A new research study using artificial intelligence aims to help hospitals identify the right bed for the right patient when they leave Emergency Departments.
Professor Darren Green from Northern Care Alliance NHS Foundation Trust will lead the project, which is coordinated and supported by Health Data Research UK with funding from the National Institute for Health and Care Research.
During winter, hospital emergency departments are often overfull with very sick patients and teams may feel pressured to discharge people to create capacity for new admissions. This creates risk for patients, especially for frail and vulnerable people, and those who do not speak English.
A tool called National Early Warning Score 2 (NEWS2) has shown potential to help identify high risk patients. It is already in general use to help clinicians understand who may be very sick and need intensive care.
But NEWS2 accurately identifies only one in four people who need intensive care and is not fully checked for deciding who does NOT need a bed, even though it’s widely used for this purpose. Other kinds of patient information, such as age, prescriptions and blood tests and diagnosis can be used to help decide about patient care but this requires too much time in an emergency situations.
This study will use artificial intelligence models, incorporating all this available data for each patient, to identify:
- Whether certain groups of patients can be prioritised for admission to specific wards very early on after their arrival in the Emergency Department
- Which patients are at risk of imminent deterioration that has not been picked up by NEWS2
- Potentially unsafe discharges or outpatient care decisions in respect of early readmissions
- The extent to which vulnerable patient groups are more significantly affected by winter pressures such as trolley waits, inappropriate discharges and undetected deterioration
- This new information will help clinicians to more easily identify high risk patients that current approaches can’t.
Professor Green, who is a Consultant in acute medicine and nephrology at Salford Royal, said: “Hospital records contain huge amounts of data that can support us in decision making or flag up when a decision may carry risk. Artificial intelligence tools can see things that may not be obvious to us at the bedside. This is particularly important during busy periods when hospitals are under extreme pressure.”
The work is one of 16 projects nationally covering a range of data-driven approaches to pin-point pressures in the health care system, understand their causes and develop ways to overcome or avoid them. They apply lessons from the pandemic on how to drive rapid-response research that generate results fast and have a direct impact on health policy and clinical care.
Published August 2023:
Stelios Logothetis, Darren Green, Mark Holland, Noura Al-Moubayed: Predicting acute clinical deterioration with interpretable machine learning to support emergency care decision making (Scientific Reports).