Intolerance of uncertainty, anxiety sensitivity, and health anxiety during the COVID-19 pandemic: Exploring temporal relationships using cross-lag analysis

Abstract

Intolerance of uncertainty (IU) and anxiety sensitivity (AS) have been widely discussed and explored as factors that may contribute to health anxiety. We propose that IU and AS are salient issues for many during the COVID-19 pandemic, and may play a role in the development or exacerbation of health anxiety during the pandemic. Studies have examined links between IU and AS with health anxiety during the pandemic, but these relationships have not been tested together using a longitudinal study design. In the present study, measures of IU, AS, and health anxiety were collected from 301 adults at two time points 6 months apart during (early stages of) the COVID-19 pandemic using an online survey platform. Cross-lagged analysis was utilized to simultaneously estimate cross-sectional and longitudinal associations between these three variables. Robust cross-sectional associations were observed, and IU prospectively predicted changes in both health anxiety and AS. No other statistically significant prospective associations emerged. Present findings support the putative role of IU in health anxiety, suggesting that some observed links between AS and health anxiety could be driven by shared variance with IU. IU may be an important factor to monitor and target in health anxiety interventions during the pandemic.

1. Introduction

Health or illness anxiety is a common problem, affecting up to 6 % of the general population during the course of their lives, and up to 20 % of medical outpatients (Tyrer, 2018). Health anxiety is characterized by excessive worries about health and fears of developing a disease, resulting in hypervigilance for signs of illness and maladaptive information seeking and safety behaviors. The impact of health anxiety on well-being and service utilization can result in substantial healthcare costs – for example, a naturalistic longitudinal study by Fink, Ørnbøl, and Christensen (2010) found that individuals with severe health anxiety used 41–78 % more healthcare per year.

There has been extensive research exploring cognitive and behavioral factors theorized to be associated with risk for excessive health anxiety (see Asmundson, Abramowitz, Richter, & Whedon, 2010; Asmundson & Fergus, 2019; Salkovskis & Warwick, 2001; Taylor, 2004). Two factors that have been widely discussed, examined, and shown to be linked with health anxiety in the literature are intolerance of uncertainty and anxiety sensitivity (e.g., Fergus & Bardeen, 2013; Fetzner et al., 2014; Gerolimatos & Edelstein, 2012; Horenstein, Rogers, Bakhshaie, Zvolensky, & Heimberg, 2019; Norr, Albanese, Oglesby, Allan, & Schmidt, 2015; Wright, Lebell, & Carleton, 2016). Although definitions of intolerance of uncertainty (IU) have changed over time, recent and influential work defines IU as “an individual's dispositional incapacity to endure the aversive response triggered by the perceived absence of salient, key, or sufficient information, and sustained by the associated perception of uncertainty” (p.31, Carleton, 2016). IU may contribute to distress in responses to uncertainty triggered by the experience of potential risks (e.g., exposures) and symptoms of disease or illness, which in turn can lead to problematic ways of coping with feelings of uncertainty with known links to health anxiety, such as information seeking or avoiding medical appointments/tests (e.g., Baerg & Bruchmann, 2022; Bottesi, Marino, Vieno, Ghisi, & Spada, 2021). Anxiety sensitivity (AS) is a trait characterized by aversive emotional reactions to feelings or symptoms of anxiety. AS may contribute to the tendencies to be hyper-vigilant to bodily sensations and experience distress when such sensations occur, which are established aspects of the health anxiety cycle (e.g., see Warwick & Salkovskis, 1990; Abramowitz , Olatunji, & Deacon, 2007).

