756. Computational modelling of cognitive flexibility reveals novel dissociations across compulsive and non-compulsive disorders

Abstract

Abstract Background Impairments in cognitive flexibility cause dysfunction across a wide range of neuropsychiatric disorders, including depression, schizophrenia, and autism spectrum disorders (ASD). However, the differences in the underlying computational processes remain poorly understood. Aims & Objectives We examined reinforcement learning (RL) processes underlying cognitive inflexibility in 436 patients compared to 377 healthy control participants. The patient groups included individuals with schizophrenia, major depressive disorder (with and without substance use disorder (SUD)) and ASD in children, adolescents and adults. We predicted that the parameters representing ‘stickiness’, which is the tendency to repeat an action regardless of previous outcomes, can help dissociate these disorders. Method We applied RL models using hierarchical Bayesian modelling to probabilistic reversal learning data from six different patient groups. Following selection of the best-fitting model by comparing six RL models, we examined how value-free ‘stickiness’ parameters, value-based learning rates and reinforcement sensitivity (inverse temperature) differ between these groups. Results Adolescents and adults with ASD showed lower stimulus stickiness compared to adolescent and adult control groups, respectively (both 0 ∉ 95% HDI). Stimulus stickiness in children with ASD was unaffected (0 ∈ 75% HDI). Individuals with schizophrenia demonstrated higher stimulus stickiness but this was accompanied by lower side stickiness compared to control participants (both 0 ∉ 95% HDI). Stimulus stickiness in patients with depression was unaffected (0 ∈ 75% HDI), however, patients with depression and SUD had higher stimulus stickiness compared to control participants (0 ∉ 75% HDI). Reward and punishment learning rates were higher in children with ASD compared to an age-matched control group (both 0 ∉ 95% HDI) and lower in adults with ASD compared to adult controls (both 0 ∉ 95% HDI) but were unaffected in the other groups. Reinforcement sensitivity was lower in all three ASD groups, schizophrenia, and depression with SUD, compared to their respective control groups (0 ∉ 75% HDI). Discussion & Conclusions In previous studies, we reported higher ‘stickiness’ in SUD and decreased ‘stickiness’ in obsessive compulsive disorder and gambling disorder (Kanen et al., 2019; Zuhlsdorff et al., 2024). Taken together, this suggests that accounting for value-free influences underlying cognitive flexibility using RL parameters provides novel dissociations transdiagnostically and has implications for psychiatric treatment and classification.