Fig. 1.

Structure of the compartment model. (A) High-level structure indicating possible transitions between various illness phases and hospitalization compartments. The delay between death and its reporting to the healthcare authorities is added in order to account for the lower number of deaths observed during weekends. Infectious phases are coloured with different colours encoding our a priori beliefs on the average number of secondary cases generated in a day by an individual in the corresponding compartment. We remark that such a number depends on the degree of infectiousness as well as the total number of inter-personal contacts. For example, on the one hand symptomatic individuals are more infectious than asymptomatic ones, but on the other they are much less likely to encounter other people due to their health condition. It is thus difficult to determine a priori which phase generates more secondary cases and as a first guess they are assigned the same colour in the figure. (B) Detailed structure of the compartment model. Each compartment is split into several sub-states in order to have Erlang-distributed transition times between compartments, and to explicitly model the testing process by tracking individuals reported to the health care authorities on a parallel but separate branch. (C) Time-dependent parameters (here the viral transmission reduction due to NPIs is used as an example) are modelled by splines which can be encoded inside the parameter vector by their values at the grid points.