We propose that both IU and AS are particularly salient during the COVID-19 pandemic, which may contribute to increased health anxiety during the pandemic (see also Freeston et al., 2022 and Funkhouser, Klemballa, & Shankman, 2022). The pandemic has involved a great deal of uncertainty for all, including uncertainty about how the pandemic will evolve and the risks associated with different circumstances and behaviors, which may have led to substantial distress and maladaptive coping in individuals with high levels of IU (for supporting evidence about specific maladaptive coping behaviors, such as information seeking and reassurance seeking, see Bottesi et al., 2021, Jagtap, Shamblaw, Rumas, & Best, 2021, and Sauer, Jungmann, & Witthöft, 2020; see also Taha et al., 2014 for similar findings from the H1N1 pandemic). Likewise, increased attention to bodily sensations as potential signs of illness has been reinforced (e.g., by daily symptom monitoring/reporting requirements), which may have caused even greater distress amongst individuals with elevated AS (see Asmundson & Taylor, 2020). Consistent with our hypotheses, IU has been linked with a range of problems and symptoms in studies conducted during the pandemic (Akbari et al., 2021, Bakioğlu et al., 2021, Beck and Daniels, 2022, Bottesi et al., 2021, Del-Valle et al., 2022, Di Blasi et al., 2021, Fitzgerald et al., 2022, McCarty et al., 2022, Mertens et al., 2020, Parlapani et al., 2020, Rettie and Daniels, 2021, Sauer et al., 2020, Sohrabzadeh-Fard et al., 2021; Saulnier et al., 2022; Smith, Twohy, & Smith, 2020; Tull et al., 2020; Valle et al., 2020; Wheaton, Messner, & Marks, 2021), including health anxiety (Bottesi et al., 2021, Rettie and Daniels, 2021, Sohrabzadeh-Fard et al., 2021, Tull et al., 2020) and fear of COVID (Baerg & Bruchmann, 2022; Bakioğlu et al., 2021; Çelik et al., 2022; Mertens et al., 2020; Satici, Saricali, Satici, & Griffiths, 2020). Likewise, some recent studies have reported links between AS and problems/symptoms during the pandemic (Manning et al., 2021, Morriss et al., 2021, Rogers et al., 2021, Saulnier et al., 2022; Warren et al., 2021), including COVID fear/anxiety (Çelik et al., 2022; Moghadam, Choukami, & Mousavi, 2021; Rogers et al., 2021; Saulnier et al., 2022; Shabani et al., 2022; Warren et al., 2021). Of note, Saulnier and colleagues (2022) also examined and found support for an interactive effect of IU and AS predicting COVID-related worries and safety/avoidance coping behaviors.

Importantly, only a few previous studies during the pandemic have examined the relationships between IU, AS, and/or health anxiety prospectively (Di Blasi et al., 2021, Paluszek et al., 2021, Tull et al., 2020), and no studies have tested these relationships simultaneously. Notably, a recent study published by Paluszek and colleagues (2021) found that AS (as well as disgust sensitivity) predicted the “COVID stress syndrome” (including contamination fears), yet only IU was tested as a hypothesized moderator of those relationships and a significant moderating effect of IU was not observed in that study. Moreover, the present study builds upon previous work by Tull and colleagues (2020), which examined prospective associations between IU and health anxiety, but not AS, in a sample of adults in the United States across a one-month period. Testing prospective associations between IU and AS at the same time in prospective studies would be valuable to confirm if one or both variables can predict changes in symptoms of health anxiety over time, at the same time accounting for cross-sectional links (which are well established), temporal stability, and shared variance.

To this end, the present study examined cross-sectional and prospective associations between IU, AS, and health anxiety using cross-lagged panel analysis in an international sample. Specifically, we analyzed online survey data collected from adults at two assessment time points over 6 months during the early stages of the COVID-19 pandemic. Examining changes in IU, AS, and health anxiety over a 6-month period may allow for greater understanding of the interplay between these variables over time, building upon previous cross-sectional and prospective work published during the COVID-19 pandemic. Based on the extant literature, we hypothesized that both IU and AS would both independently predict later levels of health anxiety, after accounting for initial levels of health anxiety and the cross-sectional relationships between these variables.

