Machine Learning of Physiological Waveforms and Electronic Health Record Data to Predict, Diagnose, and Treat Hemodynamic Instability in Surgical Patients

项目摘要/摘要

Project Summary / Abstract If one could accurately predict who, when and why patients develop cardiorespiratory instability (CRI) during surgery, then effective preemptive treatments could be given to improve postoperative outcome and more effectively use healthcare resources. But signs of shock often occur late once organ injury is already present. The goal of this proposal is to develop, validate, and test real-time intraoperative risk prediction tools based on electronic health record (EHR) data and high-fidelity physiological waveforms to predict CRI and make the databases of intraoperative data and waveforms used for these developments freely accessible. This is extremely relevant because although 5.7 million Americans are admitted to an Intensive Care Units (ICU) in one year, more than 42 millions undergo surgery annually. Previous and ongoing studies conducted in the ICU and in the step down unit have built the architecture to collect real-time high-fidelity physiological waveform data streams and integrate them with patient demographics from the EHR to build large data sets, and derive actionable fused parameters based on machine learning (ML) analytics as well as display information in real time at the bedside to drive clinical decision support (CDS) in the critical care setting. The goal of this proposal is to apply these ML approaches to the complex and time compressed environment of high-risk surgery where greater patient and disease variability exist and shorter period of time is available to deliver truly personalized medicine approaches. The work will be initiated using an already existing annotated intraoperative database from the University of California Irvine including EHR and high-fidelity waveform data. This operating room database already exists and needs only to be extracted. This data will be used for the initial training and development of the ML model that will then be tested on prospectively collected University of California Los Angeles and University of Pittsburgh Medical Center databases. Simultaneously, this approach will use existing knowledge of CRI patterns derived from previous step down unit / intensive care unit cohorts, MIMIC II data, University of California Irvine data, and animal studies to create smart alarms and graphic user interface for clinical decision support based on functional hemodynamic monitoring principles. The next step will then leverage the focus on the issues and strengths of the intraoperative environment, some of which can be listed as: 1) Known patients characteristics before surgery to define pre-stress baseline, allowing functional hemodynamic monitoring stress evaluations, preconditioning, and other preoperative calibrations, 2) High degree of direct observation and data density during most phases of surgery allowing close semi-autonomous monitoring and titration of novel treatment algorithms early, 3) Defined stages in the initial part of surgery (induction, intubation, skin incision) allowing ML approaches to build large common relational database registries, and 4) Defined surgical procedure and stressors (anesthesia induction, intra-abdominal air insufflation, and other surgery-specific interventions), which will alter the impact of CRI on measured variables.