[2409.18878] Suicide Phenotyping from Clinical Notes in Safety-Net Psychiatric Hospital Using Multi-Label Classification with Pre-Trained Language Models

View a PDF of the paper titled Suicide Phenotyping from Medical Notes in Security-Internet Psychiatric Hospital Utilizing Multi-Label Classification with Pre-Skilled Language Fashions, by Zehan Li and 9 different authors

View PDF

Summary:Correct identification and categorization of suicidal occasions can yield higher suicide precautions, lowering operational burden, and bettering care high quality in high-acuity psychiatric settings. Pre-trained language fashions supply promise for figuring out suicidality from unstructured scientific narratives. We evaluated the efficiency of 4 BERT-based fashions utilizing two fine-tuning methods (a number of single-label and single multi-label) for detecting coexisting suicidal occasions from 500 annotated psychiatric analysis notes. The notes had been labeled for suicidal ideation (SI), suicide makes an attempt (SA), publicity to suicide (ES), and non-suicidal self-injury (NSSI). RoBERTa outperformed different fashions utilizing a number of single-label classification technique (acc=0.86, F1=0.78). MentalBERT (acc=0.83, F1=0.74) additionally exceeded BioClinicalBERT (acc=0.82, F1=0.72) which outperformed BERT (acc=0.80, F1=0.70). RoBERTa fine-tuned with single multi-label classification additional improved the mannequin efficiency (acc=0.88, F1=0.81). The findings spotlight that the mannequin optimization, pretraining with domain-relevant information, and the only multi-label classification technique improve the mannequin efficiency of suicide phenotyping. Key phrases: EHR-based Phenotyping; Pure Language Processing; Secondary Use of EHR Knowledge; Suicide Classification; BERT-based Mannequin; Psychiatry; Psychological Well being