he elderly man couldn’t breathe and he didn’t know why.

He wore an oxygen mask when the EMTs wheeled him in. His lungs were too diseased for the mask to help much, and he was too confused to follow questions from the team in the emergency room at Boston’s Brigham and Women’s Hospital.

And so, while the man’s adult daughter watched in anguish, he wheezed and hyperventilated in terror. His ER doctor, Ziad Obermeyer, figured the man might live two more hours without aggressive intervention, so he asked the daughter if her father would want to be put on a respirator. She didn’t know.


Obermeyer had at least 10 patients waiting for him, some of them acutely ill, and he’d been on duty for hours. This patient was a stranger to him, as were the relatives.

It’s a surprisingly common theme in American emergency rooms. Family members arrive with patients who are verging on death but can’t articulate whether they want potentially life-saving medical procedures that might also carry the risk of debilitating consequences. They have never had the talk.

Obermeyer, it turns out, is developing a novel approach so that patients — and their clinicians — avoid that kind of scenario. The idea is to help doctors know when patients have begun a decline toward death and should have a conversation about end-of-life choices — preferably before they reach the ER.

“It’s an incredibly hard place to have that conversation,” Obermeyer said of the ER. “You have about 10 minutes to make a really important decision, and you’re doing it in a noisy place where the doctor has 15 other patients they’re worried about.”

Doctors typically broach the subject of patients’ end-of life choices an average of 33 days before they die, and often when they’re in crisis, according to research by Dr. Jennifer Mack, a pediatric oncologist at the Dana-Farber Cancer Institute and Boston Children’s Hospital.

Sometimes doctors put off those conversations because neither they nor their patients want to give up hope. Other times, however, doctors don’t engage in the conversations simply because they don’t realize death is so near for their patients.

Oncologists, cardiologists, and other specialists can often predict a patient’s rate of decline based on a specific disease, Obermeyer said. But patients, particularly those who are elderly, often suffer from more than one serious illness that make it more difficult to predict when they’re near death. So a pulmonologist might treat someone’s pneumonia, for instance, without recognizing it signals a broader decline.

Obermeyer, along with Harvard economics professor Sendhil Mullainathan, wants to help clinicians better determine when patients begin their decline. And when patients do deteriorate, the clinicians can begin to grapple with questions about the care the patients want to receive.

Even patients who have pondered those questions with their families or physicians earlier in their lives might come to the issue with new perspectives when confronted with the realities of the choices they face.

For now, Obermeyer and Mullainathan are mining millions of electronic health records that detail patient illnesses, treatments, and outcomes in the final years of life. They believe that the patterns they detect will allow them to better predict the future for other patients.

The idea was all-but-unthinkable a decade ago, but advances in Big Data (the ability to store massive amounts of data cheaply) and “machine learning” (the ability to mine that data for intelligence) have put this wisdom within reach.

Three years ago, Obermeyer received a National Institutes of Health grant designed to assist “exceptionally creative scientists who propose highly innovative approaches to major challenges in biomedical research.”

He said his research has already yielded promising results, with published data coming soon. The next step is to establish a system to alert doctors that a patient’s condition might warrant an end-of-life conversation.

“I imagine the man I saw would’ve had a super-high probability of dying, based on the algorithms we’re using now,” Obermeyer said. “And I think if he’d been brought to someone’s attention, whether his primary care doctor or the person taking care of him in the hospital at an earlier point, I think things would’ve gone better.”

In the case of the man Obermeyer saw in the ER, the patient’s death went as well as it could have. Obermeyer spoke by phone to the man’s son, who had the authority to make medical decisions.

Obermeyer explained the situation: Many elderly patients need a tracheostomy after being removed from a ventilator, and patients with tracheostomies often have a hard time eating, so they need a feeding tube as well. What Obermeyer also knew, but didn’t say, is that of every 100 Medicare patients who are on ventilators for more than three days, fewer than 10 will go home, and a vast majority will end up in long-term care facilities because of complications.

The son decided against a ventilator.

By the time the son and another sister arrived, the medical team had given the man morphine to relax his breathing. “The morphine took the edge off the situation in the room,” he said. “He could hold his children’s hand and they could talk to him.”

The man died not long after. Later, the son shook Obermeyer’s hand and said it was the best thing for the father.

“It’s hard to describe how I felt,” Obermeyer said. “On the one hand I was glad we hadn’t put in a breathing tube. But it’s hard to classify that as a good death.”

Sign up for Morning Rounds

A daily dose of health and medicine news — and a finalist for Digiday’s best email newsletter.

Recommended Stories