Machine Learning Predicts Social Outcomes in Psychosis
Machine learning outperformed human experts in predicting social outcomes a year later in people at high risk of psychosis or with recent-onset depression. Researchers published their findings online in JAMA Psychiatry.
The multisite study included 116 people in clinical high risk of psychosis states, 120 people with recent-onset depression, and 176 healthy control participants followed for 18 months.
Using clinical baseline data, machine learning predicted 1-year social functioning outcomes with 76.9% accuracy in people at high risk of psychosis and 66.2% in patients with recent-onset depression, according to the study. Using structural neuroimaging, machine learning achieved 76.2% accuracy in people at high risk of psychosis and 65% in patients with recent-onset depression.
Brain Less Responsive to Social Rewards in Schizophrenia
In combined models, machine learning predicted 1-year social outcomes with 82.7% accuracy in people at high risk of psychosis and 70.3% in people with recent-onset depression.
If validated further, researchers believe the predictive models could help personalize prevention of social impairment in patients.
“By being able to better predict what will happen to people at high risk of psychosis or with recent onset of depression over time, we are able to provide individualized treatment to clients when they first present to mental health services and potentially improve their social functioning,” said researcher Stephen J. Wood, PhD, of Orygen, the National Centre of Excellence in Youth Mental Health in Parkville, Australia.
—Jolynn Tumolo
References