Identifying Psychosis Episodes in Psychiatric Admission Notes via Rule-based Methods, Machine Learning, and...

Area of Science:

• Medical Informatics
• Computational Linguistics
• Psychiatry

Background:

• Accurate psychosis diagnosis is vital but challenging due to symptom variability and underreporting.
• Electronic Health Records (EHRs) often lack detailed clinical information for retrospective psychosis identification.
• Stigma and diminished insight further complicate early and accurate diagnosis of psychotic episodes.

Purpose of the Study:

• To evaluate NLP algorithms for psychosis detection in psychiatric admission notes.
• To compare rule-based, machine learning, and pre-trained language models.
• To assess the impact of keyword pre-selection on model performance.

Main Methods:

• Analysis of 4,617 psychiatric admission notes (2005-2019).
• Application of rule-based algorithms, machine learning (XGBoost with TF-IDF), and pre-trained language models (BlueBERT).
• Evaluation of keyword pre-selection strategies to streamline data for NLP models.

Main Results:

• XGBoost with keyword-selected TF-IDF features achieved an F1 score of 0.8881.
• BlueBERT achieved a comparable F1 score of 0.8841.
• Both NLP models significantly outperformed ICD code-based detection (F1 score 0.7608), and keyword pre-selection enhanced performance.

Conclusions:

• NLP techniques, particularly when combined with keyword pre-selection, offer a powerful approach to enhance psychosis detection in clinical notes.
• This study demonstrates the potential of NLP to improve diagnostic accuracy and efficiency in psychiatric care.
• Findings provide a foundation for future research applying NLP to psychosis identification within EHR data.