000; Fredrickson & Levenson, 1998; Ong et al., 2006) and creates upward spirals of improved emotional well-being (Fredrickson & Joiner, 2002). Perhaps having high positive affect during times of negative affect protects individuals from later increasing SAD, GAD, or MDD symptom severity.

This study possesses several strengths. Rather than using a dichotomous outcome, the present study considered individuals presenting with a wide range of SAD, GAD, and MDD symptom severity spanning from non-existent to severely distressing and impairing symptomatology. Furthermore, rather than relying on questionnaires, we assessed both symptom severity and chronic stress using semi-structured interviews conducted by trained interviewers rather than relying upon self-report which can be biased by factors such as current mood state. In addition, we were able to establish temporal precedence between our three variables of interest, ensuring that observed disorder-related symptom severity did not precede chronic stress. Our variables were measured over a two-year period (see Statistical Approach), making a strong case for the durability of effects from positive/negative affect and chronic stress on SAD, GAD, and MDD symptom severity.

Additionally, our study possessed some limitations. Our measurement of positive affect was framed “over the past week.” While this specific questionnaire has demonstrated high stability over time (Kendall et al., 2016), future studies may benefit from using measures that are specifically created to capture trait versions of positive affect, such as the Positive and Negative Affect Schedule – Expanded Form (PANAS-X). Psychometric testing of the PANAS-X reflected consistently higher stability of both higher (e.g., positive affect) and lower-order (e.g., sadness) scale coefficients when items were queried using general instructions than when the same items were probed “over the past week” (Watson & Clark, 1999). Second, while a strength of our study was the use of clinical interview to measure symptom severity, our study would have benefited from using a multi-modal approach, including behavioral, physiological, or neurological data to measure positive/negative affect, anxiety, and depression. Third, chronic interpersonal stress and SAD symptom severity may possess a generative relationship, similar to that of stress generation observed in depressive disorders (Hammen, 1991). This hypothesis states that behaviors of depressed individuals – influenced by negative, depressive beliefs and expectations – contribute to the occurrence of negative life events. Chronic interpersonal stress continuation has been demonstrated in anxiety disorders broadly with neuroticism implicated as a partial contributor in this relationship, where continued stress elevation may be attributed to behaviors that correspond with high neuroticism (Uliaszek et al. (2012). However, stress generation has yet to be examined specifically in individuals with SAD. Thus, further statistical analyses employing longitudinal data, such as cross-lagged analyses or bidirectional associations, is warranted to determine the dynamics of this relationship. Lastly, our analyses were limited to SAD, GAD, and MDD. Future studies could examine the effects on other anxiety disorders.

Overall, this study highlights an exciting line of research that deserves further attention. If positive affect can indeed attenuate the effects of chronic interpersonal stress on SAD and MDD symptom severity, efforts to increase positive affect (e.g., Craske, Meuret, Ritz, Treanor, Dour & Rosenfield, in submission) in cognitive-behavioral therapy may result in further symptom reduction and protection from the deleterious effects of chronic interpersonal stress in the future.