ICP Elevation Alerting Based on a Predictive Model Hosting Platform

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): Recurring acute ICP elevations occur frequently and unpredictably among severe brain injury patients. ICP elevation can cause cerebral ischemia and lead to deadly brain herniation if untreated. Hence, prompt recognition and treatment of rising ICP are critical in managing severe brain injury patients. However, existing protocols in most neurocritical care units are reactive where bedside nurses, in response to simple threshold-crossing alarms, have to check numerical display of ICP on monitors to manually establish whether the alarm is a true one before initiating treatment. Acute ICP elevation is accompanied by distinctive ICP pulse morphological changes. By utilizing ICP pulse morphological metrics as input, we can accurately recognize precursors to ICP elevation to alert nurses and free them from a cognitively demanding process of establishing whether a consistent ICP elevation triggers the alarm. We therefore propose to deploy a previously developed accurate ICP elevation prediction model on an open-source model hosting platform to monitor continuous ICP signals and alert bedside nurses. Using this alerting system, we will further investigate the principal physiological abnormalities associated with acute ICP elevation showing different precursory ICP patterns prior to onset of elevation. We will pursue the following three aims: 1) To develop an alerting system for ICP elevation based on a model hosting platform; 2) To investigate whether the ICP alerting system helps nurses more efficiently manage ICP. 3) To detect consistent physiological abnormalities associated with acute ICP elevation. Our long-term goal is to advance intensive care monitoring so that continuous signals from monitors are fully explored to integrate with the rest of clinical data in an electronic medical record (EMR) system to enhance clinical decision making. This project represents an effort piloting a platform-based approach towards overcoming translational barriers that impede the process of making advanced predictive analytics available at point of care. Therefore, broad impacts from this project are related to future efforts at leveraging this open model hosting platform to facilitate the translation of additional predictive models in other ICUs.