Hot Autumn

 

Not too many years ago,  "hot autumn" had a precise meaning in Italy. It was a way of indicating trade union uprisings/demands.

In our day, we are no longer preparing to see workers' revolts. Still, the hot autumn we are preparing to experience is linked to the virus accompanying our daily lives for 18 months now.

Premise: All the indicators, from the incidence to the hospital beds occupancy, are decreasing. A big difference from a year ago, when the first two weeks of September showed a rapid epidemic growth.

Yet, everyone is worried about the arrival of autumn given schools starting and the resumption of indoor activities. And here is that "hot autumn" returns to make its way into the common lexicon. But what will really happen in the next few weeks? Long-term forecasts are always full of uncertainty, making them is a gamble given the variables that come into play. However, we can say something without fear of being wrong: "this autumn will be different from that of 2020".

It is plausible that the reopening of schools, the full resumption of production activities, and public transport under pressure, will help change the infection trend again this year. However, something has changed. Something substantial has changed. Last year the population was fully susceptible to the virus, unprotected by vaccines. We know how the evolution of the epidemic curve depends on three crucial factors:

1. The average number of contacts of each individual.
2. The probability that an infected person will infect another person by increasing infectivity.
3. The time of infectivity of the disease.

Last year we could only check the first term. And to contain the epidemic, we adopted the closure of activities, which made the curve go down slowly.

With the introduction of vaccines, we can affect the likelihood of infecting and the time of infectivity, that is, how long one remains contagious. A study published in Nature (https://www.nature.com/articles/d41586-021-02187-1) tells us that thanks to vaccines, the infectivity time has been halved, from two to a week, and the probability of infecting is very low. Then, although the Delta is more contagious than the original strains, it has a shorter latency time (the incubation period of the disease in which it is spreadable but without symptoms). From an average of six days, it is now four (https://www.nature.com/articles/d41586-021-01986-w). Thanks to the vaccines, the descent of the curve took place much faster.

So are we out of it? No!

We are in that phase in which it is necessary to understand which of the components driving the epidemic will have the most significant weight. Let's imagine one of those beautiful scales with two plates. On the one hand, the number of contacts per person will increase, with the resumption of productive activities and the school (with everything that revolves around the school). On the other hand, the share of vaccinated people will continue to increase. We move towards the so-called population immunity, significantly reducing the likelihood of getting infected in the vaccinated population (the data already demonstrate this). Which way the scales will hang is early to say, but at least in this hot autumn, we have something to counterbalance the increase in the number of contacts.

And then we always remember that "Unreliable predictions about COVID ‐ 19 infections and hospitalizations make people worry" (www.dx.doi.org/10.1002/jmv.27325)

To open or not to open?

 

To open or not to open? "Calculated risk", but by whom and, above all, how? To assess the risk, someone should tell us what the expectations are, with their confidence intervals, for deaths, hospitalizations, change in GDP, number of failed businesses with and without scheduled openings.

Then reading the newspapers, the feeling is that someone is no longer a researcher or a scientist or a politician, but a "closure" or "opener" regardless of the data.

Let's clarify: the novelty is the possibility of passing in the yellow area, slightly lightened by the dinner at the restaurant in open spaces and by the school in the presence (also in orange). It is not an "everything open".

The restrictions are and will always be based on the analysis of data available at the regional level. Making decisions at the provincial level would be the way forward, but who controls them? Seeing Sardinia in the red zone and Calabria in orange raises more than a few doubts about the monitoring system.  But this is another, unfortunately, usual, story of skills that are not used.

How will the infection evolve if the yellow colour is assigned to a region still at risk? We know that we often close late and open too early. We know that in the yellow area, there is the possibility of an increase in infections.

The questions we ask ourselves are two.

1. How much will seasonality affect the evolution of the contagion? We cannot know this based on last year's experience: last year we were in a very different situation. We reopened in the summer after an extended total closure.

2. If the contagion were to spread more in absolute terms, how much will the vaccinations do now affect reducing deaths and hospitalizations? We can quickly expect an increase in the incidence mainly reserved for younger ages, who currently seem to run a greater risk than a year ago.

Analyzing the age distribution of hospitalizations, the infection network, the length of hospitalization, here is what could answer both questions. The data? They are not public. Many are not collected correctly or are not registered at all.

We also ask ourselves how many deaths we will still have to count every day. In this sense, vaccination is protecting the most vulnerable and the elderly. It is still too early to draw conclusions, but the lethality rate has dropped, and the examples of Israel and the UK bode well. Of course, with 300 deaths a day, it is difficult to speak of a return to normal. Given the vaccination rates, it still takes a couple of months to reach a sufficient proportion of first doses, plus the time for immunity, plus the time to see an effect on the death curve.

So? Really, after 15 months spent analyzing the evolution of the epidemic, are there those who are crying out for the re-establishment of the yellow zone? Ok, it will be a slightly different yellow zone. However, let's face it, even the red areas of these last few weeks have been much less red than in the past; indeed, the only red colour was that the restaurants and schools were closed.

The boys/children are often asymptomatic and are sent to school without any fear. It would be enough to do swabs continuously to reduce the risk. Too many costs, too much time? There are ways to reduce costs and time. Process the swabs in groups. Develop risk priority screening strategies. Modulate the screening frequency based on the observed incidence. There is a way to save without reducing the effectiveness of monitoring.

