Extubation Decision Making with Predictive Information for Mechanically Ventilated Patients in ICU

Abstract

Weaning patients from mechanical ventilators is a critical decision in intensive care units (ICUs), significantly affecting patient outcomes and the throughput of ICUs. In this study, we aim to improve the current extubation protocols by incorporating predictive information on patient health conditions. We develop a discrete-time, finite-horizon Markov decision process (MDP) with predictions on future information to support the extubation decision. We characterize the structure of the optimal policy and provide important insights into how predictive information can lead to different decision protocols. We prove that adding predictive information is always beneficial, even if the physicians overtrust the predictions as long as the prediction accuracy satisfies certain conditions. Using a comprehensive dataset from an ICU in a tertiary hospital in Singapore, we compare the performance of different policies and demonstrate that incorporating predictive information can reduce ICU length of stay (LOS) by up to 9.4% and, simultaneously, decrease the extubation failure rate by up to 18.9%. The benefits are more significant for patients with poor initial conditions at ICU admission. Furthermore, simply optimizing LOS using a classical MDP model without incorporating predictive information leads to an increased extubation failure rate by up to 6%. Both our analytical and numerical findings suggest that predictive information is most useful in identifying patients who can benefit from continued intubation to execute personalized and delayed extubation.

Keywords: Medical Decision Making; Predictive Information; Markov Decision Process; Intensive Care Unit; Mechanical Ventilation