Approaches to Daily Monitoring of the SARS-CoV-2 Outbreak in Northern Italy

Abstract

Italy was the first European country affected by the Sars-Cov-2 pandemic, with the first autochthonous case identified on Feb 21st. Specific control measures restricting social contacts were introduced by the Italian government starting from the beginning of March. In the current study we analyzed public data from the four most affected Italian regions. We (i) estimated the time-varying reproduction number (R_t ), the average number of secondary cases that each infected individual would infect at time t, to monitor the positive impact of restriction measures; (ii) applied the generalized logistic and the modified Richards models to describe the epidemic pattern and obtain short-term forecasts. We observed a monotonic decrease of R_t over time in all regions, and the peak of incident cases ~2 weeks after the implementation of the first strict containment measures. Our results show that phenomenological approaches may be useful to monitor the epidemic growth in its initial phases and suggest that costly and disruptive public health controls might have had a positive impact in limiting the Sars-Cov-2 spread in Northern Italy.

Keywords: COVID-19; epidemiology; infectious disease; outbreak analyses; public health.

Figure 1

Time-dependent reproduction number R_t in the regions Lombardy, Veneto, Emilia Romagna and Piedmont, from March 3rd to April 30th. Black solid line: estimate of R_t, gray areas: 95% confidence intervals, dotted line: threshold for outbreak extinction.

Figure 2

Five-day Generalized Logistic Model (GLM) forecasts of SARS-CoV-2 new infections in Lombardy, Emilia Romagna and Veneto (observed data: Feb. 25th to April 30th), and Piedmont (observed data: Feb. 28th to April 30th). Empty circles represent new observed cases, the vertical dashed line indicates where the real observations stop, the red continuous line the best prediction of the epidemic in the following 5 days, the red dashed lines the 95% confidence bands, and the blue lines the bundle of models estimated by the prediction algorithm. Bootstrap size was set to 100.

Figure 3

Evolution of the epidemic predictions in Lombardy based on the Generalized Logistic Model (GLM). An increasing amount of epidemic data (black circles) are used, starting from Feb. 25th until March 21st (day of the total lockdown) and then extending the data by 5 days until April 30th. Empty circles represent observed cases, the vertical dashed line indicates where the real observations stop, the red continuous line the best prediction of the epidemic up to May 5th (day 70 of the epidemic), the red dashed lines the 95% confidence bands, and the blue lines the bundle of models estimated by the prediction algorithm. Bootstrap size was set to 100.

Figure 4

Five-day Generalized Logistic Model (GLM) forecasts of SARS-CoV-2 deaths in Lombardy (observed data: Feb. 25th to April 30th), Veneto and Emilia Romagna (observed data: Feb. 26th to April 30th), and Piedmont (observed data: March 5th to April 30th). Empty circles represent deaths, the vertical dashed line indicates where the real observations stop, the red continuous line the best prediction of the epidemic in the following 5 days, the red dashed lines the 95% confidence bands, and the blue lines the bundle of models estimated by the prediction algorithm. Bootstrap size was set to 100.