Machine Learning-Triggered Reminders Improve End-of-Life Cancer Care

A study published last week in JAMA Oncology highlights how researchers at Penn Medicine have developed a machine-learning (ML) intervention designed to predict cancer mortality risk and deliver electronic nudges to spur conversations with patients about their end-of-life preferences.

According to the press release, these conversations are a key part of a patient's care plan, and clinicians must work to develop a care plan in line with the patient's goals before they become too ill. Depending on the outlook for their disease, one patient may prioritize treatment that will extend their life for as long as possible, while another may prefer a care plan designed to minimize pain or nausea once cancer advances to an incurable stage.

To help encourage clinicians to start these conversations in a timely manner, the researchers developed a tool to predict six-month mortality and deliver 'nudges' to providers.

“This study demonstrates that we can use informatics to improve care at end of life,” said senior author Ravi B. Parikh, MD, an oncologist and assistant professor of medical ethics and health policy and medicine in the Perelman School of Medicine at the University of Pennsylvania and associate director of the Penn Center for Cancer Care Innovation at Abramson Cancer Center, in the press release. “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. The problem is that we don’t do it enough, and it can be hard to identify when it’s time to have that conversation with a given patient.”

The research team had previously demonstrated the predictive performance of the algorithm, pairing it with behavior-based 'nudges' to initiate serious illness conversations with high-risk patients. These nudges, which were delivered in the form of emails and text messages over 16 weeks, tripled the rates of these conversations.

This study expanded on that research, including 20,506 patients treated for cancer at several Penn Medicine locations, with a total of more than 40,000 patient encounters evaluated over a 24-week follow-up period.

During that time, the researchers found that rates of advanced care planning conversations nearly quadrupled among high-risk patients, rising from 3.4 percent to 13.5 percent. Rates of potentially harmful therapy at end-of-life decreased by 25 percent, with the use of chemotherapy or targeted therapy in the final two weeks of life decreasing from 10.4 percent to 7.5 percent among patients who died during the study.

The research team also found that the increase in conversations about goals of care was observed in patients who weren’t flagged by the algorithm as high-risk, suggesting that the nudges caused clinicians to change their behavior across their practice. Further, the increase was observed across all patient demographics but was larger among Medicare beneficiaries, suggesting that the intervention may help rectify disparities in serious illness conversations.

But, according to Parikh, more work must be done to encourage these conversations.

“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.”

The researchers intend to expand the research even further by pairing artificial intelligence (AI) algorithms with a prompt for early palliative care referral and using the algorithm for patient education.