An optimized mathematical model for cancer patient care planning in the COVID-19 era.

e18663 Background: The COVID-19 pandemic has threatened to collapse hospital and Intensive Care Unit (ICU) services, and it seems to limit the care of oncologic patients. The objective was to develop a mathematical model designed to predict the hospitalization and ICU admission demands due to COVID-19 to forecast the availability of hospital resources for the scheduling of oncological surgery and medical treatment that require hospitalitation or possible use of ICU services. Methods: We have implemented a SEIR model designed to predict the number of patients requiring hospitalization and ICU admissions for COVID-19. We evaluated the model using the number of cases registered in the hospitals of the province of Granada (Spain), that altogether cover 914,678 inhabitants. Calibration was performed using data recorded between March 15 and September 22, 2020. After that, the model was validated by comparing the predictions with data registered between September 23 and November 7, 2020. Besides, we performed a predictive analysis of scenarios regarding different possible sanitary measures. Results: Using patient registered data we developed a mathematical model that reflects the flow among the different sub-groups related to COVID-19 pandemics (Table). The best algorithm that fitted the disease dynamics was Particle Swarm Optimization, that minimized the difference between model output and real data used to calibrate the model. The validation phase showed the accuracy of the predictions, especially concerning trends in hospitalizations and ICU admissions. The different scenarios modelled on November 10, 2020 allowed us to predict the evolution of the pandemic until July 1, 2021, and to detect the peaks and valleys of disease prevalence. Conclusions: The mathematical model presented provides predictions on the evolution of COVID-19, the prevalence and hospital or ICU care demands. The predictions can be used to detect periods of greater availability of hospital resources that make it possible to schedule the oncologic surgery and intensify the care for oncologic patients. Furthermore, our model can be adapted to other population by recalibrating the model according to demographic data, the local evolution of the pandemic and the health policies. [Table: see text]