Independent academic researchers confirm that Kinsa’s granular illness signal provides highly accurate local illness estimates and forecasts, enabling better state-level public health surveillance.
Last month, Kinsa was proud to be featured in a poster presentation during Infectious Disease Week. This poster marks the second report published by academics from the University of Iowa regarding Kinsa data.
The poster presentation showed that Kinsa’s deidentified, timestamped and geolocated signals are an accurate predictor of localized influenza detection and therefore can improve state-level ILI forecasts.
The results of this study also concluded that using Kinsa’s signal can inform clinical and public health activities at a more local, granular level, such as ramping up staffing during outbreaks, optimizing orders for necessary vaccines and prescription drugs, and predicting outpatient volumes and healthcare utilization.
Prior to Infectious Disease Week, epidemiologist Dr. Aaron Miller and Dr. Phil Polgreen and team published “A Smartphone-Driven Thermometer Application for Real-Time Population- and Individual-Level Influenza Surveillance” in the peer-reviewed Journal of Clinical Infectious Disease. It concluded that “[Kinsa’s] thermometer readings capture real-time ILI activity at a population level and can also be used to generate improved forecasts,” and that our “thermometer counts were highly correlated with national ILI activity.”
Over 10 million temperature readings, along with respective de-identified symptom, age and gender data points were studied alongside state-reported ILI activity from 46 states. Kinsa’s signals alone were able to improve forecasting accuracy in 43 of 46 states, with 8 states depicting a very large increase (>50%) in forecasting accuracy. Furthermore, all states saw a considerable improvement in forecasts of ILI activity when Kinsa’s thermometer readings were incorporated into baseline models of existing surveillance data- a 25.5% median improvement.
The multiple dimensions of Kinsa’s signals and thermometer readings, most notably the real-time availability, can not only accurately track and improve real-time influenza activity at a national and state level, but also allow for more granular and longer term forecasts that can enable better local surveillance efforts during infectious illness outbreaks.
Thank you to the brilliant team at the University of Iowa for taking the time to study our unique data signal.