Cognitive subgroups among individuals with serious mental illness and differences in global functioning

Background

Individuals with serious mental illness (SMI), including bipolar disorder (BD) and schizophrenia (SZ), frequently exhibit impairments in both global functioning (GF) and neurocognition. Previous studies have identified transdiagnostic cognitive clusters; however, their association with detailed psychosocial and clinical outcomes remains insufficiently understood. In particular, it is unclear to what extent neurocognitive profiles relate to short-term versus long-term aspects of functioning.

Methods

This study included 300 euthymic individuals with SMI (BD: n = 148; SZ: n = 135; schizoaffective disorder: n = 17) from the multicenter PsyCourse cohort. Neurocognitive performance was assessed using a standardized test battery covering attention/processing speed (Trail Making Test A, Digit Symbol Test), executive function (Trail Making Test B, Digit Span backward), short-term memory (Digit Span forward), and crystallized intelligence (MWT-B). Global functioning was operationalized across three domains: clinician-rated functioning (Global Assessment of Functioning, GAF), subjective quality of life (WHOQOL-BREF), and clinical outcomes (e.g., illness course variables, treatment status). GF variables were further divided into short-term (GAF, WHOQOL-BREF, Clinical Global Impression) and long-term indicators (e.g., illness duration, number of episodes, number of hospitalizations). Hierarchical cluster analysis using Ward’s method on z-transformed neurocognitive scores was conducted to identify subgroups, followed by k-means validation and discriminant analysis. Cluster differences were examined using analyses of (co)variance and chi-square tests, controlling for age, sex, and education where appropriate.

Results

A four-cluster solution was identified and replicated across methods, with high classification accuracy (92.0% cross-validated). Clusters reflected distinct neurocognitive profiles: (1) above-average performance, (2) average performance, (3) selectively impaired reduced attention/processing speed, and (4) selectively impaired crystallized intelligence. Clusters differed significantly in most aspects of neurocognitive performance (p < .001). Regarding outcomes, clusters with poorer neurocognition (3 and 4) showed significantly lower GAF scores and higher illness severity, as well as greater work impairment and higher rates of inpatient treatment. In contrast, clusters with relatively preserved cognition (1 and 2) demonstrated better short-term functioning. When looking at clusters 1 and 2, crystallized intelligence and attention/processing speed (as measured with TMT-A) were the only domains both above average and with higher scores than clusters with less favorable outcomes. Despite these differences, clusters did not differ in long-term GF indicators, including illness duration, number of episodes, number of hospitalizations, or number of suicide attempts. Diagnostic composition varied across clusters, with a higher proportion of SZ in cognitively impaired clusters and BD in better-performing clusters. Education was better in clusters 1 and 2, although half of individuals from across all clusters were unemployed.

Conclusions

Neurocognitive subgroups in SMI are robust across diagnostic categories and show meaningful differences in current functioning and symptom severity. Crystallized intelligence and attention/processing speed may represent key factors for short-term functioning. However, these profiles appear to have limited relevance for long-term illness trajectories. These findings highlight the importance of targeting cognitive and environmental factors to improve functional outcomes, while suggesting that long-term prognosis may be influenced by additional mechanisms beyond neurocognition.

References

No references