Pre-hospital prediction of adverse outcomes in patients with suspected COVID-19: Development, application and comparison of machine learning and deep learning methods

•Rapid and accurate assessment of suspected COVID-19 patients is required to identify at-risk patients.•Our machine learning models provided better predictive performance than existing triage methods item There were some differences in the features selected between the machine learning algorithms.•Oxygen saturation, a patient’s level of consciousness and frailty were the most important features.•Our models could be deployed to provide accurate predictions of patient outcomes.