Associating functional recovery with neurocognitive profiles identified using partially ordered classification models.

摘要

Neurocognitive deficits are a core feature of schizophrenia and a significant cause of functional disability. However, targeting these deficits with new treatment approaches will only yield functional improvements if those cognitive operations that are responsible for different dimensions of functional recovery can be identified. A major challenge is that conventional neuropsychological tests, the most practical tools for broadly sampling cognitive functions in treatment trials, are polyfactorial, so that task performance is influenced by multiple cognitive operations. Hence, it is difficult to pinpoint exactly for which cognitive operations a low scoring subject may have poor functionality. We have previously applied in a neuropsychological test battery administered to 220 patients having schizophrenia or schizoaffective disorder, Bayesian statistical methods (yielding partially ordered sets, or posets) designed to mimic the expert analysis of a neuropsychologist by classifying patients into discrete groupings or "states" each having a unique cognitive profile. Here, we report on the association of attributes describing these states (viz. working memory, capacity for divergent thinking, cognitive flexibility and psychomotor speed) with two domains of functional outcome (work/education and residential functioning) rated up to 18 months later. After multiplicity correction, only working memory was associated with work/education outcome. While working memory was not associated with residential outcome, the remaining three attributes were. These findings suggest that different neurocognitive operations may be responsible for different outcome domains. Findings support the use of the poset methodology for clarifying patterns of relationships between discrete neurocognitive attributes and domains of functional outcome.