Forecast of the occupancy of standard and intensive care unit beds by COVID-19 inpatients

This paper proposes a methodology to forecast the number of hospital beds required by COVID-19 inpatients in mild and in critical conditions. To that end, a compartmental model is extended to include the number of critical inpatients, which require hospitalization in intensive care units (ICUs). The model parameters are tailored by using a data-driven approach and a computational methodology for numerical optimization. A multi-objective cost function is adopted, representing the match between the model output and the observed data for four variables, namely the total number of cases, demises, hospitalizations and ICU beds. Results for different regions of the Brazilian state of Sao Paulo are presented. The results show that the model represents well the training data and is able to predict the required health system resources.