Number of tests, number of patients admitted, fraction of those tested positive in hospital: what are the data that give us a real picture of the situation of the epidemic? Can forecasts based simply on number of people testing positive really allow us to make adequate forecasts? Never before has data played such a central role in our lives as at this time: knowing how to read them properly and interpreting them correctly by identifying the parameters to be considered is fundamental.
This is what emerges from the document signed by a group of scientists and scientists involved in a data analysis project on coronavirus in Friuli Venezia Giulia. The letter signed by the researchers describes some limitations and possibilities for development regarding the collection and analysis of data in Friuli Venezia Giulia and in general. The initiative was promoted by SISSA.
"We, a group of physicists, computer scientists, data scientists and statisticians of the research institutes of Trieste" as it is written in the document "we have collaborated with the Friuli Venezia Giulia region in the analysis of the regional aggregated data on the outbreak of Covid19. This experience has allowed us to see firsthand what can be read in the data and what are the limitations of the various techniques behind the forecasts that we hear every day".
Data on positive test results are strongly influenced by many factors, explain the scientists in the letter. And some information provided may lead to inaccurate assessments. Others, however, such as the numbers of hospitalized (a datum that is stabilizing), or the fraction of people testing positive admitted to a hospital compared to all the positive (in constant descent from a week) give a clearer picture of the situation. The social structure of the population is equally important: particularly in small regions.
"A significant part of the increases comes from situations such as retirement homes, environments where the virus spreads easily, with unfortunately devastating effects but not representative of the virus circulation in the general population. Analyses that simply extrapolate predictions from positive numbers give apparently clear results but risk being strongly misleading, especially for long-term forecasts, because they ignore the uncertainty and incompleteness inherent in these data."
The researchers close the letter with a recommendation: "The Covid19 pandemic has taken all Western countries by surprise; the sacrifices we are making now are necessary to return to a level where future infections can be controlled. Let’s not be unprepared a second time. We hope that the experience gained at this stage will lead to closer cooperation between health authorities and the scientific community to develop appropriate data collection strategies, supported by informed communication, transparent and responsible for analyses and models".
The full text of the letter (in Italian) is available here
Photo by Markus Spiske via Unsplash