VA is already using AI to identify and assist veterans in crisis, officials say

SDI Productions/Getty Images

The use of AI capabilities to review the agency’s surveys to veterans has helped save a couple thousand lives and helped another couple thousand retired servicemembers receive housing, according to VA’s chief veterans experience officer.

The Department of Veterans Affairs is leveraging artificial intelligence to better identify veterans at risk of suicide or homelessness, according to VA officials. 

VA has been focused on efforts to reduce the number of veteran suicides, which remain higher than among most other U.S. populations. The department’s ongoing efforts have included the use of predictive tools to identify veterans at the highest risk of self-harm, as well as ways of using AI to scan clinicians’ records and progress notes. 

VA’s AI inventory has identified over 100 use cases where the emerging technologies can be used to amplify the department’s work on behalf of veterans, including new and enhanced programs for identifying retired servicemembers at risk of suicide and those facing homelessness or drug addiction. 

During nonprofit VetsinTech’s launch of its new Vets in AI program on Thursday, VA Assistant Under Secretary for Health Carolyn Clancy noted that the department’s Recovery Engagement and Coordination for Health — Veterans Enhanced Treatment program, or REACH VET, “was originally a statistical model and now has become an AI.”

REACH VET first began as a pilot program in 2016 and used a predictive algorithm to identify veterans who are in the top 0.1% tier of suicide risk. Veterans in this category are pinpointed and then connected with VA coordinators to receive more targeted support. The program is active at 28 VA sites and identifies roughly 6,700 veterans each month who need help, according to data shared by Rep. Mariannette Miller-Meeks, R-Iowa, during a House hearing earlier this week.

VA’s crisis line also allows third-party individuals to reach out with concerns about veterans who may be at risk of self-harm. Clancy said, for example, that if someone sees a Facebook post from a veteran they don’t know expressing suicidal thoughts, then the crisis line can locate that person “90-plus percent of the time.” But AI presents an opportunity for VA to better identify veterans in need.

“We're working with a variety of contractors who are using very different models of AI to try to figure out, more upstream, what signals are you seeing,” Clancy said. “It's a very different thing when you make an appointment. What about people who are noticed by family members who are not doing as well? So I think that's going to be very, very interesting.”

Chief Veterans Experience Officer John Boerstler also said the department is applying AI to the surveys it sends out to veterans, which has also resulted in more targeted assistance. 

“If a veteran uses the free textbox on a survey to say, ‘I'm so frustrated, I'm willing to give up right now,’ or ‘I'm going to kill myself,’ or ‘I'm in a bad housing situation,’ or ‘I'm living on a friend's couch,’ all of these phrases will pick up in the natural language processing tool and then automatically send an alert to the veterans crisis line or the housing for veterans crisis line, which are both 24/7,” Boerstler said.

He added that VA has “saved a couple thousand lives and we've been able to put a couple thousand folks into permanent supportive housing by getting on the front end and identifying those moments of crises.”