Predictive modelling of major trauma PROMs using machine learning University of Leicester

Every year across England and Wales, 16,000 people die after injury and many tens of thousands are left disabled for life.

In the early stages after injury, some patients will have a good chance of recovery while others (for example, those with severe head injuries) will need a higher level of therapy input. However, for a large group of patients in between these extremes, it is difficult to know what the outcome will be.

There is already a well-developed system for collecting national data about major trauma patients – the Trauma Audit and Research Network (TARN) – and a more recent system for major trauma PROMs.

This project from the University of Leicester and partners took major trauma PROMs data to the next stage by linking it to the data held within TARN to create a model to predict PROMs.

The research found that, using information available soon after injury, a mathematical model can predict what a patient will say about their health when asked six months later. There is a 70% accuracy for predicting outcomes. This model can potentially be used to audit the quality of care in a hospital by asking ‘Did the hospital achieve the patient outcomes that were predicted?’. Hospitals that perform worse than expected can then be identified and improvement steps taken.

The researchers also compared different mathematical methods of making a prediction, which increased understanding of the potential strengths and weaknesses of each method when applied to patient data.

The work has provided a new quality measure for national clinical audits, allowed therapists to target interventions, enhanced patient involvement in their recovery and enabled researchers to identify high-risk groups for intervention studies.

Contact details

For more information about this project, please contact Timothy Coats, Professor of Emergency Medicine, University of Leicester.