Read more

February 02, 2022
2 min read
Save

Self-report scale, EHR data may improve suicide risk identification

Data from a brief patient self-report scale and electronic health records may help clinicians better identify patients at high suicide risk, according to results of a prognostic study published in JAMA Network Open.

“Recent studies suggest that applying machine learning (ML) methods to electronic health records (EHRs) can improve clinicians’ ability to identify patients at high risk for suicide,” Matthew K. Nock, PhD, of the department of psychology at Harvard University, and colleagues wrote. “However, critics note that such models have many more false positives than true positives and fail to detect meaningful proportions of the patients who go on to die by suicide.

electronic medical record
Source: Adobe Stock

“A separate line of research has suggested that patient self-reports and behavioral data obtained during clinical encounters may help improve clinicians’ ability to identify patients at high risk of suicide attempt,” they added. “Models that combine information across all these data sources might be the most effective at prediction.”

Nock and colleagues aimed to determine a method by which to best predict suicide attempts within 1 and 6 months of ED presentation for psychiatric conditions. They examined 1,818 patients (55.9% men; median age, 33 years) in this population who presented to an ED between Feb. 4, 2015, and March 13, 2017. They defined suicide attempts 1 month and 6 months following ED presentation using a combination of data from EHRs and patient 1-month (n = 1,102) and 6-month (1,220) follow-up surveys. Via ensemble machine learning, they developed predictive models and a risk score for suicide.

Results showed 137 of 1,102 (12.9%) patients with a suicide attempt within 1 month and 268 of 1,220 (22%) within 6 months. Clinicians’ assessment alone did not significantly outperform chance at suicide attempt prediction, with externally validated AUC of 0.67 for the 1-month model and 0.6 for the 6-month model. Models based on EHR data had somewhat higher prediction accuracy, which was best when using patient self-reports, particularly when combining patient self-reports with EHR and/or clinician data. Researchers noted similar performance for a model that used 20 patient self-report questions and an EHR-based risk score. A total of 30.7% of patients were classified as having highest risk attempted suicide within 1 month of their ED visit, which accounted for 64.8% of all 1-month attempts, in the best 1-month model. Further, a total of 46% of patients classified as having highest risk attempted suicide within 6 months of their ED visit, which accounted for 50.2% of all 6-month attempts, in the best 6-month model.

“The results of this prognostic study suggest that suicide risk assessments made using EHR-based and self-report–based risk scores may yield relatively accurate and clinically actionable predictions about the risk of suicide attempts by patients after presenting to an ED,” Nock and colleagues wrote. “These results highlight the need for tests of the implementation of such risk assessment tools to target preventive interventions.”