Prospective study using artificial neural networks for identification of high-risk COVID-19 patients

The COVID-19 pandemic caused a major public health crisis, with severe impacts on global health and the economy. Machine learning (ML) has been crucial in developing new technologies to address challenges posed by the pandemic, particularly in identifying high-risk COVID-19 patients. This identification is vital for efficiently allocating hospital resources and controlling the virus's spread. Comprehensive validation of these intelligent approaches is necessary to confirm their clinical usefulness and help create future strategies for managing viral outbreaks. Here we present a prospective study to evaluate the performance of state-of-the-art ML models designed to identify high-risk COVID-19 patients across four clinical stages. Using artificial neural networks trained with historical patient data from Mexico, we assess the models' accuracy across six epidemiological waves without retraining them. We then compare their performance against neural networks trained with cumulative historical data up to the end of each wave. The findings reveal that models trained on early data can effectively predict high-risk patients in later waves, despite changes in vaccination rates, viral strains, and treatments. These results suggest that artificial intelligence-based patient classification methods could be robust tools for future pandemics, aiding in predicting clinical outcomes under evolving conditions.