AI's Role in the Push for Suicide Prevention

Janae Sharp
SEPTEMBER 10, 2018

How Tech, AI Can Support Suicide Prevention

Social context is critical to tailoring interventions. Social media use also provides insight into suicide risk, making it a useful addition to risk modeling.

>> READ: The Population Health Problems Facing Urban and Rural America

Resources to help understand social media and how it affects suicide risk are increasing. Google recently altered searches for suicide-related issues to include crisis numbers, and Facebook is working with AI to promote suicide prevention. Providers should have access to social media-facilitated data for decision support. “One effort to better understand if social media can help prevent suicide is, in which people can donate their anonymized social media data or that of their loved one who died by suicide” says Julie Cerel, Ph.D., professor in the university of Kentucky college of social work and president of American Association of Suicidology.

But the lack of access to care may create confusion about social solutions in underserved populations such as Utah. When presented with an underserved population, healthcare providers need better risk assessment tools and greater resources to combat climbing suicide rates. Healthcare providers need a better way to assess risk with the sometimes-fragmented data available at the point of care. Much of the information relevant to elevated risk for suicide is not located in the electronic health record (EHR) and not typically seen. The HBI Spotlight Suicide Attempt model improves risk stratification despite scarce resources.

The data that we have don’t always include how people feel about social support and the like — and clinicians might not have that at the point of care. We get insights from the data that we have. There is other research that indicates that these pieces are important in predictive models, but hospitals and payers aren’t always going to have that. We have a way to predict the risk through information that is readily available at the point of care, such as EHR data and insurance claims data. The problem we are trying to solve is: How can we get insights from data that are readily available at the individual level, so we can explore some of the more personal areas?

For data scientists, outside experts can help interpret what we’ve found and connect the pieces. Take the HBI Spotlight Suicide Attempt Model, whose data suggested that a lower respiratory risk index was associated with an increase in suicide attempt risk. I also saw that some of the language barrier did not have a strong impact on suicide risk. I asked Cerel what that might mean. She says much of the data related to differences in rural or urban settings. Many rural white patients have a lack of mental health support, coupled with greater access to lethal means through firearms. Rural suicide statistics models, then, should be viewed within the lens of decreased access to mental health care and increased access to guns and other lethal means.

Data science can help provide decision support to providers and identify risk that isn’t immediately clear. AI models increase our understanding of the risk of suicide. The more we know, the better providers will be at connecting patients with resources. In areas like Utah, where the need doesn’t meet the demand for mental health support, we have many opportunities to improve. Better data science models will improve care, but they are only part of the puzzle to combat our mental health crisis. We can make things better, but much work is left to be done.

If you’re feeling suicidal, talk to somebody. Call the National Suicide Prevention Lifeline at 1-800-273- 8255; the Trans Lifeline at 1-877-565-8860; or the Trevor Project at 1-866-488-7386. Text ‘START’ to the Crisis Text Line at 741-741. If you don’t like the phone, connect to the Lifeline Crisis Chat at

Get the best insights in healthcare analytics directly to your inbox.

In the U.S., Healthcare Data Access Is a Scavenger Hunt
Researchers Develop New Suicide Risk Prediction Model from Massive Cohort
Why C-Suites Need to Get a Grip on Physician Burnout

Become a contributor