About the 9.5% COVID19 lethality observed in Italy and about unsolved problems to be considered before thinking about a massive test policy

Lots of people are asking questions about the high lethality rate in Italy, currently at 9.5% and growing. I start by claiming that the common belief that a large number of asymptomatic are not tested and therefore not showing in the number is only partially true. There are many unsolved problems before we can think about having large test on population. Keep reading to understand the details.

Confirmed cases of COVID-19 per million inhabitants in Italy by province[source:wikipedia]

First, let’s define some terminology:

lethality = reported deaths / current tested positives

mortality = reported deaths / final number of positives

The former is computed during the infection, while the latter is computed post-infection when the estimations are more accurate. Lethality is an estimation made during the infection always based on incomplete and biased data. The bias is progressively reduced over time. Mortality is the final number, expressed only at the end.

Then, let’s consider the following aspects:

Define the baseline. Whenever you talk about high or low the first question is what is the baseline. WHO declared the COVID19 lethality is 3.4%. However, this is an estimation based on the information available on 3rd March using data available from China. Please remember that the WHO spokesperson called the number “the current global snapshot”. Please, also remember that the WHO reported China 3.4 percent on average but 5.8 percent in Wuhan, Hubei province where the virus first emerged, and 0.7 percent in other areas in China. Therefore, this baseline would likely be updated with new updated information and new data coming from additional countries.

Define a sample/test policy. COVID19 is a highly contagious disease, with an R0 value around 2.5 in China during early stages of the epidemic. This means that every person will likely contaminate 2.5 people causing the observed exponential growth. Estimates suggest that about 80% of people with COVID-19 have mild or asymptomatic disease, 14% have severe disease, and 6% are critically ill. In simple words, lots of people get the disease and they go undetected. According to the Guardian, “The small town of Vò, in northern Italy, where the first coronavirus death occurred in the country, has become a case study that demonstrates how scientists might neutralise the spread of Covid-19… We were able to contain the outbreak here, because we identified and eliminated the ‘submerged’ infections and isolated them,” Andrea Crisanti, an infections expert at Imperial College London, who took part in the Vò project, told the Financial Times. “That is what makes the difference. ‘‘If these people had not been discovered,” said the researchers, they would probably have unknowingly infected other inhabitants. .. The percentage of infected people, even if asymptomatic, in the population is very high,” wrote Sergio Romagnani, professor of clinical immunology at the University of Florence, in a letter to the authorities. “The isolation of asymptomatics is essential to be able to control the spread of the virus and the severity of the disease”.

Therefore, one question frequently asked is, why we do not test everyone?

Testing people with no symptoms is very difficult with current logistics. Firstly, It takes 2 days to get the results. However, this is now improving but for limited samples. Secondly, testing rooms or spaces must be separated to avoid the risk of becoming another source of infection. However, this is improving with drive-through tests for limited samples. Thirdly, tests themselves are limited in number and still very expensive. If you want to know more about testing please have a look to COVID19 testing on wikipedia.

With the above three conditions in mind, today it is virtually impossible to test on a massive scale and therefore ad-hoc testing policies need to be put in place.

Italy has made more than 250.000 tests from the beginning of the outbreak up to 23rd March.

At the beginning the policy was to give a sample to all the suspect cases, and then this evolved into giving a sample to all the people with suspected symptoms.

Other countries have adopted different policies consisting in using telco data and payments to rapidly track back all the people who had potential interactions with positive cases regardless of the symptoms.

It is clear that some proportion of ad hoc testing needs to target people for whom there are no symptoms or particular reasons to think they are infected. However, understanding what the proportion is remains an unsolved problem. This is essential to the purpose of understanding a sound proof baseline.

Define a policy for dealing with asymptomatic people. There is no doubt that proactive defence is a need to deal with the virus. However, even assuming that there were a way to test everyone in a not-so-near future, then what would be the right way to separate positives and asymptomatic people from the rest of the population?

Remember that estimates suggest that about 80% of people with COVID-19 have mild or asymptomatic disease, 14% have severe disease, and 6% become critically ill.

Imagine a family with 2 asymptomatic kids and 2 parents in their 50’s with typical diseases developed at this age such as obesity, hypertension, diabetes, and all the other all diseases we learned about bringing the life expectancy to 75+ in the Western world but now considered extremely dangerous conditions in case of COVID. Would you separate the kids from the parents?

