The Early Psychosis Screener (EPS): Quantitative validation against the SIPS using machine learning.

Area of Science:

• Psychiatry and Mental Health
• Machine Learning Applications
• Clinical Psychology

Background:

• Early psychosis detection is crucial for intervention and improving outcomes.
• Current screening methods may have limitations in accessibility or efficiency.
• Identifying individuals at clinical high risk (CHR) is a key step in early intervention.

Purpose of the Study:

• To develop and validate a self-report early psychosis screener (EPS) using machine learning.
• To identify highly informative self-report items for early psychosis detection.
• To compare the performance of the developed screener against existing measures and clinician-administered tools.

Main Methods:

• Machine learning, including spectral clustering and Minimum Redundancy Maximum Relevance, was applied to a large item pool.
• Data were collected from 229 individuals screened using the Structured Interview for Psychosis-risk Syndromes (SIPS).
• A 26-item Early Psychosis Screener (EPS-26) was developed and its performance evaluated using receiver operating characteristic analysis.

Main Results:

• Machine learning identified two item groups related to psychosis/mania and depression/anxiety/functioning.
• The 26-item EPS demonstrated high accuracy (AUC=0.899) in differentiating clinically low risk (CLR) from CHR individuals.
• The EPS-26 outperformed the Prodromal Questionnaire-Brief Version (PQ-B) (AUC=0.834) and showed comparable performance to the clinician-administered SIPS.

Conclusions:

• The self-report EPS-26 is a promising tool for screening individuals at risk for psychosis.
• This screener may aid in reducing the duration of untreated psychosis.
• Further validation against actual conversion to psychosis is ongoing.