(PS10-C61) Risk Estimation Bias in Anxiety and Depression: Parsing Subjective and Objective Risk for Low Probability Negative Events

Foundational cognitive models hypothesize that persons with anxiety and depression are prone to risk estimation bias: they overestimate their risk for negative outcomes to a greater degree than the general, non-clinical population. A key limitation in the literature is that most research paradigms do not measure true risk estimation bias, wherein a person’s subjective risk estimate is directly compared to reality (i.e., their objective risk). This prevents empirical confirmation of the theory that anxiety and depression relate to higher risk estimation bias (i.e., estimates being further from reality compared to estimates in healthy populations), rather than just higher subjective risk estimates. We sought to test this hypothesis and explore whether measuring risk estimation bias provides added utility above measuring subjective risk. We also   tested the specificity of these associations by considering general anxiety and depression symptoms alongside stress-response symptoms more closely related to the feared negative outcome.

A sample of 319 Florida residents was assessed three times from April 2020 - February 2021. Participants reported anxiety and depression symptoms and pandemic stress responses. At the final assessment (February 2021), participants rated their subjective risk for COVID-19 exposure, from 0-100. Risk estimation bias was calculated by comparing this subjective estimate to an objective, individualized measure of exposure risk based on participant behaviors and concurrent COVID cases in their unique zip area. Reflecting the low base rate probability of COVID exposure, average objective risk in our sample was low (M=3.48, SD=5.62, Range=0–30.22). Subjective risk estimates and risk estimation bias were therefore highly correlated (r=0.99, p<.001). Separate general linear models revealed that early pandemic symptoms of anxiety (β=0.22, p< .001), depression (β=0.14, p=.015), and pandemic-related stress response (β=0.24, p< .001) predicted higher risk estimation bias after controlling for demographic covariates. Only pandemic-related stress remained significant (β=0.19, p=.008) when these constructs were examined in the same model. To assess the utility of our bias measure, we tested an alternative model regressing subjective risk on symptoms. The pattern of results for the effect of symptoms on subjective risk estimates was identical to the results for risk estimation bias, indicating that assessing true risk estimation bias did not actually provide incremental utility in our study.

This study is one of the first to empirically demonstrate that for low-probability events like contracting COVID, anxiety and depression are associated with inflated subjective risk estimates, but not necessarily with greater bias in those estimates. Clinically, our results highlight that existing cognitive debiasing techniques might benefit from focusing on reducing subjective risk estimates regardless of their objective reality Our results also suggest that more outcome-specific symptoms impact risk estimation processes above and beyond the effect of general affective symptoms.