A new study analyzing the health records of 1.3 million patients of the Veterans Health Administration has developed a method to fill gaps in detecting a history of self-harm. History of self-harm is the best predictor of future self-harm and suicide risk, but clinicians’ dependence on diagnosis codes often comes up short in finding this important information in years of medical records.

New Method Developed

The new method, called Positive Unlabeled Learning Selected Not At Random (PULSNAR), ‘learns’ from patients with a diagnostic code and estimates how many similar patients might be among those without one. This approach can help flag patterns for review, potentially identifying patients who may need closer attention.

The research team has also published a related study using this approach to detect under-coded opioid use disorder and plans to apply the method to detect PTSD, depression, bipolar disorder, and sleep disorders.

Original reporting: KOAT Albuquerque — read the source article.