Predicting patients’ risk of needing to return to hospital

Unplanned readmission to hospital reduces patients’ quality of life and puts extra pressure on the NHS.

Some patients are readmitted to hospital because they don’t have the health or social care they need at home.

One solution to this problem is ‘virtual wards’. This means patients live in their own home but receive the same healthcare they would receive in hospital. Healthcare professionals assess the patient’s condition every day, either via a home visit or a video call, and can perform tests and administer treatments.

Using virtual wards, it is sometimes possible to discharge patients from hospital earlier. This can benefit the patient, as well as relieving pressure on beds within the NHS.

Project aims

The aims of this project are to:

1. Create a computer programme, optimised to work in South West England, that can predict the risk of a patient needing to be readmitted to hospital
2. Use this computer programme to:
   • Work out how virtual wards affect the risk of patients needing to be readmitted to hospital
   • Work out, for individual patients, the optimum number of days for them to stay in hospital before being discharged to a virtual ward

In the first part of this project, we will create a computer programme to predict the risk of patients needing to be readmitted to hospital. This programme will use machine learning. This means it will use information about what causes patients to be readmitted to hospital to identify patterns. It can then use these patterns to predict the risk of individual patients being readmitted.

The programme will predict a patient’s risk of being readmitted to hospital within 30 days of being discharged. It will base the risk on the patient’s:

• Socio-demographics
• Health and social care use before hospital admission
• Long-term health conditions diagnosed before hospital admission
• Main diagnosis
• Length of stay in hospital

and the hospital department the patient is admitted to.

In the second part of this project, we will apply the computer programme to patients admitted to virtual wards in the Bristol, North Somerset and South Gloucestershire healthcare system. It will predict the risk of patients being readmitted to hospital had they not been admitted to a virtual ward. 

We will also note the proportion of patients actually readmitted to hospital, after being discharged to a virtual ward. By comparing these numbers we will calculate how virtual wards affect the risk of patients being readmitted to hospital.

We will also investigate how the length of time patients spend in hospital affects their risk of being readmitted.

What we hope to achieve

By the end of this project, we hope to have a computer model that can predict patients’ risk of being readmitted to hospital in South West England.

We will have evaluated how well this model could help make decisions about how soon to discharge patients from hospital in the Bristol, North Somerset and South Gloucestershire healthcare system.

Following this project, we plan to seek further funding to further develop the computer programme.

In the long term, we hope it will help healthcare staff and patients make joint decisions about when patients should leave hospital, how they will be supported at home and their risk of being readmitted.