Construction of a Machine Learning-Based Risk Scoring Tool for Post-traumatic Acute Hypotension

Acute hypotension episodes (AHE) are the most common adverse events in post-traumatic emergencies, and the occurrence of acute hypotension is very dangerous for the life safety of critically ill patients. This can lead to fainting or shock, irreversible organ damage or even death. This study uses the mimic IV database to screen key indicators for acute post-traumatic hypotension using the XGBoost method, to select vital signs indicators based on timeliness, economy and convenience of clinical application to construct a simple hypotension scoring tool, and to specify the rationale and process for the construction of the scoring tool. The scoring of the condition can also assist healthcare professionals in the rapid diagnosis of the cause of the patient’s condition. The use of specific scores to indicate the extent of a patient’s condition not only facilitates patient understanding, but also allows the system to be used more frequently by doctors in the clinical setting. The results show that a risk of traumatic haemorrhagic shock scoring tool based on five vital signs - systolic blood pressure, diastolic blood pressure, heart rate, respiratory rate and blood pressure - can accurately identify patient outcomes and differentiate the severity of injuries, providing some clinical decision support.