Analysing recovery from pandemics by Learning Theory: the case of CoVid-19

Romney B. Duffey; Enrico Zio

In the above graph of Figure 4 , the infection reduction/recovery rate for Italy was estimated to occur in about 30 days, by assuming it was a timeframe similar to that of China. The present prediction is, then, that there is at least a 30-day recovery timeframe as the horizon for the infection rate to get down to a minimum achievable level of say 50 per day (as for China); infection rates should be below 10% of the peak (i.e. about 600 per day) in about 3 weeks from the peak. But this prediction can be effectively monitored and should be updated as more data arises. In fact, at present, only very few days have elapsed since the seemingly reached peak rate, so it is useful to look at a future trajectory. This can be done as in Figure 6 , by plotting the data simply changing the denominator of the non-dimensional elapsed time after the peak rate N* from the 30 days (27 March to 27 April) assumed in analogy to China, to say 20 days (27 March to 6 April) or 40 days (27 March to 6 May) for estimated reduction to the (non-zero) minimum (i.e. 50-100 per day). This sensitivity analysis allows to verify, based on objective data, whether or not any change or improvement in countermeasures is warranted, and by how much. The calculations can assume the infection rate peaked either on about 24 March or March 29, the latter being shown as giving the fewest data points (see Figure 6 above)