AI helps terminal cancer patients make the most of their final days
Encouragement from an AI quadruples the rate of end-of-life conversations between doctors and terminal cancer patients — and those conversations appear to affect how patients choose to live out their final days.
End-of-life care: For people with advanced cancer, there is a window of time between when it becomes likely that their cancer is terminal and when they are no longer well enough to make decisions about their medical care.
During this period, it’s vital that doctors talk to their patients about their two primary options for end-of-life care.
“Many patients would rather have quality of life than quantity of life.”Amy R. MacKenzie
One is to undergo chemotherapy or another systemic treatment until the last possible moment — this may allow the patient to live slightly longer, but it also often includes side effects that may decrease their quality of life and land them in their hospital for more of their final weeks.
The other is to opt primarily for therapies designed to reduce pain and make the patient comfortable for however much time they have left, called palliative care.
“We often assume patients want to live as long as possible,” said Amy R. MacKenzie, an oncologist who chaired a session on improving end-of-life conversations during the 2021 ASCO Annual Meeting, who wasn’t involved in developing this new AI.
“Although that is often true, many patients would rather have quality of life than quantity of life,” she continued, “but we won’t know the answer if we don’t ask the question.”
The challenge: Talking to a patient about end-of-life care can be tough, though, and the vast majority of cancer patients die without their doctor ever broaching the unpleasant but important topic.
“Communicating with cancer patients about their goals and wishes is a key part of care and can reduce unnecessary or unwanted treatment at the end of life,” said Ravi B. Parikh, an associate director at the Penn Center for Cancer Care Innovation.
“The problem is that we don’t do it enough,” he continued, “and it can be hard to identify when it’s time to have that conversation with a given patient.”
AI nudges: In 2019, Parikh led the development of an AI algorithm that analyzes more than 500 variables about a cancer patient to predict their likelihood of dying in the next 6 months. It then assigns a priority rating for an end-of-life care conversation to each patient.
The following year, his team published the preliminary results of a clinical trial testing whether the AI could increase the rate of these important conversations between doctors and cancer patients deemed “high priority.”
“Patients benefit when their health care clinicians understand each patient’s individual goals and priorities for care.”Ravi B. Parikh
During the trial, some doctors were sent weekly emails that included a list of up to six high-priority patients on their schedules for the next week. Other doctors didn’t receive any notifications.
The email intervention nearly quadrupled the rate of these important talks: in the control group, just 3.4% of encounters with high-priority patients included a discussion of end-of-life care, while the rate was 13.5% in the AI intervention group.
What’s new? The Penn team has now published the final results of the trial, which ended up involving more than 20,000 cancer patients during its initial 16-week intervention period (when doctors were sent notifications) and a 24-week follow-up period.
According to the new paper, published in the journal JAMA Oncology, only 7.5% of high-priority patients in the AI intervention group who died during the study underwent chemo or another systemic therapy in the final two weeks of their lives. In the control group, the rate was 10.4%.
This suggests that the AI notifications not only increased the rate of end-of-life care conversations, but that those conversations allowed some patients to choose to change the course of their treatments, potentially enjoying a better quality of life in their final days.
Looking ahead: The Penn team has already expanded the use of its algorithm to all University of Pennsylvania Health System oncology practices and plans to test pairing it with a prompt to refer patients to palliative care specialists.
“While we significantly increased the number of dialogues about serious illness taking place between patients and their clinicians, still less than half of patients had a conversation,” Parikh said.
“We need to do better because we know patients benefit when their health care clinicians understand each patient’s individual goals and priorities for care,” he continued.
We’d love to hear from you! If you have a comment about this article or if you have a tip for a future Freethink story, please email us at [email protected].