Imagine a situation when a family gets the disease and they also take care of the grandparents. Would you stop taking care of the elders? These are all unsolved problems today.

Keep in mind that the sample/test policies will have a significant impact on the observed mortality rate when the disease is over. Specifically it will impact the denominator.

Define your attribution policy. It has been observed that COVID19 has a higher lethality for the elderly. It has been estimated that <1% of people in their 20s die from the virus, while the rate increases to >15% for 80-year-olds. Different countries have different attribution policies. In certain countries they attribute a death to COVID if and only if COVID was diagnosed in absence of any pre-existing disease such as diabetes, heart problems, obesity, and others. Other countries consider a COVID death also if COVID was diagnosed with other diseases.

Keep in mind that the way in which your attribute death will have a significant impact on the final mortality rate. Specifically it will impact the numerator.

Consider the Demographics. In Italy, 22.75% of the population is 65+ years old meaning that there are at least 12 million elderly. This number is significantly more than the 5,955 thousand persons above 65 years reported in Hubei in 2017.

Consider when the lockdown started. Italy went in full lockdown when there were 800 deaths. China went on full lockdown at 80 deaths. With the initial growth rate of about 26%, this is equivalent to 13 critical days of delay. The impact of this delay has been devastating.

Consider the max capacity for Intensive Care Units (ICU). Critically ill patients need help with oxygen and frequently with ICUs. In three major areas in Italy (Bergamo, Brescia and Piacenze) ICUs are close to max capacity and soon we will run into a situation where people will not be able to get critical support. Up to this moment, this situation has been avoided by moving critical cases to other regions which are in less critical situations. However, the national capacity in Italy is also running out.

So, now trying to answer the questions “why the high lethality rate in Italy, currently at 9% and growing?” The answer is complex but I think we should take into account the following factors:

1. Baseline is the 5.8 observed in Wuhan, Hubei;
2. Italy has adopted a large scale testing policy but so far no track back policy;
3. Italy attributes deaths to COVID even if COVID is only one cause of the death;
4. Italy has an older population;
5. Italy had a critical delay in imposing total lockdown (˜13 days compared to China);
6. Italy is close to the maximum capacity for ICU;
7. We are still talking about estimated lethality, and that’s an estimate progressively updated with more confidence as additional data becomes available. The mortality will be declared at the end of the infection.

In short, the above 7 factors should be considered when the lethality of 9.5% observed in Italy (http://tinyurl.com/covid-italy) is compared with the 3.4 percent on average observed in China and the 5.8 percent observed in Wuhan.

Unsolved problems to be considered before thinking about a massive test policy

Increasing the number of tests will certainly increase the confidence on our numbers but first we need to solve a few unsolved problems:

1. Massive scale testing is simply not practicable as of today. Tests are too expensive, they require too much time, and there are still logistic problems in terms of isolated testing areas. Hopefully, these problems will be solved during the next few months but it is unlikely that a 100% coverage will be achieved.
2. Even assuming 100% coverage a non-negligible part of the population will be negative today and possibly positive tomorrow because the situation is in constant evolution. Can we test 100% of the population and have everywhere results updated in real time, with no delay and infinite accuracy?
3. Therefore, it is clear that some proportion of ad hoc testing needs to target people for whom there are no symptoms or particular reasons to think they are infected. However, understanding what is the proportion is still an unsolved problem.
4. Additionally, it is clear that we need to effectively learn how to perform backtracking of positive cases in order to find all the people who came in touch with positives up to 14 days before each single positive case is detected. This methodology worked pretty well in South Korea, Singapore, and China. However, it is still an unsolved problem how to adopt these techniques in the Western world.
5. Even assuming that there were a way to test everyone in the not-so-near future, then what would be the right way to separate positives and asymptomatic people from the rest of the population? Would you separate families and close-relatives? Is society ready to deal with the economic and social implications of a massive-scale adoption of separation, and for how long? These are unsolved problems.
6. Countries are progressively self-isolating. However, how long can this last? There is a clear risk that if one nation successfully applies a set of policies to contain the virus but other nations don’t then the virus will impact the former nation sometime in the future. It is imperative to have more collaboration in order to set up common protocols and common rules for the sharing of data.

We are all in this together.

We human beings