Toward Automatic Risk Assessment to Support Suicide Prevention.

Area of Science:

• Public Health
• Artificial Intelligence
• Mental Health Research

Background:

• Suicide remains a significant global public health concern with persistent high mortality rates.
• Traditional methods for assessing suicide risk have proven largely ineffective for medical specialists.
• Advancements in machine learning offer new possibilities for predicting suicide risk through data analysis.

Purpose of the Study:

• To evaluate an artificial intelligence (AI) driven, referral-centric methodology within the National Health Service (NHS).
• To ascertain if statistically significant findings can be obtained from historical suicide data.
• To explore potential strategies for the application and exploitation of the developed methodology.

Main Methods:

• Utilized data from patients who died by suicide between 2013-2016.
• Incorporated both structured patient data and free-text clinical notes.
• Employed advanced machine learning and text mining techniques for analysis.

Main Results:

• The pilot study indicates machine learning's potential in predicting suicide risk.
• Identified individuals at higher risk of suicide at the point of mental health service referral.
• The methodology shows promise despite limitations in sample size and absence of non-suicide cases.

Conclusions:

• Machine learning demonstrates potential as a tool for suicide risk prediction.
• This AI-based approach can inform clinical practice by identifying at-risk individuals.
• Further research is warranted to refine the methodology and address study limitations.