Situational Uncertainty Interacts with Anxiety Sensitivity and Distress Intolerance to Predict Anticipated Worry and Preparation for a Hurricane

Abstract

Background:

There is evidence for the impact of emotional intolerance on reactivity to stressors, but it is unknown whether the level of situational uncertainty may moderate this relationship. We examined whether situational uncertainty moderated the relationship between emotional intolerance and anticipated anxious responding to hurricane forecasts, considering three aspects of emotional tolerance: anxiety sensitivity, distress intolerance, and hurricane-specific distress intolerance.

Methods:

Participants (N=358) were Florida residents who experienced Hurricane Irma. Participants were presented with two hypothetical storm forecasts that varied in level of uncertainty: 5-day forecast (high uncertainty) and 3-day forecast (low uncertainty). Participants rated their anticipated worry and preparation for each forecast.

Results:

Significant interactions between forecast uncertainty and both anxiety sensitivity and hurricane-specific distress intolerance emerged on anticipated worry, such that there was a stronger relationship in the high uncertainty condition. Forecast uncertainty also moderated the relationship between anxiety sensitivity and anticipated preparation in the same direction. There were no significant interactions between forecast uncertainty and distress intolerance on either anticipated worry or preparation.

Conclusions:

Specific aspects of emotional intolerance appear to have a stronger influence on anticipated worry and preparatory behavior in high uncertainty situations. These findings suggest that distinct emotional tolerance factors may influence these responses.

Introduction

Emotional Intolerance

Emotional intolerance is a cognitive vulnerability that renders people to be sensitive to, and unable to cope with, their internal thoughts and emotions. Emotional intolerance is thought to be a transdiagnostic risk factor for psychopathology (Bernstein et al., 2009), and several studies have implicated emotional intolerance in both the development and maintenance of mood and anxiety disorders (Hayes et al., 1996; Keough et al., 2010). This impaired ability to tolerate emotional discomfort is thought to amplify the experience of negative emotions and increase maladaptive avoidance strategies, thus leading to the development of psychopathology over time (e.g., Bernstein et al., 2009). High emotional intolerance has also been shown to predict anxiety-related outcomes following exposure to a stressor, such as a natural disaster (McLaughlin & Hatzenbuehler, 2009; Viana et al., 2018; Vujanovic et al., 2017). Importantly, emotional intolerance is a multifaceted, higher-order construct consisting of two distinct but related components: anxiety sensitivity (AS) and distress intolerance (DI; Bernstein et al., 2009; Mitchell et al., 2013). AS, the fear of anxiety-related sensations, and DI, the inability to endure negative emotional states, are thought to be lower-order factors that contribute to one’s overall level of emotional intolerance. Although AS and DI are related, they also demonstrate differential associations with different symptom domains (Bernstein et al., 2009). Further research is needed to explore the differential relationships of these two emotional intolerance constructs with other forms of psychopathology, such as worry.

AS, or the tendency to misinterpret anxiety-related sensations as threatening, has been studied extensively in relation to anxiety symptoms. People with higher levels of AS have an elevated fear of anxiety-related sensations and their consequences (Keough et al., 2010). Moreover, AS both increases the risk of developing anxiety-related disorders (e.g., Paulus et al., 2018) and maintains anxiety symptoms (e.g., Allan et al., 2018; Ehlers, 1995). AS has been associated with increased post-traumatic stress and anxiety symptoms following a traumatic event (e.g., Viana et al., 2018), and it appears to be an important risk factor for anxious responding following a hurricane (e.g., Hensley & Varela, 2008).

Another aspect of emotional intolerance is DI, the perceived inability to withstand negative emotional and physical states. High levels of DI have been linked to the development and maintenance of various psychopathologies, including substance use disorders (Brown et al., 2005), anxiety disorders (Laposa et al., 2015), and personality disorders (Bornovalova et al., 2011). Persons with higher DI are thought to be more likely to respond negatively to stressors and stress-related contexts, in turn leading to maladaptive coping strategies such as worry and avoidance (Zvolensky et al., 2010). DI also predicts worry and anxiety-related symptoms following a natural disaster, such as a hurricane (e.g., Cohen et al., 2016; Marshall-Berenz et al., 2010). Although DI has been traditionally conceptualized as a general construct, studies have found evidence for additional domain-specific components of DI and stress the importance of a multi-modal assessment of DI (McHugh & Otto, 2011). DI has been captured with a variety of self-report and behavioral measures, such as the Distress Intolerance Index (DII; McHugh & Otto, 2012) or the Willingness to Pay DI measure (WTP-DI; McHugh et al., 2011). The WTP-DI allows for the assessment of DI in specific domains and is modestly correlated with other measures of more global DI captured by the DII (McHugh et al., 2011). Research indicates that the various measures of DI capture related, but distinct aspects of the general construct (McHugh et al., 2011), though few studies have examined multiple DI measures simultaneously. Specific measures of DI appear to demonstrate stronger associations with specific clinical outcomes compared to measures of general DI (Sirota et al., 2010), however further research is needed to compare the effects of specific versus general DI.

