Aggregated nocturnal cough data from the Sleep Cycle smartphone app—covering January 2023 to January 2026 across England—demonstrated strikingly strong associations with established respiratory illness indicators. Population-normalized metrics (coughs per user and coughs per hour of sleep) achieved raw national correlations of ~0.95 with NHS 111 acute respiratory infection triage calls, and prewhitened correlations above 0.55, confirming the relationship extends beyond shared seasonality. Crucially, both normalized metrics peaked one week before COVID-19 PCR positivity, and coughs per hour of sleep peaked one week before influenza PCR positivity—suggesting genuine early-warning capability.
This preprint, not yet peer-reviewed, represents a meaningful advance in passive, scalable surveillance infrastructure. Traditional respiratory surveillance systems suffer from inherent delays: patients must seek care, specimens must be processed, and data must be reported. A smartphone-based signal requiring no active user participation could partially bypass these bottlenecks. The 0.95 correlation with NHS 111 calls is remarkable, though it warrants scrutiny—Sleep Cycle's user base skews younger and wealthier, potentially introducing demographic bias that limits representativeness. Regional analyses appeared promising but weren't quantified in detail in the abstract. The one-week lead time, if validated prospectively, could meaningfully accelerate public health responses during emerging waves. The finding that unnormalized total cough counts performed poorly underscores the importance of methodological rigor in digital surveillance. Confirmatory prospective studies with demographic breakdowns and multi-country replication should be priorities before operational deployment.