The rt index is not reliable | The HuffPost

The recently published data on the Rt reproduction factor is being discussed throughout Italy. We read that the country’s Rt factor has dropped, with great general satisfaction, from 1.72 to 1.43. Unfortunately, however, the Rt calculation in recent weeks has become increasingly unreliable and the decrease in the estimated Rt is completely irrelevant, in fact this decrease is a worrying symptom of the cracks that are opening in the monitoring system: it is not a signal that only authorizes optimism.

This unreliability of the Rt value is clearly written in official documents. If last week we had an Rt that had a 95% probability between 1.45 and 1.83, this week it is between 1.08 and 1.81. As for him exit poll, there is no certainty, and the data is communicated in the form of “fork”As the journalists who present the results on election night have learned well. Unfortunately, in the journalistic communication of epidemiological data the bifurcations disappear and remain only in official communications.

In one week the width of the fork doubled, and this increase is extremely worrying. The summary of the last press release is that the increase in the number of daily infected could have remained almost constant or almost doubled or have done something in between: dramatically uncertain conclusions, on which I fully agree.

If the data, that is, the numbers used to calculate Rt, are not reliable, their estimated value is not reliable. How did we get to this situation? In April, the procedure chosen for the calculation of Rt was perfectly suited to the health situation at the time and was based on the assumption that the data from the national follow-up were reliable. Sadly, as of October 20, this is no longer the case. For reasons that are not easy to identify precisely, but probably related to the excessive number of cases, the monitoring system does not provide accurate information on the evolution of the epidemic.

The Rt calculation system is based on the number of people who develop symptoms, but if, due to various inefficiencies, this number is incorrect, the estimated value of Rt also becomes an incorrect number. This fact is well known to all statisticians who have tried to calculate by themselves the evolution of Rt from the official data of the different countries. The doubling of the amplitude of the bifurcation in the country’s RT in just one week reflects this situation well.

The indicators of epidemic growth are numerous: the number of people who develop symptoms, the number of people who test positive, the percentage of people who test positive, their health status at the time of the swab, age distribution, emergency calls, the number of people entering and leaving hospitals and intensive care, the number of deaths, etc …

Each of these indicators provides us with information on epidemic growth from a different point of view. In the best of worlds, they should be consistent with each other. If this consistency ceases to exist, relying on a single indicator, the one that was the most reliable six months ago, is completely arbitrary.

Therefore, it is absolutely necessary to use the different indicators in combination to obtain reliable Rt estimates. I’m sure the insiders are perfectly aware of this. But it is not easy to do: it takes a coordinated scientific effort to understand which is the most reliable method. But scientific consensus can only be reached through public discussions and this can only happen if all this data is made public. There is a lot of talk about artificial intelligence, neural networks, big-data, but the data that are available on the epidemic are gods microdata.

As the COVID-19 commission of the Accademia dei Lincei clearly recalled, in the absence of transparency, any conclusion becomes debatable at the scientific level and, therefore, also at the political level. What is the point of deciding to open or close the Regions based on an unreliable number, with enormous uncertainty?

But, where do we see that on October 20 the data for the number of symptomatic patients cease to be reliable? Just look at the available charts with an expert eye www.epicentro.it During the last 30 days.

Above is the graphic taken from board of November 13.

the green represents the number of positive cases registered on each day, while blu represents the number of positive people who experienced the first symptoms on that date. Data for the number of positive cases (green) before November 13 is not indicated. As clearly stated on the website, the data on positives after November 5 (green) and on symptomatic (blue) after November 30 are provisional. Thus you should avoid looking at the data after these two dates.

It is an extension of the previous figure where the provisional data have been eliminated and only the final data remain.

These are the same data as the previous figures for a longer period.

As you can see with the naked eye (data in the graph cannot be downloaded!) The relationship between symptomatic and positive cases is approximately constant over time: symptomatic cases increase in the same way as positive cases. However, in the week of October 19, the symptomatic patients stop growing despite the large increase in positives, and also in the following week this difference (in proportion) becomes greater and greater. The slowdown – almost to the end – in the growth of symptoms as of October 20 does not tell us anything about the growth of the epidemic and it does not make sense to use it to deduce a lower value of Rt in recent days: it only tells us that the flow of information in this regard has deteriorated, that starting from the periphery reaches the center.

To better understand where the flow has stopped, a region-by-region analysis must be performed, but this task is extremely difficult as regional data cannot be downloaded. Looking only at the national data, it is impossible to have a reliable assessment of Rt.

I would like to close on a note of very cautious optimism: it is absolutely true that regardless of Rt, there are several signs that could indicate a slowdown in epidemic growth, although it is difficult to reach a quantitative conclusion.

• On November 14, for the first time, not only did the weekly average of new positives decrease, but also the number of new positives was lower than that registered last week.

• The number of people in hospitals and recovery rooms is increasing at a slow rate, as it would be clearer if separate input and output data were available.

These signs are ambiguous as it is difficult to extract the resource saturation effect of a real slowdown. Fortunately, there are also less ambiguous signs, but they could be one simple statistical fluctuation:

• Calls to 118 in the Milan area for respiratory diseases have remained constant at 500 a day for about ten days, but have dropped to 400 for two days. If it is not a statistical fluctuation, it is a very interesting positive trend.

• The number of deaths that previously doubled each week is increasing more slowly, approximately one and a half times a week, and this trend is strengthening.

Without a crystal ball and without the help of the octopus Paul, it is difficult to know if in the coming days the decline in the weekly average number of cases will be more and more decisive and if the number of deaths will continue to follow the curve of cases. It must also be said that these perhaps positive signs are partly offset by a decidedly negative signal: the increase in the average age of those infected, a bad sign that indicates that the infections of many young people escape the monitoring system.

You have to put all this information together to get a coherent picture of the situation. We have to invent a new methodology to accurately estimate Rt, a methodology that is used not to make predictions for the future, but to understand the development of the epidemic and determine the effect of the various containment measures already taken. This information is essential to get out of the epidemic peak as quickly as possible with as little damage as possible. This can only be achieved by involving the entire scientific community starting with the timely publication of all relevant data in digital format.