Statistics Shows That Early Lockdown Measures Are Our Best Weapons Against Covid

Statistics Shows That Early Lockdown Measures Are Our Best Weapons Against Covid

NEW STUDY BY LU AND BORGONOVO, USING SENSITIVITY ANALYSIS TECHNIQUES, DEMONSTRATES THAT POLICY VARIABLES ARE MUCH MORE RELEVANT THAN THE INTRINSIC FEATURES OF THE PANDEMIC IN CONTAINING THE NUMBER OF INFECTIOUS

Emanuele Borgonovo and Xuefei Lu singled out the time of lockdown introduction as the key variable in reducing the number of COVID-19 infectious, in a paper that combines a standard epidemiological model (SEIR: Susceptible, Exposed, Infectious, Recovered), machine learning techniques and sensitivity analysis.
 
In a paper for the COVID Crisis Lab, Borgonovo, Full Professor at Bocconi University, and Lu, PhD in Statistics at Bocconi and now Assistant Professor at the University of Edinburgh, used publicly available data for the progression of the pandemic in Italy up to 20 April 2020 and estimated the relative importance of six factors acting as parameters of the SEIR model:
 
protection rate (the rate at which the susceptible population becomes insusceptible due to activation of public health policies such as the imposition of social distance measures or provisions for wearing face masks, the introduction of contact tracing apps, etc.),
infection rate (the parameter controlling how often a susceptible-infected contact results in a new infection, that can be reduced by measures such as ‘social distancing’),
average latent time (the period between the time an individual is infected and the time at which the individual becomes infective),
quarantine rate (the rate at which the infectious portion of the population can be isolated from the rest of the population)
number of initially infected individuals,
time of intervention (the date at which the intervention took place).
 
The sensitivity analysis (a set of methods that allows the measurement of the effects of a change in one or more model input variables on the model’s output) highlighted that policy variables such as intervention time and quarantine rate are much more important  than the intrinsic features of the pandemic. Time of intervention turned out to be 4 times more relevant than quarantine and 8 times more important than the initial number of infectious and infection rate. Protection rate and latent time play an even smaller role.
 
The scholars have also been able to estimate the time lag between the issuance of the lockdown and the full effect of the measure: 5 days. These results are in accordance with discussions in current economic research.
 
“This study”, concludes Professor Borgonovo, “confirms the strength of sensitivity analysis in obtaining insights useful to the decision-makers. Not only does it say that policy variables are the key drivers of pandemic containment, it also shows that there isn’t much interaction between the variables, i.e., that a change in one of them displays its own effects irrespective of the changes in the other variables”.
 
Xuefei Lu and Emanuele Borgonovo, “Is Time to Intervention in the COVID-19 Outbreak Really Important? A Global Sensitivity Analysis Approach”, arXiv:2005.01833v1.

by Fabio Todesco
Bocconi Knowledge newsletter

News

  • The Star That Makes Algorithms Work Better

    A new paper tackles the topic of the shape of the solution space and marks a leap forward in understanding how to design faster and more accurate algorithms  

  • Breaking Barriers: Gender Lens Investing reshaping financial landscapes

    Innovative research explores role of gender lens investing in sustainable growth  

Seminars

  October 2023  
Mon Tue Wed Thu Fri Sat Sun
            1
2 3 4 5 6 7 8
9 10 11 12 13 14 15
16 17 18 19 20 21 22
23 24 25 26 27 28 29
30 31          

Seminars

  • Francesco Decarolis - Competition in Digital Markets

    FRANCESCO DECAROLIS - Bocconi

    Alberto Alesina Seminar Room 5.e4.sr04, floor 5, Via Roentgen 1

  • Unsplittable Flow on a Path Joint work with Tobias Mömke and Andreas Wiese

    FABRIZIO GRANDONI - Istituto Dalle Molle di Studi sull'Intelligenza Artificiale

    Room 3-E4-SR03