Abstract:
|
Predicting the time to a clinical outcome for patients in intensive care units (ICU's) helps to support doctor's treatment decisions. The time to an event of interest could be, for example, survival time or time to recovery from a disease within the ICU. We developed methodology that extends current survival analysis models and improves the time-to-event prediction accuracy by systematically identifying similar patients based on relevant patient information and making personalized predictions. Both time-fixed and time-dependent covariates can be incorporated in our proposed model, and censoring and competing risks are considered as well. Our method is validated in the Multi-Parameter Intelligent Monitoring for Intensive Care (MIMIC-III) database, and we will also present properties of our methodology through a comprehensive simulation study.
|