RESEARCHERS have discovered that a machine learning algorithm can improve end-of-life care for patients with cancer. Electronic ‘nudges’, which are delivered to clinicians with an algorithm that predicts the risk of mortality, have raised rates of conversations with patients regarding end-of-life care preferences fourfold.
A randomised clinical trial also discovered that reminders triggered by machine learning decreased the use of aggressive chemotherapy significantly, as well as reducing other systemic therapies for end stage patients. These treatments have previously been linked to side effects, leading to unnecessary hospital stays and poor quality of life.
Senior author of the study, Ravi B. Parikh, Perelman School of Medicine, University of Pennsylvania, and Associate Director of the Penn Center for Cancer Care Innovation, Abramson Cancer Center, Philadelphia, Pennsylvania, USA, commented: “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.”
Parikh and colleagues used a previous machine learning algorithm, which can identify high-risk patients with cancer who could pass away within the next 6 months, and added a behavioural-based ‘nudge’ system, in the form of text messages and emails, to induce clinicians to have conversations about end-of-life care with patients. An intervention at the 16-week period was found to triple the rates of such discussions.
The study consisted of 20,506 patients who were being treated for cancer at several Penn Medicine clinics. There were more than 40,000 patient encounters in the cohort, allowing the largest study with an intervention of this kind to be carried out in serious illness care in oncology. After a follow-up period of 24 weeks, conversation rates with high-risk patients rose to 13.5% from the previous 3.4%. The use of chemotherapy or targeted therapies decreased from 10.4% to 7.5% among those who died during the course of the study. Other end-of-life metrics, such as intensive care unit use and hospice enrolment, were not affected.
The study highlights the importance of using artificial intelligence within oncology and constitutes the first randomised trial of a machine learning-based behavioural intervention in the field. The study size has since been enlarged to include all oncology practices within the University of Pennsylvania Health System.