Talk to the Veterans Crisis Line now
U.S. flag
An official website of the United States government

Office of Research & Development

print icon sign up for VA Research updates
header

Automated way to identify Veterans at risk of suicide?

thumbnail
San Diego VA nurse LeAnn Shipp views a mock patient's electronic medical record. VA researchers are developing automated methods to extract information from the records that can be used to improve care. One example: identifying Veterans at risk of suicide. (Photo by Kevin Walsh)
Clinical clues—San Diego VA nurse LeAnn Shipp views a mock patient's electronic medical record. VA researchers are developing automated methods to extract information from the records that can be used to improve care. One example: identifying Veterans at risk of suicide. (Photo by Kevin Walsh)

Investigators with VA and the University of Washington say sophisticated searches of the free text in VA patients' electronic medical records may be a way to identify those at risk of suicide. The researchers presented their findings at the 46th annual Hawaii International Conference on System Sciences, held last month.

The team used two data sets, on 10,000 and 100,000 Veterans, respectively, and developed search terms to query clinicians' notes for indicators of suicide risk. They homed in on clinical notes indicating past suicide attempts. This is the strongest signal of risk, according to studies. Next are depression, bipolar disorder, schizophrenia, and alcoholism.

One of the challenges was weeding out records that contained the words "suicide attempt" or similar phrases but that were simply part of clinicians' documentation of suicide screenings, with no positive findings.

Eventually, the researchers came up with query terms that were reasonably precise in finding actual suicide attempts as documented in the notes. Using the smaller data set, they zeroed in on 2,210 documents—belonging to 597 patients—that contained significant language indicating past suicide attempts. A psychiatrist manually reviewed the records and classified 402 of the patients as "true positives." In other words, the automated method was about 80 percent accurate.

The researchers wrote, "We show the potential for text search to readily identify high-risk individuals who have attempted suicide at some point in their lives." They plan to further refine the search methodology to yield even more precise results.

Currently, VA securely stores electronic medical records on nearly 10 million Veterans, current and past patients. The records contain some 1.2 billion documents in all. The records are made available to authorized VA researchers with strict privacy safeguards in place.

The suicide study was part of a larger VA research initiative aimed at using natural language processing to extract clinically meaningful information from doctors' and other clinicians' free-text entries in Veterans' medical records. Free text is more difficult to "mine" than structured data—such as checkboxes or radio buttons—but researchers are using Google-like technology and other tools to teach computers how to recognize key phrases.



Questions about the R&D website? Email the Web Team

Any health information on this website is strictly for informational purposes and is not intended as medical advice. It should not be used to diagnose or treat any condition.