NHS adopts AI software to forecast A&E admissions

Hospitals are to start using artificial intelligence (AI) software to predict daily A&E admissions weeks in advance.

The software, which analyses data from a number of sources that include 111 calls, local Covid-19 infection rates, traffic levels and the weather, is being introduced in 100 NHS hospital trusts. Trials have shown that the software had an “impressive” ability to forecast daily admissions, broken down by age, up to three weeks in advance.

The software also takes into account bank holidays such as New Year’s Eve, which tends to be a very busy time for A&E.

A report in The Times said that on days predicted to be quiet, managers will be encouraged to free up A&E staff to prioritise elective care and deliver more routine tests and operations to tackle the backlog. If one hospital is forecast to have a particularly quiet day, it could lend staff to a neighbouring trust with high predicted admissions that day. Similarly, hospital managers will also be able to increase bed capacity or the number of staff on call if they are forecast to have particularly high demand.

The hope is that by using the software to plan ahead and share workforce, it will be possible to bring waiting lists down.

Addressing Covid-19 backlogs

The technology has been tested in nine NHS trusts. It was used during the pandemic to help to predict how many beds, ventilators and frontline staff hospitals would need each day.

Professor Stephen Powis, NHS national medical director, said: “Pressures remain high, but staff are determined to address the Covid-19 backlogs that inevitably built up throughout the pandemic, and while that cannot happen overnight, harnessing new technologies like the A&E forecasting tool to accurately predict activity levels and free up staff, space and resources will be key to helping deliver more vital tests, checks and procedures for patients.”

A letter announcing the introduction of the A&E admissions forecasting tool, co-developed by the NHS with Faculty, an AI company, is being sent to trusts today. It says that “sophisticated modelling techniques” will support “planning for non-elective and elective activity”.

Breaking down admission forecasts by age allows staff to plan for specific bed needs, such as for paediatric patients or for elderly patients.

The letter says that experts are working to expand the tool to include “weather data as a leading indicator”. Severe spells of weather can be a predictor of a rise in A&E admissions.