Alert fatigue: when the alarm goes unheard

An ICU nurse starts the night shift with six monitored patients. In the first hour, the monitors sound dozens of times: a loose electrode, a patient who moved, a saturation that dropped and came back on its own. On the computer, every prescription she checks opens an alert window, almost always about interactions the team already knows. Near three in the morning, an alarm sounds at bed 4 and she finishes the note before going to check. This time, the alarm was real. The scene is fictional and was written for teaching.

This text is about alert fatigue, the phenomenon in which repeated exposure to frequent alarms and warnings, or ones that need no action, leads to a slower or absent response. The aim is to look at what studies show about electronic health record alerts and monitor alarms, and at what helps reduce the excess. If you play the case of the day on NEXO, you will recognize the challenge: amid many clues, reasoning has to separate what is noise from what changes management.

Key points

• Alert fatigue appears when an excess of barely relevant warnings leads to desensitization and a slower response.
• In several services, most medication alerts were dismissed by physicians.
• Repeated alerts for the same patient reduced the chance of acceptance.
• Dismissing alerts becomes a habit, and many were closed in under three seconds.
• In a Brazilian coronary care unit, on average more than 10 alarms sounded per hour.
• Reducing repetitions, adjusting thresholds and prioritizing important alerts appear as strategies in the studies.

What alert fatigue is

A scoping review of 32 publications found varied definitions, but most described alarm fatigue as a phenomenon in which repeated exposure to frequent alarms, or ones that need no action, leads to sensory overload, emotional strain and gradual desensitization of professionals (Michels et al., 2025). The result is a higher risk of delayed or inadequate response and of patient harm.

In the electronic health record, the concept is similar. Alert fatigue is a phenomenon in which insignificant warnings or recommendations from the decision support system appear in excess and professionals start dismissing them without considering their importance (Sutton et al., 2020). A study on habit adds that it results from an overload of alerts that leaves the clinician desensitized to the following alerts (Wang et al., 2022).

Two mechanisms

The authors of a study on clinical trial alerts describe two related factors (Embi and Leonard, 2012). The first is the irrelevance of the alerts and cognitive overload, also called alert overload. The second is the decline in response to a specific type of alert as the clinician is exposed to it repeatedly, gradually becoming desensitized.

A primary care study tested hypotheses linked to these two mechanisms: cognitive overload from the volume and complexity of work, and desensitization from repeated exposure to the same alert over time (Ancker et al., 2017). The results help separate what weighs more, as will be seen below.

Why the topic matters for clinical reasoning

Alerts and alarms are clues. Well-trained reasoning weighs each clue by what it adds, not by how it appears. When most clues change nothing, the brain learns to ignore them, and the one that matters goes along with them. The text on base rate in diagnosis explains why a rare sign loses value when it appears amid many false positives.

There is a direct parallel with diagnosis. When a screening test yields many false positives, those who read the results learn to distrust all of them, and the true positive starts to be received with the same distrust. Something similar happens with alerts: the problem stops being the machine and becomes the expectation it created in those who use it. That is why the solutions described below change both the system and the team's habit.

It is also worth remembering that fatigue sets in gradually. The decision to ignore alerts rarely appears on the first day of work. The response gets slower as each irrelevant warning teaches that the next one will probably also be irrelevant, and it is this silent learning that the studies below try to measure.

Alerts in the electronic record: the scale of the problem

The numbers on alert dismissal repeat across different countries and services. In a South Korean academic emergency department, 611 physicians handled 101,186 visits, and almost two thirds of the 102,887 medication alerts were dismissed (Yoo et al., 2020). The authors identified 21 significant risk factors for dismissal and note that this behavior is considered the main factor preventing the system from preventing adverse events.

In a US hospital, opioid allergy alerts, the most frequently triggered in the prescribing system, were dismissed at a growing rate, reaching 89.7 per 100 alerts in 2013 (Topaz et al., 2015). The rate was high even for severe reactions, whether immune-mediated or life-threatening. Alerts with an exact match between allergy and medication were dismissed less often than those matched only by group.

Why physicians dismiss alerts

One study gathered the reasons for dismissing drug interaction alerts at 10 sites in the United States, with different electronic records (Wright et al., 2019). The researchers found 177 distinct reasons, and the number of reasons offered at each site ranged from 3 to 100. Many sites offered reasons that did not even apply to interactions. Three categories accounted for 78 per 100 dismissals: will monitor or take precautions, no clinical importance, and benefit greater than risk (Wright et al., 2019).

In the opioid study, about one third of the reasons pointed to irrelevant alerts (Topaz et al., 2015). The authors argue for less interruptive interfaces and more precise alerts to reduce fatigue. A review on decision support follows the same line and notes that physicians with more experiential knowledge tend to use the system less and dismiss its alerts more (Sutton et al., 2020). The same review cites studies in which up to 95 per 100 alerts had no consequence, with physicians who often disagreed with them or distrusted them (Sutton et al., 2020).

