CRII: SCH: A Computational Framework to False Alarm Suppression in Intensive Care Units

Northern Arizona University, Flagstaff AZ

Investigators

Abstract

False alarms are widely considered the number one hazard imposed by the use of medical technologies. The Emergency Care Research Institute named alarm hazards as the number one of the 'Top 10 Health Technology Hazards' for 2012, 2013 and 2015. Healthcare providers are usually overwhelmed with 350 alarm conditions per patient per day, of which 80-99% are meaningless or false. These false alarms can be due to several factors such as patient movement, malfunction of individual sensors and imperfections in the patient-equipment contact, resulting in alarm fatigue among healthcare providers and the possibility of missing a true life-threatening event lost in a cacophony of multiple alarms. These false alarms can also cause patient anxiety, inferior sleep structure and depressed immune systems. Thereby, alarm safety has been determined as a national patient safety goal by The Joint Commission, which accredits and certifies nearly 21,000 health care organizations and programs in the United States. This project will develop a multifaceted framework to reduce the false alarm rate in Intensive Care Units (ICUs) by integrating principles from information theory, game theory, graph theory and signal processing. The alarms in ICUs are mostly created based on the measurements made by individual machine/monitors, while the majority of the alarms produced by these individual machines are considered false. The majority of current methods to suppress the false alarm rate attempt to design new monitors or create more accurate sensors. These methods are often tailored to specific devices or datasets and the significant intrinsic correlations among the extracted features from different sensors are overlooked in these methods. A real-time and accurate yet general method will be developed to reduce the number of false alarms while avoiding the suppression of true alarms through integrating information from a variety of devices and considering the non-linear correlations and mutual information among the features collected from these devices using a new game theoretic approach. The performance of this proposed method will be evaluated using PhysionNet's publicly available MIMIC II dataset considering the three vital signals of ECG, PLETH, and APB. The proposed false alarm detection method can potentially save many patients' lives and significantly reduce medical costs. This work can advance the research program and practice of teaching at the newly established Northern Arizona University School of Informatics, Computing and Cyber System (SICCS) by developing a new course in game theoretical optimizations, integration of this research in several undergraduate-level course modules and training of graduate students.