Situational Uncertainty

Although the association between emotional intolerance and anxious responding has been well-documented, it is unclear to what extent situational factors may moderate this relationship. One particularly relevant situational factor is the level of situational uncertainty. Situational uncertainty, defined as an event that is marked by unpredictability, ambiguity, and a lack of information (e.g., Brashers et al., 2000), is associated with increases in worry and other responses to stress (Maissi et al., 2004; Sarinopoulos et al., 2010). Stressful situations occurring under uncertain conditions lead to more worry as compared to more certain conditions (Einstein, 2014), and explicitly uncertain situations, defined as scenarios that specifically involve a sense of uncertainty, have been found to be more anxiety-provoking compared to more certain situations (Reuman et al., 2015). Within the clinical literature, excessive anticipatory worry and maladaptive behavioral responses to ambiguous, yet potentially threatening situations is a core feature across anxiety disorders (Grupe & Nitschke, 2013). Grupe and Nitschke (2013) proposed the Uncertainty and Anticipation Model of Anxiety (UAMA), which suggests that processes involved in responding to uncertain threat function maladaptively in anxiety disorders. Based on this model, we expect that uncertainty would increase anxious responding, particularly for individuals with high levels of emotional intolerance who are already predisposed to anxiety. Further research is needed to understand how situational uncertainty impacts the association between emotional intolerance and anxious responding.

Limited research has explored whether situational uncertainty may influence the relationship between emotional intolerance and how a person cognitively and behaviorally responds to a stressor. Rosen and Knäuper (2009) examined the interplay between intolerance of uncertainty and situational uncertainty in terms of worry and information seeking in a sample of undergraduate students. In this study, they manipulated students’ uncertainty around whether they may have a fictional sexually transmitted infection by providing information sheets that either 1) were designed to increase their uncertainty about whether they had this disease (High Uncertainty) or 2) were designed to reduce their uncertainty about having the disease (Low Uncertainty). The authors also induced participants to be either higher or lower in intolerance of uncertainty using a linguistic manipulation and providing false feedback on their tolerance of uncertainty. The authors found a significant interaction between intolerance of uncertainty and situational uncertainty, such that persons in the high uncertainty condition who were induced to also have high intolerance of uncertainty worried significantly more than those with low intolerance of uncertainty. Although situational uncertainty appears to strengthen the relationship between intolerance of uncertainty and anxious responding, this interaction with other emotional intolerance constructs has not been established.

AS is thought to involve both the misappraisal of a sensation as well as uncertainty regarding the consequences of that sensation or change (Carleton et al., 2007). There is some evidence to suggest that AS may affect both cognitive and physiological responses to uncertainty, as AS has been associated with differences in startle responses to unpredictable (Lejuez et al., 2000; Yartz et al., 2008; Zvolensky et al., 2000) but not predictable bodily threat (Nelson et al., 2015). However, no study has explored the effect of situational uncertainty in relation to both AS and DI. It is important to include both lower-order factors of emotional intolerance to examine whether the pattern of relationships remains consistent across multiple aspects of emotional intolerance. As clinicians often rely on different interventions to reduce either AS or DI (e.g., Keough & Schmidt, 2012; Wright et al., 2020), it is important to understand whether both constructs similarly impact anxious responding under uncertain conditions to better inform clinical practice. Additionally, research could benefit from examining these associations within the context of real-life situations, which would increase the ecological validity and salience of the situational uncertainty. The uncertainty surrounding an approaching hurricane is an ideal tangible scenario in which to consider the relationship between AS, DI and varying levels of uncertainty.

Hurricanes: An Example of Situational Uncertainty

Hurricanes and their accompanying storm forecasts represent a naturalistic example of everyday situational uncertainty during hurricane season (May – November). The unexpected and uncertain nature of hurricanes is captured across both storm severity and expected landfall, and can lead to heightened levels of worry and preparatory behaviors as a storm approaches (Meyer et al., 2013; Smith et al., 2009; Villegas et al., 2013). While anxiety and preparatory behaviors are both normative and expected for a storm that has a high likelihood of making landfall, each hurricane season is also marked with many examples of inappropriate anxious responding and maladaptive preparatory behaviors that present public health challenges (Demuth et al., 2016; Wong-Parodi & Feygina, 2018). For example, individuals with worse mental health were more likely to prepare for a hurricane and evacuate, irrespective of whether they lived in an evacuation zone (Wong-Parodi & Feygina, 2018). This over-response may become maladaptive since it often results in increased stress and worry that is not necessarily commensurate with their risk. Further research is needed to better understand the predictors of worry and behavioral responses to impending hurricanes. While emotional intolerance has been shown to predict both worry and behavioral responses in clinical samples (Kertz et al., 2015; Zvolensky et al., 2010), it is unclear whether emotional intolerance may predict heightened worry and preparatory behaviors in relation to a hurricane.

Furthermore, an increasing body of literature points to high variability in responses to viewing hurricane forecasts (e.g., Meyer et al., 2013; National Research Council et al., 2006), which may be partially explained by individual differences across cognitive and emotional domains that may influence responses to hurricane warning messages (e.g., Demuth et al., 2018; Villegas et al., 2013). Some individual difference factors that may affect how people respond to hurricane forecasts and warning messages include level of knowledge and experience with previous hurricanes, the amount of resources (e.g., socio-economic status), one’s social network, and perceptions of risk (Dash & Gladwin, 2007; Villegas et al., 2013). As recent projections indicate that hurricanes will continue to grow in strength and frequency (Bhatia et al., 2019), it is important to identify individual differences in responding to the uncertain threat presented by the storms. Gaining this insight may also have a broader impact on better understanding the associations between uncertain threat, emotional tolerance, and subsequent emotional and behavioral responses.

The primary visual aid for hurricane forecasts provided by the National Hurricane Center is the cone of uncertainty, which is a graphic depicting the path of a tropical storm that includes a center line and an error cone of varying width for where the hurricane is predicted to make landfall. The width of the error cone is determined using forecast errors over the past five years, with a wider cone of uncertainty demonstrating a more uncertain predicted path and a narrower cone demonstrating greater certainty (Cox et al., 2013). As a hurricane approaches, the cone of uncertainty narrows to reflect the more certain path of the upcoming storm. As such, hurricane forecasts present a naturalistic situation within which to consider how the level of uncertainty may influence associations between affective risk factors and subsequent emotional and behavioral responding. To our knowledge, no study has examined the interaction between emotional intolerance and situational uncertainty within the specific context of responding to an upcoming hurricane.