When dismissal causes harm

Dismissal is not harmless. A study cited in the research on habit estimated that, in the United States alone, about 57.6 million alerts for adverse drug events were dismissed in 2014, of which about 5.5 million were dismissed inappropriately, with about 196,600 resulting adverse events (Wang et al., 2022). The alert that matters is usually among many that do not.

The review on decision support also notes that errors with drug interactions are common and preventable, and that up to 65 per 100 hospitalized patients are exposed to one or more potentially harmful combinations (Sutton et al., 2020). The problem, therefore, is separating the signal from the noise, not turning off the signal.

Repetition, complexity and habit

The primary care study found that, on average, a quarter of medication alerts and a third of clinical reminders a physician received were repeats for the same patient in the same year (Ancker et al., 2017). Alert acceptance was associated with work complexity and repetitions, but not with workload volume.

The chance of accepting a reminder fell 30 per 100 for each additional reminder per visit and 10 per 100 for each five-point increase in the proportion of repeated reminders (Ancker et al., 2017). The authors concluded that physicians became less likely to accept alerts as they received more of them, especially repeated ones, and that reducing repetitions for the same patient is a promising target.

The decline over time

A randomized trial tracked trial recruitment alerts in the medical record (Embi and Leonard, 2012). Response rates showed a significant downward trend over time, falling with each period, with p < 0.0001 (Embi and Leonard, 2012). The authors treated the finding as a form of alert fatigue and argued for adjusting the frequency and format of these notices (Embi and Leonard, 2012). The example is useful because a recruitment alert is almost never urgent for the patient in front of the doctor. Even so, it competes for the same attention as safety alerts, and a drop in response to one type of notice can spread to the others.

The habit of closing the window

A cohort study with 1152 physicians exposed to a specific alert in the medical record measured the strength of the dismissing habit (Wang and colleagues, 2022). Dismissing alerts increased the strength of this habit, and habit strength was associated with more dismissals after the first. About 72.5 out of every 100 alerts were dismissed less than 3 seconds after appearing, and 13.2 out of every 100 in less than 1 second (Wang and colleagues, 2022).

In less than 1 second, it is hard to read the alert. The authors cite field observations and interviews in which physicians developed the habit of dismissing alerts, with many closed without relevant cognitive consideration (Wang and colleagues, 2022). The text on diagnostic uncertainty discusses the value of pausing a few seconds before deciding.

ICU monitor alarms

A Brazilian observational study tracked a coronary care unit at a cardiology hospital with 170 beds (Bridi and colleagues, 2014). Over 40 hours of observation, 426 alarms were recorded: 227 came from multiparameter monitors and 199 from other equipment, such as infusion pumps, dialysis pumps, mechanical ventilators and intra-aortic balloons. The average was 10.6 alarms per hour (Bridi and colleagues, 2014).

The authors describe the mechanism in clear terms. When many alarms mask the clinically important ones, some end up disabled, silenced or ignored by the team, and the safety of critically ill patients is compromised (Bridi and colleagues, 2014). Disabling alarms, failing to program them or configure them according to the patient's condition, and setting the volume very low appear as research subjects in the same paper.

The study also recalls a paradox. Alarms from equipment made to protect the patient increased noise in the unit, distractions and work interruptions, and created a false sense of safety (Bridi and colleagues, 2014). The conclusion reinforces the importance of properly configuring the physiological variables, volume and parameters of monitor alarms in the ICU routine.

Customized monitors

To reduce alarms without necessary action, settings may need to be adjusted for each patient, but nurses often fail to do this because of competing demands and alert fatigue itself (Ruppel and colleagues, 2018). A clinical ICU tested alarm customization software and compared two months before and two months after implementation.

Medium-priority alarms for heart rate, respiratory rate and blood pressure fell significantly, and their duration also decreased (Ruppel and colleagues, 2018). Oxygen saturation alarms stayed the same. After the intervention, nurses reported spending less time on alarms without necessary action, and the proportion of those who said alarms got in the way of work and of those who had witnessed an important alarm being ignored fell (Ruppel and colleagues, 2018).

What nursing staff report

A systematic review on alarm fatigue in the ICU gathered studies with 389 nurses (Lewandowska et al., 2020). The professionals found the alarms annoying and too frequent, disrupting care and reducing trust in the system. The review cites studies in which false or clinically unimportant alarms reached 85 to 99 out of every 100 alarms.

Noise also weighs in. The same review cites a project in which the noise level in the ICU ranged from 47 to 77 decibels, with false alarms among the main sources, and notes that noise disrupts communication and increases staff stress (Lewandowska et al., 2020). Prolonged exposure to noise was identified as a factor in professional burnout among intensive care nurses.

