Machine learning derived notifications for impending episodes of hemodynamic instability (HD) and respiratory failure (RF) events are interesting because they can alert clinicians in time to intervene before these complications occur.
Do machine learning alerts, telemedicine system generated alerts or biomedical monitors have superior performance for predicting episodes of intubation or administration of vasopressors?
A machine learning (ML) algorithm was trained to predict intubation and vasopressor initiation events among critically ill adults. Its performance was compared to biomedical monitor alarms and telemedicine system alerts.
Machine learning (ML) notifications were substantially more accurate, precise, with 50-fold lower alarm burden than telemedicine system (TS) alerts for predicting vasopressor initiation and intubation events. ML notifications of internal validation cohorts demonstrated similar performance for independent academic medical center external validation and COVID-19 cohorts. Characteristics were also measured for a control group of recent patients that validated event detection methods and compared TS alert and (BM) biomedical monitor alarm performance. The TS test characteristics were substantially better, with 10-fold less alarm burden, than BM alarms. The accuracy of ML alerts (0.87-0.94) was in the range of other clinically actionable tests; the accuracy of TS (0.28-0.53) and BM (0.019-0.028) alerts were not. Overall test performance (F-scores) for ML notifications were more than 5-fold higher than for TS alerts which were higher than those of BM alarms.
Machine learning derived notifications for clinically actioned HD and RF events represent an advance because the magnitude of the differences of accuracy, precision, misclassification rate, and pre-event lead time are large enough to allow more proactive care and have markedly lower frequency and interruption of bedside clinician work flows.

Copyright © 2023. Published by Elsevier Inc.