The Current Study

The overarching aim of the current investigation was to test the moderating role of situational uncertainty on the relationship between emotional intolerance and anticipated worry and preparatory behavior, specifically within the context of a hypothetical hurricane forecast. A community sample of participants who had experienced Hurricane Irma in 2017 was assessed. All participants were shown two realistic forecasts for a hypothetical hurricane that differed in the level of situational uncertainty: a 5-day forecast with a large cone of uncertainty, and a 3-day forecast with a narrow cone of uncertainty. Our primary aim was to examine whether the effect of three different emotional intolerance constructs on anticipated emotional and behavioral responses would be moderated by situational uncertainty. We hypothesized that higher levels of emotional intolerance would predict greater anxiety, as indicated by increased anticipatory worry and preparatory behavior, in line with previous research (Intrieri & Newell, 2020; Macatee et al., 2015). We further hypothesized that forecast uncertainty would moderate the relationship between emotional intolerance and worry and preparation, such that emotional intolerance be a stronger predictor of anticipated anxious responding under conditions of high uncertainty (5-day forecast) compared to low uncertainty (3-day forecast), in line with the findings of Rosen and Knäuper (2009).

Method

Participants

Participants (N = 394) were Florida residents who had experienced Hurricane Irma in September 2017. The current study was part of a larger study designed to explore the mental and physical health responses to Hurricane Irma (see Broos et al., 2021 for more details). One of the inclusion criteria for the broader study was having experienced Hurricane Irma in 2017. As the impact of Hurricane Irma was felt across the entire state of Florida, we sampled participants through two sources: South Florida community members and Florida residents recruited through Amazon’s MechanicalTurk (MTurk) system. As individuals throughout Florida have experience viewing and interpreting hurricane forecasts during hurricane season, we decided to include participants from the entire state. Consistent with previous research (Arditte et al., 2016), participants were removed prior to analysis if they failed to complete at least four out of five total validity checks or if they completed the survey in less than 60% of the projected time (n = 36).

The final sample included 358 participants between the ages of 19 and 79 years (M = 36.3, SD = 12.98). Out of this sample, 63 were recruited from the South Florida community and 295 were recruited from MTurk. Approximately half of the participants (56%) identified as female (n = 201). The sample was racially and ethnically diverse, with 73% identifying as Caucasian/White, 9% as African American/Black, 9% as Hispanic/Latino, 5% as Asian American, 3% as multiracial, and 1% as Native American/American Indian. Demographic statistics for each sample, as well as statistical comparisons between them, are displayed in Table 1. MTurk participants tended to be more male and tended to report lower levels of anticipated worry and preparation behavior for the 3-day forecast (the comparison was trending for the 5-day forecast). Given differences between samples, all analyses controlled for sample.

Table 1.

Comparison of Demographic Characteristics and Study Variables Between the Community and Amazon MTurk Samples

	Overall (n = 358)	Community (n = 63)	MTurk (n = 295)	Statistical Test
Demographics
Age	36.3 (12.98)	35.95 (14.32)	36.38 (12.7)	t = −.02
Gender (Female: n, %)	201 (56%)	45 (71%)	156 (53%)	χ2 = 6.52*
Race (n, %)
Caucasian	263 (73%)	41 (65%)	222 (75%)	χ2 = 2.25
African American	34 (9%)	6 (10%)	28 (9%)	χ2 < .01
Latino	32 (9%)	9 (14%)	23 (8%)	χ2 = 1.95
Asian American	17 (5%)	6 (10%)	11 (4%)	χ2 = 2.68
Multiracial	9 (2%)	1 (2%)	8 (3%)	χ2 = .006
Native American	3 (1%)	0 (0%)	3 (1%)	χ2 = .002
Emotional Intolerance Factors
WTP-DI	3.17 (1.96)	3.19 (1.82)	3.17 (1.99)	t = −.09
DII-S	7.14 (6.77)	7.0 (6.2)	7.17 (6.91)	t = −.19
ASI-3S	9.27 (8.91)	7.87 (7.99)	9.59 (9.1)	t = −1.49
Hurricane-Related Outcomes
5-day Worry	4.68 (1.63)	4.98 (1.39)	4.61 (1.68)	t = 1.85+
3-day Worry	5.23 (1.64)	6.02 (1.01)	5.06 (1.7)	t = 5.95**
5-day Preparation (n, %)	285 (79.6%)	56 (88.9%)	229 (77.6%)	χ2 = 3.39+
3-day Preparation (n, %)	315 (87.9%)	62 (98.4%)	253 (85.8%)	χ2 = 6.49*

Note. All values are means and standard deviations unless otherwise noted. Abbreviations: ASI-3S = Anxiety Sensitivity Index-Short Form; DII-S = Distress Intolerance Index-Short Form; WTP-DI = Willingness to Pay-Distress Intolerance.

⁺p < .10.

^*p < .05.

^**p < .01.

Procedure

The larger study was approved by the Institutional Review Board at the University of Miami. The sample was recruited approximately four to eight months after Hurricane Irma, between January 2018 and May 2018, through several methods. Flyers were distributed throughout the South Florida community, which included instructions for how to access the survey online. Online listings were posted to advertising portions of Craigslist, Facebook, and Amazon, as well as on the laboratory website. Participants recruited through these avenues were classified as community participants. Concurrently, online participants were recruited through MTurk, which allowed for a more representative sampling of Florida residents than community sampling alone. Standard procedures were followed for recruiting MTurk samples (e.g., Behrend et al., 2011), and participants’ IP addresses were screened to ensure they were located in Florida.

