An algorithm used by the NHS to predict which patients are most likely to deteriorate is inaccurate when used on Covid-19 patients, a new study shows. The algorithm has ben used for years and is endorsed by NHS England. It gives each patient a score which predicts how their condition will change. King's College London researchers studied how good the algorithm was at predicting the health outcomes of 1,276 Covid-19 patients admitted to hospital in March and April 2020. The data reveals 'poor-to-moderate' accuracy for spotting patients who are at risk of being admitted to ICU or dying two weeks later. It performed 'moderately' in the short term, three days later, the study authors say.