Correcting under-reported COVID-19 case numbers

Alexander Lachmann

The exponential growth curves indicate an extreme rise in fatalities world-wide. Data from China, Italy, and Iran suggest that extreme measures such as curfews and lockdown on national level can reduce the growth rate of COVID-19 over time. But even then the rate only decreases over a period of more than 30 days resulting in a significant number of new infections. It should not be assumed that the number of deaths drops quickly after such measures are taken. By extrapolating the cumulative deaths it is possible to project daily hospitalizations. The model proposed here does assume that only a fraction of the population is affected by the virus and growth limiting effects from a shrinking susceptibility pool are not the dominant factor. In cities like New York city we estimate that a substantial fraction of the population was infected or will be infected. The idea of flattening the curve (14) was not successful in New York City. This study suggests that the current reporting of COVID-19 cases significantly underestimates the true scale of the pandemic. The lack of testing complicates the estimation of the true CFR and causes significant misinformation. This study has leveraged information derived from a well-tested subpopulation (South Korea). With testing capacities of 20,000 tests daily, it has the largest and most accurate coverage compared with all other countries as of writing. The low falsenegative rate in detecting COVID-19 infections leads to the lowest CFR compared with all other countries except Germany (1.49) with major case count. By applying the parameters estimated from this benchmark country, the proposed method can adjust global COVID-19 case numbers. This method is limited in its ability to predict the exact number of cases accurately. The requirement of predicting the exponential growth phase of the viral spread for a 23 day period will lead to high uncertainties, specifically to the onset of behavioral change. This method tries to apply a best case scenario in which lockdown measures are taking effect on the growth rates of cumulative deaths immediately. The method relies on the assumption that deaths by COVID-19 are detected and reported reliably. False-negative rates can have a distorting effect on the case adjustment. This is especially true if the benchmark country does not adequately report deaths from COVID-19. Additionally, the assumption of a globally similar CFRs is untested and should be applied with caution. The major finding in this study is a reproducible exponential decay of the growth rate. This exponential decrease is suggesting multiple doublings of viral infection after lockdown measures are taken. The model does not take population saturation into account such as SIR models. In the case of New York city, were a substantial proportion of the population might be infected, the model can over-estimate the growth rates. The predicted exponential decay function EDG(t) can be combined with SIR models to more accurately predict COVID-19 dynamics. This method suggests that due to the fast exponential growth of true case counts, most modern healthcare systems are not able to track the changes with sufficient and constant coverage. The model predicts a spike in hospitalizations at a rate that will add significant burden to healthcare ICU capacity in the next two weeks. It also highlights the importance of publicly accessible real-time data and the relevance of combining global healthcare efforts.