This appeared last week:
How 'cranky comments' can help spot CDS alert errors
Partners HealthCare researchers show how they mine override comments to detect areas where clinical decision support systems could be improved.
By Mike Miliard
December 31, 2018 01:47 PM
It's an all too common occurrence for clinical decision support alerts to misfire or otherwise malfunction. And that's a common frustration for clinicians. New research in the Journal of the American Medical Informatics Association shows how data scientists at Partners HealthCare are making use of provider feedback to spotlight and fix CDS errors.
WHY IT MATTERS
Rule-based alerts in the electronic health record could help notify physicians of risks such as adverse drug interactions or nudge them to add relevant information to the EHR.
Rule-based alerts in the electronic health record could help notify physicians of risks such as adverse drug interactions or nudge them to add relevant information to the EHR.
But such alerts can also malfunction, break, or might be triggered for the wrong reasons or patient populations. That could cause alert fatigue or pose safety risks.
The variety and types of these misfires can vary widely - as can clinicians' responses to inaccurate CDS alerts.
"In prior work, we developed a taxonomy of CDS alert malfunctions based on 68 cases of alert malfunctions collected from healthcare institutions in the United States," wrote Partners HealthCare researchers in JAMIA.
They also researched the ways in which those errors were found.
"Although the most common discovery method in the sample was user reports via traditional channels such as the help desk, safety reports, and contacting CDS maintenance personnel, five malfunctions were discovered by reading free-text override reasons written by providers when overriding alerts," they said.
"These malfunctions were discovered at our own institution, by chance, when a student researcher was reviewing alert overrides and noticed that some comments seemed to express strong frustration and disagreement with the alerts," researchers explained.
The researchers pointed to several instances where clinicians made note of the inaccuracies - often using colorful and indignant language:
- "BPA misfiring: no potassium on file, but there was a K done!"
- "Inappropriate warning as K is 4.3"
- "He is on beta blocker!"
- "You are stupid"
- "This is an inappropriate rec"
- "cyclosporine is eye drops!"
More here:
All software products should have an instant feedback mechanism for ideas. Comments etc.
The ADHA especially needs such feedback!
David.
No comments:
Post a Comment