This recent article on machine learning in digital epidemiology, published in Infection Control and Hospital Epidemiology got me thinking of an article we published nearly a decade ago. Our article on a clinical productive model for catheter related bloodstream infection prediction from the electronic medical record is available in open access format here. The publication admittedly lacks immediate clinical application, and like machine learning for infection prevention, it is not quite ready for prime time.
Machine learning in infection prevention could result in smart, individualized patient algorithms which allow for EMR based real-time central line use tracking, documentation of appropriate catheter dressing, CHG bathing and awareness of patient specific risk factors. This would then generate individualized patient centered alerts and automated central line stop-orders (requiring provider override) in the face of escalating bloodstream infection risk.
Individualized daily risk assessment, improved catheter maintenance and decreased catheter overuse likely means fewer catheter related bloodstream infections.
That would be cool.
