Mathematical model optimized for prediction and health care planning for COVID-19

Hospitalizations from covid-19: a health planning tool

OBJECTIVE

Estimate the future number of hospitalizations from Covid-19 based on the number of diagnosed positive cases.

METHOD

Using the covid-19 Panel data recorded in Spain at the Red Nacional de Vigilancia Epidemiológica, Renave (Epidemiological Surveillance Network), a regression model with multiplicative structure is adjusted to explain and predict the number of hospitalizations from the lagged series of positive cases diagnosed from May 11, 2020 to September 20, 2021. The effect of the time elapsed since the vaccination program starting on the number of hospitalizations is reviewed.

RESULTS

Nine days is the number of lags in the positive cases series with greatest explanatory power on the number of hospitalizations. The variability of the number of hospitalizations explained by the model is high (adjusted R2: 96.6%). Before the vaccination program starting, the expected number of hospitalizations on day t was 20.2% of the positive cases on day t-9 raised to 0.906. After the vaccination program started, this percentage was reduced by 0.3% a day. Using the same model, we find that in the first pandemic wave the number of positive cases was more than six times that reported on official records.

CONCLUSIONS

Starting from the covid-19 cases detected up to a given date, the proposed model allows estimating the number of hospitalizations nine days in advance. Thus, it is a useful tool for forecasting the hospital pressure that health systems shall bear as a consequence of the disease.