Data-driven clustering of mental health symptoms and brain functional connectivity signatures in transdiagnostic psychiatric inpatients
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background
Psychiatric disorders are notoriously heterogenous, often rendering diagnostic efforts challenging, and leading to poor therapeutic outcomes. The growing emphasis on ‘transdiagnostic’ approaches in psychiatry aligns well with the National Institute of Health-devised Research Domain Criteria (RDoc) framework that seeks to enable precision psychiatry. Here, we sought to identify transdiagnostic subgroups, which share behavioral and neural abnormalities, in a large population of psychiatric inpatients treated at a tertiary psychiatric hospital.
Methods
The sample included 1,571 patients across five units within the Depression & Anxiety Disorders Division at McLean Hospital who completed self-assessments as part of an ongoing quality-of-care improvement project. Self-assessments included validated measures of depression, anxiety, anhedonia, trauma, personality and substance misuse as well as broader screening questions. First, we identified naturally occurring transdiagnostic subgroups based on the self-reported questionnaires using consensus clustering to implement partition-around-medoids. In a subsample, we ascertained brain functional connectivity differences between the resultant subgroups using a multi-granular approach, i.e. whole-brain to connection-wise.
Results
In a k=2 partitioning solution, the first transdiagnostic cluster; C1 (N=809) had consistently higher means across all questionnaires compared to the second cluster; C2 (N=762). QIDS total score (P<0.0001, effect size = 0.54) and BASIS-24 derived total score (P<0.0001, effect size = 0.53) showed the largest difference between the groups. In follow-up imaging analysis (N=26) functional connectivity (FC) differences were observed connection-wise (P<0.0001) and between functional networks (P<0.0001), with C1 showing stronger FC than C2. Cluster-by-sex analyses revealed that females had higher BASIS-24 derived depression/functioning (C1|C2; P<0.0001 | P<0.0001), and QIDS (C1|C2; P<0.0001 | P<0.0001) scores compared to males. This was coupled with greater intermodular FC in males than females in C1 (P<0.0001) and C2 (P<0.0005).
Conclusions
These results suggest that inpatients with transdiagnostic symptoms show biobehavioral abnormalities, underpinned by sex differences. Behaviorally, this is a function of acuity, i.e., severity of psychopathological symptoms, and not diagnosis. Biologically, the dysfunction captured here may span across various brain networks, rather than a singular region.