And at the end of it all, if we don't carry out checks, we can stay here and talk as much as we want, but there will never be anyway to contain the contagion.

From the Red Zone

 

"If he has a science degree and says what you want, then that's ok." This sentence contains what happens daily in Italian communication. One has a specific opinion, and one is looking for someone with a proper scientific title ready to support it.

Our work is not like that. We are not here to support the views of some at the expense of others. To be authoritative, research must leave out personal opinions; otherwise, it loses credibility. We, too, and the results of our analyzes are often questioned. It is mostly part of the game. Talking, telling, doing research exposes researchers to criticism and comments.

The highly unpleasant part is that some comments, even from esteemed colleagues, are not carried at the scientific level but go beyond the academic domain to become personal attacks.

Our firm stance on managing epidemic data may reasonably dislike some and can be manipulated by others.

But we are not on that level. In both cases, our contribution is of a practical nature. It wants to lead to targeted interventions for better management of the situation, which remains an emergency. The controversy, the political manipulation, is of no interest to us.

Someone asks us, "what are you proposing?", It is easy to criticize, much more difficult to build.

In recent months, our scientific works (results verified and scrutinized by other technicians) have been (and are being) published in influential journals in the statistical field and beyond. We have built forecasting models that can (could have been) used by policymakers. We are not the only ones (RobBayes, Covistat). The concrete proposals are manifold. For example, unemployed people could be hired to resume contact tracing. Implement rapid monitoring campaigns in schools, administer salivary swabs to students, and process them with the pooled sample testing technique. In this technique, five (or more) swabs are analyzed together in the laboratory.  If the pooled sample is negative, all the individual swabs are negative, and no further analysis is required; if the pooled sample is positive, the five swabs are examined again one by one to identify the positive (s). (a technique that saves a lot of time and money, and that we had proposed, unheard, already a year ago).

Furthermore, it would be appropriate to use university laboratories so as not to further burden public health. Establish surveillance samples managed by Istat, which has all the skills to select and manage them. Again employing the same personnel trained ad hoc and hired for the occasion in the operational management. Create a network of regional laboratories that, in turn, sequencing a large number of positive swabs weekly to determine the effective spread of the variants. In short, the concrete proposals are not lacking at all: for months, the statisticians, all together, tried to raise their voices and made themselves heard. Our group is just a vehicle for this voice.

Yet, some still think it is all based on personal interest, a longing for visibility. Those who know us know that we have a strong sense of institutions. We identify ourselves with the state with the institutions we work for. Like everyone else, we can't wait for this moment to pass to return to our lives, made up of dinners, soccer, sea, mountains, family, tango, aikido, and above all, travel.

And for today from the red zone (of shame), that's all.

A year has gone by

 

Weekly incidence per 100thousand inhabitants starting from October 2021

A year has gone by since the start of the pandemic. We managed to set up a research group starting from a Facebook chat without  (almost) ever meeting in person. We set up monitoring and forecasting models, verifying their reliability and accuracy, day by day, before making them public. We had them evaluated and commented on by other colleagues, who felt they were valid for publication in prestigious scientific journals. We started national and international collaborations, made a ton of seminars, talked to journalists, and made friends with many beautiful and new people. In short, we started "a pandemonium" to study, propose, imagine solutions and actions in this terrible moment. And we are only 5. In Italy, there are many people of great competence on various subjects (sample design, for example), and they, like us, think, propose, imagine. Brains at work everywhere, some international excellences involved, a truly extraordinary scientific debate. Still, something is not working. There really is something wrong.

With all this intelligence, expertise, and tons of ideas available, what is the only solution currently available to "manage" the epidemic? Lockdown…. Despite this is  (see the plot above) the weekly incidence situation in the Italian regions. The CTS and consequently the government use Rt estimated using Lombardy information at the beginning of March 2020 as a starting point and other methodological "subtleties" that we better not comment on. There is no statistical culture. There is a lack of adequate skills to analyze data. We have proposals. All of us statisticians have been working since March 2020 to make them available to decision-makers. Are you sure that there is no better approach than the use of Rt and a deterministic flow chart to decide on the everyday life of all of us?

Birthday

 

A year ago - on March 3, 2020 - StatGroup-19 was born. It was born spontaneously, without the awareness of a future, not even near. Then, listening to people's confusion and doubts, we tried to understand and explain the epidemic. And a year has passed, quickly. It's time for birthdays, and birthdays are also moments for reflection. A year has passed in which we have studied, written, discussed, criticized, collaborated, published. An intense, tiring, tough year for us as well as for everyone. But there is also the satisfaction of having tried to give a fact reading. To explain reality complexity, foresee and understand how to manage the situation.

It's been a year, and how are we? Reading posts and some newspapers, it seems to be March 2020. But we are in 2021:

• We have vaccinated almost 1 million people over 80s.
• We have targeted closures.
• We know much, much better the virus.

What does not change is the inability to communicate correctly by both the institutions and some newspapers. The pursuit of the headline is paroxysmal, and now if you don't shout "we'll all die", you won't be published. This crying to call "Wolf, wolf", for which the cries were the same in August, and October 2020, did not help perceive fundamental nuances. We are always in the same positions. No, we will not all die. We will have to be careful for a long time to come. We now have all the tools to do this. We will get out of it. Will the economy also emerge? The social fabric? If we can seize the opportunities we have, we could grow instead of collapsing.

