[klue07]

> Computers analyzed more than 30,000 different risk factors, including traditional ones like age and mental-health history. But some surprising issues—such as gastrointestinal problems, infections and injuries like rib fractures—were tied to increased risk.

> Dr. Nock says that some of these may be self-inflicted or related to impulsive behavior. Using the historical data, the approach was able to detect 45% of suicidal acts an average of about three years before the event.

> A study using similar methods looking at suicide risk among U.S. Army soldiers in the year after hospitalization for psychiatric issues was published in 2015 in JAMA Psychiatry. The algorithm was able to predict about 53% of the suicides in that high-risk period.

I wonder what the false positive rates on these are.