Haskayne researcher promotes patient safety and nurses’ well-being

Nurses in intensive care units may hear as many as 400 alarms a day, per patient. Because most are false alarms, there’s a risk they may stop responding to the warnings, jeopardizing patient safety. This “cry wolf effect” can lead to danger, or death, for patients — in at least one reported case, a patient died because some of their alarms had been deactivated.

“More than 90 per cent of the alarms can be false-positive alarms, meaning that there's an alarm and nurse goes and checks, but the patient is fine. There is nothing wrong,” says Dr. Hossein Piri, PhD, assistant professor at the Haskayne School of Business.

“This mistrust to the alarms and warning systems reduces the alarms' credibility and creates a ‘cry wolf effect’ among nurses.” 

In a study published in Operations Research, “Individualized Dynamic Patient Monitoring Under Alarm Fatigue,” Piri and his colleagues Dr. Woonghee Tim Huh and Dr. Steven Shechter of UBC Sauder School of Business and Dr. Darren Hudson, an ICU physician at the Department of Critical Care Medicine, University of Alberta, present an algorithm that solves alarm fatigue. The algorithm, embedded into smart alarm systems, personalizes the alarms for patients, making them more accurate and reducing the frequency of false alarms.

Patients coming into the ICU have different underlying conditions. For example, a healthy athletic person has a lower resting heartbeat than someone who smokes. If a person with a normal heart rate of 100 rises to 105, that’s “probably okay,” says Piri. “But if your heart rate is normally around 60 beats per minute and it suddenly goes to 100 beats per minute, then that is probably dangerous.”

As nurses collect more information about a patient, they feed it into the algorithm which then sets an upper and lower safety threshold for a particular patient’s blood pressure, respiratory rate, heart rate, and other physiological indicators. “If these physiological variables go beyond the range, then an alarm should sound,” says Piri. 

Before sounding an alarm the algorithm also considers the “cry wolf feedback loop,” recognizing that repeated false alarms have a risk to patient safety.

“If the alarm turns out to be false, it has a risk of creating mistrust among clinicians and in the future there is a possibility that they may not be responsive to the alarm, in which case a patient might be in a real danger.”

The researchers used data from ICU patients at the University of Alberta Hospital and Piri shadowed nurses on ICU units at St. Paul Hospital in Vancouver to watch how they respond to alarms. “The number of harmful events that a patient's going to experience is going to be much lower with our proposed algorithm,” he says.  

Reducing the number of alarms in an ICU has the added benefit of creating a healthier workspace for nurses. “When there are fewer alarms, there is less noise,” he says.

“In addition to patient safety, the other contribution of our paper is that it is providing a more ergonomic workspace for nurses.”