Computational phenotypes underlying effort-based decision-making and negative symptoms in a transdiagnostic severe...

Effort-based decision-making (EBDM) impairments predict negative symptoms across multiple psychiatric diagnoses. However, it is unclear whether equifinality is present and different disorders reach the same clinical endpoint of negative symptoms via different mechanistic EBDM processes. This study used computational modeling to isolate processes underlying EBDM in a large severe mental illness-spectrum sample. The Effort Expenditure for Rewards Task, negative symptom measures, and neuropsychological tests were administered to 920 participants: schizophrenia (SZ; n = 147), first-episode psychosis (FEP; n = 54), bipolar disorder (n = 53), depressive disorder (n = 37), clinical high-risk for psychosis (CHR; n = 231), other clinical (n = 99), and healthy control groups (HC; n = 299). Computational modeling identified whether participants' EBDM behavior was best fit by models indexing full or partial subjective value (use reward magnitude and/or probability) or bias (failure to use reward magnitude and probability). Best fitting models significantly differed across diagnostic groups. SZ and FEP were best fit by the bias model and less likely to use reward magnitude and probability to guide EBDM. The CHR, other clinical, depressive disorder, and HC groups were best fit by the full subjective value model and were more likely to use reward magnitude and probability, while the bipolar disorder group's behavior was more variable. Across groups, participants best fit by the bias model had the greatest negative symptoms and cognitive impairments. Results indicate mood and psychosis-spectrum disorders differentially approach EBDM. Equifinality in the pathway to negative symptoms was not supported; those with difficulty utilizing reward and probability information had the greatest negative symptoms, independent of diagnosis.