2. Materials & methods

2.1. Participants

The initial sample consisted of 308 participants who completed both initial and follow-up assessments as part of a larger study examining the relationship between symptoms of anxiety and distress, putative risk factors, and coping/safety behaviors during the early stages of the COVID-19 pandemic (for more information, see Church, Bounoua, Rodriguez, Bredemeier, & Spielberg, 2022). Participants were recruited from countries wherein governments recommended COVID-related guidelines similar to those recommended in the US, including Canada, the European Union, and Australia. Due to missing data on all key study variables, seven participants (2.2 %) were removed from the sample. The final sample consisted of 301 adults (42.9 % female, ages 18–65, mean age = 30.9). Most participants were from European countries (89.3 %), and the remaining participants were in the US (6.7 %) or other countries (4.0 %). The only study eligibility requirements were: 1) fluency in English; and 2) residence in a country wherein the government had recommended COVID-related guidelines similar to those recommended in the US, including Canada, the European Union, and Australia. Approximately 3.5 % of the participant sample reported having contracted COVID-19 at the time of data collection.

2.2. Measures

The Intolerance of Uncertainty Scale-short version (IUS-12; Carleton, Norton, & Asmundson, 2007) was used to measure responses to uncertainty, the future and ambiguous situations. The IUS-12 is composed of 12 items and has two factors; ‘Prospective IU’ (7 items; e.g., “I always want to know what the future has in store for me”) and ‘Inhibitory IU’ (5 items; e.g., “When I am uncertain I can’t go forward”) (Carleton et al., 2007, McEvoy and Mahoney, 2011). Participants rate each item using a 5-point Likert scale ranging from 1 (“not at all characteristic of me”) to 5 (“entirely characteristic of me”). As demonstrated by the total and subscale scores, the IUS-12 has good internal consistency as well as convergent and divergent validity (Carleton et al., 2007, McEvoy and Mahoney, 2011). Because the two factors are highly correlated (rs >0.62; see Hale et al., 2016 for evidence supporting a bifactor or hierarchal factor structure for the IUS) and we did not have unique hypotheses about them, total scale scores were used in analysis for the current study. Internal consistencies (Cronbach’s alpha) in the current sample were good (time 1: α = 0.87; time 2: α = 0.89).

The Anxiety Sensitivity Index-3 (ASI-3; Taylor et al., 2007) is an 18-item self-report questionnaire that measures anxiety sensitivity, or the reflection of one’s tendency to view anxiety-related sensations as aversive. There are three six-item subscales: physical (“It scares me when my heart beats rapidly”), cognitive (“It scares me when I am unable to keep my mind on a task”), and social concerns (“I worry that other people will notice my anxiety”). Items are rated using a five-point Likert scale and range from 0 (very little) to 4 (very much). Like the IUS-12, the ASI-3 subscale scores were strongly correlated (rs >0.48), consistent with the observed hierarchal factor structure of the ASI-3 (consisting of a single “higher order” factor; Rodriguez, Bruce, Pagano, Spencer, & Keller, 2004). Thus, total scale scores from the ASI-3 were used to test a priori hypotheses described above. Evidence supports the internal consistency and factorial validity of the ASI-3 scale, as well as the convergent and discriminant validity of scores (Jardin et al., 2018, Kemper et al., 2012, Taylor et al., 2007). Internal consistencies (Cronbach’s alpha) in the current sample were excellent (time 1: α = 0.90; time 2: α = 0.91).

The Short Health Anxiety Inventory (SHAI; Salkovskis, Rimes, Warwick, & Clark, 2002) is an 18-item short version questionnaire derived from the original 64-item Health Anxiety Inventory (HAI; Salkovskis et al., 2002) to assess symptoms over the past 6 months. Example items include “I am often afraid that I have a serious illness” and “I frequently have images of myself being ill”. Each item of the SHAI has four response options that allow individuals to choose what best reflects their feelings over the past 6 months. When creating the SHAI, Salkovskis and colleagues used 14 items from the full HAI with the highest item-total correlations from a sample of patients with hypochondriasis. The remaining 4-items of the SHAI were developed to measure theoretical consequences of having a serious illness. In the present study, the 14 items related to fears of becoming ill and the 4-item negative consequences (of illness) subscale were summed to create total SHAI scores. Research has shown that the SHAI possesses internal consistency, as well as convergent, divergent, factorial and criterion validity that is supported by existing evidence (Abramowitz et al., 2007, Alberts et al., 2013). Internal consistencies of the 18-item SHAI (Cronbach’s alpha) in the current sample were excellent (time 1: α = 0.90; time 2: α = 0.90).

