The art of “DIVI-nation” – predicting tomorrow’s ICU capacities from
today’s infection numbers
Abstract
Preventing the health system from collapse has been repeatedly stated as
one of the main objectives for the German containment policy for SARS
COV 2. The exact relation between infections recorded in the public
surveillance system maintained by the German center for disease control
(RKI) and data on hospital occupation published by the German
association for intensive care an emergency medicine (DIVI) has not been
analyzed to date. Using a stepwise approach as described in the paper a
linear regression model based on recorded infections with known disease
onset was found to be the most suitable predictor for the number of ICU
patients with a positive test for SARS COV 2 one month later. The model
showed an excellent model fit with nearly 90% explained variance and
reliable prediction of the maximum when applied to data beyond the
construction dataset. Still, the number of additional patients with a
diagnosis of COVID 19 does not necessarily mean a reduction of ICU
capacities in the same dimension. Based on a examination of
interrelations between parameters published in the DIVI registry it is
concluded that a temporary reorganization of hospital care for SARS COV
2 positive patients would probably help to mitigate the risks coming
with increasing infection rates.