We will continue to read the data, make short-term predictions, and write posts every now and then.

Happy birthday to us and to those who read StatGroup-19 😉

Scientists and stage

Scientists are creatives, a bit like artists. As artists, they live in very competitive environments, developing great determination during their working lives. And, again, like artists, many develop a hypertrophic ego. Scientific rigour takes a back seat; professional ethics is bent for the benefit of oneself and the headline in the newspaper. One becomes accomplices of low-grade journalism, calling "unequivocal evidence" something that is ONLY a correlation. Thundering, later on, against an equally pompous colleague, shouting "correlation is not causation" (in English, which is more like modern culture).

The scientist is human, subject to desires and passions, and, above all, to responsibilities. Particularly now, in a moment where the confusion is at its peak. A time in which the ego-maniacs, by shouting everything and the opposite of everything, have made science an opinion.

Lockdown: yes, no, maybe. The problem is why.

On this cold February Sunday, the most prominent Italian national newspapers' reading provides insights that we didn't see in a while. After the lottery on ministers of the new government, the catch click-phrase is back: "let's close everything, national lockdown, the English variant leaves no other choice". The situation is known to everyone. Skytg24 tells the state of the epidemic every day in "The Numbers of the Pandemic". We are in a stalemate. The curves of the various indicators are slowly falling. Everyone's attention must be maximum because it takes a little distraction to see the situation worsen again.

The graph we report shows a simple classification of the Italian regions according to two thresholds: intensive care occupation (above 30%) and incidence (above 250 per 100 thousand inhabitants). If a region exceeds one of the two thresholds, it becomes orange, if it crosses both red. Otherwise, it remains yellow. Nothing complicated, a simple representation, which, monitored weekly, provides some information on the epidemic's progress. We have seen worse times.

Today we talk about the lockdown, again. Why? With what goal? In analyzing the various opinions, it is precisely the goal that does not seem clear, nor even explicit. Let's try to imagine what objectives can be achieved through a lockdown.

Objective 1: local extinction of the epidemic. You want to eliminate the virus from the area of ​​interest. For this purpose, a total lockdown is useful, to get up to zero cases per day for a sufficient period (one / two complete incubation cycles, i.e. 14/28 days). Strict isolation and quarantine measures are obviously fundamental. Stringent epidemiological surveillance is subsequently indispensable, particularly on entrances from outside the territory.

Objective 2: mitigating the effects of the spread of the epidemic, with the minimization of deaths. The lockdown is not necessary, except as a last resort. A good Test, Track and Treatment / Isolation / Quarantine (TTT) strategy is sufficient. When the TTT capacity is lost in an area (there are now standardized indicators). A prompt and brief lockdown (limited to that area) is necessary until control of the situation is resumed.

Objective 3: mitigating the effects of the epidemic's spread to avoid the collapse of the health system. In this case, too, the lockdown is not necessary, except as a last resort. Serious monitoring (we have the indicators) of the pressure on the health system, linked to the definition of risk levels of collapse, generally avoids the lockdown.

Here http://dx.doi.org/10.23812/21-3-E it says "It has been noted that the spreading rate of the British variant could be greater than 70% of cases compared to the normal SARS-CoV -2 virus, with an R index growth of 0.4 ." As it is written, this sentence admits a substantial lack of strong evidence, obtained through well-designed studies and analyzed with solid statistical methods. We would not like to accept the idea that if one cannot explain something, one can blame the current virus variant.

One, none, one hundred thousand or two, is better than one?

 

What color are we today? It is the question we ask the first person we see every morning. As in the children's game "witch commands color," we stopped asking ourselves the reason for the choice of color, almost resigned. We talked about 21 indicators, designed and built to decide what we can and cannot do. Too many, redundant, and based on too old data to provide a timely response in an emergency. We discovered what Rt was and that a few decimals of Rt are precious to remain "yellow" and not pass "orange." As if there was only one way to estimate Rt. As if fixing the generation time needed to estimate Rt at the estimated values ​​for Lombardy in February made some logical sense. Rt, like the North Star of Italian epidemiology. Perhaps the only thing that everyone knows how to calculate given there is EpiEstim in R that does it for us.

In our opinion, decisions are best made on observed and updated data. Remember that data heterogeneity is already considerable (public data are anything but "clean"). Then why make everything even more uncertain by estimating values ​​whose reliability we do not know?

Suppose our aim is to block an area when the pressure on the health system is potentially unsustainable. In that case, we must use indicators that take it into account and indicators that allow us to understand the infection's state. There are many ways to follow, but we have been taught that we must start from simple things and, if not sufficient, move towards more complex approaches.

So let's start from here: not just one indicator but two. For example, let's take the weekly incidence (on the population residing in a region) and the average employment of the weekly intensive care units compared to the number of beds. With these two indicators on a graph, establishing, as an example, the two thresholds (30% for intensive care and 250 infections per 100 thousand inhabitants) already used by the government or proposed by the CTS, we obtain what is shown in the graph: Marche, Friuli, Trentino and Veneto in the red zone, 5 regions in the orange zone and the rest yellow.

