The word ‘triage’ has evolved from the French word ‘trier’ which was the term used by Baron Dominique Jean Larrey during the Napoleonic wars to manage the influx of injured soldiers for a few beleaguered doctors. Today, triage is the ultimate in front line medical care, providing both the medical practitioner and patient with the coordinates they need to ensure that care is provided to the right person, on time. However, while we are several hundred years and at least three centuries ahead of the origination of the word ‘triage’, doctors and medical centers remain under pressure and triage, especially in the recent pandemic, has never been more critical.
Against this backdrop artificial intelligence (AI) has become an increasingly powerful tool for emergency room triage. The algorithms and intelligence of AI have evolved significantly over the past few years to become more reliable and are designed to support physicians as they juggle increasingly challenging workloads. The science that sits behind the algorithm is deep and highly complex. This is not just technology and data, it’s deep learning, neural networks and machine learning. It’s using data sets and insights to create algorithms that are capable of identifying the different layers of patient triage so that physicians can ensure that patients are accurately categorized and cared for.
This requires significant volumes of clean data that can be used to ensure that AI is not just capable, but also of value in an emergency room setting. This means that AI in triage has to follow rigorous process, testing and modeling to get the best results.
In a recent study by the American College of Surgeons in Science News, the authors discussed how AI played a significant supporting role, helping clinicians triaging post-operative patients for intensive care. The algorithm discussed in the study included 87 clinical variables and 15 specific criteria and the AI triaged 41 of the 50 patients perfectly. The accuracy rate was around 82% which indicated, according to the study, that AI could be a solid partner in supporting physicians during the triage process.
AI entered into the daily clinical work of the radiology department at Brussels University Hospital to provide essential support for the practitioners managing extraordinary volumes of images. The solution is easy to use and fast, which is why it has been implemented on the emergency ward of the hospital. It has helped re-define department efficiency metrics and workload management.
AI initially experienced some teething problems – it was still finding its feet amidst extensive data sets and limited understanding around what was needed for it to thrive. Today, this dynamic has changed. Companies that specialize in AI for the medical industry have solid and proven reputations alongside measurable results and tangible benefits. The implementation of AI in any medical situation has become highly specialized and certified.
Another study, this one taken from the Scandinavian Journal of Trauma, Resuscitation and Emergency Medicine, looked at the use of AI in triage in ER specifically. The algorithm was developed using the data from nearly nine million patients with 2604 EMS run sheets from two hospitals in Korea. The result was that the AI was capable of predicting critical care at a confidence interval of 95% outperforming the Emergency Severity Index and the National Early Warning Score. In essence, AI provided essential help to professionals in the ER.
There are multiple applications of AI in triage in ER that range from apps designed for the patient to built-in AI algorithms that can be used to triage and manage patient care on the front line. Many of these have evolved to accommodate the restrictions introduced by COVID-19, and this has only served to further enhance the value of AI in the triage setting.
When looked at under the light of extreme pressure, demanding schedules and even the rigorousness of the pandemic, AI has immense potential in reducing the burden on the ER and the ICU. It can be used in telemedicine for proactive and predictive triage for remote patients or to keep patients remote until triage moves them to a different level of urgency, thereby limiting the spread of infection and ER influx. It can be used to manage patient fear as they are provided with high-level insight and support from remote locations without further risk to themselves or others.
This can also potentially ensure faster and more accurate triage, reducing pressure on the medical professional and allowing for patients in need of urgent care to receive it faster, thereby reducing the pressure on the ER. In addition to the pressure of the pandemic, AI in emergency room triage is designed to support the medical sector in all types of emergencies. As it evolves it may become increasingly capable of providing a solid foundation for triage in ER, one that has the potential to minimize risk, improve accuracy, and reduce the burden on the ICU.
Aidoc has been working with multiple medical institutions and professionals to ensure the development of a rich array of algorithms that can be used to assist in the triage of patients. The company’s C-Spine solution has been used by radiologists to ensure expedited treatment for their patients and the company is continuously investing into new ways of providing medical professionals with critical support.