Browse Teach Curate

Machine Learning Prediction of ICU Readmission Risk in Intracerebral Hemorrhage Patients

Using cohorts extracted from the MIMIC-III and MIMIC-IV critical-care databases, clinical, laboratory, and demographic features can be combined with imputation and class-balancing preprocessing to train artificial neural network, XGBoost, and random forest classifiers that estimate a hemorrhagic-stroke patient's risk of ICU readmission. Comparing these models on AUROC, sensitivity, and specificity yields a decision-support framework and identifies the variables most predictive of readmission.

Dr. Theopolis Text

Machine Learning-Based Prediction of ICU Readmissions in Intracerebral Hemorrhage Patients

Builds machine-learning models to predict whether a patient with intracerebral hemorrhage (bleeding within the brain) will be readmitted to the ICU, a marker of clinical severity and resource use. Co…