All participants first agreed to participate in the study and completed an electronic written informed consent document, stored separately from the remaining study materials. Next, participants completed demographic information, both forecast conditions, and questionnaires, which took approximately 20 to 30 minutes. Community participants who completed the survey were compensated with a $15 e-gift card, while MTurk participants were compensated $4 for their time.

Uncertainty Conditions: Hurricane Storm Forecasts

Participants were shown two hypothetical storm forecasts for a fictitious storm “Gabrielle,” identical in style to those released by the National Oceanic and Atmospheric Administration (NOAA) and National Weather Service (see Figure 1). All participants viewed both forecast conditions for a within-subject comparison. Participants were instructed to imagine that the forecast shown was being issued by the National Weather Service for a fictional storm Gabrielle. In an ecologically valid manner, participants first viewed the 5-day forecast, in which the cone of uncertainty was larger and the likelihood the storm would hit Florida was smaller (High Uncertainty condition). They next viewed the 3-day forecast, in which the cone of uncertainty was smaller, and the storm’s path clearly included Florida, indicating a higher likelihood of the storm making landfall (Low Uncertainty condition). The order of the two conditions (more uncertain followed by less uncertain) was designed to simulate the actual experience of viewing forecasts for a real hurricane, and thus the order was not randomized between participants. For each forecast condition, participants were asked several questions to assess their anticipated cognitive and behavioral responses based on that respective forecast.

Figure 1.

Two realistic forecasts for the hypothetical Storm Gabrielle, based on actual National Hurricane Center forecasts from Hurricane Irma in 2017. The panel on the left depicts a 5-day forecast that presents inherently greater levels of uncertainty with respect to the storm path. The panel on the right depicts a 3-day forecast where the cone of the projected storm path has narrowed and become more certain (i.e., less uncertain).

Measures

Distress Intolerance Index-Short Form (DII-S).

The DII-S is a 6-item version of the 10-item original DII (McHugh & Otto, 2012). The DII assesses self-reported distress intolerance, or the extent to which participants perceive themselves to be unable to manage negative emotions (e.g., “I can’t handle feeling distressed or upset”). Responses are recorded on a 5-point Likert scale (0 = Very Little; 4 = Very Much). Given procedural limitations on survey administration length, we reduced the number of items by examining factor loadings from the original DII manuscript, and consulting with the first author of that measure (K. McHugh). See the Supplement for further information on how the DII-S was created. Total scores on the DII-S range from 0 to 24, with higher scores indicating higher distress intolerance. The DII-S was reliable in the current sample (α = .95).

Anxiety Sensitivity Index-3 Short Form (ASI-3S).

The Anxiety Sensitivity Index-3 (ASI-3; Taylor et al., 2007) captures levels of anxiety sensitivity, defined as the intensity of distress one feels when experiencing physical symptoms of anxiety (e.g., “When I notice that my heart is beating rapidly, I worry that I might have a heart attack”). The ASI-3 has three subscales, including physical, cognitive, and social concerns. Responses are recorded on a five-point Likert scale (0 = Very Little; 4 = Very Much). For the current study, given survey length limitations, we reduced the ASI-3 to 9 items by selecting the three items with the highest item-total score correlations from each subscale. The Supplement contains further information on how the ASI-3S was modified. ASI-3S scores range from 0 to 36, with higher scores indicating higher levels of anxiety sensitivity. The ASI-3S was reliable in the current sample (α = .93).

Willingness to Pay – Distress Intolerance (WTP-DI, McHugh et al., 2011).

The WTP-DI is a single-item measure that captures domain-specific DI. For the current study we modified the traditional WTP-DI to assess how much an individual would pay to avoid the distress they felt leading up to and during Hurricane Irma. Responses were recorded on a seven-point scale (0 = 0% of My Monthly Income; 7 = More Than 15% of My Monthly Income). Higher scores indicate higher levels of intolerance of hurricane-specific distress. The concurrent and discriminant validity of this measure has been established in previous studies (McHugh et al., 2011), and the WTP-DI is significantly correlated with other measures of distress intolerance, including the Distress Tolerance Scale and the Frustration Discomfort Scale.

Previous Hurricane Experience.

Previous hurricane experience was measured with one item that asked participants if they had personally experienced a hurricane prior to the 2017 season. Responses were coded as “Yes” or “No”.

Anticipated Worry.

For each forecast condition, participants were asked how much worry they would experience if the hypothetical storm was real (“If Gabrielle were a real storm, how worried would you be about its threat?”). Participants responded on 7-point Likert scale (1 = Not at All Worried; 7 = Extremely Worried). Previous research supports the validity of a single item measure of worry and anxiety (Schroder et al., 2019; Young et al., 2015).

Anticipated Preparatory Behavior.

For each forecast condition, participants were asked if they would take any action to prepare for the hurricane based on the forecast (“If Gabrielle were a real storm, would you begin taking actions to prepare for it now?”). Participants responded with a binary response, indicating either yes or no (1 = Yes, 0 = No). If participants responded Yes, they were presented with several different options of preparatory behaviors and were asked to check all the actions they would anticipate doing to prepare for this fictional storm. These behavioral options were based on the hurricane preparation guidelines issued by the state of Florida, and some examples included: Fill up car with gas; Buy extra water, food, and batteries; Evacuate my home; Put up storm shutters. As these behavior options did not hang together and sometimes even contradicted other options (e.g., choosing to evacuate would probably make choosing the other behaviors less likely), a total score of the behavior items was not appropriate. Thus, we decided to use the binary response (Yes Preparation or No Preparation) as our outcome variable.

