As part of our Applied Analytics scientific theme, this seminar on Predicting emergency admissions in Scotland: findings and lessons learned  will be led by James Liley.

This seminar will be recorded and later available via our YouTube channel.

Overview

SPARRA (Scottish Patients at Risk of Admission and Readmission) is a long-term initiative by Public Health Scotland (PHS) to predict the risk of emergency hospital admission (EA) on the basis of electronic health records (EHRs). In a remarkable feat of logistics enabled by strong medical record-keeping in Scotland, risk scores are computed for essentially the entire Scottish population and distributed monthly to general practitioners for the patients in their care.

PHS has partnered with the Alan Turing Institute, a network of researchers in artificial intelligence, to develop a new version of SPARRA (version 4) which uses cutting-edge machine learning methods to optimise performance. I will describe the process we followed to develop this tool, and run through our findings, some of which give insight into the epidemiology of medical emergencies in Scotland.

A generally challenging aspect of this work is protection and privacy of data, and I will discuss how this was achieved while maintaining scientific reproducibility. Finally, I will discuss ongoing theoretical work arising from the SPARRA model in which we consider general difficulties in updating predictive scores. (James Liley)

Click here to register