Machine learning model predicts suicide risk

Hyperaxion Apr 30, 2020

A new computer model is able to predict a person’s suicide risk through electronic health records. This model can predict suicidal behavior for up to two years in advance.

It is true that computers cannot replace health professionals in the identification of mental disorders, but they can identify high-risk patients who, currently, have gone unnoticed by the health system.

Researchers have increasingly pointed to Artificial Intelligence and machine learning models as a potential aid in identifying individuals with a higher risk of suicide. In 2017, a study suggested that a model might be able to identify the “neural signature” of someone with suicidal tendencies, through the results of an MRI scan.

Machine learning model predicts suicide risk
(Credit: Cristina Barba / Flickr).

This new study builds on previous work to develop a computer model that predicts the risk of suicidal behavior by analyzing a person’s health records. The scientific article was published on March 25 on the JAMA Network Open.

The researchers trained this pre-existing model with a data set from five different health centers, which covered 3.7 million American patients. According to the New Atlas, the records contained almost 40,000 suicide attempts and the algorithm was able to predict 38% of those attempts, on average, 2.1 years before they occurred.

Ben Reis, from Boston Children’s Hospital, pointed out that while some suicide risk indicators were not surprising, such as drug addiction or pre-existing mental health conditions, others were unexpected, including the use of HIV drugs and rhabdomyolysis, a muscle condition.

The next step is to refine the model, incorporating extra data, such as clinical notes from healthcare professionals. The ultimate goal is to help doctors identify patients with suicidal tendencies.

Comments

Leave a Reply

avatar
  Subscribe  
Notify of