Automated detection and prediction of atrial fibrillation during sepsis

7. 摘要/项目总结

7. ABSTRACT / PROJECT SUMMARY We propose the “Automated detection and prediction of atrial fibrillation during sepsis” study to develop automated technologies capable of accurate atrial fibrillation (AF) detection and prediction during sepsis. Sepsis is a life-threatening, dysregulated response to infection and the most common illness leading to hospitalization in the United States, affecting ~1 million Americans yearly, and is associated with 50% of all hospital deaths. With the exception early antibiotic and fluid use, few therapies improve outcomes among septic patients; new treatment strategies are greatly needed to improve survival. New-onset AF is a common dysrhythmia among critically ill patients with sepsis, affecting up to 1 in 3 septic patients and conferring increased short- and long-term risks stroke, heart failure, and death. Prevention of AF or its complications may improve sepsis outcomes by reducing AF-related morbidity and mortality. Although several evidence-based treatments have shown efficacy in treating and preventing AF in certain high-risk subgroups (e.g., AF prevention following cardiac surgery), studying application of these therapies among critically ill patients with sepsis has been hampered by two major factors: 1) we lack validated automated mechanisms to detect AF and facilitate real-world AF research in large clinical databases, and 2) we cannot presently predict which patients with sepsis will develop AF. Our project will leverage the unique resources of the recently released Multiparameter Intelligent Monitoring in Intensive Care (MIMIC III) database. MIMIC III links continuous ECG and pulse plethysmographic waveforms to a wealth of time-varying clinical and hemodynamic data. Our project will develop and validate state-of-the art automated AF detection algorithms using waveform data from critically ill patients. Automated AF detection would enable expedited clinical treatment of AF, identification of subclinical AF, and will catalyze the study of AF in emerging electronic health record waveform databases. We will develop innovative automated AF prediction capabilities using state-of-the-art waveform analysis algorithms and machine learning methods in critically ill patients. Automated algorithms that identify patients at high risk for developing AF in the near-term would enable targeting of preventative therapies and potentially usher in a new era of AF prevention for critically ill patients. AF prevention and treatment facilitated through our project will allow targeting of novel, AF-based mechanisms of poor outcomes during and following sepsis.