Science fiction or clinical reality: a review of the applications of artificial intelligence along the continuum of trauma care

Abstract

Artificial intelligence (AI) and machine learning describe a broad range of algorithm types that can be trained based on datasets to make predictions. The increasing sophistication of AI has created new opportunities to apply these algorithms within within trauma care. Our paper overviews the current uses of AI along the continuum of trauma care, including injury prediction, triage, emergency department volume, assessment, and outcomes. Starting at the point of injury, algorithms are being used to predict severity of motor vehicle crashes, which can help inform emergency responses. Once on the scene, AI can be used to help emergency services triage patients remotely in order to inform transfer location and urgency. For the receiving hospital, these tools can be used to predict trauma volumes in the emergency department to help allocate appropriate staffing. After patient arrival to hospital, these algorithms not only can help to predict injury severity, which can inform decision-making, but also predict patient outcomes to help trauma teams anticipate patient trajectory. Overall, these tools have the capability to transform trauma care. AI is still nascent within the trauma surgery sphere, but this body of the literature shows that this technology has vast potential. AI-based predictive tools in trauma need to be explored further through prospective trials and clinical validation of algorithms.

Keywords: Artificial intelligence; Machine learning; Trauma.

Fig. 1

Overview of major types of machine learning. Overview of different types of machine learning (ML): ML is shown as a subset of artificial intelligence (AI). Within ML, there are three subtypes: supervised learning, unsupervised learning, and reinforced learning. Supervised learning is task-driven and uses labeled data to predict a predefined outcome. Unsupervised learning is data-driven and is used to find trends/outputs in unlabeled data. Reinforced learning is environment-driven and uses interaction with the environment to learn

Fig. 2

Overview of major types of supervised learning. Continuum of complexity of supervised learning algorithms: While not all types of supervised learning algorithms are shown here, four major illustrative examples—logistic regression, decision tree, random forest, and artificial neural network—are shown along a qualitative continuum from least to most complex. The diagrams are meant to provide a visualization of the algorithm processes whereby the blue circles and the orange squares represent different outcomes

Fig. 3

Trauma outcome predictor (TOP) example screen shot. This screenshot of the TOP interface shows how clinicians can input variables based on clinical assessment to predict mortality after blunt injury. The differences between the left and right panels are due to the algorithm’s ability to adjust the questions asked based on answers to previous questions; in this case, the differences in GCS answers prompt the algorithm to diverge in its input variable requirements. Reprinted from Surgery, Vol 171/6, El Hechi M, Gebran A, Bouardi HT, Maurer LR, El Moheb M, Zhou D et al. Validation of the artificial intelligence-based trauma outcomes predictor (TOP) in patients 65 years and older, Page 1689., Copyright (2022) with permission from Elsevier and the original authors