Cough sounds captured overnight by a consumer sleep app could give public health agencies about a week of warning before flu and COVID-19 waves, according to a study from the UK Health Security Agency and Swedish company Sleep Cycle.
The findings appeared September 3 as a medRxiv preprint. Researchers lined up weekly nighttime cough readings from January 2023 through January 2026 against England’s official surveillance data: NHS 111 respiratory triage calls, PCR positivity for flu and COVID-19, and hospital admissions.
Normalized cough counts crested one week ahead of flu PCR positivity. They led COVID-19 PCR positivity by about the same margin. Against NHS 111 acute respiratory calls, raw national correlations reached roughly 0.95.
Sleep Cycle’s app runs a machine-learning audio model on-device that listens for coughing in 10-second clips during sleep sessions. No raw audio is sent to company servers. Records are anonymized and aggregated to seven NHS England regions, with thousands of daily users contributing in each area.
The authors say the approach is not meant to replace established NHS and laboratory surveillance. Instead, passive monitoring adds a layer that does not depend on people seeking care, which varies with awareness, access and demographics, or on lab turnaround times.
UKHSA and Sleep Cycle announced the collaboration in January. The preprint recommends population-normalized metrics over raw cough counts for any future surveillance use, since shifts in the number of active users can distort totals.
