Longitudinal Associations Between Volatility Priors and Delusions in Individuals Recovering From an Acute Psychotic...

Computational psychiatry has revealed latent aspects of cognition that relate to psychiatric symptoms. Yet, to understand their clinical utility, it must be known whether these computational parameters contribute to symptom severity in a traitlike (stable) or statelike (dynamic) manner. Mounting evidence suggests that expecting more environmental volatility contributes to delusional thinking. Here we test whether volatility priors relate to delusion severity longitudinally in recently hospitalized individuals with schizophrenia spectrum disorder (SSD).

Participants with SSD (n = 75) were identified while receiving treatment for delusions on an acute inpatient unit and recruited into the study within weeks of discharge. Participants with SSD and participants recruited for nonclinical comparison (NCC) (n = 71) were followed over 6 months. Clinical delusions, paranoia, delusional ideation, and volatility priors were assessed at 6 time points. Volatility priors were estimated with a hierarchical Gaussian filter during a 3-option probabilistic reversal learning task. Associations were examined in linear mixed models.

At baseline, participants with SSD had significantly elevated volatility priors, in the context of more severe paranoia and delusional ideation, relative to NCC participants. Volatility priors, paranoia, and delusional ideation all significantly decreased over time in participants with SSD, but remained elevated over 6 months, never fully normalizing. Critically, volatility priors were significantly longitudinally associated with paranoia, delusional ideation, and clinical delusion severity, but not depression or worry. Results remained unchanged after controlling for antipsychotic medication and baseline cognitive ability.

We show evidence that expecting more volatility is a state-sensitive marker of delusion severity in psychotic disorders. Interventions targeting this computationally identified construct may improve clinical delusions.