Schwab Nguyen machine learning publication story 700x550

KGI Professor and Alumnus Publish Findings on Promising New Machine Learning Tool for Detecting and Treating Atrial Fibrillation

An estimated 5.2 million Americans are living with atrial fibrillation (AF)—an irregular, often rapid heart rate that typically causes poor blood flow—and account for 4.6 million annual emergency department (ED) visits. Though anticoagulation therapy is one of the most effective strategies to reduce cardiovascular morbidity and mortality, as many as 53 percent of high-risk patients are discharged from the ED without anticoagulation.

Studies have found that AF patients discharged from the ED without anticoagulation are 2.7 times more likely to die, suffer a stroke, or be readmitted over the next year, compared to those receiving anticoagulation.

Machine learning tools offer a potential solution for reducing misdiagnosis and treatment failure. Of these tools, the Lucia Atrial Fibrillation Application (Lucia App) by Lucia Health Guidelines shows the most promise to date.

Kim Schwab, PharmD, Assistant Professor of Clinical Sciences for Keck Graduate Institute (KGI)’s School of Pharmacy and Health Sciences, and KGI alumnus Dacloc Brandon Nguyen, PharmD ’20, in collaboration with physicians from a range of institutions including UCSD and Baylor, evaluated Lucia App’s ability to assist in accurately diagnosing and treating AF in the emergency department. They published their findings in “Artificial Intelligence Machine Learning for the Detection and Treatment of Atrial Fibrillation Guidelines in the Emergency Department Setting (AIM HIGHER): Assessing a Machine Learning Clinical Decision Support Tool to Detect and Treat Non-Valvular Atrial Fibrillation in the Emergency Department,” which appears in the Journal of the American College of Emergency Physicians Open.

As pharmacists working in the ED at Sharp Chula Vista Medical Center, Schwab and Nguyen have observed firsthand how inadequate therapy can lead to issues down the road for patients with heart conditions. Schwab first encountered this issue years ago.

“We saw how patients who were at high risk for stroke, particularly those with atrial fibrillation, were not being appropriately treated,” Schwab said.

The issue usually began long before the patient reached the emergency room.

“Even though they were seen by their primary care physician’s office, 60 percent of these patients or more were not appropriately treated,” Schwab said. “Then we would see these patients come through the emergency department after they had a stroke.”

Schwab and her colleagues at Sharp Chula Vista ED worked with the Cardiology department on a quality improvement project to determine if they could capture high-risk patients and send them home with the appropriate medication or refer them to a specialist such as an electrophysiologist.

This project proved successful—culminating in a clinical decision flow chart for use in the ED—and they published the paper in 2016 in the American Journal of Emergency Medicine. Based on the study and flow chart, Sharp cardiologist Dr. Gilanthony Ungab, who specializes in electrophysiology, developed the algorithm for Lucia App.

“Fast forward to this latest study, and we have now taken something that we created just to improve care in our hospital and turned it into a clinical decision model,” Schwab said. “This form of artificial intelligence can help any hospital, not just ours.”

To recommend anticoagulation treatment, a physician must assess the patient’s risk of stroke and bleeding by performing an electrocardiogram (ECG) and evaluating the patient’s medical history.

The Lucia App photographs the ECG to determine rhythm detection, calculates risk scores based on the patient’s inputted medical history, and then provides guideline-recommended treatment—such as blood thinners—or referral to a specialist such as an electrophysiologist or interventional cardiologist.

The purpose of the study was to determine the rate of accurate AF identification and appropriate anticoagulation recommendations in ED patients ultimately diagnosed with AF.

They performed a single-center, observational retrospective chart review in an urban California ED with an annual census of 70,000 patients. In a convenience sample of 297 AF cases, the Lucia AF algorithm performed better (98.3 percent vs 78.5 percent) than the standard of care for following national guideline-based anticoagulation.

This is the first cloud-based app intervention to demonstrate the potential for increased AF detection and anticoagulation treatment recommendations. The accessibility of a mobile (tablet or phone) app suggests new compliance opportunities could be implemented at low cost and without a significant disruption to a hospital’s daily operations.

Schwab and Nguyen emphasize that the app is not intended to replace the guidance of a cardiologist, but it can greatly assist physicians and mid-levels in making better decisions regarding a patient’s treatment.

“It’s very user-friendly,” Nguyen said. “All you need to do is take a picture of the ECG, and it measures all the relevant data.”

Nguyen, who initially completed a residency at Sharp in acute care and was then hired on as a pharmacist, has long been fascinated with technology. He believes forms of portable digital technology such as smartphones can help medicine become more accessible in our daily lives.

For Schwab, this research supports many of the concepts she is teaching in Dr. Armen Simonian‘s Pharmacy Informatics course.

“Using devices such as the Lucia App, we have a valuable opportunity to reduce medication errors and improve patient safety,” Schwab said.

Nguyen agrees that machine learning technology can save lives by lowering the rate of preventable errors such as overlooking medication interactions or allergic reactions.

“One interaction that you wouldn’t have thought twice about can actually end up making the critical difference in a patient’s outcome,” Nguyen said. “So it really can’t be overstated just how important these decision-making tools are in clinical practice.”

Next, Schwab and her team are planning a follow-up study where they will recruit patients to test the accuracy of the Lucia App against a control group.

“I think as clinicians, we can get tunnel vision where we’re just focused on treating the disease at hand,” Schwab said.

“But this project, where we’re working with different physicians and leaders who have a bigger vision for how medicine should improve over the next 20 or 30 years, has allowed me to step outside of my comfort zone and see that the impact we’re making is not just within the hospital.”

This project has helped Nguyen to realize new possibilities as well, looking beyond the niches that pharmacists typically pursue and adopting an entrepreneurial mindset.

“If you make an app or create a device that could help pharmacists, you really have the potential to reach millions of people,” Nguyen said.