A statistical mechanical study of epidemic spreading in a heterogeneous population and interventions

Ever since the beginning of the COVID-19 pandemic, empirical studies and modelling efforts contributed greatly to our current understanding of viral spreading in the human population and to the translation of that knowledge into public health policies that got us through the most difficult periods of our lives. Nevertheless, uncertainties abound when it comes to predicting the long-term trend of viral mutations or the way our immune system responds to novel strains of the virus, let alone other infectious pathogens on the horizon. In the face of these challenges, one way to enhance our preparedness is to shorten the decision time for targeted and decisive response through
a combination of non-pharmaceutical interventions in precise and novel ways. Model construction and calibration, theoretical analysis, and simulations can help to achieve
such a goal.

The proposed project aims to address the above issues following a step-by-step research plan. To construct a model with the desired attributes, we start from a generic description of the daily routine of individuals in the form of path trajectories. Infection
rates are assigned to contacts between the susceptible and infected that depend not only on the individuals involved, but also on the facility/location where contacts take place. We then introduce mean-field type approximations that allow transition from the agent-based model to a network model with facility/location as basic nodes. Dynamical equations for infection risk at these nodes based on transport statistics will be derived. We move on with analytical treatment and numerical exploration of mechanisms that synergistically determine the epidemic growth rate and validate them against synthetic populations from agent-based models. Case studies will be carried out to differentially
assess various policy components and public response, including the vaccination drive and wide distribution of RAT kits, in steering the 5th Omicron wave in the spring and summer of 2022 in Hong Kong.

We anticipate that the pattern of infection risk propagation revealed by our study will inspire novel intervention concepts at the network level. It will also highlight the degree of precise intervention required at different stages of a pandemic. Our modelling framework may serve as a template for the next generation of epidemic models that perform epidemic forecast and location-specific infection risk prediction through systematic integration of human activity and contact pattern, as well as disease characteristics in real time.