Analysis

We tested a series of hierarchical linear and logistic regression models using R (version 3.6.3). Prior to assessing effects of DII-S, ASI-3S, and WTP-DI on anticipated worry and preparation, we tested unconditional models to determine intraclass correlation coefficients (ICCs) for each outcome of interest. Next, we tested conditional models for each outcome, incorporating the effect of each of the three emotional tolerance variables (i.e., DII-S, ASI-3S, and WTP-DI) in interaction with condition (i.e., 3-day or 5-day forecast) to predict the outcome of interest. Given that participants were recruited from two different data sources, we controlled for the effect of sample (MTurk or community participant) in all conditional models. In addition, in light of the influence of past exposure to hurricanes on individual responses and attitudes (e.g., Demuth et al., 2016; Lindell, 2013), we controlled for previous hurricane experience in all models.

An example of the conditional models (here, using ASI-3S as the predictor and anticipated worry as the outcome) is depicted in Equations 1–4 (Equation 1: Level 1 model; Equations 2–3: Level 2 models; Equation 4: reduced form model). In the conditional models, the predicted anticipated worry for a given individual at a given time point ($Y_{ti}$) is a function of the intercept or expected value of anticipated worry for the 5-day forecast (${\beta }_{0i}$), the effect of condition on anticipated worry for a given individual (${\beta }_{1i}$), and unexplained variance in anticipated worry for a given individual at a given time point ($e_{ti}$; Equation 1).

$$Y_{ti}={\beta }_{0i}+{\beta }_{1i}{Condition}_{ti}+e_{ti}$$ (Equation 1)

${\beta }_{0i}$ is a function of the overall intercept (expected value of anticipated worry symptoms when all predictors are at zero; ${\gamma }_{00}$), the fixed effect of ASI-3S on anticipated worry (${\gamma }_{01}$), the fixed effect of sample on anticipated worry (${\gamma }_{02}$), the fixed effect of past hurricane experience on anticipated worry (${\gamma }_{03}$), and the random effect of individual ($u_{0i}$; Equation 2).

$${\beta }_{0i}={\gamma }_{00}+{\gamma }_{01}{ASI}_i+{\gamma }_{02}{Sample}_i+{\gamma }_{03}{PastHurricane}_i+u_{0i}$$ (Equation 2)

${\beta }_{1i}$ is a function of the fixed effect of condition on anticipated worry (${\gamma }_{10}$) and the fixed effect of the interaction between ASI-3S and condition on anticipated worry (${\gamma }_{11}$; Equation 3).

$${\beta }_{1i}={\gamma }_{10}+{\gamma }_{11}{ASI}_i$$ (Equation 3)

Equation 4 contains the reduced-form model.

$$\begin{gathered} Y_{ti}=\left({\gamma }_{00}+{\gamma }_{01}{ASI}_i+{\gamma }_{02}{Sample}_i+{\gamma }_{03}{PastHurricane}_i+u_{0i}\right)+\left({\gamma }_{10}+{\gamma }_{11}{ASI}_i\right){Condition}_{ti}+e_{ti} \\ {=\gamma }_{00}+{\gamma }_{01}{ASI}_i+{\gamma }_{02}{Sample}_i+{\gamma }_{03}{PastHurricane}_i+{\gamma }_{10}{Condition}_{ti}+{\gamma }_{11}\left({ASI}_i{\mathrm{x}Condition}_{ti}\right)+u_{0i}+e_{ti} \end{gathered}$$ (Equation 4)

Given that the anticipated preparation variable was measured as a binary yes or no, the effects of emotional tolerance and condition on anticipated preparation were modeled using mixed effects logistic regression, and we reported odds ratios (ORs).

Results

Preliminary Analyses

See Table 2 for correlations among study variables. All three emotional intolerance variables were significantly correlated with anticipated worry for the 5-day forecast, but only DII-S and WTP-B were correlated with anticipated worry for the 3-day forecast. Only ASI-3S was significantly associated with anticipated preparation for the 5-day forecast, and none of the emotional intolerance variables were correlated with anticipated preparation for the 3-day forecast.

Table 2.

Correlations Among Study Variables

	1	2	3	4	5	6	7
1. WTP-DI	--
2. DII-S	.28**	--
3. ASI-3S	.25**	.63**	--
4. Anticipated Worry (5-day)	.29**	.15**	.17**	--
5. Anticipated Worry (3-day)	.19**	.12*	.05	.69**	--
6. Anticipated Preparation (5-day)	.10+	.04	.16**	.52**	.34**	--
7. Anticipated Preparation (3-day)	.06	.02	0.00	.32**	.53**	.46**	--

Note. Abbreviations: ASI-3S = Anxiety Sensitivity Index-Short Form; DII-S = Distress Intolerance Index-Short Form; WTP-DI = Willingness to Pay-Distress Intolerance.

⁺p < .10.

^*p < .05.

^**p < .01.

Unconditional Models

The unconditional models examining the fixed effect of condition indicated significant differences in both outcomes from the 5-day to 3-day forecast (anticipated worry: β = .16, 95% CI [.12, .20], p <.001; anticipated preparation: OR = 3.50, 95% CI [1.86, 6.60], p <.001). For both models, the high intraclass correlation coefficients (ICCs) indicated that a substantial proportion of variability was explained by within-person effects (anticipated worry ICC = .69; anticipated preparation ICC = .72).

Anticipated Worry

We examined three separate models – one for each emotional tolerance variable – with anticipated worry as the outcome. In each model, primary predictors of interest included the emotional tolerance variable, condition, and their interaction. All models were tested controlling for cohort and previous hurricane experience.

DII-S Model.