It must be said that the situation is complicated. For example, Calabria carries out less than 1500 swabs per week per thousand inhabitants, bringing up the country's rear. It seems that it does not carry out surveillance activities. Maybe two indicators based on observed data are not enough, but at least we know exactly why we are yellow today and orange tomorrow.

Vccines and Communication

 

A brief note on how mass communication is often approaching very delicate issues, such as vaccines, in a very rough way.

Two essential definitions:

Incidence: the ratio between the number of new cases observed over a fixed time-window and the number of people in the reference group (or population).

Prevalence: the ratio between the number of active cases in a given instant of time and the number of people in the reference group (or population)

How is the effectiveness of a vaccine measured? It is measured by the attributable risk. The attributable risk is estimated as (incidence in the placebo group - incidence in the vaccinated group) / (incidence in placebo).

Corriere della Sera (https://www.corriere.it/dataroom-milena-gabanelli/covid-vaccino-quante-probabilita-ci-sono-chi-guarito-ricontagiare-immune-immuni/f1fd386a-42c5-11eb-a388-78033ff67873-va.shtml) offers us an interesting analysis, many numbers, many comments, little statistics. At first glance,  it is to be trusted, especially since it is promoted by Milena Gabanelli. Still, there is something wrong. What's the point of comparing the vaccine's estimated efficacy (the attributable risk) with the number of reinfections (the incidence)? The answer is simple, it doesn't make any sense. It's like comparing pears with fried potatoes.

Even if you want to derive the attributable risk of reinfection, you have to be very careful.

 Indeed, incidence always refers to a specific interval of time. Remark that the 1.8% reinfections reported by the Corriere are not referred to the same interval of time of vaccine evaluation. The attributable risk calculation only makes sense as long as the incidences are calculated over the same time interval, with similar circulation and transmissibility of the virus (therefore in the same geographical area as a minimum). The incidence is not intrinsic to the virus but results from the interaction between it and the population.

The very question posed by the Corriere makes little sense.

A vaccine is evaluated in terms of risks (of adverse events), costs, and benefits. The Pfizer vaccine, which is approved, and the Moderna, which is being approved, are safe and effective vaccines. The benefits in health, social and economic terms are so much greater than the risks that it is useless to even discuss them. Only those who have prejudices, or delusions of conspiracy, can think the opposite. EMA and AIFA are independent bodies that evaluate benefits and risks and then make informed decisions based on scientific evidence.

The Sweden model has largely failed (https://doi.org/10.1093/ectj/utaa025). The way to return to normal is the vaccine. We have just undertaken it, it will take time, but we are sure that we will get out of it.

Open Letter

 Since the beginning of the pandemic, the Italian Statistical Society (SIS) has repeatedly offered its expertise to help decision-makers and scientists to manage and study the situation. The former never listened. Yet, hundreds of scientific works have shown how such skills were essential in understanding and predicting events related to the pandemic.

In light of this premise, the SIS invites all civil society to sign the following open letter. To sign go here

Fight against COVID-19: high-quality data is needed for analysis and adequate skills to analyze it

The emergency due to the COVID-19 pandemic has highlighted the fundamental importance of the availability of reliable data and high skills in analyzing them to allow us to understand the pandemic, predict its evolution, prepare tools for both health policy and economic policy to face it, and evaluate the effects of the choices made.

It is increasingly evident that it is vital to offer competent support for a data collection inspired by quality criteria. We need to integrate available information using statistical criteria that protect this quality. And it is even more evident that, alongside the collection of high-quality data, there is a need to reclaim space for the scientific skills necessary to analyze them.

Why accessible data

To a large extent, the data necessary to construct adequate information are already collected by government agencies and bodies. Still, they are not made available to the scientific community. Confidentiality issues, and further unknown considerations, turn raw data into inaccessible information.

Currently, the available data are collected with the declared purpose of surveillance. Still, suppose the quality, the comparability between geographical areas, and the fundamental defining aspects are not guaranteed. In that case, any analysis of these data will be limited to monitoring the status quo, producing more projections than predictions. To study the epidemic's progress in detail, information is needed as detailed as possible to follow the individual pathways of contagion and clinical evolution.

On an aggregate level, the figures updated daily by the Civil Protection are available to all. We recognize and much appreciate the enormous work of data collection and dissemination carried out by this Agency. However, we note how, at this point in the evolution of the pandemic, what has been made available by the Civil Protection is no longer sufficient to make the government's decision-making mechanism and the scientific understanding of the evolution of the pandemic itself transparent.

In particular, based on this data, it is not possible to carry out some crucial activities.

• Reproduce the quantitative bases of institutional decisions. This emerged in all evidence as regards the recent division of the country into three zones. How indicators are defined and constructed, and the criteria for determining final decisions must be transparent. The disaggregated data with which these indicators are fed must be made available. Only in this way can the scientific community be able to evaluate the methodologies used.

• Ex-post assessment, quantitatively and rigorously, of the effects of decisions. An example of fundamental importance in this area is the choice of whether or not to close schools. Many researchers are trying to give a rigorous evaluation of the "school" effect; however, numerous scientific research on the subject does not yet provide shared conclusions. They are all based on aggregate data analysis.

• Understanding still obscure aspects of the phenomenon. The Italian scientific world is rich in skills that could usefully investigate essential elements of the phenomenon based on disaggregated data in collaboration with the institutions and agencies involved in managing the epidemiological crisis.

