And Then You're Dead

The Autopsy / Case File · Medical Error & Preventable Care

Alarm Fatigue

A hospital room full of monitors is supposed to be one of the safest places a sick body can be — every vital sign watched, every dangerous change announced the instant it happens. What that same room actually produces, hundreds of times a day, is noise. And a nervous system trained to tune out noise doesn't always catch the one alarm that mattered.

A widely cited 2012 review of the research found that a single ICU patient can trigger more than 700 monitor alarms in a 24-hour period — and that 80 to 95 percent of those alarms don't reflect anything requiring clinical action at all. The Joint Commission's own sentinel-event database recorded 98 alarm-related incidents between January 2009 and June 2012; 80 of them ended in a patient's death. The FDA has separately received reports tying 566 hospital deaths between 2005 and 2008 to alarm-related device failures or missed responses. More than a decade after the Joint Commission first raised a formal national alarm about it, alarm-related hazards are still turning up on the country's top patient-safety watchlists.

And Then You're Dead · September 2026 · 9 min read

Join the discussion →
Close-up of a hospital vital signs monitor displaying heart rate, oxygen saturation, and blood pressure readings, with a red tachycardia alarm banner across the top
A vital signs monitor displays a tachycardia alert during a training scenario at Naval Hospital Pensacola. Multiply this single alarm by every device on every patient on a unit, and the number of alerts a single clinician hears in a shift climbs into the hundreds.

The term "alarm fatigue" describes a specific, well-documented failure of human attention under a specific kind of load: when a warning sound goes off constantly and almost never means anything urgent, the people responsible for responding to it — reasonably, almost inevitably — start responding more slowly, or muting it, or tuning it out entirely. It isn't a failure of diligence. It's what happens to any nervous system exposed to hundreds of false alarms a day, in a profession where missing the one real alarm can be fatal.

Eighty to Ninety-Five Percent Noise

The scale of the underlying noise problem is what makes alarm fatigue close to unavoidable without a deliberate system fix. A 2012 integrative review published in Biomedical Instrumentation & Technology, synthesizing more than a decade of prior research, found that an ICU patient connected to standard monitoring equipment can generate upward of 700 alarms in a single day — and that somewhere between 80 and 95 percent of those alarms don't require any clinical intervention at all. Most are triggered by a patient rolling over, a sensor slipping, a brief and clinically meaningless dip picked up by an oversensitive default threshold. The same review helped establish alarm fatigue as, in ECRI Institute's words at the time, the single largest medical device technology hazard hospitals faced.

The multiplying effect across a real unit is what turns an individual patient's 700 alarms into an environment. A nurse covering four or five monitored patients on a busy shift isn't hearing one patient's alarms — she's hearing the combined, overlapping alarm load of every bed she's responsible for, on top of infusion pump alerts, bed alarms, and call lights. Under that volume, distinguishing the alarm that means a patient is actually decompensating from the alarm that means a pulse oximeter slipped off a finger becomes a genuinely difficult perceptual task, not a matter of paying closer attention.

0
monitor alarms a single ICU patient can generate in 24 hours, with 80-95% requiring no clinical action, per a 2012 integrative review of the research
0
of 98 alarm-related sentinel events reported to the Joint Commission between January 2009 and June 2012 resulted in the patient's death

Where the Deaths Get Counted

Two separate federal-adjacent tracking systems have each put a number on the toll, using different methodologies and different time windows, and both point the same direction. The FDA's device-adverse-event reporting system recorded reports tying 566 hospital deaths between 2005 and 2008 to alarm-related device issues — a voluntary reporting system, which means the true count is almost certainly higher, since underreporting is a well-documented problem across FDA device-safety data generally. Separately, the Joint Commission's own sentinel-event database — which captures the more serious incidents accredited hospitals are expected to report — logged 98 alarm-related events between January 2009 and June 2012. Eighty of those 98 resulted in a patient's death; 13 resulted in permanent loss of function; 5 resulted in an unexpected extended hospital stay or additional care. That is a strikingly high proportion of alarm-related incidents ending in the worst possible outcome, and it's exactly the pattern that pushed the Joint Commission to act.

Almost every alarm that goes off in a hospital is telling the truth about something. Most of the time, what it's telling the truth about doesn't matter. The system has no reliable way to tell a clinician, in the half-second before she decides whether to look up, which kind of alarm this one is.

A Decade on the Watchlist

In April 2013, citing its own sentinel-event data directly, the Joint Commission issued a formal Sentinel Event Alert on medical device alarm safety — a rare step reserved for hazards judged serious and systemic enough to warrant a national warning to every accredited hospital in the country. The following year, alarm system management became a National Patient Safety Goal, requiring accredited hospitals to establish alarm-management policies, identify the alarms most critical to manage, and set up mechanisms for staff education — a mandate, not a suggestion. ECRI Institute's independently published Top 10 Health Technology Hazards list had already ranked clinical alarm hazards as the single top hazard in the country in both 2012 and 2013, and continued to call it "the number one medical hazard" in its 2014 edition.

The ranking has moved since then — reflecting real progress from smarter default thresholds, alarm-management protocols, and central-station alarm routing that filters what reaches an individual clinician — but the underlying hazard hasn't disappeared from the list entirely. By 2020, alarm fatigue had fallen to fifth on ECRI's ranking, still inside the top ten more than seven years after the Joint Commission's original alert. ECRI's 2025 list, topped that year by risks from AI-enabled health technologies, still includes "dangerously low default alarm limits on anesthesia units" at number six — a narrower, more specific version of the same root problem the 2013 alert described: alarm thresholds set generically, for an average patient, rather than customized for the one actually in the bed, producing exactly the kind of nuisance alarm volume that trains a busy clinician to tune out the signal along with the noise.

What Actually Changes the Number

The fixes hospitals have adopted since the 2013 alert aren't exotic. Widening default alarm thresholds so a brief, clinically irrelevant dip doesn't trigger a full alert; customizing limits to the individual patient rather than leaving every monitor on a generic factory default; routing alarms through a secondary notification system, like a smartwatch or pager, that reaches the specific nurse assigned to that bed instead of broadcasting to an entire unit; and building in a short delay before a borderline reading escalates to a full alarm, so a sensor slip has a few seconds to correct itself before it counts. Individually, none of these interventions sounds dramatic. Together, hospitals that have implemented alarm-management programs following something close to this playbook have reported reductions in overall alarm volume in the range of 50 percent or more, without any corresponding rise in missed critical events — evidence that the original volume of alarms was mostly excess, not mostly necessary.

What hasn't changed as easily is the underlying economics of who bears the cost of getting it wrong in either direction. A hospital that sets alarm thresholds too loose risks the exact sentinel-event pattern the Joint Commission documented — a real deterioration lost in the noise. A hospital that sets them too tight, out of caution after a bad outcome, risks recreating the same noise problem it was trying to solve. Getting the balance right, unit by unit and patient by patient, takes ongoing clinical judgment and staff time that an alarm-management policy on paper doesn't supply by itself — which is part of why a hazard the Joint Commission formally flagged in 2013 was still showing up, in a narrower form, on ECRI's hazard list twelve years later.

And Then You're Dead

Every alarm in that room is doing its job — sounding, exactly as designed, the instant a number crosses a line someone programmed into it. The failure isn't in any single alarm. It's in the six hundred and ninety-nine that came before it, each one training a tired, overloaded nervous system to expect nothing.

That's it. That's the whole thing.

Comments

Loading…

Comments are anonymous and moderated.

Sources

Photo: U.S. Navy, "Vital signs monitor display," by Petty Officer 1st Class James Stenberg, via Wikimedia Commons (public domain).