Play Live Radio
Next Up:
0:00
0:00
Available On Air Stations
Watch Live

Science & Technology

New tools in the fight to stop a mysterious killer called sepsis

Sepsis is a disease that is deadly and still somewhat of a mystery. At UC San Diego, doctors and med-tech engineers are trying to crack the code with new diagnostic techniques and artificial intelligence. KPBS sci-tech reporter Thomas Fudge has the story.

Steve Israel was a healthy man in his early 40s when, one day, he noticed symptoms that made him think he had the flu. Then he started having trouble breathing, so he went to an urgent care clinic.

A Temecula resident, he was sent to the local hospital. Then, as things got worse, he was medevaced to UC San Diego Medical Center. Just before that, he said his wife got some disturbing advice at Temecula Valley Hospital.

“The medical folks there told my wife, make sure you say goodbye before he gets on the helicopter. So the way she interpreted that was, we don’t know if he’s going to be alive when you get down to San Diego and when the helicopter gets there,” he said.

Advertisement

Israel had gotten sepsis. He says his lungs soon collapsed and his kidneys and gallbladder also failed. The thing that saved his life, at UCSD, was a machine that oxygenates the blood when your lungs stop working.

Put simply, sepsis is an infection, and a physical reaction to it, that gets out of control. The infection is often bacterial but it can also be viral or fungal. Every year in the U.S. 1.7 million people get it and 270,000 die from it.

The groups of people most at risk include infants, the elderly and people with weak immune systems. Though some are seen to be at greater risk, sepsis remains something of a mystery.

Israel still has no idea how he got it seven years ago, or what kind of infection caused it. Nothing about his age of medical history made him a likely candidate.

One man who played a crucial part of Steve’s story was Dr. Gabriel Wardi. He was the Temecula emergency room doctor who ordered the helicopter flight to UCSD, where he now is director of the school’s hospital sepsis program.

Advertisement

Wardi said with sepsis, speed of diagnosis is a matter of life or death.

“Because the earlier we catch sepsis, and the earlier we start some of the life-saving therapies, the better people do,” Wardi said. “And that way we are able to start the therapies like IV antibiotics, and give patients some fluids to kinda prevent that worsening infection.”

Trying to spot and prevent sepsis

At UCSD, Wardi and colleagues created and use an AI system to spot people at high risk of sepsis.

“As soon as someone checks into our triage system, our machine learning model starts to ingest data from them. What does that mean? It means their vital signs. Their comorbidities. The medications that they take at home. And it starts to generate a risk score,” he said.

If the risk score is high enough, the system sends out an alert to the emergency department staff. UCSD deep learning researcher Aaron Boussina was on the team that created the AI model.

“If a patient walks into the emergency department, there’s about a 5% occurrence rate of sepsis,” Boussina said. “If you get one of these alerts, your risk is ten to 20 times higher.”

Dr. Gabriel Wardi listens to a conversation in a building corridor at UC San Diego on Feb. 28, 2024
Dr. Gabriel Wardi listens to a conversation in a building corridor at UC San Diego on Feb. 28, 2024

Wardi said their AI system at UCSD’s two hospitals “fires,” sending out sepsis alerts, about a dozen times a day, and four to five of those patients will get sepsis during their hospitalization. He said he believes every month for to five lives have been saved since the system began to be used.

But even if you identify a patient who’s likely to get sepsis, you still can’t be sure what’s causing it, and what the best treatment would be.

That’s where another diagnostic tool comes in: Melting DNA to identify whether the infection is bacterial, viral or fungal. A sample is taken from a person, it’s placed under a microscope and heat is applied to it.

Stephanie Fraley is a bioengineer, also at UCSD, who has created a new technology she calls MeltRead. Sitting in her lab, she entwines two of her fingers to mimic a double helix, and says when a DNA sample melts, that twisted double helix actually unravels.

The amount of heat that requires, and the way the DNA unravels, identifies the pathogen that caused the infection.

“We get a signature from that. We call it a melt signature, and that maps to sequences of pathogens that we already have melted and know what they look like,” she said. “And so we’re able to match melting signatures from a sample to melting signatures that we’ve seen before.”

She says using MeltRead instead of waiting for a blood culture saves a lot of time, taking hours not days. Again, with sepsis, time is of the essence. Fraley still awaits Food and Drug Administration (FDA) approval to use her technology in a clinic. She and others have founded a company to commercialize MeltRead when the time comes.

When sepsis remains a mystery

For all the research and the efforts to diagnose sepsis, Steve Israel still has no idea how he got it seven years ago or what kind of infection it was.

When he reflects on his experience, he thinks of the good fortune he had, having a medical team that knew about sepsis and took his condition seriously … like the physician's assistant at the urgent care clinic who called an ambulance to send him to the hospital.

“That person could have sent me home and said, ‘Here’s some medication. Here’s a breathing treatment,’” Israel said. “I think if that didn’t happen I would have died in my living room at home a couple hours later.”

But now he’s still alive, living in Temecula with his wife and three kids.