Can artificial intelligence save lives? That’s the question researchers at Vanderbilt Medical Center and Florida State University have been trying to answer.
Last year, a team of researchers created an algorithm that could potentially identify suicide-prone individuals. The machine-learning algorithm uses information like age, gender, diagnostic history, zip code, and prescriptions to assess if someone is at risk of suicide.
The data scientists gathered over 5,000 electronic medical records of adult patients who had previously self-harmed or attempted suicide. The algorithm correctly predicted 84% of patients who would attempt suicide in the next week, and was 80% accurate in predicting a suicide attempt within the next two years.
The researchers are now working alongside mental health specialists, ethicists, and computer scientists to figure out how the algorithm will operate in a clinical care setting. If someone is at risk of suicide, how should clinicians intervene? What’s the most effective approach in telling someone they’re likely to commit suicide? And with such high stakes, what effects can an inaccurate prediction have on a patient? We discuss.
If you are in need of support, please call the National Suicide Prevention Lifeline, 1-800-273-8255, for free and confidential help 24 hours a day, seven days a week.
Colin G. Walsh, M.D., assistant professor and lead data scientist of the suicide risk prediction effort at Vanderbilt University Medical Center