Abstract 11370: Track and Trigger System Based on Patient Analytics for Cardiac Arrest Risk Identification

Abstract

Introduction: Cardiac Arrest (CA) is a serious, yet not infrequent complication in the cardiacvascular intensive care unit (CVICU). A recent quality improvement project, Cardiac Arrest Prevention (CAP), showed a decrease in CA by applying a bundle to pre-identified patients at higher risk. Despite this success, CA still occurs among “low” risk patients. Automated risk analytics can offer patient-specific tools to identify risk periods in real time. Hypothesis: Methods: Time series data including vitals and laboratory values were collected (6/2019 - 8/2020). These data were fed into a risk analytics algorithm [Etiometry Inc., Boston, MA] which uses a model of patient physiology to compute risk indices for various physiologic variables. Risk indices examined in this study include IDO2 (likelihood of SvO2 < 40%), HLA (likelihood of lactate > 4 mmol/L), and IVCO2 (likelihood of PaCO2 > 50 mmHg). Various criteria using these risk indices were retrospectively evaluated for categorizing risk status. Once criteria were met a 12-hour “watcher period” was established after a 1-hour assumed response period (figure 1A). CAP status was assigned on a daily basis. Results: 665 encounters for 484 patients were included. 128 (19.2%) neonatal (<28days) encounters and 537 (80.8%) older patients, for a median age of 7.7 (IQR 2, 46) months. There were a total of 29 CA, of which 11 (38%) were captured by CAP. Figures 1B and 1C show two versions of Track and Trigger Systems (TTS) with their capture of CA events compared against CAP and when combined with CAP. TTSv1 and TTSv2 alone were able to capture 31% and 41% of the CA events. When combined with CAP these numbers increase to 55% and 62% respectively. Conclusions: A TTS based on risk indices could provide a patient/pathophysiologic status specific tool to increase our ability to identify patients at risk of CA when used in conjunction with CAP criteria. Further validation of these data in other cohorts and prospective application of the tool is warranted.