Data use register

Unplanned readmission to hospital is bad for the patient’s quality of life and puts extra pressure on the NHS.

Reasons patients are readmitted include being discharged to a setting that isn’t suitable or without appropriate health or social care support.

There are many statistical scores to identify the patients at greatest risk of being readmitted. However, these scores don’t account for a patient’s health and social care needs before being admitted or the support they receive after being discharged. These are both linked to readmission risk.

We aim to produce a machine learning model to predict a patient’s risk of readmission, optimised for South West England. This will enable evidence-based conversations between health and social care professionals and patients on post-discharge support and how to reduce readmission risk.

The model will be applied to different patient groups, including patients receiving early supported discharge through virtual wards. For this group of patients, we will use the model to measure the effect of virtual wards on readmission risk, and understand how long patients should stay in hospital before early supported discharge, depending on patient characteristics.