Boston Children's Hospital, in collaboration with MITRE, deployed machine learning algorithms in its pediatric cardiac catheterization lab that predict the risk of patient harm before procedures, allowing clinical staff to assess risk levels and make pre-procedure decisions. The framework, called GRACE (Generating Risk Reduction Analytics for Complex Cardiac Care Environments), was used the day before catheter procedures.

The GRACE/GR2AC3E initiative applied supervised machine learning techniques to electronic health record data from patients undergoing congenital cardiac catheterization at Boston Children's Hospital, using data from 2019 to 2020. The models analyzed patient-level, procedural-level, and system-level characteristics to predict clinically meaningful adverse events. A published peer-reviewed study (JSCAI, December 2024) described the methodology and performance of LASSO and Random Forest models. The framework enables staff to classify patients as high-, medium-, or low-risk the day before a procedure and plan accordingly.