Building a Real-Time COVID-19 Surveillance System with R
summary
Global dataset includes over 20 million daily reports from ~200K participants and the Sweden arm has ~140K participants with ~20K reported tests. R is used end‑to‑end to process raw reports, fit a penalized logistic regression on the ~20K test-labeled records, and produce daily regional prevalence estimates. Outputs feed an interactive Shiny dashboard and the covidsymptom package for aggregated data. Validation against hospital admissions shows good alignment and enables predicted hospitalizations 7 days ahead after correcting bias with time-weighting and selective exclusion of immediate-report downloads.
AI-written summary · about
chapters (8)
- 0:00COVID Symptom Study intro
- 3:30Data collection & demographics
- 6:03Symptom analysis & modeling
- 10:00Model refinement & adaptation
- 11:47Predicting hospitalizations
- 13:15Research tools & resources
- 14:43Conclusion & future apps
- 15:04Q&A
related talks
32mData saves lives: The power of crowdsourcing in a pandemic
16mAutomating the Dutch National Flu Surveillance for Pandemic Preparedness
26mData with a mission: a COVID-19 API case study
42mBuilding data driven solutions for COVID 19 Contact Tracing & Vaccine Management
14mDigging a Pit of Success for Your Organization
26mAccelerating COVID-19 insights for clinical use with intelligent search
01h 06mInformation saves lives
05mPivot