Background
The Emergency Pre-Hospital Artificial Intelligence in Transfusion and Trauma Induced Coagulopathy (EmPHATTIC) study is a novel implementation study which commenced in January 2019. The study explores the effect of introducing a machine learning tool to help clinicians make decisions for injured patients. The tool has been designed to help emergency clinicians identify high risk patients before they arrive in hospital.
The tool is designed to give a clinician personalised information about a patient within the first few minutes of treatment. The tool was trained and validated on 1000 patients from the ACIT study. It quantifies a patient’s risk of developing trauma induced coagulopathy (TIC) and needing a blood transfusion. TIC is a clotting dysfunction that affects about 1 in 4 seriously injured patients, and exacerbates bleeding, transfusion requirements, morbidity and mortality.
The tool currently has a web-based interface. Clinicians can enter information about a patient and then the tool uses a machine learning model to produce an output of risk. The specific type of machine learning algorithm we use is a Bayesian Network. The tool was designed by C4TS researcher Zane Perkins with assistance from computer scientists at QMUL.
Study objective
The objective of EmPHATTIC is to assess the tool's potential to improve patients’ care. We have shown retrospectively that the tool is accurate. What we need to understand now is whether the tool can retains that accuracy when used very early after injury and whether it enhances decision making above the decision standard the clinicians already make.
Method
Clinicians involved in the study will not be shown the tool's prediction, so that differences between clinician predictions and those generated by the tool, including the amount of blood transfusion needed, can be measured. These differences will be recorded after the clinicians have performed the first assessment of the patient (primary survey) and before hospital arrival.
The study will enrol clinicians across two air ambulance sites; London’s Air Ambulance and Kent, Surry and Sussex Air Ambulance. Patients will not be required to undergo any additional procedures. The study started in January 2019 and will run until August 2019.
Next steps
If EmPHATTIC results suggest machine learning was the potential to improve patient outcomes, C4TS plans to conduct follow on studies. The subsequent study will require approval from the Medicines and Healthcare Products Regulatory Agency to give the tool’s result to clinicians during patient care in a randomised control trial.
For more information, contact Max Marsden