Participants with higher DII-S reported greater overall anticipated worry (β = .11, 95% CI [.01, .21], p = .027). The fixed effect of condition remained significant after adding in the additional predictor, with anticipated worry increasing from the 5-day to 3-day forecast (β = .17, 95% CI [.13, .22], p < .001). There was not a significant interaction between condition and DII-S (β = −.01, 95% CI [−.05, .03], p = .70).

WTP-DI Model.

Participants with higher WTP-DI reported greater overall anticipated worry (β = .24, 95% CI [.14, .34], p < .001). As with the previous model, the fixed effect of condition remained significant after adding in the additional predictor, with anticipated worry increasing from the 5-day to 3-day forecast (β = .17, 95% CI [.13, .22], p < .001). There was a significant interaction between condition and WTP-DI (β = −.06, 95% CI [−.10, −.01], p = .012), such that the relationship between WTP-DI and worry was stronger in the 5-day forecast compared to the 3-day forecast (see Figure 2A). Follow-up simple slopes analyses indicated that WTP-DI was a stronger predictor of worry in the 5-day forecast (B = .25, SE = .05, p < .001) compared to the 3-day forecast (B = .16, SE = .05, p < .001).

Figure 2.

Emotional intolerance as a function of forecast condition (low versus high uncertainty) on anticipated worry. Panel A depicts the interaction for Willingness to Pay – Distress Tolerance (WTP-DI), whereas panel B depicts the interaction for Anxiety Sensitivity Index 3 – Short (ASI3-S).

ASI-3S Model.

Individuals with higher ASI-3S reported greater overall anticipated worry (β = .11, 95% CI [.01, .21], p = .002). Consistent with the findings for DII-S and WTP-DI, the fixed effect of condition remained significant after adding in the additional predictor, with anticipated worry increasing from the 5-day to 3-day forecast (β = .17, 95% CI [.13, .22], p < .001). Finally, there was a significant interaction between condition and ASI-3S (β = −.05, 95% CI [−.10, −.01], p = .014), such that the relationship between ASI-3S and worry was stronger in the 5-day forecast compared to the 3-day forecast (see Figure 2B). Follow-up simple slopes analyses indicated that ASI-3S was a significant predictor of worry only in the 5-day forecast (B = .03, SE = .01, p = .002), as ASI-3S did not predict worry in the 3-day forecast (B = .01, SE = .01, p = .27).

Covariates.

Across all three models, participants in the MTurk sample generally reported lower overall anticipated worry (DII-S: β = −.42, 95% CI [−.67, −.17], p = .001; WTP-DI: β = −.45, 95% CI [−.70, −.20], p < .001; ASI-3S: β = −.45, 95% CI [−.70, −.20], p < .001). Previous hurricane experience was not significantly associated with anticipated worry in any of the models (DII-S: β = .04, 95% CI [−.21, .28], p = .78; WTP-DI: β = .05, 95% CI [−.18, .28], p = .67; ASI-3S: β = .10, 95% CI [−.14, .35], p = .41).

Anticipated Preparation.

We next examined three separate models – one for each emotional tolerance variable – with anticipated preparation as the outcome. In each model, primary predictors of interest included the emotional tolerance variable, condition, and their interaction. All models were tested controlling for cohort and previous hurricane experience.

DII-S Model.

Participants with higher DII-S did not report overall higher likelihood of anticipated preparatory behaviors (OR = 1.00, 95% CI [.92, 1.09], p = .93). The fixed effect of condition remained significant (OR = 3.74, 95% CI 1.46, 9.58], p = .006). There was not a significant interaction between DII-S and condition (OR = 1.00, 95% CI [.91, 1.09], p = .92).

WTP-DI Model.

There was a significant fixed effect of condition on anticipated preparation (OR = 4.49, 95% CI [1.43, 14.14], p = .01), as well as a trending main effect of WTP-DI being associated with a higher likelihood of anticipated preparatory behaviors (OR = 1.29, 95% CI [.98, 1.71], p = .07). We did not observe a significant interaction between WTP-DI and condition (OR = .88, 95% CI [.64, 1.22], p = .45).

ASI-3S Model.

Participants with higher ASI-3S were more likely to report anticipated preparation (OR = 1.10, 95% CI [1.02, 1.19], p = .01). Consistent with the findings for DII-S and WTP-DI, the fixed effect of condition remained significant after adding in the additional predictor, with anticipated preparation more likely overall in the 3-day than 5-day forecast (OR = 9.03, 95% CI [3.16, 25.81], p < .001). In addition, there was a significant interaction between ASI-3S and condition (OR = .90, 95% CI [.83, .97], p = .009), such that the relationship between ASI-3S and anticipated preparation was stronger in the 5-day compared to the 3-day forecast (see Figure 3). Follow-up simple slopes analyses indicated that ASI-3S significantly predicted anticipated preparation only in the 5-day forecast (B = .10, SE = .04, p = .01), as ASI-3S was not a significant predictor in the 3-day forecast (B = −.01, SE = .04, p = .79).

Figure 3.

Anxiety Sensitivity Index 3 – Short (ASI3-S) as a function of forecast condition (low versus high uncertainty) on anticipated storm preparation.

Covariates.

In all three models, participants in the MTurk sample were less likely to report anticipated preparation (DII-S: OR = .10, 95% CI [.02, .56], p = .008; WTP-DI: OR = .13, 95% CI [.03, .57], p = .007; ASI-3S: OR = .10, 95% CI [.02, .52], p = .006). Anticipated preparation was not significantly associated with previous hurricane experiences in any of the models (DII-S: OR = .49, 95% CI [.13, 1.93], p = .31; WTP-DI: OR = .76, 95% CI [.24, 2.39], p = .64; ASI-3S: OR = 1.00, 95% CI [.28, 3.57], p = .99).

