Friday, September 9, 2016

Electronic Health Records Help Predict Suicidal Behavior


The extensive databases in Denmark, Sweden, and Finland have long been mined by researchers looking for the patterns in health and illness that can improve care.

Now a study of more than 1.7 million Boston-area patients suggests the value of long-term health histories embedded in electronic health records in the United States for predicting risk of suicidal behavior.

Researchers used data on 16,588 case subjects from Massachusetts General Hospital and Brigham and Women’s Hospital. Their model predicted suicidal behavior with a 33 to 45 percent sensitivity and 90 to 95 percent specificity three to four years in advance, wrote Yuval Barak-Corren, M.S., of the Predictive Medicine Group at Boston Children’s Hospital Informatics Program, and the Israeli Institute of Technology in Haifa, Israel, in a study posted today in the AJP in Advance.

The study was released during National Suicide Prevention Week, whose goal is to promote the understanding and prevention of suicide and support those who have been affected by it.

Mental illness including substance abuse was associated with higher risk, as expected, but so were a number of infections and injuries.

This model is not “a specific quantitative prediction of suicide risk,” wrote Barak-Corren and colleagues. “Rather, we envision an alert system by which patients exceeding thresholds of predicted risk could be flagged as at relatively higher risk to encourage clinicians to conduct more targeted assessments of suicide risk.”

For more in Psychiatric News on prediction of suicidality, see “Two-Part Assessment May Help Predict Suicidal Behavior, Study Finds.”

(Image: pandpstock/iStockphoto)

Disclaimer

The content of Psychiatric News does not necessarily reflect the views of APA or the editors. Unless so stated, neither Psychiatric News nor APA guarantees, warrants, or endorses information or advertising in this newspaper. Clinical opinions are not peer reviewed and thus should be independently verified.