1/6 New model (details below), estimating the Belgian epidemic from hospitalization, death (and case) data. Current Re estimate = 1.4 (1.3 - 1.5). Estimated deaths, continuing current behaviour, until Jan 2021: 1900 (1300 - 2700), with a peak of 300 hospitalizations per day. 
2/6 Zoomed in version. This model more reliably fits both deaths (within two age-groups) and hospitalization data, something which recent model fits didn't manage.
3/6 Until now, the model assumed that the hospitalization rate (HR) was proportional to IFR, but that assumption wasn't valid. The figure below shows estimates based on the Belgian epidemic until June 14th, using covid-19.sciensano.be/sites/default/…
4/6 Clearly, hospitalization rate (HR) is less sensitive to age than IFR: younger infected people are proportionally more hospitalized. The resulting estimated HR (<65y) over time, based on estimated change in demography of the infected population, fluctuates less than IFR.
5/6 The new model more accurately fits hospitalizations and deaths, but estimates a higher reproduction number, and a substantially higher number of infected Belgians (< 65y): 220K (150K - 330K): it estimates that today more Belgians are infected compared to the peak in March.
6/6 Despite the massive testing effort, an increasing amount of infections are not being diagnosed, as the epidemic is causing mostly infections in younger populations that are much harder to find.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
