Five Questions With: Freddy Lippert Talks AI at Pinnacle

This year’s Pinnacle EMS Leadership Conference will be held August 9–13 at the J.W. Marriott Desert Ridge Resort in Phoenix. Aimed at EMS chiefs, administrators, medical directors, managers, educators, and innovators, the show bills itself as a “participatory educational experience” for EMS leaders. In addition to a cutting-edge education program reflecting the industry’s biggest issues, Pinnacle connects attendees, faculty, and sponsors in unique roundtable discussions for true collaboration and open dialogue.

Among this year’s faculty is Freddy Lippert, MD, CEO of Copenhagen Emergency Medical Services in Denmark and an associate professor at the University of Copenhagen’s Department of Clinical Medicine. A founding member of the European EMS Leadership Network and Global Resuscitation Alliance, Lippert is the recipient of many honors, including a listing in the 2014 Kraks Blå Bog (Krak’s Blue Book) of those who have made lasting contributions to Danish society.

Lippert will speak at Pinnacle on the potential of artificial intelligence in EMS (Copenhagen EMS is an early pioneer), as well as the latest EMS research and its impact. Here he provides some background on his system’s use of AI. Register for Pinnacle by July 9 for early-bird savings.

EMS World: Copenhagen EMS uses AI to speed detection of cardiac arrest and the process of dispatching help. What went into establishing that?

Lippert: We have worked continuously to improve survival from out-of-hospital cardiac arrest in Denmark for a decade and managed to have a fourfold increase in survival and a bystander CPR rate of 80%. We wanted to improve further. Through research we realized there is a potential to recognize more cardiac arrest cases when citizens call our emergency number (1-1-2 in Europe, like 9-1-1 in the U.S.).

When did you institute that? What have the outcomes been to date?

We started a collaboration with Corti, a tech company with expertise in AI. First we tested an AI model on our historical data and found that AI could detect 10% more cardiac arrests than our call-takers—this was published in Resuscitation.¹ Then we conducted a prospectively randomized controlled trial and found the same results, published in JAMA Network Open.² AI has now been fully implemented in our dispatch center with the expectation of improving survival further.

How might U.S. EMS systems realize similar advantages?

We have seen a huge interest in our results from other European EMS agencies but also from the U.S., Australia, and Asia. There is a huge potential for saving more lives using AI as an additional decision support for call-takers. Our model has been tested in Sweden and in Seattle, demonstrating that this a solution that also applies to other countries, languages, and systems.

What other aspects of a resuscitation system have to accompany this?

The AI solution can be added to any call center as an add-on or integrated into the CAD system. It is an additional support to call-takers to provide an alert in case of a cardiac arrest pattern in the conversation. The focus should be implementation and changing the work process of the call-taker and the interaction of AI and humans. Humans call-takers are proud professionals, and they often believe they are right in most cases, but they are not. If you can combine other healthcare data into the AI model, you will get full benefit of AI.

What other areas of EMS could benefit from AI?

AI is going to change dispatch processes completely in the near future. I am convinced, based on our experiences, that CAD systems will integrate AI as decision support in dispatch—not only for cardiac arrests but also for time-critical cases like myocardial infarction, stroke, respiratory distress, and sepsis. AI might also help create a risk profile of the caller and their criticality and thereby make sure the right resources are being used for those in need. AI can also be used to monitor the quality of the call and give direct feedback to the individual call-taker.

References

1. Blomberg SN, Folke F, Ersbøll AK, et al. Machine learning as a supportive tool to recognize cardiac arrest in emergency calls. Resuscitation, 2019 May; 138: 322–9.

2. Blomberg SN, Christensen HC, Lippert F, et al. Effect of Machine Learning on Dispatcher Recognition of Out-of-Hospital Cardiac Arrest During Calls to Emergency Medical Services: A Randomized Clinical Trial. JAMA Netw Open, 2021 Jan 4; 4(1): e2032320.

Carol Brzozowski is a freelance journalist and former daily newspaper reporter based in South Florida. Her work has been published in more than 200 media outlets.