Start website main content

  • Embeds

Covid-19 in Italy, a statistical study published in Scientific Reports sheds light on the reasons behind the heterogeneity of mortality rates across Italian regions during the first pandemic wave

Publication date: 19.08.2021
Image for italy-graph_0_0.jpg
Back to Sant'Anna Magazine

Scientific Reports published the results of a study conducted by Francesca Chiaromonte, professor of statistics and coordinator of the EMbeDS Department of Excellence, with a team of Italian researchers working at the Sant’Anna School of Advanced Studies, the Pennsylvania State University and the Université Laval.

The first wave of the pandemic unfolded with strikingly different mortality patterns across Italian regions; among the strongest statistical predictors of such patterns are mobility, positivity rates, availability of primary care, and the size of potential contagion hubs in schools, workplaces and hospitals. These are the main findings of a study published in Scientific Reports by a group of Italian researchers working at the Sant’Anna School of Advanced Studies, the Pennsylvania State University and the Université Laval.

During the first wave of the pandemic, the researchers started monitoring epidemic data released by Italian authorities and attempted to correlate it with Google data on human mobility, as well as various publicly available sources of information on socio‐demographic, infrastructural and environmental factors. Most group members are young statisticians and data scientists busy developing techniques and algorithms in an area of statistics, Functional Data Analysis, that studies data in the form of curves or surfaces – so they began applying their own methods to the data, characterizing epidemic curves and exploring differences across various regions of the country.

“Quite regrettably” says Francesca Chiaromonte, Professor of Statistics at the Sant’Anna School and at Penn State, “the data available to the public and to the scientific community fall far short of what would be needed to run analyses that can unequivocally inform policy. This was true in the first half of 2020 and, while some progress has been made, is still true today.” The COVID-19 epidemic in Italy has brought, in addition to unprecedented death and suffering, a sobering awareness of the current limitations in the way authorities gather, process, and make available the massive data that could help researchers and policy makers understand complex phenomena and shape effective responses. “Epidemiological data have been imperfect and imperfectly distributed, the most readily available mobility information has come from Google, and proxies capturing potentially relevant facets of our demography, health, infrastructure and environment are hard to get and merge from central or local governmental authorities or statistical offices” adds Francesca, “the issue is not that the data does not exist; it exists, but there is no reliable and integrated mechanism or platform to make it all systematically available to researchers who wish to study it.”

Notwithstanding these limitations, using data available at the resolution of Italian regions, and leveraging their sophisticated statistical tools, the group was able to pinpoint some important and significant patterns. “We characterized heterogeneous and staggered epidemics in different areas of the country, recapitulating and quantitating what policy makers, scientists, and the public saw unfolding during the months of February, March and April 2020” says Marzia Cremona, who after obtaining a doctorate in Mathematical Models and Methods in Engineering at the Politecnico di Milano and pursuing her postdoctoral research at Penn State is now an Assistant Professor in data science at the Université Laval, “we found an extreme, ‘exponential’ pattern unfolding in Lombardia and the worst hit areas of the north, and a milder, ‘flattened’ one in the rest of the country, including Veneto, where cases appeared concurrently with Lombardia but aggressive testing was implemented early on.”

The study documented strong associations of COVID-19 mortality with mobility and positivity. “These associations persist when using models that control for other factors” says Tobia Boschi, who obtained his Laurea Magistrale in Mathematical Engineering from the Politecnico di Milano and is now a PhD student in Statistics at Penn State, “so our results, along with those of other studies in Italy and elsewhere, do support a role for mobility as a key modulator of the epidemic and for positivity as an important monitoring variable.”

The results of the study also support a role for factors such as distributed, primary health care – which appears to mitigate mortality, and hospitals, schools and workplaces – which appear to aggravate the epidemic working as ‘contagion hubs’. “Of course these findings need to be confirmed and fine-tuned on higher resolution data, but evidence from multiple studies has been accumulating over the last year, and it could inform decision making — for instance, on short- and medium-term investments to boost distributed health care, or strategies to ‘pod’ patients, students or employees'' says Lorenzo Testa, an Allievo of the Sant’Anna School with a background in Economics, who is now pursuing a Laurea Magistrale in Data Science and Business Informatics from the University of Pisa.

“We are already extending our study to cover a broader time span and compare different waves of the epidemic, to see what predictors and factors appear to retain a similar role, or how statistical associations change across subsequent waves” says Jacopo Di Iorio, who also obtained a doctorate in Mathematical Models and Methods in Engineering at the Politecnico di Milano, spent a year as a postdoctoral researcher at the Sant’Anna School and will transition to a postdoctoral position at Penn State this fall. “Our work demonstrates how Functional Data Analysis techniques can offer original and useful insights when applied to this type of data, from Italy and elsewhere around the world” adds Jacopo.

As they continue their own studies, these young researchers are keen on making the tools, algorithms and analysis pipelines they developed available to the scientific community. “I am proud of the quality of the statistical and computational training they received in top Italian universities, and of their willingness to expand their horizons through international training and collaborations” says Francesca Chiaromonte. “At Sant’Anna, we are trying to create a community and provide resources for this new generation of Italian computational and data scientists through the EMbeDS Department of Excellence, which I coordinate”, she adds. “Let’s give them data, and space; they have a concrete chance to make things better.”

Links to the Sant'Anna School website: