Scaling analysis of COVID-19 spreading based on Belgian hospitalization data

Bart Smeets; Rodrigo Watte; Herman Ramon

Since the current scaling behavior of H a still closely follows the algebraic growth regime, it is very difficult to make accurate predictions on when the inflection point away from this regime will occur. As a rule-of-thumb, reliable prediction capacity does not extend a period of about 3 days. The SIR-X model predicts that H a will start to plateau around 40 days after the initial day. The parameters from the SIR-X model fit suggest a very low value for the containment rate of both infected and susceptible individuals κ 0 . In other words, containment measures have only a weak effect on removing healthy individuals from the susceptible population. On the other hand, the removal , compared to predictions from the SIR-X model, assuming that 15% or 20% of hospitalized patients require intensive care. (d) Longer-time extrapolation assuming 17.5% ICU shows a peak of ICU patients around April 20th. Different curves show varying ICU retention times. The shorter the ICU retention time, the lower and earlier the ICU peak will be. rate of symptomatic individuals is much higher, leading to a strongly decreased effective reproduction number, and a moderately high quarantine probability. These SIR-X model parameters are somewhat similar to values estimated for the Beijing region of China [4] . When extrapolating the number of deaths using the SIR-X model, the predicted death toll due to COVID-19 will exceed 10 3 by April 3rd. The SIR-X model for accumulated hospitalizations is compatible with the current number of patients in intensive care when assuming that between 15% and 20% of hospitalized patients need ICU treatment, and that the average retention time in ICU is around 12 days. Extrapolation with these parameters predicts a peak in number of ICU patients around April 15th, with the number of ICU patients exceeding the capacity of the Belgian healthcare system of 2650 beds.