Your smartphone's cough data is tracing air pollution to the bedroom

A new analysis of nighttime cough sounds, captured passively by a popular sleep-tracking app, has produced one of the largest real-world datasets yet connecting everyday air pollution to disrupted sleep — and it suggests that even pollution levels that meet international safety guidelines may still be enough to trigger more coughing after dark.

Researchers at the University of Cambridge's Department of Computer Science and Technology, working with Tsinghua University in China and the Swedish sleep-technology company Sleep Cycle, mined cough recordings gathered by phones left on nightstands in 32 cities across 12 countries over roughly 500 consecutive nights. The work, appearing this week in the journal Communications Health, is among the first to use smartphone-based passive sensing rather than hospital records or self-reported symptoms to track a population's respiratory health night by night, city by city.

The numbers behind the cough

The team, led by Dr. Tong Xia of Tsinghua University and Professor Cecilia Mascolo of Cambridge, used an AI model built into the Sleep Cycle app to detect and count coughs directly on users' devices, then pooled the anonymized counts by city and matched them against local fine-particulate pollution readings, weather data and influenza surveillance figures. The dataset drew on tens of thousands of users' recordings collected over an 18-month stretch.

The central finding: every additional 10 micrograms per cubic meter of PM2.5 — the fine particulate pollution that penetrates deep into the lungs — was associated with a 2.4 percent rise in nighttime coughing. The pattern held even in cities where daily pollution stayed within the limits set by the World Health Organization, and it survived after the researchers statistically adjusted for flu season, which they had expected to erase the signal. It did not. A natural stress test arrived unplanned: when wildfire smoke sent Los Angeles's particulate levels soaring in January 2025 — downtown monitors recorded daily PM2.5 concentrations reaching the "Unhealthy" category, as documented separately by an American Chemical Society-backed analysis of that event — nighttime coughing in the city rose in step with the smoke.

From clinic charts to nightstand microphones

Air pollution is already established as a major public health burden, linked to roughly 7 million premature deaths a year worldwide, and the WHO tightened its recommended exposure limits in 2021 after concluding that harm occurs at levels once considered safe, according to the organization's global air quality guidelines. But most of that evidence comes from hospital admissions, mortality statistics and long-term cohort studies — coarse instruments that miss the early, low-grade symptoms, like an occasional cough, that precede more serious illness by months or years.

Consumer sleep-tracking apps, almost by accident, have built exactly that kind of dataset. Sleep Cycle's app already analyzes users' breathing and movement sounds each night to estimate sleep quality; repurposing the same audio stream to flag coughing required no new hardware and no new data collection, only new analysis. Dr. Mikael Kågebäck of Sleep Cycle emphasized that the underlying privacy design made the project possible in the first place: coughs are detected and counted on each person's own device, so no audio recording ever leaves the phone — only the resulting counts are pooled for research.

Who the data describes

The 32 cities in the study span a deliberately wide range of pollution levels and climates, giving the researchers statistical power that single-country studies typically lack. Professor Mascolo noted that most public health studies of this kind have data from only one country, so a dataset spanning a dozen makes the association far harder to dismiss as a local quirk of weather, culture or reporting habits. Because participants are drawn from an app's existing user base rather than a randomly sampled population, the findings do not claim to represent every resident of every city studied — smartphone owners who track their sleep skew toward wealthier, more health-conscious users.

"This suggests that there may be no such thing as a safe threshold for exposure to air pollution, although our research is purely observational," said Professor Cecilia Mascolo of the University of Cambridge.

Dr. Xia, the study's lead author, said he had expected the signal to disappear once flu season was accounted for. "I was surprised we could still see the signal at all," he said. "I thought that adding in the confounding factor of flu season would make the link in the data disappear, but it didn't." He added that a cough, whether a one-off or something more persistent, is a sign of irritation or inflammation in the lungs — "a sign of something that hasn't become chronic yet, but might in future."

What happens next

The authors are explicit that their work shows an association, not proof that pollution directly causes each additional cough; unmeasured factors that move together with pollution, such as humidity, pollen or indoor heating patterns, cannot be fully ruled out from observational data of this kind. They are now working to extend the model to more cities and to test whether short-term spikes in coughing can serve as an early warning signal for local health authorities, arriving faster than hospital admission data during pollution events such as wildfires or winter smog episodes.

For now, the practical takeaway for anyone living in a city is narrower but still useful: a run of nighttime coughing that coincides with a smoggy or smoky stretch of weather is not necessarily "just a cough" to be waited out, and checking local air quality readings — most cities now publish them in real time — is a reasonable first step before assuming a cold or allergies are to blame.