Psychosis

Can machine learning better stratify patients with early psychosis?

Because symptoms vary widely between patients, a more specific method is needed to stratify individuals with psychosis — beyond clinically high risk and recently begun, a study published in JAMA Psychiatry† To better identify and categorize symptoms, researchers developed a machine learning model to “re-draw boundaries” within early psychosis.

The study involved 749 participants and 10 sites in Europe and the United Kingdom. The researchers divided the participants into 4 groups: clinically high risk of psychosis (CHR-P), recently started psychosis (ROP), recent-onset depression (ROD), and healthy control participants. They validated and compared the results, analyzed premorbid risk factors, 18-month longitudinal disease course, and polygenic risk scores for schizophrenia (PRS). They measured clinical data and biological data obtained from gray and white matter MRI scans and cerebrospinal fluid measurements.

The researchers found depression symptoms in 51% of the CHR-P and ROP and in 26% of a high-functioning subgroup. They also found an underlying genetic risk of early disease. These and other discoveries surrounding genetic cognitive risk factors support the use of “preventive neurodevelopmental strategies and provide a homogeneous target that could increase the possibility of translation,” the researchers believe.


read more

The researchers admitted that it is difficult to classify individuals with early disease because symptoms can change quickly. They also stated that the patterns found “could be further explored for more sensitive underlying dimensions and subgroups.”

In conclusion, “associations between the modalities were not straightforward and may indicate relatively independent preventive targets,” the researchers said. “The results provide important context beyond the severity of positive symptoms in these groups and their associations with brain volume reduction.”

Reference

Dwyer DB, Buciuman MO, Ruef A, et al. Clinical, brain and multilevel clustering in early psychosis and affective phases Published online, 2022 May 18. JAMA Psychiatry† 2022;10.1001/jamapsychiatry.2022.1163. doi:10.1001/jamapsychiatry.2022.1163

Leave a Comment

Your email address will not be published. Required fields are marked *