Tool Identifies Patients in Need of Serious Illness Conversations
April 1, 2023 2 minutes read
Text messages generated by a machine-learning tool resulted in clinicians engaging in more serious illness conversations with high-risk patients.1
“There are a lot of barriers to serious illness conversations,” notes Ravi B. Parikh, MD, one of the study authors and assistant professor in the department of medical ethics and health policy and department of medicine at the University of Pennsylvania.
Patients may be unwilling to engage on a particular clinical visit, or there may be cognitive bias on the part of physicians who think it will take too long. Physicians might assume the patient has enough time left to put off the discussion until later. “Patients often miss opportunities to explore their goals and wishes and tradeoffs because doctors don’t initiate it in a timely fashion,” Parikh says.
Parikh and colleagues conducted a randomized clinical trial that included 20,506 patients with cancer. High-risk patients were identified with a validated machine-learning algorithm to predict six-month mortality. Physicians received weekly emails comparing their rates of serious illness conversation to their peers, weekly lists of high-risk patients, and text messages to prompt doctors to the need for a serious illness conversation before encounters with those patients.
Among patients included in the intervention, rates of receiving chemotherapy at the end of life were 25% lower compared to a control group. The intervention resulted in an overall increase in serious illness conversations from 3.4% to 13.5%. “The conversations quadrupled after the intervention and stayed at an elevated level for close to six months,” Parikh reports.
For some physicians, the performance reports were a strong motivating factor. Those physicians realized they were not engaging in as many conversations as they should be. Also, physicians were not documenting the discussions to inform other providers. “There was a recognition of the problem, and willingness to accept this nudge,” Parikh says.
While all clinicians improved their conversation rates, most did not change too much. “The increase was mainly driven by a small number of physicians who really bought into the intervention,” Parikh observes.
Other physicians doubted the algorithm’s ability to correctly identify high-risk patients. “There were some false-positives, where patients got flagged who weren’t actually high-risk in the eyes of the physician,” Parikh says. To engender trust, clinicians need to know why the algorithm is generating a prediction. “The more we can get to that, the more impact we are going to have,” Parikh says.
REFERENCE
1. Manz CR, Zhang Y, Chen K, et al. Long-term effect of machine learning-triggered behavioral nudges on serious illness conversations and end-of-life outcomes among patients with cancer: A randomized clinical trial. JAMA Oncol 2023; Jan 12:e226303. doi: 10.1001/jamaoncol.2022.6303.
Text messages generated by a machine-learning tool resulted in clinicians engaging in more serious illness conversations with high-risk patients.
You have reached your article limit for the month. Subscribe now to access this article plus other member-only content.
- Award-winning Medical Content
- Latest Advances & Development in Medicine
- Unbiased Content