Abstract
Objective: Clinical notes contain information not present elsewhere,including drug response and symptoms, all of which are highly important whenpredicting key outcomes in acute care patients. We propose the automaticannotation of phenotypes from clinical notes as a method to capture essentialinformation, which is complementary to typically used vital signs andlaboratory test results, to predict outcomes in the Intensive Care Unit (ICU). Methods: We develop a novel phenotype annotation model to annotate phenotypicfeatures of patients which are then used as input features of predictive modelsto predict ICU patient outcomes. We demonstrate and validate our approachconducting experiments on three ICU prediction tasks including in-hospitalmortality, physiological decompensation and length of stay for over 24,000patients by using MIMIC-III dataset. Results: The predictive models incorporating phenotypic information achieve0.845 (AUC-ROC) to predict in-hospital mortality, 0.839 (AUC-ROC) forphysiological decompensation and 0.430 (Kappa) for length of stay, all of whichconsistently outperform the baseline models leveraging only vital signs andlaboratory test results. Moreover, we conduct a thorough interpretabilitystudy, showing that phenotypes provide valuable insights at the patient andcohort levels. Conclusion: The proposed approach demonstrates phenotypic informationcomplements traditionally used vital signs and laboratory test results,improving significantly forecast of outcomes in the ICU.