Skip to main content

StrokeClassifier: An Automated, EHR-based Ischemic Stroke Etiology Classification System

Yale Innovation Summit
05/3106/01/2023

StrokeClassifier: An Automated, EHR-based Ischemic Stroke Etiology Classification System

There are nearly 676,000 ischemic stroke events per year in the United States. One-quarter of these strokes are recurrent and each stroke can be associated with significant morbidity and mortality. Unlike heart attacks which are almost categorically due to atherosclerosis, there is a wide variety of causes of a patient's ischemic stroke. Evidence-based therapies can target specific sources of stroke that a patient may have to reduce the risk of recurrent stroke. However, these therapies are currently underutilized in the U.S. due to factors such as diagnostic uncertainty. Nearly 40% of ischemic strokes are diagnosed as cryptogenic. In order to converge onto the cause of stroke, a clinician needs to decipher a large amount of data collected during the encounter. Only one out of 6 ischemic stroke patients are treated by a trained vascular neurologist in the U.S. Our clinical decision support tool is a solution that takes the input of data gathered in the electronic health record during the stroke hospitalization and outputs the likely stroke etiology and guideline-recommended therapy for an individual patient.