Discussion

The primary objective of the current study was to examine the relation between three different emotional intolerance constructs and anxious responding, defined as anticipated worry and anticipated preparatory behavior. This study also tested the moderating role of situational uncertainty, using a novel experimental paradigm. As hypothesized, all three emotional intolerance constructs were linked with higher reported anticipated worry overall, but only anxiety sensitivity (AS) predicted higher likelihood of anticipated preparation. Importantly, forecast uncertainty moderated the relationship between emotional intolerance and anticipated worry for both AS and hurricane-specific distress intolerance (DI), such that this relationship was stronger in the 5-day forecast (high uncertainty) condition. Forecast uncertainty did not moderate the relationship between DI and anticipated worry. The results for behavioral preparation were more mixed, and, contrary to our hypotheses, forecast uncertainty only moderated the relationships between AS and anticipated preparatory behavior. Together, these findings suggest that a person’s level of emotional intolerance may predict their cognitive and behavioral responses to uncertainty in a hurricane context, and that distinct emotional tolerance constructs influence these responses.

As expected, higher levels of emotional intolerance predicted higher levels of anticipated worry. These data are consistent with previous research that has linked emotional intolerance with higher overall worry (e.g., Floyd et al., 2005; Kertz et al., 2015). Although the effect of emotional intolerance on worry was consistent across models, the effect on behavioral preparations was mixed. Of the three emotional intolerance constructs, only AS predicted increased likelihood of preparation (hurricane-specific DI had a trending significant effect). This finding is in line with past research on the relationships between behavioral responses, situational uncertainty, and emotional intolerance, which collectively suggest that higher levels of uncertainty and emotional intolerance would predict behavioral change to alleviate those stressors (Reuman et al., 2015; Rosen & Knäuper, 2009). Surprisingly, DI did not predict increased likelihood of preparation and the effect of hurricane-specific DI was only trending, contradicting previous research on how individuals with higher distress intolerance may engage in maladaptive behavioral responses to reduce distress (Keough et al., 2010). In contrast to the extant literature, our findings suggest that preparatory behavior was largely not associated with either the individual level of emotional intolerance or the level of situational uncertainty. The discrepancy in our findings between worry and preparation behaviors may point to ways in which emotional intolerance uniquely contributes to cognitive mechanisms and maladaptive behavioral coping strategies across anxiety disorders (Carleton, 2012; Grupe & Nitschke, 2013). Further research is necessary to elucidate the specific role of emotional intolerance in contributing to the behavioral components of various anxiety disorders.

Importantly, our results indicated that the effect of AS and hurricane-specific DI on anticipated worry depended on level of uncertainty: both AS and hurricane-specific DI had a stronger effect on worry in the high uncertainty condition (5-day forecast). Individuals higher in AS and hurricane-specific DI also exhibited less of a change in worry as the hurricane approached (i.e., at the 3-day forecast), suggesting that high levels of emotional intolerance may prevent the appropriate modulation of worry as the level of risk changes. These data are consistent with previous research that has linked emotional intolerance with higher overall worry (e.g., Floyd et al., 2005; Kertz et al., 2015), and suggest that the effect of emotional intolerance on worry may be particularly salient under high uncertainty conditions. Extending this literature, the current study is the first to examine the interplay of situational certainty and other aspects of emotional tolerance, including AS and DI, in predicting worry and behavioral responses. We also found that uncertainty moderated the relationship between AS and anticipated preparation in the same direction, suggesting that this moderating effect of uncertainty may influence both cognitive and behavioral responses. These findings are consistent with previous research examining the interplay of emotional intolerance and situational uncertainty on physiological responses (Nelson et al., 2015; Rogers et al., 2019). Additionally, recent theories have suggested that a core feature of anxiety disorders may be excessive anticipatory responding to unpredictable and uncertain situations (Grupe & Nitschke, 2013). Our results are in line with these theories and suggest that persons higher in emotional intolerance may be experiencing an increase in worry when faced with uncertain situations.

Situational uncertainty moderated the effects of both AS and hurricane-specific DI on worry but did not moderate the relationship with DI, highlighting that despite their interrelatedness (Mitchell et al., 2013), these constructs relate distinctly to affective responding under uncertain conditions. In prior literature, both AS and DI have been linked with maladaptive stress responding (Zvolensky et al., 2010). Although forecast uncertainty moderated the relationship between both AS and hurricane-specific DI and anticipated worry, it did not moderate the relationship between the additional measure of distress intolerance (i.e., the Distress Intolerance Index) and worry. Our findings indicate that general DI predicts anticipated worry regardless of the level of uncertainty, while the effect of AS and specific DI depends on uncertainty. In fact, AS only significantly predicted anticipated worry in the uncertain condition, signifying that AS may be particularly important under conditions of uncertainty. This finding suggests that it may be the interpretation of anxiety-related sensations as aversive (Keough et al., 2010), more so than a perceived inability to withstand negative emotions or physical states (Brown et al., 2005; Stamatis et al., 2020), that predicts a tendency to worry about an uncertain future event. Future research is needed to explore the differential relationship of AS and general DI on anxious responding under conditions of uncertainty.

