Dynamical modeling, calibration and robustness analysis of COVID-19 using Italian data

Abstract

Italy was the first European country severely hit by the COVID-19 pandemic. In late February and March 2020, the number of people requiring hospitalization and mechanical ventilation has soared, putting a strain on the Italian health system. In the absence of pharmaceuticals therapies, the government implemented a set of mobility restrictions for transmission containment. Starting from the need of predicting hospitalization and ICU rates for the Umbria region in Italy, we propose the application of a computational framework to model the epidemic and analyze the effects of the imposed lock-down. We calibrate a compartmental model of COVID-19 clinical progression using a Bayesian method called Conditional Robust Calibration (CRC) against the daily epidemiological data. Then, we perform a robustness analysis on the calibrated model, in order to quantify the influence of model parameters on the hospital capacity and to draw possible scenarios of different containment measures. CRC confirms the hypothesis of underestimation of new positive cases and highlights how identifying presymptomatic transmission is crucial for reducing the contagion. Moreover, our results show the central importance of the lock-down timeliness and intensity, in order to curb the contagion and avoid a relapse.