Impact of case definition on early detection systems for COVID-19
About the Project
Detecting infectious diseases, such as COVID-19, early can accelerate case isolation and break chains of infection. We can train a model to do this but there is no clear definition of when a case starts, meaning the ground-truth labels (used for training and evaluation) are often inconsistent between studies. A hypothesis is that case definition, i.e., the time-point at which an individual changes labels from healthy to infected, has a significant effect on systems designed to identify individuals.
This project will use home testing kit study data to analyze and compare the effect of changing the case definition on detection through existing models (or new models, depending on project progress) and consider whether the evaluation metrics reflect real-world impact. If progress allows, the project will seek to find the optimal case-onset point (or some function of accuracy vs early detection) for an early-detection system.
- Shannon Rumsey
- Edward Ho
- Nealson Setiawan
- Jen Park
- Chunting Zheng
- Arinbjörn Kolbeinsson, Evidation
- Eric Daza, Evidation
- Megan Elcheikhali, UCSB
About Evidation Health
Evidation measures health in everyday life and enables anyone to participate in ground-breaking research and health programs. Built upon a foundation of user privacy and control over permissioned health data, Evidation's Achievement platform is trusted by millions of individuals—generating data with unprecedented speed, scale, and rigor. We partner with leading healthcare companies to understand health and disease outside the clinic walls. Guided by our mission to enable and empower everyone to participate in better health outcomes, Evidation is working to bring people individualized, proactive, and accessible healthcare—faster. Founded in 2012, Evidation Health is headquartered in California with additional offices around the globe. To learn more, visit evidation.com, or follow us on Twitter @evidation.