IC3 uses electronic health records to help predict patient health status

2022 IC3 Highlight

In a Frontiers in Digital Health article in November, Intelligent Critical Care Center (IC3) researchers—including assistant director for AI research Benjamin Shickel, associate director for research Tezcan Ozrazgat-Baslanti, and IC3 co-directors Azra Bihorac and Parisa Rashidi—published the results of their study to examine how transformer modeling can predict ICU patient acuity through electronic health record (EHR) tokenization. For EHRs, “tokenization” involves replacing sensitive patient data with a non-sensitive identifier or “token.”

As described in the article “Multi-dimensional patient acuity estimation with longitudinal EHR tokenization and flexible transformer networks,” the clinical AI framework created with EHR tokenization was more accurate compared to baseline machine learning models for predicting six mortality and readmission outcomes.

Read the Frontiers in Digital Health article here.