Regional probabilistic situational awareness and forecasting of COVID-19

Mathematical models and statistical inference are fundamental for surveillance and control of the COVID-19 pandemic. Several aspects cause regional heterogeneity in disease spread. Individual behaviour, mobility, viral variants and transmission vary locally, temporally and with season, and interventions and vaccination are often implemented regionally. Therefore, we developed a new regional changepoint stochastic SEIR metapopulation model. The model is informed by real-time mobility estimates from mobile phone data, laboratory-confirmed cases, and hospitalisation incidence. To estimate locally and time-varying transmissibility, case detection probabilities, and missed imported cases, we present a new sequential Approximate Bayesian Computation method allowing inference in useful time, despite the high parametric dimension. We test our approach on Norway and find that three-week-ahead predictions are precise and well-calibrated, suitable for real-time surveillance. SignificanceWe developed a regional infectious disease spread model focussing on operational usefulness in real time. The model is informed by near real-time mobile phone mobility data, laboratory-confirmed cases, and hospitalisation incidence. The model is used to estimate reproduction numbers and provide regional predictions of future hospital beds. Regional reproduction numbers are important due spatio-temporal heterogeneity due to for example local interventions. We assume different regional reproduction numbers for different periods of the epidemic. We propose a new calibration method to estimate the reproduction numbers and other parameters of the model, tailored to handle the increasingly high dimension of parameters over time. The model has been successfully used for local situational awareness and forecasting for the Norwegian health authorities during COVID-19.