Suicide prevention ‘more feasible’ using AI-powered screening alerts

Key takeaways:

• Interruptive CDS led to higher rates of decision to screen for suicide risk vs. noninterruptive CDS.
• The AI makes implementing suicide prevention efforts more feasible for busy clinics, a researcher said.

An AI system greatly improved clinicians’ decisions to screen for suicidality among those considered to be at a higher risk for a suicide attempt, results of a randomized study published in JAMA Network Open showed.

Suicides have risen exponentially in the United States in the last several years. Such deaths totaled almost 50,000 in 2022, a 3% increase from the previous year.

"Most people who die by suicide have seen a health care provider in the year before their death, often for reasons unrelated to mental health," Colin Walsh, MD, MA, an associate professor of biomedical informatics, medicine and psychiatry at Vanderbilt University Medical Center, said in a press release. "But universal screening isn't practical in every setting.”

Walsh and colleagues developed the Vanderbilt Suicide Attempt and Ideation Likelihood model (VSAIL), “to help identify high-risk patients and prompt focused screening conversations,” he said.

According to the release, the VSAIL model analyzes routine information from electronic health records to determine a patient’s 30-day risk for a suicide attempt.

In the single-center randomized clinical trial, the researchers assessed and compared two alert approaches among patients identified as high-risk by VSAIL:

• interruptive clinical decision support (CDS), which involved automatic pop-up alerts that interrupted clinician workflow; and
• noninterruptive CDS, which displayed an “elevated suicide risk score” in the patient’s electronic chart.

Hovering over the noninterruptive CDS’ icon resulted in a pop-up identical to the interruptive CDS.

The trial included 561 participants with 596 clinician encounters that occurred between Aug. 17, 2022, and Feb. 16, 2023.

The researchers found that 42% of encounters in the interruptive CDS arm resulted in clinicians electing to screen, compared with 4% of encounters in the noninterruptive CDS arm (OR = 17.7; 95% CI, 6.42-48.79).

Interruptive CDS also led to higher numbers of decision to screen vs. the baseline rate of suicide risk assessment in the same clinical settings a year before (22% vs. 8%).

The researchers identified some study limitations, which included possible leakage of suicide risk assessment. The trial was also not powered to determine changes in the rates of suicide attempts or deaths, although neither such events occurred in either arm during a 30-day follow-up period.

The researchers noted that interruptive CDS “tends to be more effective in prompting behavior,” but some disadvantages of it, such as fatigue experienced by the clinicians from the frequent alerts, “counterbalance its effectiveness.”

“[AI] makes it more feasible for busy clinics to implement suicide prevention efforts," according to Walsh.

"Health care systems need to balance the effectiveness of interruptive alerts against their potential downsides," he said in the release. "But these results suggest that automated risk detection combined with well-designed alerts could help us identify more patients who need suicide prevention services."

References:

• AI tested for alerting clinicians of suicide risk at three VUMC clinics. Available at: https://news.vumc.org/2025/01/03/ai-tested-for-alerting-clinicians-of-suicide-risk-at-three-vumc-clinics/. Published Jan. 3, 2025. Accessed Jan. 22, 2025.
• Walsh C, et al. JAMA Netw Open. 2025;doi:10.1001/jamanetworkopen.2024.52371.