New artificial intelligence prediction model may reduce hospital readmission rates

By Brittany Otis
botis@health.southalabama.edu

Artificial intelligence may be a useful tool for providers to better predict patient outcomes. It was beneficial for Amod Amritphale, M.D., the director of cardiovascular research and an interventional cardiologist at USA Health, who used computer algorithms to learn more about patients who suffered a stroke. Public data showed that some of those patients were readmitted to the hospital within 30 days, even after undergoing a surgical procedure to open a narrowed carotid artery.

Carotid arteries are blood vessels located on both sides of the neck that deliver blood to the head and brain. The procedure, known as carotid artery stenting (CAS), is usually performed as a preventative method or after a stroke.

Amritphale launched a study to describe the rates and causes of unplanned readmission following CAS using a national healthcare database. His findings were published in the April 2021 edition of Advances in Therapy.

“The last thing any doctor wants to see is a patient back in the hospital, especially during a short period of time, with adverse or negative outcomes,” Amritphale said. “I wanted to understand more about the outcomes of patients who underwent CAS or cardiovascular procedures and harness the power of artificial intelligence to develop a strong prediction model.”

Amritphale explains that if 100 patients underwent CAS and 10 people came back to the hospital within a short period following their discharge, he would compare those 10 patients with the remaining 90 people who were not readmitted. “I would take that data and say ‘What made these 10 people come back to the hospital?’ So, the next time I see someone like them, I can tell that patient they are at risk for readmission and discuss ways to prevent it,” he said.

The artificial intelligence tool also will help patients.

“If patients know beforehand that they are at high risk to come back to the hospital, they will pay closer attention to their bodies,” Amritphale said. “They will focus more on diet and exercise or if they are experiencing symptoms, they will know to call their doctor or go the emergency room sooner.”

The novel model presented in this study boasts 79 percent capability to accurately predict at-risk patients.

Amritphale hopes physicians use artificial intelligence tools to help diagnose patients, which could make a difference in healthcare. “Money being spent on healthcare could have been saved and used towards prevention of cardiovascular diseases if we use this method,” he said. “This will better help providers as we strive to create the best outcomes for our patients.”

Collaborators on the study include researchers from USA Health, the National Institutes of Health, The George Washington University and the University of California, Los Angeles. Advances in Therapy is an international journal that covers the use of therapies, devices and surgical techniques across all therapy areas.