by Salvatore Rinzivillo (ISTI-CNR), Joakim Sundnes (SIMULA) and Karin Rainer (AGES)
The epidemic emergency created by the rapid spread of SARS-Cov-2 drew attention to the methods and models that allow us to understand, predict and control the diffusion of infectious diseases. A thorough knowledge of the diffusion dynamics of viruses can help us conduct effective tracking of the transmission chain, precisely identify and assess restrictive measures, and promptly counteract local outbreaks. Mathematical models and simulation systems based on these models provide a means of obtaining such an understanding and enable evaluation and comparison of multiple mitigation approaches.
The ongoing pandemic has fostered cross-collaboration among multiple disciplines, including mathematics, physics, epidemiology, medicine and computer science, as well as social science, ethics, and law. The wide availability of data collected through multiple Big Data sources, and the advances in the field of data science and artificial intelligence (AI), have facilitated the development of data-driven, holistic epidemiology as a tool to analyse the disease dynamics and the assessment of the associated risks. In this special issue, we review recent advances in model-based and data-driven epidemiology that can help us understand and predict the ongoing global pandemic.
Situational monitoring for decision support
One of the main countermeasures to reduce the diffusion of the SARS-Cov-2 is the enforcing of social distancing, both with personal protections and movement restrictions. To balance the need for segregation with the general needs of society and the continuation of normal activities, contact tracing methodologies have been proposed, with the objective of quickly identifying the potential risk of contagion after the identification of an infected person. This kind of tracing can be implemented at different levels of granularity.
At the individual level, we can exploit mobile devices to reconstruct the social contacts of each person, by means of dedicated apps capable of estimating reciprocal distance on the basis of communication signals (Kondylakis et al, Faye). An app installed on the individual device may also be exploited to enforce communication with the healthcare system, providing direct dialogue with healthcare professionals, doctors, or assistance in the self-assessment of symptoms (Petrellis). The effectiveness of these solutions strongly depends on the penetration of their adoption within the population. The promotion of their usage should comprise a balanced evaluation of the challenges and opportunities for these apps, identifying trusted entities, supported by the scientific community, to develop dependable and secure applications. Moreover, a campaign to inform the population should be promoted, as well as guarantee the protection and oblivion of the data when it is no longer needed.
At the population level, communication infrastructures can provide country-wide indicators of population mobility, which is a vital input to predictive models of pandemic spread. Mobile phone data can be aggregated to develop indicators such as stay-at-home or mobility indices, providing a timely pulse of the situation and a measure of the effectiveness of the implemented solutions (Gaal et al). Also, earth observation images may offer a view to monitoring the effectiveness of the lockdown measures, by estimating the movement of vehicles from times of images (Duro et al).
The data that is actively collected by dedicated personnel should be certified by ensuring internal constraints and soundness (Carvalho). The data collected with these approaches and others during the course of the pandemic help to reconstruct and model the dynamics of the diffusion and to compare with well-known spreading models (Kuruoglu and Li). Such data-driven analysis makes it possible to reason about the external impact of the pandemic on society since it goes beyond the healthcare sphere and has a profound impact on economics (Lo Duca and Marchetti) and other dimensions of our society (Ponsard and Nihoul).
Simulation systems and forecast
Simulation systems provide an estimate of the evolution of the epidemic in different scenarios, evaluating the impact of policy implementation and actions, like enforcing restrictions or introducing sanitary improvements, e.g., new treatments or vaccines.
The simulation systems may model the single individual and their social structure (Rossetti et al) evaluating different diffusion models and parameters. The system may also incorporate external layers, like mobility infrastructure, economic processes (Emrich and Popper), geography, and spatial structures (Zimeras).
The simulation systems suffer from high computational complexity to model the decision and actions of each synthetic agent. The impact of specific actions and phenomena, such as the airborne dispersion of droplets (Henry, et al), can be modelled as a specific parameter.
Alternative methods exploit a simulation process based on compartments, i.e. potential states of sub-populations. These approaches simulate the aggregate behaviour of a group of people to derive indicators for relevant indices (e.g., number of IBUs, replication number Rt) (Szederkényi et al). In the article by Edeling verification techniques are applied to estimate the validity of the provided indicators.
Many of these models are well known in the literature and may potentially be enhanced by injecting observations collected from the data and models learned by machine learning algorithms (Veiber). Recent approaches (Coro) also take into account the influence of climate on the local diffusion of the virus.
This issue of ERCIM News focuses on research surrounding the current epidemic emergency: all the contributions address different actions to understand and counteract the spread of the virus. On one hand, the current COVID-19 epidemic has brought to the fore a wide field of research activities that have been developed in Europe over recent decades. An existing network of research institutes was able to pull together and collaborate to provide efficient tools to fight this emergency.
On the other hand, new research challenges have emerged: we highlight here two main topics “situational monitoring for decision support” and “simulation systems and forecast”. There is a clear and strong focus to embed evidence collected from the data into a repertoire of modelling theories that now require strong support from the patterns emerging from data collections, in particular for those diffusion dynamics that have not been clearly understood. A viral agent like SARS-Cov-2, that exploits social interactions among individuals to spread in the population, calls for more efficient simulation engines, that should be capable of managing the computational complexity of thousands or millions of synthetic agents, to explore potential scenarios derivable from policy actions.