Carnegie Mellon to use anonymized Sleep Cycle cough data in five-year outbreak study

Carnegie Mellon University’s Delphi Research Group has launched a five-year collaboration with Sleep Cycle to analyze deidentified nighttime cough and breathing signals and test whether those signals can improve respiratory disease surveillance. The data, drawn from Sleep Cycle’s public Cough Radar visualization and provided to researchers in privacy-preserved form, will feed epidemiological modeling and forecasting efforts aimed at earlier detection of influenza, respiratory syncytial virus and SARS-CoV-2.

Sleep Cycle will deliver “deidentified research data related to coughing and breathing” and trends derived from what the company calls “anonymized, differentially private data” from its Cough Radar, a public tool that aggregates nightly coughing intensity across regions. Sleep Cycle described itself as “the world’s leading AI sleep technology company” and said its proprietary audio-based cough-detection technology and data-science work have shown that nighttime cough behavior can correlate with real-world viral activity. The company said relevant research insights will be announced continuously during the program.

The principal investigator at Delphi, Professor Roni Rosenfeld, framed the study as an evaluation of where sleep-based signals might add value to existing public-health data streams. “This research will evaluate the utility of Sleep Cycle-derived cough and breathing signals for epidemiological surveillance applications,” Rosenfeld said. “Our goal is to rigorously assess where these indicators can add value alongside existing public health data streams. Bolstered monitoring could lead to earlier detection of seasonal and emerging respiratory disease outbreaks, allowing health officials to react faster and safeguard the public health.”

Delphi’s announcement, posted by Aaron Aupperlee on the university site, says researchers will explore whether nighttime cough and breathing patterns can provide earlier visibility into respiratory disease activity and support forecasting models. The stated geographic focus of the collaboration is the United States; the analysis will use aggregated signals rather than individual-level reporting, according to the institutional materials.

The partnership raises both opportunity and questions for public-health practice. If validated, sleep-derived signals could complement clinic-based reporting, lab testing and wastewater surveillance by offering a near-continuous, population-scale view of nocturnal respiratory symptoms. Early detection of rising cough intensity in a region could give health officials additional lead time to target testing, messaging or vaccination campaigns.

The public announcements leave key technical and ethical details unspecified. The statements use multiple privacy descriptors, including deidentified, anonymized and differentially private, but they do not describe the exact data fields to be shared, the geographic resolution of “regions,” whether any raw audio or device identifiers will ever be accessed, or whether the collaboration has human-subjects review. The funding source for the five-year program and any plans to share datasets or code publicly were not detailed in the materials released by the partners.

The project tests a broader shift in which consumer health apps supply aggregated behavioral signals to academic researchers and public-health systems. Over the next five years, Delphi and Sleep Cycle will publish their findings and assess whether nocturnal cough and breathing metrics can move from a promising signal to a reliable component of outbreak monitoring, while the unresolved privacy and transparency issues will shape how widely those signals are adopted.