Why adequate skills

Statistical skills are currently in high demand and very difficult to find around the world. They have become increasingly exclusive and rare given the ever-increasing demand, reinforced by the current COVID-19 emergency. For example, Pfizer, a pharmaceutical company at the forefront of vaccine development and distribution, will only share its data in research groups where a biostatistician conducts the analyzes. In Italy, the data currently collected in the wake of the emergency is affected by many problems and high variability. Therefore, they need, even more than other biomedical data, specific skills to correctly deal with elements of confounding, imbalance, and high variability. All these aspects cannot be managed correctly without having advanced statistical skills.

Timely and effective, methodologically reliable, and shared answers are obtained when the right skills are involved in collecting and validating data and the same analysis. The scientific process requires numerous steps, in each of which specific skills are necessary for a correct construction of the information tools.

• Definition of the problem. First of all, it is essential to define what needs to be observed to answer the questions of containment, monitoring, and forecasting of the epidemic and its impact in the social and economic sphere. Diversified skills are needed in this process. Highly multidisciplinary teams, within which scientists from different areas can interact, are necessary to address all aspects of the problem. In this phase, on the one hand, the primary data required for the analyzes must be defined and, on the other, the construction and implementation of harmonization protocols between the different data sources.

• Management of databases. Specific computer and statistical skills are required to construct and manage data archives with massive flows of information. The data must not only be stored/saved but, above all, validated quickly to give timely answers and to ensure public access.

• Information analysis. In this phase, the ability to define and develop models capable of grasping the underlying characteristics of the phenomenon of interest, highlighting potential causal relationships, defining specific estimation procedures for unknown quantities and indicators, and building predictions that take into account the uncertainty that accompanies each estimate.

• Sharing of information. Different analysis models need to be compared, for example, in terms of predictive ability, interpretability, and robustness. To this end, it is desirable to establish periodic meetings, at least twice a week, between the researchers who develop the models and the institutions that could use them, openly and transparently, to share the best solutions.

• Dissemination of information. We are supporters of access to data by the entire scientific community. Accepting this request would allow greater transparency on the part of politics. It would enable civil society to obtain reliable and certifiable information. However, accessibility must be accompanied by an incisive and growing promotion of quantitative culture in all areas, starting with communication operators and political decision-makers.

It should be noted that this document asks for access to detailed data, and this access is not new to the national information system. In fact, on matters of an economic nature, the information is available in great detail. This point allows interested parties to analyze and process any type of issues (for example, data produced by ISTAT, Bank of Italy, Chambers of Commerce).

It should be strongly emphasized how the right skills are fundamental for analyzing such a complex phenomenon as the COVID-19 pandemic. The enormous variability observed at global, national, and regional levels must be incorporated into the assessments that lead to political and economic decisions. Knowing how to distinguish between association and causal relationships concerning observations and variables included in the analysis models is fundamental to avoid decisions based on random variations and/or spurious effects.

Science, Statistics and Democracy

 This beautiful monologue by Alessia Ciarrocchi at #propagandalive gave us the starting point for a reflection. The Science of which Alessia Ciarrocchi tells us, the one that embodies democratic values, was born a few centuries ago when Galilei revolutionized how we looked at nature. Galilei formalizes something that moved in the thought of his time: the scientific method.

Since then, studying reality has followed a rigorous path based on the observation of a phenomenon, the formulation of a possible explanation of the same, which must then be validated by collecting (appropriately) observations and analyzing the results of the observation (experiment). From these results, it is decided whether to accept, reject, or partially modify the explanation given and then start again with the same procedure.

In this description of the scientific method, it is immediately apparent where Statistics enters. It is the pillar of this research path: In part, it enters into the definition of the mathematical model that describes the initial explanation, provides the methods for designing the observation, and the correct techniques to analyze the result in the light of the formalization given by the mathematical model initial.

Statistics is not a specific science understood as a homogeneous body of knowledge of a field of reality. Still, it represents the methodological foundation of the whole of Science, and for this reason, its role becomes central today. Statistics allows the quantitative verification of political decisions, allows for dismantling the hoaxes, in short, allows the control of democratic values, especially now, at the time of the "data society."

Data, decision makers and statistics

 As StatGroup-19, we always had a goal: to not lose our temper. Ok, we didn't make it. Last night we lost it!

----------

In the   17/11/2020  episode DiMartedì broadcast, yet another approximate and, in some ways, tragicomic debate on the "data" of COVID19 took place. This time, we went very far, bringing the discussion on the methodological reasons behind using some indicators (such as Rt) to define the Italian regions' levels of epidemiological criticality. To discuss the issue, journalists and experts (no data experts). The debate was characterized by a complete absence of statistical culture, thus becoming "comical" at times for those who know the subject. At times it was "tragic" because it brought to light, even more, the cultural chasm that is engulfing everyone from decision-makers to commentators in this very delicate moment for the country. The absence of quantitative culture and numerical illiteracy are genuinely a social emergency, now amplified by the need to make decisions based on numerical scientific evidence.

But let's get to the evening theater:

Prologue: the eternal contradiction. In the common imagination, it is recognized that statistics is a fundamental tool for explaining and predicting the behavior of complex phenomena, such as an epidemic. Being statistics a subject of study in many university courses, it is also true that everyone thinks they know it. Therefore it is often considered unnecessary to involve the community, the people, who are dedicated to studying it professionally. This is also true when dealing with issues at the highest levels of criticality. Knowing some basic notions of Statistics does not make statisticians or data analysis experts. Just as knowing how to drive a car does not make Formula 1 drivers.

