Monday, November 27 | 9:40 a.m.-9:50 a.m. | M3-SSIN02-2 | Room S404
If you have considered that a “STAT” designation on imaging orders could be overutilized, misused, or obscure the urgency of the order, you will want to hear how an automated system has been designed to prioritize some STAT orders to the top of the queue.
During this Monday session presented by Renaid Kim, attendees will see how machine-learning models classified STAT imaging orders based on the likelihood of medical or surgical intervention required within 24 hours if positive. Out of 500 outpatient STAT imaging studies, a process of order exclusion using free-text clinical indications left 351 imaging studies for evaluation.
For each of the imaging studies, three different radiologists evaluated the indication and assigned an integer score between 1 (most likely) and 5 (least likely), with a simple majority determining the final score. A fourth radiologist adjudicated studies with interrater score difference ≥ 2.
The project tested Lazy Predict’s library of classifiers to demonstrate that the AdaBoost Classifier model achieved the best performance on the test set with the area under the receiver operating characteristic curve of 0.778 and an F1 score of 0.773.
Kim et al concluded that machine-learning algorithms can use free text clinical indications to identify studies that should be prioritized to prevent delays in lifesaving interventions. Stop by this session to learn more.