The dissociation of results using DI and hurricane-specific DI highlights the importance of measurement in capturing distress intolerance (McHugh et al., 2011; Stamatis et al., 2020). Specifically, whereas forecast uncertainty did not interact with distress intolerance, measured by the DII-S, it did interact with the WTP-DI—a more behaviorally-focused measure of distress intolerance that is hurricane-specific. This finding aligns with previous literature indicating that the WTP-DI correlates only modestly with other self-report measures of distress intolerance (McHugh et al., 2011). Notably, the DII was developed using items from three distinct emotional tolerance measures (McHugh & Otto, 2012), including the Distress Tolerance Scale (Simons & Gaher, 2005), the Frustration-Discomfort Scale (Harrington, 2005), and the Anxiety Sensitivity Index (Reiss et al., 1986). In contrast, the WTP-DI represents an arguably cleaner measure of distress intolerance, in that it involves hypothetical behavioral responses the participant would take to avoid an aversive situation (McHugh et al., 2011). Further, the WTP-DI carries the benefit of adaptability to a specific scenario: for our purposes, the WTP-DI asked specifically about the amount individuals would be willing to pay to avoid distress associated with their previous experience with Hurricane Irma. Consequently, it is perhaps unsurprising that this hurricane-specific measure of distress intolerance was more predictive of storm-related worry in the context of uncertainty. In future studies, it will be important to extend this paradigm using behavioral assessments of DI, such as the cold pressor task (Zvolensky et al., 2001), along with other self-report measures.

While previous studies have manipulated uncertainty in the laboratory, our study introduced a new experimental paradigm that attempted to simulate exposure to a real-world stressor in a naturalistic way. Rosen and Knäuper (2009) examined the moderating role of situational uncertainty on intolerance of uncertainty by manipulating the amount of information available on a fictional infection for which the participants may have tested positive. A neuroimaging study manipulated uncertainty by presenting uncertain or certain cues preceding neutral or unpleasant pictures (Sarinopoulos et al., 2010). However, the tasks used in these uncertainty manipulations may not be representative of how individuals experience uncertainty in daily life. Our paradigm recreated experiences representative of how people experience hurricane-related situational uncertainty. By presenting participants with a realistic simulation of a hurricane forecast, which is a familiar image of uncertainty for Floridians during the annual hurricane season, we aimed to capture cognitive and behavioral responses to uncertainty in a more naturalistic manner than other experimental paradigms. Further, the behavioral response options used in this design were based on the hurricane preparation guidelines issued by the state of Florida, adding to the paradigm’s real-world validity.

Limitations

There were several limitations of the current study. Although we measured three aspects of emotional tolerance, we did not measure intolerance of uncertainty, another emotional intolerance construct thought to be relevant to experiences of uncertainty. As intolerance of uncertainty is considered to be a distinct component of emotional intolerance (e.g., Bardeen et al., 2013; Norr et al., 2013) and may therefore demonstrate a different relationship with situational uncertainty, future studies should also include assessments of intolerance of uncertainty to determine the specificity of these relationships. Additionally, although participants had to report that they experienced Hurricane Irma to be eligible for this study, we were unable to confirm the level of exposure to Hurricane Irma for each participant. Another limitation is the unknown validity of the forecast paradigm used. Though we created the forecast paradigm to appear identical to the forecasts used by the National Hurricane Center, we do not know how our manipulation relates to the experience of viewing a hurricane forecast for a real storm. More research is needed on the validity and utility of using this paradigm to manipulate uncertainty. Related to the forecast paradigm, the lack of randomization for the order of the forecasts is another limitation. While the order of the forecasts (5-day then 3-day) was ecologically valid and designed to simulate the real experience of viewing hurricane forecasts as a storm approaches, it is possible that order effects may have biased the results in our study. For example, individuals who perceive uncertainty to be more threatening may experience heightened arousal following the 5-day forecast (High Uncertainty), which may have biased their reporting in the 3-day forecast (Low Uncertainty).

Another set of limitations of the current study concerns the measurement of anticipated worry and preparatory behavior. Participants were asked to rate their anticipated worry and preparatory behavior; however, we did not assess their actual responses to a hurricane. Previous studies have identified biases in affective forecasting in both healthy (e.g., Wilson & Gilbert, 2005) and clinical samples (e.g., Thompson et al., 2017). Given that these biases may be magnified in those with higher emotional intolerance and internalizing symptoms (e.g., Mathersul & Ruscio, 2020), our outcome variables may have been impacted by forecasting bias. Another limitation was the reliance on a single-item assessment of anticipated worry, introducing potential measurement issues for this construct (McIver & Carmines, 1981). Although previous research supports the use of a single-item measure of worry (e.g., Schroder et al., 2019), future research should replicate our findings using full-length validated measures of worry. A final limitation is the binary measurement of preparatory behavior used in our study. As binary outcomes are less sensitive to changes (e.g., Schmitz et al., 2012), it is possible that using a binary outcome for anticipated preparation impacted our overall findings for this outcome. Further research is needed to explore the influence of situational uncertainty using continuous measures of behavioral responding.

Conclusions

The current study explored the moderating role of situational uncertainty on the relationship between emotional intolerance and anticipated worry and preparatory behavior in response to hypothetical hurricane forecasts. Emotional intolerance appears to be a stronger predictor of heightened worry under conditions of high uncertainty, suggesting that individuals high in emotional intolerance may experience inappropriate increases in worry in response to uncertainty. People with higher anxiety sensitivity may tend to report higher levels of worry and behavioral responses when they are in situations that are uncertain. Future research should explore whether increased reactivity to uncertainty is one mechanism linking emotional intolerance with internalizing psychopathology. Future research should also explore whether these findings generalize to other uncertain and anxiety-provoking situations. The current findings suggest that situational uncertainty may exacerbate the effect of emotional intolerance on anxious responding to the situation, such as increased worry or over-preparation for the situation. As the hurricane season is a period of fluctuating uncertainty, individuals high in emotional intolerance may be at higher risk for increased anxiety during this time.

Data Availability

The data that support the findings of this study are available from the corresponding author, KT, upon reasonable request.