Multimodal machine learning workflows can predict psychosis in patients with clinical high-risk syndromes and recent-onset depression, according to a recent publication in JAMA Psychiatry (2020;e203604. doi:10.1001/jamapsychiatry.2020.3604).
“Diverse models have been developed to predict psychosis in patients with clinical high-risk states. Whether prediction can be improved by efficiently combining clinical and biological models and by broadening the risk spectrum to young patients with depressive syndromes remains unclear,” wrote Nikolaos Koutsouleris, MD, Department of Psychiatry and Psychotherapy, Ludwig-Maximilian-University, Munich, Germany, and colleagues.
This study aimed to evaluate whether psychosis transition can be predicted using multimodal machine learning that integrates clinical and neurocognitive data, structural magnetic resonance imaging (sMRI), and polygenic risk scores (PRS) for schizophrenia.
The main outcomes and measures included accuracy and generalizability of prognostic systems.
A total of 668 individuals were included in the analysis: 343 patients (167 with clinical high-risk syndromes and 167 with recent-onset depression) and 343 healthy volunteers.
A balanced accuracy of 73.2% was attained by effectively ruling out but ineffectively ruling in psychosis transition. In contrast, algorithms showed high sensitivity but low specificity.
A cybernetic risk calculator combining all algorithmic and human components predicted psychosis with a balanced accuracy of 85.5%. In comparison, an optimal prognostic workflow produced a balanced accuracy of 85.9% at a much lower diagnostic burden by sequentially integrating clinical-neurocognitive, expert-based, PRS-based, and sMRI-based risk estimates as needed for the given patient.
“These findings suggest that psychosis transition can be predicted in a broader risk spectrum by sequentially integrating algorithms’ and clinicians’ risk estimates,” concluded Dr Koutsouleris and colleagues.—Lisa Kuhns
