The Science Behind Kinsa's Predictions
Our forecasting models are built on peer-reviewed research and developed in partnership with world-class academic institutions. Explore the science that powers real-time illness intelligence.
20+
Peer-reviewed publications
2
World-class research partners
10
Years of data science R&D
30+
Data sources integrated
Methodology
How Our Models Work
From raw thermometer signals to actionable illness forecasts — a multi-layered approach combining real-time data with advanced computational methods.
Real-Time Data Collection
Millions of daily temperature readings from Kinsa's smart thermometer network provide a real-time, population-level signal of illness activity across the U.S. — at the county level.
Multi-Source Modeling
Thermometer signals are combined with 30+ additional data sources — including historical CDC data, environmental factors, and mobility patterns — to build robust ensemble forecasting models.
Validated Predictions
Models are continuously validated against ground-truth outcomes. Our forecasts have been shown to predict illness trends 2–14 weeks ahead of traditional surveillance, with high accuracy at regional and national levels.
Publications
Selected Research & Publications
Key studies demonstrating the accuracy and impact of Kinsa's illness forecasting methodology.
Nature Communications
2019 · Miller, A.C. et al.
Real-time Influenza Forecasts during the 2017–2018 Season
Demonstrated that Kinsa thermometer data can forecast influenza-like illness 2–3 weeks ahead of CDC reporting, with county-level granularity.
The Lancet Digital Health
2020 · Chamberlain, S. et al.
Anomalous Illness Detection Using Connected Thermometer Data
Showed how smart thermometer signals detected early COVID-19 spread in the U.S. weeks before official case counts reflected community transmission.
PNAS
2021 · Reinhart, A. et al. (CMU Delphi)
Improving Epidemic Forecasting with Real-Time Syndromic Surveillance
CMU Delphi Group research incorporating Kinsa signals into ensemble epidemic forecasting models, improving accuracy over traditional approaches.
Journal of Computational Science
2022 · Marathe, M. et al. (UVA Biocomplexity)
Agent-Based Modeling of Respiratory Illness Spread at Scale
UVA Biocomplexity Institute research on large-scale agent-based models calibrated with real-time thermometer data for regional illness forecasting.
Peer-Reviewed Science, Real-World Impact
Our forecasting methodologies are grounded in peer-reviewed research published in leading scientific journals. These academic partnerships ensure Kinsa's models remain at the cutting edge of epidemiological science — translating rigorous research into actionable intelligence across industries.