Natural language processing — part of the technology that makes Google work — could help VA detect suicide risk among Veterans.
That’s the idea behind ongoing research at the VA Puget Sound Health Care System and the University of Washington, along with other VA sites.
One study used the technology to find past suicide attempts in VA’s electronic medical record system. Past attempts are the most compelling sign of future risk, say researchers — about twice as strong as the next-best predictor, major depression. Studies show that a past suicide attempt raises the odds of an eventual completed suicide by 40 times.
“The electronic medical record system stores a very large body of clinical notes,” explains Ken Hammond, MD. “We’ve shown that we can use search engine technology to more easily identify those Veterans who have attempted suicide at some point in their lives. That can help us prevent future attempts.”
Hammond recently retired from VA as a psychiatrist but continues to help conduct the research.
…researchers developed an automated text search that was about 80 percent accurate…
His group used data on more than 100,000 Veterans. They developed search terms to query the free text — such as doctors’ notes — in patients’ records. Their goal was to zero in on red flags indicating past suicide attempts.
One challenge was ruling out instances in which “suicide attempt” or a similar phrase appeared as part of the documentation of a suicide screening, but with no indication that the patient had in fact attempted suicide.
The researchers developed an automated text search that was about 80 percent accurate, compared against manual checking of each record by a psychiatrist.
They presented the results earlier this year at the annual Hawaii International Conference on System Sciences.
Hammond, together with VA colleagues in Salt Lake City and Boston, is looking at other suicide risk factors that may show up in Veterans’ records, especially in the notes of mental health providers. One example is childhood abuse.
The work is part of a larger VA research project on natural language processing and the electronic medical record. VA securely stores electronic medical records on some 14 million current and past patients. The records contain some two billion documents in all. The data is available to authorized VA researchers, with strict privacy safeguards in place.
For more information on VA research, visit www.research.va.gov.