Our study demonstrates a possible real-word application for the transformative power of Large Language Models in psychiatry: based on a standardized information extraction pipeline, a psychiatry-native model delivers highly relevant data to the classifying algorithm. In clinical practice, this predictive score might be useful to allocate scarce preventive resources more efficiently to minimise the necessary application of coercive measures.

Shared authorship of poster: Guillermo Calvi and Falk Gerrik Verhees. Poster presentation at the conference: Falk Gerrik Verhees