Modeling framework for managing epidemiological scenarios

Computational epidemiology is an interdisciplinary research area that aims to develop and use computational models to understand and control the spread of a disease in a population. Computational models and algorithms have been successfully used to evaluate different intervention strategies, including decisions in the field of public health such as the definition of vaccination and antiviral timings, as well as interventions aimed at limiting social contacts like the closure of schools, public spaces, and the suspension of socio-economic activities. In the last twenty years, different pandemics such as Dengue, SARS, MERS, swine flu, Ebola, Zika, and COVID-19 have highlighted the difficulties in managing such a global emergency. In particular, the outbreak of COVID-19 represents the most striking example of a global pandemic that has seen the proliferation of epidemic studies that analyzed various aspects related to the spread of infectious diseases. Although these studies entail several limitations and assumptions, they played a fundamental role in supporting policy-makers in the choice of management strategies for such emergencies. In particular, a continuous monitoring system helps anticipate, manage, and simulate the dynamics of health crisis medical needs. The goal is to advance the preparedness and response capabilities of territorial public health agencies in emergencies. To achieve this goal, it is necessary to establish a comprehensive, integrated platform, to provide early warnings, monitoring, and forecasting tools to public health response agencies and local healthcare services for anticipating medical needs. We can distinguish three main core objectives: i) to early detect the onset of a health crisis, such as the spread of cases due to specific pathologies, ii) to assess the impact of the health crisis at the territorial level, and iii) to forecast scenarios integrating macro- and micro-level models of different types. Moreover, a continuous monitoring system is crucial for socio-economic stability by enabling early detection of issues, optimizing resource allocation, and facilitating informed decision-making. By providing real-time data and fostering data-driven innovation, it contributes to increased productivity, competitiveness, and overall economic resilience. The sheer amount of data to handle brings several challenges, including data heterogeneity, time dependence, temporal/spatial multi-resolution, incompleteness, and inaccuracy. These aspects need to be taken into account while developing incremental and scalable data analysis techniques to learn the information needed to design and validate epidemiological models. The aim of this Ph.D. thesis is the development of modeling tools, techniques, and guidelines useful to provide timely information on the future effects of political interventions on the progress of the contagion and to assist the actions of policy-makers by allowing them to evaluate and compare different actions to the possible outcome. We begin with a general introduction to computational epidemiology and continuous monitoring systems, followed by a discussion of the contributions and the related state-of-the-art. The focus then shifts to each specific contribution. First, we present an agent-based model designed to evaluate the efficacy of various non-pharmaceutical interventions and vaccination strategies within a well-defined closed environment (we focused on a generic middle school as a case study). This model assesses the infection risks associated with school activities and examines the effects of potential control strategies following the introduction of infectious cases. Next, we introduce Sybil, an integrated framework that combines two different types of approaches: a data-centric approach (a machine learning-based predictive model) and an analytical approach (a variant-aware compartmental model) to enhance prediction accuracy and explainability taking advantage of the relative stability of disease characteristic indices to make future projections. Following this, we analyze the relationship between government-imposed restrictions, mobility patterns, and the effective reproduction number Rt of COVID-19. Using data from major platforms, we examine how movement trends respond to policy changes, revealing cultural differences in social compliance between Northern and Southern European countries, which influence disease transmission dynamics. Then, we present Forge4Flame (F4F), a user-friendly dashboard that addresses a key limitation of agent-based models: the challenge of constructing realistic environments, particularly when modeling complex spaces such as offices, schools, supermarkets, and more. This tool automatically generates the necessary code, simplifying the modeling process without the need for any computational expertise. Finally, the concluding chapter summarizes the key findings and outlines potential directions for future research.