One of the parameters for understanding the severity of an infection #8217 is its lethality, i.e. the ratio of the deceased to people infected, read in another way lethality is the probability of dying that has a person infected.
It seems like a fairly easy calculation and yet it gets very complex, during a’epidemic, have the’exact death count and, especially, l’exact number of people infected.
The 3 August l’ISTAT  published the first results of SARSCOV2's ’&Seroprevalence Survey in Italy. They've been tested 64660 People, choosing a representative sample of the Italian population.
The initial goal was #8217 to test a much larger population but many citizens did not join the ’ initiative. The reduced sample may affect the quality of the data, however, l’ISTAT declares the #8217;confidence interval that turns out to be quite narrow, index of the good reliability of their results.
The estimate of the infected subjects reported by the #8217;ISTAT may also be affected by the fact that some infected subjects, especially asymptomatic and paucysynthetic, may not develop antibodies or lose them after a few months.
According to this consideration, cases estimated by serologicals may be lower than the real cases, It is also true that such an #8217 investigation is affected by the false positives of the #8217;examination that would lead to overestimating real cases.
Aware of all these limits we have a denominator to use in the calculation of lethality (lethality - deaths / Infected). As a numerator we can use either the number of confirmed deaths or the #8217;excess mortality .
So far we've had 35203 confirmed deaths COVID
In the March-May period 2020 Italy had an excess of mortality of 44105 deaths compared to the average of 5 previous years.
THE ISTAT estimates 1482377 infected citizens.
With this data, we can estimate:
-lethality based on confirmed deaths: 2,4%
-lethality based on excess mortality: 3%
It would be reasonable to think that the worst-affected regions experienced greater lethality and therefore one would expect a much greater lethality in Lombardy than in the other regions..
If we compare the lethality calculated by the confirmed deaths, however, we see that Lombardy has recorded a lethality slightly lower than the national average.
The regions with the highest lethality were Liguria (3,5%) Emilia-Romagna (3,4%) while Campania, Umbria, Molise, Calabria and Basilicata see a much lower lethality (1.1% to 0,7%).
The worst-affected regions may have 'lost' more deaths, as evidenced by the severe excess mortality of Lombardy (25880 deaths in 3 months).
If we look at the same graph based on the excess mortality (regions that have been over-average in previous years), the situation changes a little bit, Lombardy sees a greater lethality than the national average but the regions with higher lethality remain Liguria and Emilia-Romagna.
How to justify this regional difference of lethality?
I'm not able to give a certain explanation, I trust that most of the hospital facilities involved are well caring for, all over Italy.
One idea might be the different age of the population of the various regions, Liguria has the highest average age in Italy (48,5 years vs. 44,9 years).
If we evaluate a graph of the average age vs regional lethality we notice a certain correlation although not so strong.
Lethality by age group
Preliminary istat results estimate the number of citizens infected by age group.
Using COVID confirmed death data we can estimate lethality by age group.
So we see that in the under 18 the lethality was very low, close to 0 (4 deaths on 194093 cases), in the band 18-49 years the lethality has been of the 0,07%, in the band 60-69 the numbers go up and we see a lethality of 1.7%, in the overs 70 lethality becomes a fearsome 10,6%
The ISTAT estimate of infected cases is reliable?
We have already seen the possible limitations of a serum prevalence survey such as that of ISTAT, we have some data to validate the results?
We can try to look at the data available to healthcare professionals. Hospital staff have been subjected to much more careful and precise monitoring and examination than the rest of the population in both symptomatic and asymptomatic cases that have emerged from both tampon and serological screenings.. It is therefore likely that both deaths and cases are closer to the actual data.
The ISS  States 90 deaths in health workers on duty 29932 cases.
If we see lethality in the health care staff (orange bars), compared to the lethality estimated by THE ISTAT data (blue bars) we find almost overlapping data, good confidence index of their results
Some scientific papers suggest that antibodies may decrease rapidly in asymptomatics and paucysintomatics.
In May, right after the epidemic wave, Lombardy  carried out a massive screening of the medical staff reporting the 13,4% positives
THE ISTAT data says that they were infected on 7,5% Lombardi and also say that health care workers have a 2.1x risk of getting infected compared to other citizens. 7,5 * 2,1 = 15,75%, not far from the 13,4% reported by the Region (which would still be included in the confidence intervals).
As I wrote at the beginning, lethality is just one of the parameters to be considered for grasping the severity of an infection and an epidemic.
An infection with low lethality, high contagiousness, discreet morbidity poses a greater danger to society than a high lethality infection but low contagiousness.