Identifying Suicide Ideation and Suicidal Attempts in a Psychiatric Clinical Research Database using Natural Language...

Area of Science:

• Clinical Informatics
• Computational Linguistics
• Psychiatry

Background:

• Traditional suicide prevention research faces limitations like small sample sizes and recall bias.
• Electronic Health Records (EHRs) offer vast data, but extracting suicidality information is challenging.
• Natural Language Processing (NLP) can enhance information extraction from unstructured clinical notes.

Purpose of the Study:

• To develop and evaluate novel NLP approaches for identifying suicidality in a psychiatric clinical database.
• To address limitations in previous suicide prevention research through advanced data analysis.
• To create adaptable algorithms for detecting suicide ideation and attempts.

Main Methods:

• A rule-based NLP approach was developed to classify suicide ideation.
• A hybrid machine learning and rule-based NLP approach was implemented to identify suicide attempts.
• The methods were applied to a psychiatric clinical database, analyzing free-text clinical notes.

Main Results:

• Both NLP classifiers demonstrated good performance in the evaluation study.
• The methods accurately detected mentions of suicide ideation and suicide attempts within free-text documents.
• The developed classifiers showed high accuracy in identifying suicidality indicators.

Conclusions:

• The novel NLP approaches offer accurate and efficient detection of suicide ideation and attempts in clinical data.
• These methods overcome previous research limitations by utilizing large EHR cohorts.
• The algorithms are adaptable to other clinical datasets without requiring medical codes or additional risk factor data.