Parcellating cognitive heterogeneity in early psychosis-spectrum illnesses: A cluster analysis

Jacob J Crouse; Ahmed A Moustafa; Sophia E R Bogaty; Ian B Hickie; Daniel F Hermens

doi:10.1016/j.schres.2018.06.060

Parcellating cognitive heterogeneity in early psychosis-spectrum illnesses: A cluster analysis

Schizophr Res. 2018 Dec:202:91-98. doi: 10.1016/j.schres.2018.06.060. Epub 2018 Jul 2.

Authors

Jacob J Crouse¹, Ahmed A Moustafa², Sophia E R Bogaty³, Ian B Hickie³, Daniel F Hermens⁴

Affiliations

¹ Youth Mental Health Team, Brain & Mind Centre, The University of Sydney, NSW, Australia. Electronic address: jcro8838@uni.sydney.edu.au.
² MARCS Institute for Brain, Behaviour & Development, NSW, Australia; School of Social Sciences & Psychology, Western Sydney University, NSW, Australia; Department of Social Sciences, College of Arts and Sciences, Qatar University, Doha, Qatar.
³ Youth Mental Health Team, Brain & Mind Centre, The University of Sydney, NSW, Australia.
⁴ Youth Mental Health Team, Brain & Mind Centre, The University of Sydney, NSW, Australia; Sunshine Coast Mind & Neuroscience - Thompson Institute, University of the Sunshine Coast, QLD, Australia.

PMID: 30042029
DOI: 10.1016/j.schres.2018.06.060

Abstract

Cognitive impairment is argued to represent a core feature of psychosis-spectrum illnesses. However, within-diagnosis heterogeneity is common, and risk factors for poor cognition remain to be examined after statistically accounting for heterogeneity. Accordingly, we used a data-driven technique (cluster analysis) to empirically-derive cognitive clusters across diagnoses and examined whether concurrent substance use or a history of a neurodevelopmental/behavioral disorder differed between clusters. Data from 135 young help-seekers (aged 12-30 years) with a psychosis-spectrum illness were retrospectively analyzed. Ward's hierarchical cluster analysis classified three cognitive clusters characterized by: (1) normal-range; (2) mixed; and (3) grossly-impaired performance. Despite mostly comparable clinical and demographic measures, cluster 1 had superior socio-occupational functioning and the highest estimated premorbid IQ, followed sequentially by clusters 2 and 3. Proportions of cannabis and amphetamine users did not differ significantly across clusters, nor did rates of patients with a neurodevelopmental/behavioral disorder history. Cluster 3 was however comprised of fewer 'risky' drinkers, possibly reflecting reduced opportunity for social drinking associated with cognitive impairment. Estimated premorbid IQ predicted cluster membership (2 vs. 1 & 3 vs. 1), as did clinician-rated socio-occupational functioning and 'not being enrolled in school or tertiary education' (3 vs. 1). Our results suggest that concurrent substance use and history of a neurodevelopmental/behavioral disorder do not adequately explain cluster-level cognitive variance in this sample. Future work should integrate neurobiological measures associated with cognition (e.g. white matter integrity) to discern whether clusters reflect neurobiological subtypes better representative of pathophysiology than present symptom-based classifications.

Keywords: Cluster analysis; Neurocognition; Psychosis-spectrum; Psychosocial functioning; Substance use.

Publication types

Research Support, Non-U.S. Gov't

MeSH terms

Adolescent
Adult
Child
Cluster Analysis
Cognitive Dysfunction* / classification
Cognitive Dysfunction* / etiology
Cognitive Dysfunction* / physiopathology
Female
Humans
Male
Psychotic Disorders* / classification
Psychotic Disorders* / complications
Psychotic Disorders* / physiopathology
Young Adult