Clinical Assessments Predicted Continued Cannabis Use in Psychosis Patients with 73% Accuracy

Clinical interview data alone predicted continued cannabis use with 73% accuracy in 109 recent-onset psychosis patients. The model generalized to 73 clinical high-risk patients. Lower functioning, specific substance use patterns, urbanicity, and lack of alternative coping strategies were the strongest predictors.

Prospective study of 109 recent-onset psychosis (ROP) patients who reported lifetime cannabis use at baseline. Machine learning models tested whether clinical assessments, cognitive tests, and brain MRI could predict cannabis use between baseline and 9-month follow-up. The model was validated in 73 clinical high-risk patients.

Continued cannabis use worsens outcomes in psychosis. If clinicians can identify which patients are most likely to keep using, they can target interventions more effectively. This study shows that standard clinical assessments may be sufficient for this prediction.

The finding that brain scans and cognitive tests added nothing beyond clinical interviews is practically important: it means clinicians do not need expensive neuroimaging to identify at-risk patients. The coping strategies finding also points to a specific intervention target.

Moderate sample size (109 ROP, 73 CHR). "Continued use" was defined as any cannabis use in 9 months, not frequency or quantity. The model needs testing in larger, more diverse populations before clinical implementation.