Hospitals worldwide faced an unprecedented and uncertain influx of patients with a high-resource consumption during the first wave of COVID-19. Hospitals were required to make regional predictions on hospital admission and resource requirements to plan, prepare and respond to the crisis. This presentation aims to provide insights based on the experience and modelling efforts of Addenbrookes Hospital, based in Cambridge (UK), to predict local hospital admissions, model patient flows through the hospital and estimate subsequent use of resources. The application of a stochastic approach provided a more informative prediction of the demand for critical resources. In particular, analysis of bed occupancy during the COVID-19 surge shows that the bed occupancy does not follow a normal distribution as suggested by deterministic/epidemiological models, which has significant operational implications during the recovery phase of the pandemic. The approach offers a general simulation methodology for regional hospitals to predict a wide range of essential hospital resources during further possible waves of surges COVID-19 patients.