Act (of faith). The debate runs off quickly. The journalist Damilano tackles the issue and asks Professor Richeldi (pulmonologist and member of the CTS, ed) to explain the reasons behind the choice of "parameters and if that index over which we all squirm (Rt, ed) is still realistic. " Professor Richeldi's reply will become an eternal citation in every Statistics course, in memory of how often decisions are made without having an appropriate statistical culture. Turning to the journalist, he reveals: "it is good that we understand each other; I don't know if you knew about the Rt index last year? I personally didn't […] now if we do not trust and if we do not entrust the experts with a direction, in this country, in my opinion, we will hardly jump out". Then pressed by Floris on the quality of the data and the role of Merler (defined by the conductor as "super professor"), he closes with an epic ending: "the super professor has in his hands the data evaluating their quality,  and reliability, and how they are inserted in a multiparametric parameter (! ed) ". Here is the mirror of the moment, meaningless words used to justify unclear choices, relying on faith, faith in a single person, the only one in possession of the data.

Epilogue: Rt. So it's no good at all. Relying on experts is fine but taking a leap of faith towards what a single expert, that has been chosen as the sole repository of the statistical truth on the COVID issue claims is not acceptable from a scientific point of view. The approximation with which such an important topic as that of the scientific reasons behind the use of indicators is approached and discussed cannot pass. In this situation, some commentators also fail to understand the relevant role of statistics experts, even in the public debate. On Rt, we must be able to discuss. First of all, Rt cannot be calculated. It can be estimated that it can be given a statistical approximation accompanied by a level of uncertainty. As an estimate, it is based on a procedure (model), and different methods will provide different estimates, with different uncertainty assessments. In this context, the act of faith cannot be done. It should be explained and clarified why this indicator should be used and why the assumptions made are the best possible, as well as guaranteeing their empirical verifiability.

Conclusion. Unfortunately, it is more and more frequent to attend debates on relevant issues from a decisional perspective in which statistics are present, except to realize that the only ones absent are the statisticians. Not knowing/wanting to recognize the scientific importance of their role in studying complex phenomena, in the management of critical issues, and in decision-making processes is extremely serious. Statistics is the methodological essence on which all scientific research rests, both in the experimental and social fields. It has the same role that Philosophy has for science. It is its theoretical and conceptual foundation: research without statistics would be blind. We must therefore understand and accept that in the (also) public debate on methodological issues, the only ones who can explain and clarify doubts to decision-makers and citizens are the statisticians.

Let's make it "relative"

 

In the last few days, with the observed incidence's continuous increase, we witness a "confrontation" competition. There is a strong need among journalists and commentators to reference the first phase of the epidemic to relate what we are observing today. In many televised debates, comparisons between absolute numbers are presented, which, in general, make very little sense. These comparisons try to relate the current phase with the critical stage of the highs of late March and early April. In our opinion (and not only ours), this attempted comparison is not valid at all. In the first phase of the epidemic, cases of positive SARS-Cov-2 were intercepted through "diagnosis." In contrast, today, they are mainly found through "screening" and tracing.

Comparisons must always be made in relative terms, where the adjective "relative" is used to indicate that the quantities to be compared must be considered in "relation" (related) to the procedure from which they are derived. For example, comparing cases of COVID-19 only makes sense if these refer to the number of tampons used. But not all swabs, only those used to find new cases (the situation where 1 swab = 1 person). This is to check the differences between a diagnostic and a screening strategy.

The need to analyze relative data, or relationships between connected quantities, such as odds, proportions, rates, risks, odds, and all further processing, has always been a central theme of our posts and public interventions. In this sense, if we want to find some connection between the first wave and the current phase, we should start from the relationship between positive cases observed and swabs carried out.

In the graph, we reported this relationship, relative to the total swabs (in red) and the swabs for test only (in blue, excluding the control swabs to verify negativization to the virus). Recall that the data reported only the total swabs in the first phase of the epidemic, with no separation between test swabs and control swabs. The "red" ratio, although not correct from a statistical point of view, is still informative of the trend of the epidemic. It represents a lower limit to the "blue" ratio, which estimates the "positive rate."

Analyzing these two relationships' dynamics, the only similarity that can be guessed is between the initial phase of late February-early March and the current one. This seems to indicate that we are at the beginning of a second phase. Therefore any comparison with the maxima of the first wave is really inconsistent on a statistical level.

We are then in an initial phase of acceleration of the infection, but still manageable. Some restrictive measures can be though for specific commercial or leisure activities. Yet, they are far more sustainable than a new generalized closure (which may be the only "final weapon" if this rise does not stop). After all, in general, to citizens, it is required very little: wear a mask, respect the physical distance, wash your hands often. We can do it. Just try seriously.

Syndemic

 Covid-19: beyond the epidemiological aspect, there is a social aspect. On September 26, on #Lancet, one of the most prestigious and reliable scientific journals, there was a discussion about a part often overlooked in the various analyzes on Covid-19 (https://doi.org/10.1016/S0140-6736(20)32000 -6): is it really a pandemic? We usually take it for granted, but it could be a syndemic (a word unknown to most) in reality. Syndemic is defined by Merril Singer in the 1990s, "A syndemic or synergistic epidemic is the aggregation of two or more concurrent or sequential epidemics or disease clusters in a population with biological interactions, which exacerbate the prognosis and burden of disease."

