Progressing through my medical training, I have witnessed the progression of paper-based medical records, all the way to different forms of electronic medical records, let alone intelligence infused medical records. It is safe to say that health informatics has evaded the way we practice medicine in all its disciplines and all across health care systems.

Artificial intelligence (AI) describes the capability of a machine to learn human cognitive functions and learning. AI applications in healthcare have brought in a paradigm shift powered by data mapping, data aggregation, analytics, and algorithmic techniques that can simulate our decision making as clinicians.

Furthermore, it can predict and suggest clinical pathways, data-based prognosis, and outcomes. AI has already been incorporated in major disciplines such as genetics, diagnostic imaging, neurology, and cancer. Yet, its path into emergency medicine (EM) is still paving its way for vast integration.

EM is a unique field of medicine, as its rich with varying paces of practice, the criticality of conditions, acuity of diagnostic decisions, and a highly stressful environment. It puts their providers consistently on a stretched active clinical decision making and interventions. Hence it is worth to foresee how AI can help enhance and complement the emergency department (ED) functions and add significant benefits to the EM physician’s daily tasks.

One of the main applications of AI is triage. Efficient triage can significantly enhance patients flow, lengths of stay, resource allocations, and risk stratifications. A study published by the American College of Emergency Physicians evaluated electronic triage (E-Triage) systems based on machine learning as opposed to the Emergency Severity Index (ESI). They found out that E-Triage can more accurately classify ESI level 3 patients and highlight opportunities to use predictive analytics to support triage and decision making. (1) A lot more studies established the use of different forms of electronic triage algorithms in improving patient distribution by clinical outcomes, and improved acuity predictions.

Another application of AI was significantly noted in diagnostic imaging departments. Offering remote clinics with restricted resources access to tools for reading imaging needed for active clinical interventions. Feeding into these AI systems is a wealth of comparative studies to predict and describe abnormal studies, and enhance its predictions. Let alone how efficient it would be in a fast-paced ED, getting approximate quick predictions that can be overseen by supervising radiologists.

Additionally, AI has been used in monitoring patient’s vitals, and predicting deteriorating clinical course, requiring early resource utilization and critical decision making in a timely manner. One significant example where AI and machine learning is heavily invested in is Sepsis, and mortality prediction scores, aiding at early detection, guiding clinical course and interventions by using simple data trajectories and analysis.

Another utilization of AI in an ED setting is predictions of Acute Coronary Syndromes, predicting the urgent need for revascularization from reading 12 Lead electrocardiographs (ECGs). A Study done in Keio University Hospital developed an AI model enabled to detect patients requiring urgent revascularization within 48 hours from only 12 leads electrocardiogram. (2) This significantly helps fast pace a lot of the grey cases we see and monitor in our ED’s, especially if validated with risk stratification scores we are already utilizing.

It is worth saying that there are still some barriers to the vast adoption of AI integration to EDs as it’s still a new evolving technology, with restrictive access, ethical discussions, safety, and needed regulations.

I personally have always had a utopian vision of how far health informatics can take our clinical practice, specifically EM. Injecting machine learning and AI into healthcare curates the perfect system that could decrease lengths of stay, intelligently and safely triage our patients, predict clinical course, suggest evidence-based treatment pathways, reduce medication errors and improve clinical outcomes. A more utopian version of my vision is how such a system can help remote and restricted regions requiring extensive resources to aid the reach of its care to underserved populations. It goes without saying that most of these do exist in one way or another, some are still being enhanced, and some are under the works for the next stage. We would foresee its progress nonetheless and slow infusion into our daily practice.