Consequences for the patient

The scoping review linked alarm fatigue to delayed responses, communication failures and more stress and burnout among professionals (Michels et al., 2025). Among the direct consequences for the patient, the authors cite sleep disrupted by too many alarms. The text on interprofessional clinical reasoning shows how the team can share out monitoring instead of concentrating it on one person.

What helps reduce the excess

The scoping review lists strategies on four fronts: training programs, technical improvements such as customizing alarms, protocol adjustments and organizational interventions (Michels et al., 2025). The review on decision support proposes prioritizing and selecting the truly important alerts, those with the greatest impact, and adapting alerts to specialties and severity levels (Sutton et al., 2020).

Strategy	Example studied	Finding	Source
Reduce repetitions	Primary care	Repetitions for the same patient reduced acceptance	Ancker et al., 2017
Customize monitor alarms	Clinical ICU with software	Fewer heart rate and blood pressure alarms	Ruppel et al., 2018
Customized thresholds	Vital signs in pediatrics	Fewer heart rate alarms	Poole and Shah, 2018
Review dismissal comments	Alert rules in a hospital	62 broken rules found	Aaron et al., 2019
Review reasons for dismissal	10 sites with different records	177 distinct reasons, many irrelevant	Wright et al., 2019
Prioritize and customize	Decision support review	Alerts by specialty and severity	Sutton et al., 2020

Less interruptive alerts

The opioid study argues for interfaces that interrupt less and for more precise alerts (Topaz et al., 2015). A meta-analysis on decision support classified as interruptive any system that appears on screen and prevents the user from continuing until the alert is acknowledged (Kwan et al., 2020). The choice between interrupting and merely informing is a design decision with a direct effect on fatigue, and each service needs to measure what happens after the change.

Tailored thresholds

In pediatrics, vital signs vary greatly with age, and standard alarm thresholds are considered inadequate, generating more alarms than necessary (Poole and Shah, 2018). A study proposed customized thresholds based on percentiles and estimated that they would reduce low and high heart rate alarms by up to 50 and 44 out of every 100, respectively (Poole and Shah, 2018). In that estimate, sensitivity was 62 out of every 100 and specificity rose to 49 out of every 100 (Poole and Shah, 2018). The authors note that vital sign alarms should be a safety net in care, but that managing them during hospitalization has become a major patient safety problem (Poole and Shah, 2018).

The example shows an idea that applies to any threshold. A single cutoff for everyone ignores each patient's pretest probability, and an alarm calibrated to the average fires too often at the extremes. Adjusting the threshold to the patient means applying to the monitor the same reasoning applied to a test.

Use the comments of those who dismiss

A hospital investigated whether the comments left by users when dismissing alerts helped find faulty rules (Aaron et al., 2019). Of the rules investigated, 62 were broken, 13 could be improved and 45 worked. One example: an alert suggested replacing potassium when the test field held a non-numeric value, such as a hemolyzed sample, and doctors commented that the potassium was normal (Aaron et al., 2019). The authors note that too many inappropriate alerts make professionals lose trust in the system and start ignoring the ones that matter.

A possible routine for the team

The list below is didactic and was assembled from the strategies cited above. It does not replace local protocols.

1. Measure before changing. Count how many alerts and alarms each professional receives and how many are dismissed (Yoo et al., 2020).
2. Cut repetitions. Repetitions for the same patient are a promising target (Ancker et al., 2017).
3. Listen to the reasons. Dismissal comments reveal broken rules (Aaron et al., 2019).
4. Adjust thresholds. Setting variables, volume and monitor parameters well is part of the routine (Bridi et al., 2014).
5. Prioritize by severity. Alerts by specialty and severity reduce the noise (Sutton et al., 2020).
6. Train the team. Training programs appear among the reduction strategies (Michels et al., 2025).

What about decision support in general

A meta-analysis of 122 trials found that decision support systems increased by 5.8 points the proportion of patients who received the desired care (Kwan et al., 2020). Only 17 of the 122 trials considered alert fatigue in the design or application of the system, and the gain associated with that care was small and without statistical significance (Kwan et al., 2020). The finding shows how little the topic is still studied in the trials themselves.

Back to the scene, the nurse at bed 4 did not ignore the alarm out of carelessness. She responded as the environment trained her to respond. The text on morbidity and mortality sessions shows how to discuss events like this by looking at the system, and the text on how to organize reasoning in the medical record helps record what was seen and when. For those who decide on tests and management, the text on clinical decision-making reminds us that every clue must be weighed by what changes afterwards.

How NEXO helps train this reasoning

In NEXO, the case of the day is fictional, with a real clinical presentation and up to six progressive clues, and you propose the diagnosis by searching for the ICD-10 code.