To explore whether predictors of changes in health anxiety might be accounted for by coinciding changes in general anxiety, we measured changes in self-report tendencies to worry using the Penn State Worry Questionnaire (PSWQ; Meyer, Miller, Metzger, & Borkovec, 1990). This scale has strong evidence for reliability and validity, and is widely used in basic and applied research on etiology and treatment of generalized anxiety (Brown et al., 1992, Meyer et al., 1990, Startup and Erickson, 2006). Internal consistencies of the PSWQ in the current sample were excellent (time 1: α = 0.94; time 2: α = 0.95).

2.3. Procedures

Participants were recruited using an online crowdsourcing platform (Prolific Academic, https://www.prolific.co) in June of 2020 amidst the COVID-19 pandemic. Previous work has indicated that Prolific Academic participants are more diverse than other crowdsourcing platforms, such as Amazon’s Mechanical Turk, therefore increasing the likely representativeness of Prolific samples (Newman et al., 2021, Peer et al., 2017). Follow-up data were collected after a 6-month interval in December 2020. A total of 520 participants completed the 1st wave of data collection and were invited for the 2nd wave. Given anticipated attrition as well as potential variance in the quality of data, a larger participant sample was recruited for the time 1 collection. Additionally, fewer participants were recruited to complete data collection at time 2 due to the study constraints and the increased statistical power of repeated measures designs to test prospective relationships. For the 2nd wave, data was collected from the first 300 individuals who consented to participate. However, due to a technical error, data was collected from 301 individuals. We conducted follow-up independent samples t-tests to examine differences at baseline in key study variables. Results revealed no significant differences between those who completed the 2nd data collection and those who did not on the ASI (p = .246), IUS (p = .743), and SHAI (p = .092). Mean study completion time was approximately 41 min. Thus, to address concerns about the quality of data validity, individuals who completed data collection in less than 20 min (i.e., less than half the mean completion time) at Time 1 (n = 27) were excluded from invitation to participate in the follow-up assessment. Participants provided written consent, obtained through Prolific Academic. All procedures were approved by the University of Delaware Institutional Review Board. Study data were collected and managed using REDCap (Research Electronic Data Capture) tools, a secure, web-based software platform, hosted at the University of Delaware (Harris et al., 2009, Harris et al., 2019).

3. Calculation

We explored potential bidirectional effects using cross-lagged panel model analyses. Cross-lagged structural equation modeling was conducted using MPlus v.8 (Muthén & Muthén, 1998) to examine predictive relationships between intolerance of uncertainty (indexed via the IUS-12), anxiety sensitivity (indexed via ASI-3), and health anxiety (indexed via SHAI). Importantly, we planned to examine all three variables within the same cross-lag model in order to estimate cross-sectional and prospective relationships between each set of variables, simultaneously taking into account shared variance among the measures. As noted above, total scale scores were examined from all three measures because our a priori research questions and hypotheses related to the overall constructs rather than facets of these constructs (based on some previous studies of IU and AS during the pandemic which have not shown consistent findings with subscales of the measures; e.g., Tull et al., 2020; Warren et al., 2021). Thus, testing all subscales of the measures may reduce statistical power to address our research questions (because of lower reliability and degrees of freedom). Within the cross-lagged model, autoregressive paths were included, and the variables were allowed to covary within each time point. Of note, two participants were missing SHAI data at Time 2. In order to include all participants with at least partial data, missing data were estimated using maximum likelihood estimator with robust standard errors (MLR). Standardized results are reported below. Follow-up (sensitivity) hierarchal regression analyses were conducted including changes in PSWQ scores as a predictor in step 1, in order to explore whether any observed prospective associations might be driven by changes in generalized anxiety during the early phases of the pandemic. To examine effect sizes, standardized betas may be interpreted similar to correlation coefficients wherein small = 0.1, medium = 0.3 and large = 0.5 (Cohen, 1992).

4. Results

Descriptive statistics for the three primary measures at times 1 and 2 are presented in Table 1, and distributions for IU and AS scores (histograms) are presented in Fig. 1, Fig. 2. Notably, 30 participants (10 %) scored greater than or equal to the proposed clinical cutoff of 27 on the SHAI (see Abramowitz et al., 2007) at Time 1, and 33 participants (11 %) scored at or above the clinical cutoff at time 2, suggested elevated levels of health anxiety were not uncommon in our study sample.

