BIDSA Fellow Emanuele Borgonovo, Full Professor at Bocconi University, and Xuefei Lu, PhD in Statistics at Bocconi and now Assistant Professor at the University of Edinburgh,
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.
Borgonovo and Lu 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 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”.
Read the article on Bocconi Knowledge at this link.
Full paper available online: Xuefei Lu and Emanuele Borgonovo, “Is Time to Intervention in the COVID-19 Outbreak Really Important? A Global Sensitivity Analysis Approach”.