Using ILI surveillance to estimate state-specific case detection rates and forecast SARS-CoV-2 spread in the United States

Justin D Silverman; Alex D Washburne

The copyright holder for this preprint . Figure 3 : (a) The excess ILI estimated falls within the range of what one could expect from a US epidemic growing at the exponential growth rate defined by the growth of new deaths. Adjusting the ILI surge based on decreased care seeking in New York does not reconcile the difference bewteen the ILI surge and the epidemic curve from US growth rates, suggesting additional forces are at play. The cause of the apparent deceleration in the ILI surge is hypothesized to be due to some combination of successful interventions, faster decreases in care-seeking behavior changing than measured in New York, and/or other possibilities including faster growth and higher subclinical rates. (b) If the growth rate in the US is faster than US deaths suggest, such as a growth rate observed in Italy prior to the Italian lockdown (2.645 day doubling time), it could provide alternative explanations of the curvature of the excess ILI through a larger subclinical rate and epidemic curves near their peak at the time of the peak of the ILI surge. Serology or other measures of prevalence are needed to reconcile these alternative hypotheses. (c) More generally, the ILI surge forces a dependence between growth rate (doubling time), the clinical rate, and the lag between the onset of infectiousness and presentation to the doctor with ILI, where faster growth implies a slower clinical rate.