A seizure alarm aims to detect an event and alert someone who can respond. Evidence that it detects a particular kind of seizure does not establish that it prevents injury or sudden unexpected death in epilepsy, known as SUDEP.
Health Technology Wales’s completed 2026 appraisal found limited clinical-effectiveness evidence for wearable and bed-based technologies detecting tonic-clonic seizures. It also found insufficient evidence to determine cost effectiveness from a Welsh perspective. Those findings describe gaps in the evidence, rather than proving that every alarm is ineffective. HTW appraisal outcome
“Seizure alarm” covers different technologies
The Welsh assessment considers wearable devices and sensors used with a bed. Devices can detect signals such as movement, sometimes combined with other physiological measurements, and then send an alert. The particular seizure type and the setting in which the technology was tested matter. Technology and evidence review
A result for tonic-clonic seizure detection should not be extended to every seizure. Equally, a study conducted under hospital monitoring does not automatically establish the same performance during ordinary nights at home. The relevant comparison concerns the named device, signals, population and conditions, rather than the general appearance of a wristband or mattress sensor.
Three different meanings of “works”
When reading a device claim, it helps to separate three questions:
| Question | What the measure concerns |
|---|---|
| Did the device detect recorded seizures? | Detection performance within the tested seizure type and setting |
| How often did it alert without a seizure? | The burden of false alerts over a stated period |
| Did using the device reduce harm? | Outcomes such as injuries or deaths, requiring evidence beyond detection alone |
HTW’s report describes variation between devices and studies, including tradeoffs between detecting events and generating false alerts. Combining their figures into a single accuracy score would conceal those differences. Detection findings and limitations
The period used for a false-alert rate matters too. An alert count per night and a count per hour are different measures. Neither can be understood properly without knowing when the device was worn or in use.
The response is another part of the system
An alarm creates the possibility of notifying another person. Whether an alert reaches them, whether they are available and what happens next are further parts of the real-world process. A detection percentage cannot answer all those questions.
This is different from a general care alarm’s connection through a telephone network, but the service distinction is useful: detecting, transmitting and responding are separate steps. A working component does not describe the complete arrangement.
Epilepsy Action says there is not enough evidence that seizure alarms can guarantee safety or prevent SUDEP. It also acknowledges that some people find them helpful within a risk-reduction plan. The two statements concern different outcomes and can both be true. Patient-organisation guidance
Reassurance has value without becoming proof of prevention
The Welsh appraisal includes patient and carer perspectives describing potential reassurance alongside concerns about accuracy, false alarms, usability and access. Those experiences matter, but feeling more secure is not a measurement of reduced mortality. Patient and public involvement
For someone reading about an alarm, the useful evidence names its intended seizure type, study setting, missed events, false alerts and response arrangements. This guide does not select a device or recommend changing an individual monitoring plan. It explains why a claim about detection needs to remain distinct from a promise of protection.