Too many alarms cause care providers to ignore real emergencies
Jul 14, 2017, 5:26 AM
(AP Photo/Gerry Broome)
PHOENIX — Call it ‘the emergency that cried wolf’ or simply alarm fatigue.
“The care providers are getting too many false alarms which makes them insensitive to the true alarms,” said Fatemeh Afghah, assistant professor at Northern Arizona University’s School Of Informatics, Computing, and Cyber Systems.
Current research shows staff can get hundreds of alarms per patient, per day, she said. As high as 90 percent of them don’t actually need immediate attention.
“The current mechanisms are designed to be too conservative to make sure that we’re not going to actually miss any true alarm,” she said.
That many alarms can worry patients and be cumbersome for doctors and nurses.
After an alarm was ignored when Afghah’s own one-month-old baby was in the hospital, she realized the process could be better, even though it was just a sensor on one of her baby’s toes that got disconnected.
“What if this was real?” she said. “They know that babies move a lot, and that’s more likely to be the cause, and they have a lot of patients.”
But how can they know if this is a true emergency?
The thought stayed in her mind for several years, until Afghah started working on a solution using multiple signals from the patients monitoring devices to determine if there is one true emergency.
“There are many categories of solutions already being worked on around the world,” she said. “Some are working on designing better sensors, so you can increase [their] accuracy.”
Others are working on designing a better connection with the patient, so the sensor doesn’t report every innocuous movement.
“What we’re doing is working on better signal processing,” she said. “We are working on a computer assisted algorithm that can combine these signals from different sensors.”
The collective signal would be analyzed by the computer algorithm to determine what it means.
“It’s going to look at the previous saved data sets,” she said. “Then compare whatever they have in the memory with what it’s detecting right now and then estimate if the alarm is real or not.”