Hum-Dinger: A revolutionary adaptive alarm management system.

Observation:  Alarm fatigue is a major problem in hospitals. Hospitals use physiological monitoring systems on their sickest patients alerting healthcare professionals to patients whose health is deteriorating.

hum-dinger-final-pitchProblem:  Unfortunately, from 70% to 99% of alarms, depending on the hospital unit, do not signal any important clinical change and can be safely disregarded. When faced with these high false alarm-rates, healthcare professionals develop alarm fatigue as they are desensitized, leading to a failure of timely response to appropriate alarms.

Need:  A way to address alarm fatigue in nurses and physicians that increases the percentage of actionable alarms while improving workflow and decreasing adverse patient outcomes. In 2013, the Joint Commission made Alarm Safety a National Patient Safety Goal. As of January 1, 2016 hospitals are required to develop and implement specific policies and procedures to address Alarm Safety

Hum-Dinger aims to manage alarms by providing suggestions to medical care practitioners to adjust alarm settings on monitoring devices and other actions so that alarm fatigue is minimized. We use an adaptive aiding approach where adaption is based on machine learning. The product integrates continuous monitor data from distributed monitors and clinical data from the EHR database using the HL7 protocol. The actions are taken (including dismissal) of alarms are the machine learning training labels. Hum-Dinger exploits and drives stronger man-machine symbiosis instead of an antagonistic relationship

Machine Learning is the key in-the-loop enabling component which is used in an adaptive aiding framework. It learns from actions taken (including dismissal) of alarms along with inputs which are continuous monitor data and electronic health record data. The product frames alarm management as an optimization that includes the human factors of attention. This concept is the basis for the development along with EHR and monitor data. A smartphone app for medical care practitioners is the adaptive aiding interface that suggests actions and also records actions taken.