How should we present the epidemic curve for COVID-19?
Epidemic curves are used by decision makers and the public to infer the trajectory of the COVID-19 pandemic and to understand the appropriateness of current response measures. Symptom onset date is commonly used to date cases on the epidemic curve in public health reports and dashboards. However, third-party trackers often plot cases on the epidemic curve by the date they were publicly reported by the public health authority. These two curves create very different impressions of epidemic progression. On April 1, the epidemic curve for Ontario, Canada based on public reporting date showed an accelerating epidemic, whereas the curve based on a proxy variable for symptom onset date showed a rapidly declining epidemic. This illusory downward trend (the "ghost trend") is a feature of epidemic curves anchored using date variables earlier in time than the date a case was publicly reported, such as symptom onset date or sample collection date. This is because newly discovered cases are backdated, creating a perpetual downward trend in incidence due to incomplete data in the most recent days. Public reporting date is not subject to backdating bias and can be used to visualize real-time epidemic curves meant to inform the public and policy makers.