Reduction of false alarms in the intensive care unit using an optimized machine learning based approach

Wan-Tai M Au-Yeung,Eric M Isselbacher,Ashish K Sahani,Antonis A Armoundas

doi:10.1038/s41746-019-0160-7

Wan-Tai M Au-Yeung, Eric M Isselbacher + Show 2 more

Open Access

https://doi.org/10.1038/s41746-019-0160-7

Copy DOI

Abstract

This work attempts to reduce the number of false alarms generated by bedside monitors in the intensive care unit (ICU), as a majority of current alarms are false. In this study, we applied methods that can be categorized into three stages: signal processing, feature extraction, and optimized machine learning. At the stage of signal processing, we ensured that the heartbeats were properly annotated. During feature extraction, besides extracting features that are relevant to the arrhythmic alarms, we also extracted a set of signal quality indices (SQIs), which we used to distinguish noise/artifact from normal physiological signals. When applying a machine learning algorithm (Random Forest), we performed feature selection in order to reduce the complexity of the models and improve the efficiency of the algorithm. The dataset used is from Reducing False Arrhythmia Alarms in the ICU: the PhysioNet/Computing in Cardiology Challenge 2015. Using the performance metric “score” from the Challenge, we achieved a score of 83.08 in the real-time category on the hidden test set, which is the highest in all published work.

Highlights

In the intensive care unit (ICU), bedside monitors are used to alert healthcare providers when a patient’s physiological signals are out of normal range so that an appropriate response can be provided.In a prior study, it was discovered that 88.8% of annotated arrhythmia alarms were false positives.1 the majority of alarms do not require clinical intervention and, become a burden.1,2 Excessive numbers of false alarms cause noise disturbance, desensitization, and decreased quality of care, such that false alarms have often been listed as one of the top technology hazards.5,6Common sources of false alarms in the ICU are noisy physiologic signals that go out of range
Li et al.11 presented a framework for false alarm reduction using a machine learning approach that combined up to 114 signal quality and physiological features extracted from the ECG, photoplethysmograph (PPG), and, optionally, the blood pressure (BP) waveform
It can be observed that as the cost of false negatives (FN) increases, the true positive (TP) rate mostly increases while the true negative (TN) rate mostly decreases for all types of arrhythmia

Summary

Introduction

Common sources of false alarms in the ICU are noisy physiologic signals that go out of range. Li et al. presented a framework for false alarm reduction using a machine learning approach that combined up to 114 signal quality and physiological features extracted from the ECG, photoplethysmograph (PPG), and, optionally, the BP waveform. False alarm suppression rates were 86.4% for asystole, 100% for extreme bradycardia, 27.8% for extreme tachycardia, and 19.7% for ventricular tachycardia, with 0% true alarm suppression. These methods are promising, much improvement is still needed

Objectives

Results

Conclusion