The limits of using the game are described on the NEXO economy page. If you research the teaching of clinical reasoning, the page for researchers explains how to get in touch. Professionals who write or review cases find information under curation and creators, and institutions can learn about the proposal on the partnerships page. Training with differential diagnosis and clues and knowing how to find the right ICD-10 code complete this practice.

Frequently asked questions

Does alert fatigue affect only doctors?

It affects nurses strongly, and in the ICU professionals describe the alarms as bothersome and too frequent (Lewandowska et al., 2020).

Does an ICU have many alarms per hour?

In a Brazilian coronary care unit, there were 426 alarms in 40 hours of observation, on average 10.6 per hour (Bridi et al., 2014).

Do repeated alerts worsen the response?

The chance of accepting a reminder fell as the proportion of repeated reminders for the same patient increased (Ancker et al., 2017).

Can dismissing alerts become a habit?

Previous dismissals increased the strength of the dismissing habit and were associated with more dismissals later (Wang et al., 2022).

Can the number of alarms be reduced safely?

Personalized thresholds in pediatrics would reduce heart rate alarms while keeping the estimated sensitivity (Poole and Shah, 2018).

Take-home messages

• Alert fatigue is the desensitization caused by frequent and barely relevant warnings.
• Most medication alerts were dismissed in several services studied.
• Repeated alerts for the same patient reduce acceptance more than the workload does.
• Dismissing becomes a habit, and many alerts are closed in under three seconds.
• In the ICU, monitors and other equipment add alarms at a rate of several per hour.
• Cutting repetitions, adjusting thresholds, prioritizing by severity and listening to the reasons for dismissal appear as ways forward.
• The ignored alarm is usually a system problem, not just a problem of whoever was on shift.

Want to practice reasoning with progressive clues? Play the case of the day.

References

1. Aaron S, McEvoy DS, Ray S, et al. Cranky comments: detecting clinical decision support malfunctions through free-text override reasons. J Am Med Inform Assoc. 2019. PMID 30590557.
2. Ancker JS, Edwards A, Nosal S, et al. Effects of workload, work complexity, and repeated alerts on alert fatigue in a clinical decision support system. BMC Med Inform Decis Mak. 2017. PMID 28395667.
3. Bridi AC, Louro TQ, da Silva RC. Clinical Alarms in intensive care: implications of alarm fatigue for the safety of patients. Rev Lat Am Enfermagem. 2014. PMID 25591100.
4. Embi PJ, Leonard AC. Evaluating alert fatigue over time to EHR-based clinical trial alerts: findings from a randomized controlled study. J Am Med Inform Assoc. 2012. PMID 22534081.
5. Kwan JL, Lo L, Ferguson J, et al. Computerised clinical decision support systems and absolute improvements in care: meta-analysis of controlled clinical trials. BMJ. 2020. PMID 32943437.
6. Lewandowska K, Weisbrot M, Cieloszyk A, et al. Impact of Alarm Fatigue on the Work of Nurses in an Intensive Care Environment-A Systematic Review. Int J Environ Res Public Health. 2020. PMID 33202907.
7. Michels EAM, Gilbert S, Koval I, et al. Alarm fatigue in healthcare: a scoping review of definitions, influencing factors, and mitigation strategies. BMC Nurs. 2025. PMID 40542361.
8. Poole S, Shah N. Addressing vital sign alarm fatigue using personalized alarm thresholds. Pac Symp Biocomput. 2018. PMID 29218906.
9. Ruppel H, De Vaux L, Cooper D, et al. Testing physiologic monitor alarm customization software to reduce alarm rates and improve nurses' experience of alarms in a medical intensive care unit. PLoS One. 2018. PMID 30335824.
10. Sutton RT, Pincock D, Baumgart DC, et al. An overview of clinical decision support systems: benefits, risks, and strategies for success. NPJ Digit Med. 2020. PMID 32047862.
11. Topaz M, Seger DL, Lai K, et al. High Override Rate for Opioid Drug-allergy Interaction Alerts: Current Trends and Recommendations for Future. Stud Health Technol Inform. 2015. PMID 26262047.
12. Wang L, Goh KH, Yeow A, et al. Habit and Automaticity in Medical Alert Override: Cohort Study. J Med Internet Res. 2022. PMID 35171102.
13. Wright A, McEvoy DS, Aaron S, et al. Structured override reasons for drug-drug interaction alerts in electronic health records. J Am Med Inform Assoc. 2019. PMID 31329891.
14. Yoo J, Lee J, Rhee PL, et al. Alert Override Patterns With a Medication Clinical Decision Support System in an Academic Emergency Department: Retrospective Descriptive Study. JMIR Med Inform. 2020. PMID 33146626.