“Informed estimates claim that 80% to 99% of alarms set off in hospital units are false or clinically insignificant, representing a cacophony of sounds that do not present a real danger to patients.”

So begins a recent study published at JMIR which pointed to AI algorithms as a potential solution to the widespread problem of “alarm fatigue.”

There’s no small amount of irony connected to the phenomenon, as automatic alarms are supposed to be a means of expediting necessary, if not critical, care for patients, but the reality, the report notes, is that given the number of false alarms, “(s)ensory overload is very likely to produce an unsafe environment for patients. The desensitization of a health care team to alerts can lead to longer response times for handling anomalies as well as possibly missing life-threatening events.”

They also noted that one of the primary causes of alarm fatigue is the poor design of alarm-generating algorithms. In response, the researchers developed an AI algorithm that aims to reduce the total number of alarms delivered to providers by incorporating a notification delay approach to decide whether to deliver a unique notification to caregivers rather than several alarms for the same alarm situation.

Moreover, the algorithm decides whether to add a false alarm probability (FAP) label to the notification, and whom to notify within the group of caregivers.

For the study, the researchers applied their algorithm to patient monitoring data and vital signs recorded during 32 surgical cases where patients underwent anesthesia, and they their “algorithm by using data  . . . selected from 3 out of the 32 surgical cases in the dataset.”

The experiment demonstrated “that providing a reasoning system can reduce the notifications received by the caregivers by up to 99.3% of the total alarms generated.”

Moving forward, the researchers plan to develop algorithms that will determine whom to notify within a group of caregivers, taking into account members’ specialization level, degree of experience, availability, geolocation, and current workload conditions.

Researchers also noted that in the current project they evaluated the algorithm in an experimental environment, so any future work will focus on testing the algorithm using more realistic clinical conditions by increasing the number of patients, monitoring parameters, and types of alarm.

In short, by refining the AI algorithm, researchers expect to be able to reduce alarm fatigue and the negative consequences that accompany it.

“Nearly all studies assume that a reduction in the number of total alarms and/or false alarms will reduce alarm fatigue. Thus, by presenting these results, we expect that our algorithm can be used as a useful strategy for avoiding alert fatigue,” they said.