Covid-19 could be just that. It rarely kills on its own; it does so more often in conjunction with other diseases, ailments, and in any case, overt or hidden health problems. Most of these secondary diseases are, in large part, related to people's social status. Horton, in the #Lancet discussion, adds: "The syndemic nature of the threat we face means that a more nuanced approach is needed if we are to protect the health of our communities. [...] a syndemic approach reveals biological and social interactions that are important for prognosis, treatment, and health policy." 

Covid-19 is not just a viral disease; it is the litmus test of social diseases. Italy leaves no one behind, whether you are no-mask or no-vax, health care is public and will treat you, regardless of class. Let's not forget it….

 

From here

Keep calm and wear the mask

 

These days, we witness reasonably substantial growth in the number of cases of COVID-19 (and in the positive rate) throughout Italy. Some regions are definitely on alert (Campania); others are only on alert (Lazio, Sardinia, Liguria, for example).

However, this situation must be distinguished from what happened in March-April, when there was, as now, more than 3000 daily positives.

 In the most dramatic days, the policy of tampons was very different. We only had tampons for the symptomatic, often severe, and we could not look for the asymptomatic. Today it is very different, practically everyone gets tampons, and we add, fortunately. We have gone from 30,000 tampons at the end of March to about 130,000 today. We have been monitoring the positivity rate for weeks now, which is the ratio between posts and the number of people tested. Today we are at 6.5%, definitely up. This allows us to understand how much COVID-19 is present in the population, always remembering that, fortunately, asymptomatics are about 90% of positive test cases. A large number of swabs, an effective screening policy allow us to protect the most fragile and prevent intensive care clogging's catastrophic effect.

How can we contribute, by doing our part, to avoid the spread without control of the epidemic? Here too, it is not complicated; there are a few very simple points:

1. Wearing masks and often washing our hands 2. Avoiding gatherings and keeping distance if there are many of us 3. Isolating those who are positive even if asymptomatic 4. Often airing the rooms where we spend time, from the office to the classroom, the workshop, the bedroom, etc.

In Italy, a great help can come from the use of the app #Immuni !

We wonder if it is indispensable to use the bugbear of a new lockdown, feeding uncertainties and fears in the population. The goal must be to raise awareness and inform people. Frightening them doesn't help them understand what needs to be done. The country is still doing well. Everyone (or almost everyone) is rowing on the same side. The health system is strengthening and providing quick and effective responses. So just be careful and follow the few rules, those of common sense, that we have given ourselves and, in part, have imposed on us. Then there is the economy, and yes, that is at least as scary as the epidemic. We cannot afford a new generalized closure; it is evident, we must be aware of it. And then just put the mask (correctly). The situation is serious, but it takes very little to remain open and competitive.

Evolution of a WHO indicator: the cumulative incidence rate at 14 days

 

The 14-day incidence rate per 100,000 inhabitants is the official indicator through which  WHO evaluates the speed of the spread of the Covid-19 epidemic in the various countries of the world.

In general, the incidence is the number of new cases observed. It can be daily, monthly, etc. In this case, W.H.O. considers the new cases observed in the last two weeks and expresses them per 100,000 inhabitants, thus removing the population size effect. Being represented in units of time (14 days), this ratio is called "rate," as it describes the variation in the number of new cases as time varies.

The graph shows this speed indicator's trend from the beginning of the epidemic to today for 30 countries: when the indicator grows, the epidemic accelerates. When it decreases, the epidemic slows down.

At the moment, the ranking of these 30 countries is as follows.

Italy occupies the 13th position with 39 new cases every two weeks per 100,000 inhabitants. The worst situation is that of Israel, with the speed indicator equal to 883.

---

NB: the zeros in the ranking are approximations of decimal values

Comunicating science during the pandemic

(photo from here)

Every time we write a post about Covid-19, there is a question that invariably guides us: "How do we explain the epidemic so that even without technical education, people can be aware of what is happening?". Indeed, there is much more behind this question. There is the relationship between scientific communication, often technical and cold, and mass communication, driven by feelings and fears. Scientific journals are the place to communicate science. Impactful scientific work,  a study that shakes the scientific community, has a high degree of reliability and, generally, requires a lot of time, a lot of work, and a lot of data.

Newspapers, blogs, social networks, on the other hand, are the main places to communicate with the mass (the mass media). Obviously, the mass media cannot wait for the times of science. Mass communication is "fast" and, at times, driven by the desire for the "scoop," the sensational. In a normal situation, the two ways of communicating don't often integrate.

This epidemic has brought the two worlds closer together: the scientific scoop's sensational and research have pervaded scientific journals. Scientific research's solidity and rationality have made its way into communicators who had never dealt with science. However, this is not necessarily a good thing.

