I found an article in JAMA Network (Medical News & Perspectives) the other day which discussed a recent study on the use of Artificial Intelligence (AI) in suicide risk assessment (Hswen Y, Abbasi J. How AI Could Help Clinicians Identify American Indian Patients at Risk for Suicide. JAMA. Published online January 10, 2025. doi:10.1001/jama.2024.24063).
I’ve published several posts expressing my objections to AI in medicine. On the other hand, I did a lot of suicide risk assessments during my career as a psychiatric consultant in the general hospital. I appreciated the comments made by one of the co-authors, Emily E. Haroz, PhD (see link above).
Dr. Haroz preferred the term “risk assessment” rather than “prediction” referring to the study (Haroz EE, Rebman P, Goklish N, et al. Performance of Machine Learning Suicide Risk Models in an American Indian Population. JAMA Netw Open. 2024;7(10):e2439269. doi:10.1001/jamanetworkopen.2024.39269).
The model used for the AI input used data available to clinicians in patient charts. The charts can be very large and it makes sense to apply computers to search them for the variables that can be linked to suicide risk. What impressed me most was the admission that AI alone can’t solve the problem of suicide risk assessment. Clinicians, administrators, and community case managers all have to be involved.
The answer to the question “How do you know when someone’s at high risk?” was that the patient was crying. Dr. Haroz points out that AI probably can’t detect that.
That reminded me of Dr. Igor Galynker, who has published a lot about how to assess for high risk of suicide. His work on the suicide crisis syndrome is well known and you can check out his website at the Icahn School of Medicine at Mount Sinai. I still remember my first “encounter” with him, which you can read about here.
His checklist for the suicide crisis syndrome is available on his website and he’s published a book about as well, “The Suicidal Crisis: Clinical Guide to the Assessment of Imminent Suicide Risk 2nd Edition”. There is also a free access article about it on the World Psychiatry journal website.
Although I have reservations about the involvement of AI in medicine, I have to admit that computers can do some things better than humans. There may be a role for AI in suicide risk assessment, and I wonder if Dr. Galynker’s work could be part of the process used to teach AI about it.
Go Retire Psychiatrist 