This modelling approach allowed computing excess cases in 11 of 13 regions but did not provide a good fit in the remaining two regions where a marked influenza peak was not present (see Supplement).