Fact checked byShenaz Bagha

Read more

January 19, 2023
1 min read
Save

Public records assist model’s prediction of suicide risk after psychiatric hospitalization

Fact checked byShenaz Bagha
You've successfully added to your alerts. You will receive an email when new content is published.

Click Here to Manage Email Alerts

We were unable to process your request. Please try again later. If you continue to have this issue please contact customerservice@slackinc.com.

Clinical notes and public records were shown to improve a machine-learning model’s prediction of suicide after psychiatric hospitalization, researchers wrote in JAMA Psychiatry.

“The months after psychiatric hospital discharge are a time of high risk for suicide. A previously developed machine learning model showed that post-discharge suicides can be predicted from electronic health records and geospatial data, but it is unknown if prediction could be improved by adding additional information,” Ronald C. Kessler, PhD, of the department of health care policy at Harvard Medical School, and colleagues wrote.

Source: Adobe Stock.
Clinical notes and public records were shown to improve a machine-learning model’s prediction of suicide after psychiatric hospitalization. Source: Adobe Stock

Kessler and colleagues sought to determine whether model prediction could be improved through the addition of information extracted from clinical notes and public records.

The machine-learning models were trained to predict suicides in the 12 months after short-term psychiatric hospitalizations between the beginning of 2010 and Sept. 1, 2012. They were then tested in hospitals from Sept. 2, 2012, through Dec. 31, 2013.

The model assessed a total of 448,788 hospitalizations. Net benefit over risk horizons was highest for the model that included both clinical notes and public records. In addition, notes and record predictors also had the highest predictor class-level Shapley additive explanations values (values = 64% and 49.3%, respectively).

“The model had positive net benefit over 3-month to 12-month risk horizons for plausible decision thresholds,” Kessler and colleagues wrote. “Although caution is needed in inferring causality based on predictor importance, several key predictors have potential intervention implications that should be investigated in future studies.”