Cnr-Icar in collaboration with TU Delft University (NL), a prediction model for evaluating the trend of new positives and death cases for COVID-19 in Italy has been implemented. The proposed model, named NIPA-LD (Network Inference-based Prediction Algorithm with LockDown), is a time-discrete SIR epidemic model. According to this model, an individual can be in S (Susceptible – healthy but vulnerable  for  infection), I (Infected),  or R (Recovered –  recovered from the disease or,  unfortunately,  dead) state. Following a network-based approach, in NIPA-LD, contacts take place between nodes of a network composed by the various regional populations. This prediction algorithm starts from such a contact network for estimating the diffusion parameters of the SIR model. Specifically, the spreading and curing rates are estimated by starting from temporal data series on new positive cases (or deaths when predicting these last cases) from March 1, 2020 until a day n and then, the virus evolution for the days subsequent to n is predicted.

Differently from a classical SIR model where the spreading rates are constant, specific transmission modifiers have been applied in order to vary the spreading rates according to the closure measures imposed by the Italian Government during the various phases of the lockdown. In such a way, the model accounts for a lower infection probability if restrictive measures are active.

The results of this research work show that NIPA-LD prediction is more accurate than the one provided by a classic SIR model. Moreover, this prediction model would allow to evaluate which lockdown strategies are more appropriate to manage the trend of epidemic.

For further details, this work is freely available in the ArXiV repository at the url:

https://arxiv.org/pdf/2010.14453.pdf 

Involved partners:

• ICAR- CNR (Clara Pizzuti, Annalisa Socievole)
• TU Delft, Faculty of Electrical Engineering, Mathematics and Computer Science, NL (Bastian Prasse, Piet Van Mieghem)