A machine learning-based behavioral intervention could improve end-of-life cancer care

Electronic alerts were delivered to health care physicians based on a machine learning algorithm that predicts a quadrupling risk of death for conversations with patients about end-of-life care preferences, according to the long-term results of a randomized clinical trial published by Penn Medicine investigators in Oncology Gamma Today. The study also found that reminders generated by machine learning significantly reduced the use of aggressive chemotherapy and other systemic treatments at the end of life, which research shows is associated with poor quality of life and side effects that can lead to unnecessary hospital admissions in their final days.

For patients when cancer has advanced to an incurable stage, some may prioritize treatment that extends their lives as long as possible, and others may prefer a plan of care designed to reduce pain or nausea, depending on the outlook for their disease. Talking to patients about their diagnoses and values ​​can help clinicians develop care plans that better align with each individual’s goals, but discussions are essential before patients become too ill.

“This study shows that we can use informatics to improve end-of-life care,” said senior author Ravi Parikh, MD, an oncologist and assistant professor of medical ethics, health policy, and medicine at the university’s Perelman School of Medicine. Pennsylvania and associate director of the Penn Center for Innovation in Cancer Care at Abramson Cancer Center. “Communicating with cancer patients about their goals and desires is an essential part of care and can reduce unnecessary or unwanted treatment at the end of life. The problem is that we don’t do it enough, and it can be difficult to determine when it’s time to have that conversation with a patient. specific “.

Parikh and colleagues previously showed that a machine learning algorithm can identify patients with cancer who are at risk of dying within the next six months. They paired the algorithm with behavior-based “alerts” in the form of emails and text messages to prompt doctors to initiate serious patient conversations during appointments with high-risk patients. The preliminary results of the study, published in 2020, showed that the 16-week intervention tripled the rates of these conversations.

The study marks an important step for AI in oncology, as the first randomized trial of a machine learning-based behavioral intervention in cancer care. The study included 20,506 patients treated for cancer at several Penn Medicine sites, with a total intake of more than 40,000 patients, making it the largest study of a machine learning-based intervention focused on critical disease care in oncology.

Results published today show that after a 24-week follow-up period, conversation rates nearly quadrupled, from 3.4 percent to 13.5 percent, among high-risk patients. Use of chemotherapy or targeted therapy in the last two weeks of life decreased from 10.4 percent to 7.5 percent among patients who died during the study. The intervention had no effect on other measures of end of life, including enrollment in hospice homes, length of stay, inpatient death, or intensive care unit use at end of life.

Notably, an increase in conversations about goals of care was also observed in patients not identified by the algorithm as high risk, suggesting that the alerts caused clinicians to change their behavior across their practices. The increase was seen in all patient demographics, but was greatest among Medicare recipients, suggesting that the intervention may help correct the disparity in conversations about a serious illness.

Based on the results of this study, the research team extended the same approach to all oncology practices within the University of Pennsylvania Health System and are currently analyzing these findings. Additional plans for the research include pairing AI algorithms with a prompt for early palliative care referral and using the algorithm for patient education.

“While we’ve dramatically increased the number of conversations about a serious illness occurring between patients and their doctors, less than half of patients are still talking,” Parikh said. “We need to do a better job because we know that patients benefit when healthcare practitioners understand each patient’s individual goals and priorities for care.”

The study was supported by the National Institutes of Health (5K08CA26354, K08CA263541) and the Penn Center for Precision Medicine.