Many high-profile scientific journals have stolen the way of communicating from the mass media. They give ample space to "intellectual exercises" on COVID (in "who has the longest model" style), based on uncertain hypotheses,  on very few data and models approximate. Science, not the Daily Unknown, is full of them. In another era, scientific journals would have rejected all these intellectual exercises as inconsistent without batting an eye. At a time like this, giving space to poorly constructed studies without clear premises on the validity of the data and methods is an irresponsible act. The exploitation of these works is around the corner. We must be aware of it. The mass media do not have the skills to judge whether one research is valid or not. Somehow, they trust scientific communication, even if they do not always fully understand it. A communication distortion was, therefore, created.

From this distortion, a new mission is born, which is ours, but also of many other scientists (see the end of the post): to communicate concepts, models, results, which simultaneously have the same scientific solidity of well-done research and the typical immediacy of the phenomena that shake the masses. Communicating science through mass media is not trivial. Excellent examples are (some in Italian and some in English):

- https://barbaragallavotti.wordpress.com

- http://www.mathisintheair

- http://maddmaths.simai.eu/

-      https://cattiviscienziati.com/

-      https://www.isi-web.org/index.php/news-featured/20229-statisticians-react-to-the-news-every-week

-     https://statmodeling.stat.columbia.edu/

all blogs that take the space and time to explain things well.

Scientific communication on social media is still a very young and inconsistent creature; it does not have a well-defined identity. We are trying to give it one, starting with warning against the "ad hoc" manipulative sensationalism and elaborations that sometimes come together (see Sweden's case, a failure and not a panacea).

Let's not forget the responsibility that all of us communicators (of science and the masses) have. Better a little less media visibility and good research than an approximation and the search for the scoop to impress, sometimes upset (terrify) the masses. We reiterate that we do not cheer for science. We are not sellers of opinions, "science is not democratic"; it is competence, and this is what must be valued and defended.

 

From August to September*

 In August, we witnessed a gradual but constant increase in the infection from Covid-19. The number of new positives (in the definition of Civil Protection or daily incidence in epidemiological terminology) rose from about 250 cases at the end of July to about 1350 cases of the first week of September. The increase in daily incidence, in addition to almost tripling the positive totals numbers (in the definition of Civil Protection or prevalence in epidemiological language) going from about 12,500 cases to about 32,000 cases, it is projecting on the number of people hospitalized in intensive care.

In the first graph (Fig. 1), we show an indicator useful for controlling the trend changes in our interest quantity. The indicator is based on the crossing of two delayed moving averages: a short one (at seven days) and a long one (at 14 days). This method is widely used in the analysis of financial markets to confirm changes in the direction of the trends of the securities in which one is interested. Take, for example, two moving averages calculated on the value of a company's shares or a financial fund's value. The first of these two moving averages concerns a short duration while the second a more extended period. When the short moving average passes above the long moving average, there is a buy signal; the stock is increasing in value. Conversely, when the short moving average goes below the long average, there is a sell signal—the value drops.

 

 

In this graph, we analyze the number of hospitalized cases that occupy intensive care every day from the beginning of July to the first week of September. The crossing of the short and faster moving average (in red) from bottom to top with the longer and slower moving average (in blue) confirms a change in trend in intensive care admissions at the beginning of August, currently showing a strong tendency to increase.

 

To analyze this trend change in intensive care units' occupation, it is also interesting to look at the relationship between the number of cases admitted to intensive care and the number of patients hospitalized with milder symptoms. Suppose we consider the whole group of hospitalized people (in intensive care and with milder symptoms). In that case, this ratio represents an estimate of a crucial epidemiological indicator. That is the probability share of the "hospitalization in intensive care" event (in the epidemiological terminology odds, a term derived from the language of betting). The probability share of an event of interest is the ratio between the probabilities in favor of that event and the odds against it. To better understand how odds are used, let's think about betting. Imagine we have 100 men and 80 women; in these two groups, 90 men drank wine last week, and only 20 women did the same. The odds of drinking wine among men are 90 against 10, 9 (90: 10 = 9), while among women, it is 30 against 50, that is, it is 0.6 (30: 50 = 0.6). So if I had to bet who will drink wine during the next week between a man and a woman, I would undoubtedly prefer to bet that it will be a man. Indeed the share of men drinking wine is much higher than that of women. I can also calculate the ratio between the two quotas and see that men have an aptitude for drinking wine 15 times stronger than women (9: 0.6 = 15).

 

Building on this way of thinking, we try to analyze hospitalizations in intensive care. We are now observing the second graph (Fig. 2), where we report the respective probability share trend from the beginning of July to the first week of September. We can see that the level is still relatively stable, ranging between 5 and 9 (out of 100). The latter shows that there is still no significant change between critical and mild in the composition of total hospitalizations.

One sign of concern comes from the odds ratio. From the beginning of August, you can see a gradual increase in the level of critical hospitalizations compared to mild ones; it rises from about five against 100 to a value of about nine against 100. Between the low at the end of July and the first weeks in September, the ratio between the two quotas is approximately 1.75. In the last month, the latter means that the composition between intensive care admissions and regular ward admissions has changed, with intensive care admissions increased by approximately 75% compared to hospitalizations with mild symptoms.

*This post is born from a chat with Enrico Bucci https://cattiviscienziati.com/author/eb72enrico/

Statistical friends have great ideas

This is just a short post to share a new blog first publication  Our friends E. Ashley Steel and Peter Guttorp talk about  "Sharing statistical thinking – an essential skill for reading the news". We simply love the topic and the way they talk about it. We highly recommend reading the post!