Table 1.

Sample means (and SDs) for total scores from the three study measures at initial administration (T1) and 6-month follow-up (T2).

	T1	T2
IUS-12	34.53 (8.89)	34.18 (9.36)
ASI-3	22.05 (12.51)	21.99 (12.96)
SHAI	15.51 (7.96)	15.72 (8.37)
PSWQ	50.26 (14.91)	50.36 (15.52)

Fig. 1.

Histograms illustrating distributions of: (top panel A) Intolerance of Uncertainty Scale (IUS) total scores at Times 1 (baseline) and 2 (6 month follow-up); and (bottom panel B) Anxiety Sensitivity Index (ASI) total scores at Times 1 and 2.

Fig. 2.

Cross-lagged associations between intolerance of uncertainty, anxiety sensitivity, and health anxiety across six months of the COVID-19 pandemic Note: N = 301. Standardized beta coefficients depicted. Bolded lines represent significant cross-lagged associations of interest, * p < .05. Lines on the left (with corresponding beta coefficient values) represent cross-sectional associations between the three variables at Time 1 (baseline), lines on the right represent correlated residuals between the variables at Time 2 (6 months follow-up), and lines in the middle represent prospective (cross-lagged, from Time 1 to Time 2) associations between variables.

Results from the cross-lagged analysis are presented in Fig. 1. As anticipated based on prior research, all cross-sectional relationships between key study variables were statistically significant at both time points (ps <.01). When examining prospective associations, we found a statistically significant positive association between IU at time 1 and health anxiety at time 2 (β =[ 0.13 p <[ 0.01. Further, we found a statistically significant positive association between IU at time 1 and AS at time 2 (β = .17, p <[ 0.01. In contrast, we found no statistically significant associated between time 1 health anxiety and time 2 IU or AS (ps >.30), nor was time 1 AS statistically significantly associated with time 2 IU or health anxiety (ps >;0.10. All cross-sectional associations were in the medium to large range (βs ==0.30–0.63, and the statistically significant prospective associations were small effects. Scatterplot illustrating the associations between IU at time 1 and health anxiety at time 2 (panel A) and AS (panel B) are shown in Fig. 3.

Fig. 3.

Scatterplot of the associations between (top panel A) Intolerance of Uncertainty Scale (IUS) scores at Time 1 with Short Health Anxiety Inventory (SHAI) scores at Time 2 (6 month follow-up); and (bottom panel B) IUS scores at Time 1 and Anxiety Sensitivity Index (ASI) scores at Time 2.

Interactions between IU, health anxiety, and AS were not examined in the model because moderation was not part of our research questions/hypotheses. Instead, a post hoc (hierarchal regression) analysis was conducted to explore the interaction between IU and AS predicting health anxiety at follow-up (beyond variance accounted for by baseline health anxiety), based on some recently published findings (e.g., Paluszek et al., 2021; Saulnier et al., 2022). Results from post-hoc analysis did not support an interactive effect of key study variables (β <[ 0.01 p >[ 0.90. In the present regression model, the results for the main effects of IU and AS were very consistent with those from the primary (cross-lag) analysis described above (IU: β = .13, p <[ 0.01 AS: β = .07, p =[ 0.15. Likewise, two sensitivity analyses were conducted using hierarchal regression modeling to predict time 2 health anxiety and time 2 AS with changes in PSWQ scores entered in step 1 (in addition to time 1 health anxiety and AS). In these models, IU scores from time 1 remained a statistically significant predictor of time 2 health anxiety; β = .14, p <[ 0.01 and also time 2 AS; β = .18, p <[ 0.01

5. Discussion

The present study explored cross-sectional and longitudinal associations between intolerance of uncertainty (IU), anxiety sensitivity (AS), and health anxiety reported at two time points over 6 months during the early stages of the COVID-19 pandemic. We found that initial levels of IU predicted both health anxiety and AS at follow-up, after accounting for the shared variance among the variables. In contrast, AS and health anxiety did not prospectively predict each other or IU at follow-up when all relationships between key study variables were tested simultaneously. Although the observed prospective effects were small, present study findings provide novel insights into the relationships between IU, health anxiety, and AS in the context of an ongoing global pandemic.

