Applying data analysis to the prediction of trauma patient outcomes
Becca Stoner, a PhD student from Queen Mary University of London, is the recipient of this year’s Joint Orthopaedic Research Fellowship, co-funded by Orthopaedic Research UK and The Royal College of Surgeons of Edinburgh. An Orthopaedic Registrar by training, she will be researching the use of statistical
Becca describes how her project is designed for scenarios where there are several patients who require urgent treatment,
‘In a mass casualty incident, to save as many lives as possible you need to know in what order to treat people. There are existing scoring systems to predict mortality, but they tend to be rather simplistic. By harnessing the power of AI, we can analyse the information we have about trauma patients, for example how their heart rate is changing over time and what’s happening to their blood pressure, to make better decisions.
‘This mortality prediction model analyses information about the patient’s physiology status, as well as other information that the clinician has about how they have sustained their injuries. It then calculates the likelihood of this patient surviving the first 48 hours. This gives clinicians a clear idea of the degree of intervention that is required and at what stage. But it doesn’t simply calculate mortality, it also provides information about what life threatening problems the patient may be developing, for example whether they are likely to bleed to death through trauma-induced coagulopathy or experience rapid organ failure. This gives clinicians an idea of their main priorities as well as the likelihood of these things occurring.’
She will be using a modelling system called Bayesian Networks, which is used extensively as a predictive tool in areas such as environmental science.
‘This is a really useful way of dealing with a fast-moving situation in which you may not have access to all of the patient’s data. Unlike other models, the Bayesian Network can make predictions based on partial or incomplete information. This makes is particularly useful as the basis for a decision-making tool for clinicians experiencing the real time pressures of a trauma situation.’
Far from downplaying the expertise of clinicians, Becca sees this type of modelling as enhancing their skills at a time when there is considerable demand on the clinician’s cognitive capacity,
‘There are sceptics who question why they need a computer to tell them something that they already know, but the model isn’t trying to make decisions for clinicians, but simply provide additional information and maybe a prompt to consider other options.’
Becca sees the value of the ORUK network in supporting her work,
‘Undertaking research in AI can become computer scientist focused, so it’s important to involve clinicians when we build these tools, and in doing so, we can be involved in developing AI in other areas of orthopaedics.’