Longer than necessary antibiotic courses in patients with sepsis (an infection of the blood) may lead to side-effects like C diff diarrhoea, and increase antibiotic resistance. But not switching antibiotics promptly when patients don’t respond or get worse is harmful. In hospital, lots of blood tests are done routinely and patient status is regularly monitored, e.g. through oxygen saturation. We want to see whether we can use these routine measurements to work out what “getting better normally” looks like for patients with sepsis (whether or not a microbe causing the illness is found). We want to bring information together from lots of different measurements into a single score which could say how an individual patient was responding compared with “typical” patients doing well. This would be like the growth charts used in children to work out whether they are growing normally. We will then test using statistical methods whether this could be used to help doctors stop antibiotics faster when they are no longer needed, or switch them more quickly when this might help patients.