The prospective association between IU and health anxiety is consistent with previous work (e.g., Fergus & Bardeen, 2013; Tull et al., 2020), and supports the putative role of IU as a risk factor for health anxiety. Indeed, the consistency of rapidly emerging findings in the past year supporting the role of IU in health anxiety during the pandemic seems to highlight the robustness of this relationship (see also Korte et al., 2021). In addition to substantially increased health risks, a salient aspect of the early stages of the pandemic was the uncertainty of such health risks (in general, and in different contexts; see Freeston, Tiplady, Mawn, Bottesi, & Thwaites, 2020 and Funkhouser et al., 2022). In turn, individuals with elevated IU may have been more likely to overestimate such health risks (e.g., Bredemeier & Berenbaum, 2008; Pepperdine, Lomax, & Freeston, 2018; see Funkhouser et al., 2022) and/or engage in safety behaviors linked with health anxiety (e.g., pathological information seeking, increased attention and reactivity to physical symptoms/sensations) during times of increased uncertainty. Additional research is needed to test IU as a proposed mechanisms and explore whether the present findings pattern have persisted later in the pandemic.

In contrast, the lack of a significant prospective association between AS and health anxiety was inconsistent with our predictions and some previous work (e.g., Fergus & Bardeen, 2013; Warren et al., 2021; Wheaton, Berman, & Abramowitz, 2010; but see also Blakey & Abramowitz, 2017 and Olatunji et al., 2009). The present (null) finding will be important to replicate in independent samples, to bolster confidence that the finding is not a Type II error. Together, the current findings may suggest that previous evidence for a temporal link between AS and health anxiety may be driven by unaccounted for shared variance with IU. Indeed, the potential role of IU on the link between AS and health anxiety is bolstered by the observed prospective association between initial levels of IU and both AS and health anxiety at follow-up in the present study, suggesting that IU could contribute to the development or exacerbation of both AS and health anxiety, potentially accounting for some of the shared variance between those constructs (see also Çelik et al., 2022, O’Bryan & McLeish, 2017, O’Bryan et al., 2022, and Wright et al., 2016). Further, the prospective link between initial IU and changes in AS may suggest that individuals with elevated IU tended to become more vigilant (and reactive) to bodily sensations during the pandemic, perhaps as a way to cope with their uncertainty during the ongoing pandemic. More generally, given that both AS and IU are transdiagnostic constructs that may reflect “fundamental fears” (Carleton et al., 2007, Carleton, 2016), and given that the two are moderately correlated, we propose that both AS and IU may be important to study and examine simultaneously in studies of outcomes linked to both (see also Carleton, Fetzner, Hackl, & McEvoy, 2013). The present results build on previous studies examining the correlates and potential consequences of elevated IU and/or AS during the pandemic (e.g., Paluszek et al., 2021; Rettie & Daniels, 2021; Saulnier et al., 2022; Tull et al., 2020; Warren et al., 2021). Indeed, AS may very well contribute to symptoms of emotional distress (including health anxiety) in other ways not captured by our prospective analyses (e.g., pre-pandemic levels of AS may contribute to initial levels of health anxiety). Further, AS may contribute to other related but distinct problems during the pandemic, either independently or interactively with IU (e.g., catastrophizing; Saulnier et al., 2022).

