AI to better target suicide risk screening in routine care

AI to better target suicide risk screening in routine care
AI to better target suicide risk screening in routine care

Most people who commit suicide have consulted a health professional in the previous year, for other reasons. In this context, researchers have developed a model based on artificial intelligence (AI) to identify patients at risk. Their latest work shows that automatic risk detection, combined with well-designed clinical alerts, would enable more targeted prevention. They present their results in a press release.

A new study of Vanderbilt University Medical Center shows that artificial intelligence (AI)-driven clinical alerts could help doctors identify patients at risk for suicide, potentially improving prevention efforts in routine medical settings.

A team led by Colin Walsh, MD, MA, associate professor of biomedical informatics, medicine and psychiatry, tested whether their AI system, called the Vanderbilt Suicide Attempt and Ideation Probability (VSAIL) model , could effectively encourage physicians at three VUMC neurology clinics to screen patients at risk for suicide during regular clinic visits.

The study, published in JAMA Network Open compared two approaches: automatic pop-up alerts that interrupted the doctor’s workflow and a more passive system that simply displayed risk information in the patient’s electronic record.

The study found that interruptive alerts were much more effective, leading doctors to conduct suicide risk assessments in relation to 42% of screening alerts, compared to just 4% with the passive system.

“Most people who commit suicide saw a health professional in the year before their death, often for reasons unrelated to mental healthWalsh said. But universal screening is not practical in all settings. We developed VSAIL to help identify high-risk patients and spark targeted discussions about screening. »

Suicide has been on the rise in the United States for a generation and is estimated to claim the lives of 14.2 out of every 100,000 Americans each year, making it the 11th leading cause of death in the United States. Studies have shown that 77% of people who die by suicide had contact with GPs in the year before their death.

Calls to improve risk screening have led researchers to explore ways to identify patients who most need to be evaluated. The VSAIL model, which Walsh’s team developed at Vanderbilt, analyzes routine information from electronic medical records to calculate a patient’s 30-day suicide risk. In previous prospective tests, where VUMC patient records were flagged but no alerts were triggered, the model was effective in identifying high-risk patients, with one in 23 people flagged by the system subsequently reporting suicidal thoughts.

-

In the new study, when patients identified as high-risk by VSAIL showed up for appointments at Vanderbilt neurology clinics, their doctors randomly received interruptive or non-interruptive alerts. Research has focused on neurology clinics because certain neurological conditions are associated with an increased risk of suicide.

The researchers suggested that similar systems could be tested in other medical settings.

“The automated system only reported about 8% of all patient visits for screening, Walsh points out. This selective approach makes it easier for busy clinics to implement suicide prevention measures. »

The study included 7,732 patient consultations over a six-month period, resulting in 596 screening alerts in total. During the 30-day follow-up period, a review of VUMC medical records revealed that no patients in the two randomized alert groups experienced episodes of suicidal ideation or suicide attempt. Although interruption alerts were more effective in prompting screenings, they could potentially contribute to “alert fatigue” – when doctors are overwhelmed by frequent automatic notifications. The researchers noted that future studies should examine this concern.

“Health systems must balance the effectiveness of interruption alerts with their potential downsides, concludes Walsh. But these results suggest that automatic risk detection combined with well-designed alerts could help us identify more patients at risk. »

• According to a press release from Vanderbilt University Medical Center, January 3, 2025, spotted by the Infosuicide blog.
Risk Model–Guided Clinical Decision Support for Suicide ScreeningA Randomized Clinical Trial, C. Walsh et al., JAMA Network Open (2025). DOI: 10.1001/jamanetworkopen.2024.52371

-

--

PREV My prevention assessment: accessible tools to support those involved in “reaching out” to people in vulnerable situations
NEXT Bagnols Hospital on the front line facing the flu epidemic: wearing a mask is compulsory, internal mobilization plan, discover the list of measures taken today by management