News

AI tool predicts risk of emergency hospital visits

Experts have harnessed the power of artificial intelligence (AI) to more accurately anticipate which patients will require emergency hospital admissions across Scotland.

A medical professional cares for a patient in a hospital

Researchers employed machine learning to develop an update for the first time in 12 years for a tool used by healthcare providers to highlight individuals at high risk of urgent hospital care within the next year.  

In tests they found that SPARRAv4 – Scottish Patients At Risk of Readmission and Admission version 4 – is better able to identify emergency admissions than the previous version. 

Researchers say the AI-powered update will help healthcare providers in Scotland anticipate and plan more effectively for emergency cases and manage healthcare resources more efficiently. 

Reducing strain

Emergency hospital admissions routinely account for around half of all hospital stays in Scotland, placing tremendous strain on the healthcare system.  

Researchers from the Universities of Edinburgh and Durham worked with Public Health Scotland to develop an improved tool to help manage this growing issue. 

The team used health records from 4.8 million people living in Scotland, gathered between 2013 and 2018, and held by Public Health Scotland. The records included information that is routinely collected by healthcare providers, such as patient history, prescription details and previous hospital admissions. 

In an era where healthcare systems are under high stress, we hope that the availability of robust and reproducible risk prediction scores such as SPARRAv4 will contribute to the design of proactive interventions that reduce pressures on healthcare systems and improve healthy life expectancy.

Dr Catalina VallejosReader at the University of Edinburgh’s MRC Human Genetics Unit

Improved accuracy

Experts used machine learning techniques to analyse the dataset and developed SPARRAv4 to predict which patients might require emergency hospital care within a 12-month period.  

As well as correctly identifying more emergency admissions, SPARRAv4 was also found to be better at gauging individual patients’ level of risk of needing urgent hospital care.  

Researchers highlight that while the tool will serve as a critical aid, it will not replace the essential clinical judgement of medical professionals. 

Public Health Scotland will start promoting the updated model and engaging with healthcare professionals to encourage its widespread adoption in Scotland.  

The research, supported by The Alan Turing Institute and Health Data Research UK, is published in npj Digital Medicine.

AI and machine learning depend on large amounts of high-quality data and secure platforms. Thanks to Public Health Scotland's exceptional data curation, this research collaboration has developed a model that could greatly benefit the public. This demonstrates how big data can create tools to support medical professionals when identifying patients who might benefit from early intervention.

Dr Louis AslettAssociate Professor of Statistics at Durham University

The SPARRA model was developed to respond to a growing recognition of the need to shift from reactive healthcare to a more preventative and anticipatory approach. This has been a fruitful research collaboration between Public Health Scotland and colleagues from the Alan Turing Institute, to harness the power of Scotland’s data, through the use of innovative statistical and AI techniques to update our SPARRA model.

Dr Jill IrelandPrincipal Analyst at Public Health Scotland

Related links

Read the full paper in npj Digital Medicine

MRC Human Genetics Unit at the University of Edinburgh

Image credit: Jacob Wackerhausen via Getty Images