James W. Wheless watched the data arrive from 242 epilepsy patients wearing smartwatches during hospital monitoring, and what emerged was striking: a simple app had caught nearly every major seizure that mattered most. At Le Bonheur Children's Hospital in Memphis, Tennessee, Wheless and his team discovered that EpiWatch—an application running on standard smartwatches—detected 98% of tonic-clonic seizures with a false alarm rate so low it fundamentally changes how people living with epilepsy might protect themselves.
The stakes of this finding are profound. Tonic-clonic seizures, which involve violent muscle contractions and loss of consciousness, carry a particular danger: sudden unexpected nocturnal death in epilepsy, or SUDEP. For people who sleep alone or whose seizures remain uncontrolled, this risk is substantial. Existing wearable seizure monitors promise caregiver alerts, but many are plagued by false alarms—machines crying wolf so often that people stop wearing them or trusting their signals. The gap between a device that works and one that actually gets used is where lives hang in the balance.
In the study published in Neurology Open Access, participants with an average age of 23 were monitored continuously in a specialized hospital unit over an average of two-and-a-half days, tracked simultaneously by video-electroencephalograph (EEG)—the gold standard for seizure detection—and by the EpiWatch app. The results were unambiguous: the app detected 46 of 47 tonic-clonic seizures, achieving a 98% sensitivity rate. The single missed seizure occurred not because the app failed, but because a caregiver was restraining the participant's arm at the moment it happened.
What truly set EpiWatch apart was its restraint. Over 16,000 hours of monitoring, the app generated just 56 false alarms—a rate of 0.08 per day, meaning caregivers encountered one erroneous alert roughly every 12.4 days. This was 90% lower than competing devices, which ranged from 0.67 to 2.52 false alarms daily. Of the 242 participants, 87% experienced no false alarms whatsoever, while only 4% dealt with more than one. When false alarms did occur, more than half were traceable to mundane activities: video games, repetitive arm movements—things that might trigger a device hyper-sensitive to motion but that the EpiWatch learned to ignore.
The practical implications ripple outward. Wheless emphasized something often overlooked in medical technology: stigma. Many seizure monitoring devices are visibly medical, marking the wearer as different, potentially discouraging consistent use precisely when reliability matters most. A smartwatch, by contrast, is ordinary. Millions wear them. An app running on that familiar device carries no social burden, removing a quiet but real barrier to adoption. Combining that normalcy with fewer false alarms creates a tool people might actually keep using long-term—the condition that transforms any medical device from promising laboratory result into life-saving technology.
The research does carry limitations. All seizures occurred in the controlled environment of a hospital epilepsy monitoring unit, where conditions are far more standardized than the messy reality of daily life. Whether the app maintains its precision when people are at home, at work, or in motion remains to be demonstrated. But for people contending with uncontrolled seizures, even a 98% detection rate combined with minimal false alarms represents a genuine leap forward—the difference between a device that monitors and one that might actually prevent tragedy.