The present study has some notable strengths, including the longitudinal study design and sample size (to ensure good statistical power). Of course, the study also has limitations to consider when interpreting the results and planning further research. First, the study relied exclusively on self-report measures, which could lead to some spurious associations attributed to shared method variance. Future studies should utilize other methods to measure one or more key study constructs, such as clinical interviews/ratings. Further, the study used an internet survey platform – although online data collection has important strengths (e.g., diversifying samples, efficiency), the use of online crowdsourcing platforms carries some elevated risk of invalid responding, even when methods are used to limit or identify such data (e.g., based on short survey completion times). Further, the extent to which present study findings can be generalized to other samples, or other times remains unclear. The lack of demographic information such as participant ethnicity, occupation, or SES limits our conclusions and confidence about the generalizability of present study findings, as well as our ability to conduct further analyses across different subgroups (e.g., education, socioeconomic status, or health status) which may highlight unique differences in the relationship between key study variables. Prospective studies over longer time periods, and with more than two observation points, would permit testing more complex models of the temporal links between IU, AS, and health anxiety, including mediational models. Notably, the extent to which baseline SHAI scores from the current study captured levels of pre-pandemic vs. early pandemic health anxiety remains unclear, particularly given the timeframe assessed by the SHAI (past 6 months). Indeed, pre-pandemic sentiment of health anxiety may give support to the relationships explored in the current study by providing unique insight into the maintenance and exacerbation of health anxiety within the context of a global pandemic. Additionally, our measure of health anxiety did not specifically address COVID-specific health anxiety (which was not examined due to the lack of established measures when this study was initiated), and therefore it remains unclear the extent to which the relationship between intolerance of uncertainty and anxiety sensitivity may uniquely or differentially related to COVID-specific health anxiety. More generally, collection of baseline data prior to the onset of the pandemic would have supported more definitive conclusions about pre-pandemic relations between IU, health anxiety, and AS as well as changes in these variables associated with pandemic conditions. Further, small effects like those observed in the present study are expected and common in longitudinal research in the field, however more research is needed to explore/confirm the clinical significance of these prospective relationships. Last, but most importantly, the prospective analyses and results provide stronger evidence for theorized causal links than cross-sectional data alone but are by no means sufficient to establish causality. More prospective studies are needed, and importantly, future studies should aim to measure these constructs more times and/or over a longer timeframe, and critically, should include and examine other important variables known or hypothesized to contribute to health anxiety. Advancing research on health anxiety, particularly doing a global pandemic, could illuminate potential mechanisms driving associations between IU and increases in health anxiety.

If supported in future research, findings from the present study may have some valuable clinical implications. In particular, present study findings highlight the potential utility of focusing attention and efforts on elevated IU in programs to reduce or prevent health anxiety during the pandemic. For example, IU may be a valuable target in the treatment of health anxiety, perhaps utilizing IU-focused intervention that have been developed and tested for treating generalized anxiety (see Robichaud & Dugas, 2006 and Hebert & Dugas, 2019). Indeed established cognitive-behavioral therapy approaches for health anxiety have shown promise during the pandemic (e.g., see Sharrock et al., 2021), however there is potential for cognitive-behavioral treatment programs/protocols to be bolstered by incorporating IU-focused strategies developed by Dugas and colleagues for improving uncertainty tolerance. Measuring IU as a method for identifying individuals at risk for health anxiety could also have value, though the utility of IU as a risk factor is likely to be contingent of the magnitude of relationship between these variables. Although effect sizes observed in the present study were small, the smaller sizes could be due in part to the short timeframe of the follow-up and/or limitations of the measures. In addition to work aiming to replicate our results and addressing the limitations discussed above, future studies should aim to directly test speculative clinical implications.

6. Conclusion

In summary, the present study tested prospective links between intolerance of uncertainty, anxiety sensitivity, and health anxiety over 6 months during the early stages of the pandemic using cross-lagged analysis. Results revealed that initial levels of intolerance of uncertainty predicted levels of health anxiety 6 months later, as well as anxiety sensitivity 6 months later. Although questions remain about the nature and time course of the relationships between IU and health anxiety, present study findings may have important implications for theories and treatment of health anxiety.

Funding statement

This research did not receive any specific funding from agencies in the public, commercial, or not-for-profit sectors. During the time that this manuscript was written, KB received funding from the Patient-Centered Outcomes Research Institute (CER-2020C1-19382) and the National Institute of Mental Health (R21 MH123888-01A1). LDC received funding from the National Science Foundation (2021317035), and NB received funding from the National Institute of Mental Health (F31MH120936).

Ethical approval statement

Study procedures were approved by the University of Delaware Institutional Review Board. Participants provided written consent, obtained through Prolific Academic.

Conflict of interest

The authors have no conflicts of interest to disclose.

Data availability

Data are available from the authors by request.