Case study: AI tool improving outcomes for patients by forecasting A&E admissions

A&E departments across the country are facing unprecedented pressure dealing with patients requiring urgent care. The total number of attendances in June 2023 was 2,221,000, an increase of 1.3% on June 2022 (NHS England, June 2023).

Historically, it has been difficult for decision makers to effectively plan and make sure crucial resources are available without knowing anticipated surges in levels of demand.

Using lessons learned during the recent pandemic, NHS England data scientists worked with Faculty, a private AI organisation, to develop an A&E demand forecasting tool. This tool provides hospitals in England with anticipated A&E admissions three weeks in advance and alerts them to potential upcoming surges. This helps local and frontline NHS staff make more informed decisions on their allocation of resources by understanding when emergency demand pressures are likely to be higher or lower, ensuring there is capacity for patients when they need it.

Accounting for historical trends in the data as well as variations such as seasonality, weather and public holidays, forecasts are broken down by age so that staff can plan for specific bed needs, such as for paediatric patients, ensuring that they are provided with the relevant care they require.

NHS leaders are now able to proactively plan for surges in demand and divert resources to other areas such as elective care when demand is lower with the aim of treating patients as fast as possible.

The provision of accurate three-week ahead forecasts places the NHS in a better position to anticipate and prepare for surges in demand rather than react once they’ve started. This benefits patients as well as frontline staff who are better placed to deal with all the different situations they are faced within A&E.

The tool – which has been co-developed in collaboration with frontline, clinical and operational staff in nine pilot NHS trusts – is now available to 123 hospital trusts and has been shown to be approximately twice as accurate at predicting admissions than a baseline comparison model.

The tool has been built upon the NHS Data Platform, and NHS England is procuring a Federated Data Platform (FDP) to support health and care organisations to make the most of the information they hold. Vital tools such as the A&E forecasting tool will continue to be supported through the FDP in order to provide trusts with the information they need to understand patterns, solve problems and plan services for their local populations and patients.

An Associate Director for Performance Information at Luton and Dunstable Hospital said, “I think you’ve got something which is ground-breaking and incredibly useful”.