Nurses received 180 alarms per patient per day, of which roughly 85% were non-actionable. Alarm fatigue was a documented patient-safety issue and the interface treated every alarm as equal.
STUDY 23 / 30 · ANONYMISED · NDA · HEALTHCARE · US · 84 HOSPITALS
HEALTHCARE · US An alarm system that earns the attention it demands.
RoleLead designer + researcher
Timeline18 weeks
Team2 designers, 1 researcher, 1 PM, 3 frontend, 2 clinical SMEs
VerticalClinical monitoring · ICU and step-down
Prioritised by clinical actionability and trajectory rather than by threshold breach, with a graduated escalation that reserves the loudest signal for the rarest event.
Non-actionable alarms fell substantially while true-positive response time improved - the two moving together rather than trading off.
46ALARMS PER PATIENT / DAYfrom 180
-34%RESPONSE TIME · TRUE POSITIVEvs baseline
0MISSED CRITICAL EVENTSin 6-month study
THE ARGUMENT
Why the obvious solution was wrong.
The study matters because the product problem was reframed before the interface was polished.
An alarm that fires 180 times a day and is right 15% of the time is not a safety system. It is a noise generator that a rational clinician learns to filter, and the filtering is indiscriminate because nothing in the signal distinguishes the 15%. Every hospital knows this. The interfaces continue to alarm on threshold breach because threshold breach is easy to compute and actionability is hard.
We prioritised on trajectory and actionability rather than on crossing a line - a slowly worsening trend in a patient with a relevant history outranks a transient breach in a stable one. Escalation is graduated: an ambient indicator, then a directed notification to the assigned nurse, then a unit-level alert, then the loudest signal reserved for events that are genuinely rare. Non-actionable alarms fell by three quarters and response to true positives got faster, because the signal that remained was worth turning around for.
THE INTERFACE CRAFT
The interaction, rendered as a working product surface.
The specimen below is code-native and uses the study's own design logic. The client interface remains protected.
HEALTHCARE
ALARM PRIORITYNeeds action
Room 408 · 02:14
DESIGN DECISIONS
Positions we would defend.
Each decision names the principle and the product consequence, not a stylistic preference.
Actionability is the criterion
An alarm for a condition with no available action is not information. It is interruption, and it costs attention that a real event will need.
Reserve the loudest signal
A maximum-urgency signal used weekly is a signal that has been spent. Its value depends entirely on its rarity.
Measure fatigue as an operational metric
Alarm burden should appear on a unit dashboard, not in a root-cause analysis after harm.
PRODUCT LEADER READOUT
What transfers, and what should remain specific to this product.
A case study is useful when its operating principle travels without turning the original interface into a template.
Read the operating condition
For Clinical monitoring · ICU and step-down, the transferable lesson is not a copied screen. It is the condition the interface had to make legible: A worsening trend in a relevant patient outranks a transient breach in a stable one. The alarm reflects clinical meaning, not arithmetic. Rebuild that visibility for your own roles, risk, terminology, and operating cadence.
Protect the design rule
An alarm for a condition with no available action is not information. It is interruption, and it costs attention that a real event will need. Keep that rule in the acceptance criteria, component states, and production QA record so later visual cleanup cannot erase why the interaction exists.
Measure behaviour after ship
The evidence record is 46 for alarms per patient / day, from 180. Recreate the baseline and outcome window before rollout, segment the result by role and context, and state clearly what the measure cannot prove.
RESEARCH RECORD
The work behind the interface.
These artefacts connect the final interaction back to the evidence and product model that produced it.
Alarm burden study
Instrumented three ICUs for six weeks, logging every alarm and coding actionability with clinical review; 85% were non-actionable.
Clinician response ethnography
Observed 22 nurses across full shifts, recording the moment an alarm was consciously ignored and what preceded it.
Prioritisation model design
Built the trajectory-and-actionability model with two intensivist and validated against retrospective adverse events.
Safety review
s Full clinical safety case with the client's patient-safety committee before any change reached a live unit.
“Fewer alarms and faster response. The committee needed to see the data twice before they believed it.”
Chief Nursing Informatics Officer, hospital network · under NDA
NEXT STUDY · 24 / 30 · HEALTHCARE