Covid-19 in Italy: can we predict what is coming over the next couple of weeks?
Italy is going through dramatic days in its fight to Covid-19, and every day the new cases and new death statistics keep people worried and anxious for what may lie ahead.
The main question in Italy, now that the nation is in lockdown, is when we are going to see the peak, and what this will mean in terms of new cases and new deaths. As deaths follow new cases with a delay of a few days, we can try to make a short-term prediction based on the recent pattern observed, and we can also extend this prediction a few more days based on a possible scenario where the virus becomes more “under control” due to the measures implemented from 10 March in Italy.
This is not based on the dynamics of pandemics — I’ve seen great tools and research around to make long term predictions, and governments are using them to introduce control measures that can limit the spread and keep the health systems functioning.
I’m Italian, I moved to London in 1998, but especially in these days, my heart is with my fellow countrymen and women, who are at the forefront of this battle with Covid-19. China and South Korea managed their battles in a different way and, as most say, more promptly and with more determination than Italy. They are now in a different phase, full of confidence, while Italy is bracing for the worst, with many unanswered questions. Other countries in the rest of the world are carefully observing, as they may follow a similar path — people in Spain, the USA, Germany, France, the UK and many other countries can draw similarities, with a delay of 1 or 2 weeks, from what they see in Italy.
At the time of writing this post (21st March) Italy is already, by far, the country with the largest number of deaths from the virus (4825 vs. 3,261 in China) and it will not be long before it also registers the largest number of confirmed cases (currently 53,578 vs. just over 81,000 in China). We all hope that the tough measures put in place over the past couple of weeks in Italy to reduce the spread of the virus (basically asking everyone to isolate themselves and stay at home) will soon show their results, but at the moment we are still seeing the result of a slow initial response to the initial clusters in the north of Italy, and the lack of preparation: the virus takes time to spread (often through asymptomatic cases) and be registered. And then it takes time to kill.
I’m a software programmer and I like numbers, maths and statistics. I’m not a virologist or a doctor, but like many of us, I’ve been trying to understand what is going on, and trying to read the data to make sense of what is happening and what might come over the next few weeks and months.
The north of Italy, and Lombardy in particular, is where most of the cases have been registered. The health system is so overwhelmed, that, despite the efforts, apart from few exceptions, only serious cases, who are considered for hospitalization, are tested. When new cases are confirmed, they will unfortunately be candidate new deaths a few days later. And it doesn’t take long to go from falling ill to death: reading articles and papers on the web, the average time is around 8 to 10 days.
I looked at the stats of new cases and new deaths in Italy, and in particular at their percentage over the existing cases and past deaths to understand the pattern of growth. I noticed that the growth of these two figures (cases and deaths) aligns fairly well when new deaths are shifted by 9 days. We have about 3 weeks of data for this observation (22 February to 12 March).
Look at the blue and red lines in the chart below, representing the % growth in new daily cases and new daily deaths (smoothed using an average over a 3-day window). The red line (new death growth%) has been shifted back by 9 days. The growth of new cases has been decreasing from 30% or more in February to 13%-15% of the third week of March. The growth of new deaths is following the same pattern with a 9-day delay.
The green line represents the number of new deaths on a daily basis, and on 21 March it reached the number of 793.
Now, because deaths seem to be in lockstep with new cases from 9 days before, if this pattern continues over the next 9 days (22 — 30 March) what can we expect from the absolute number of new deaths? Look at the yellow line (Pred New Deaths) for a possible answer: numbers are going to soar to more than 2,000 new deaths per day on 30 March… a very sad prospect.
This is not scientific by any means — there’s a big “IF” on this prediction (that this pattern continues to be observed) but it’s not out of this world that this relationship between new deaths growth and new cases growth continues as is for a few more days. This will most likely not be true once the number of recoveries becomes significant (at the moment it’s only just over 10% of total cases) or once Italy manages to test significantly more people. Currently, especially in Lombardy, I understand that the vast majority of tests are done to people with serious symptoms.
But there is hope: on 10 March Italy went into a lockdown to try and reduce the spread of the virus. Experts say that results should start to show after 2 weeks, if people follow the recommendations to isolate themselves and avoid social contact as much as possible.
So I tried to project this hope into a decreasing percentage of new cases every day, starting from 14% on 22 March (pretty much as the past few days) and gradually down to 1% on 1 April and then down to 0.1% on 7 April. This is the orange line (Pred % New Cases). In South Korea, with a much smaller number of infected people, it took around 1 week to see a decrease to under 1%. In China, with many more cases, it took around 2 — 3 weeks to see a similar decrease, so hopefully this is possible, if not plausible.
So if this hope materializes in this way, what could we expect after 30 March in terms of new deaths? Look at the continuation of the yellow line after 30 March: deaths would peak at nearly 2,400 on 31 March, and then start to decrease, but would remain over 1,000 per day for another week. Sadly, at this stage, on 7 April, the total number of deaths in Italy would be over 30,000…
Again, I’m not a virologist or medicine expert. I’m just trying to make sense of the numbers available, and tried to extrapolate a pattern I noticed for a couple of weeks.
Where I’m taking the data from:
