Penn Medicine algorithm flags patients most in need of advance care planning

Andrea Park - Print  | 

A machine learning algorithm accurately predicted oncology patients' six-month mortality, thus identifying those who would benefit most from a proactive discussion about end-of-life goals and values, a new study from Penn Medicine found.

Researchers from the Philadelphia-based institution developed the algorithm, which used EHR data such as gender, age, comorbidities and lab and ECG results to predict the six-month mortality risk of cancer patients at two hospitals in the University of Pennsylvania Health System.

Of the patients flagged as "high-risk" by the algorithm, just over half died within six months, and nearly 65 percent had died within a year and a half, compared to less than 8 percent of "low-risk" patients. A panel of 15 oncologists surveyed agreed that at least 60 percent of high-risk patients would have benefited from immediate advance care planning meetings with their physicians.

"Patients oftentimes don't bring up their wishes and goals unless they are prompted, and doctors may not have the time to do so in a busy clinic. Having an algorithm like this may make doctors in clinics stop and think, 'Is this the right time to talk about this patient's preferences?'" said Ravi Parikh, MD, the study's lead author and an instructor of medical ethics and health policy at the University of Pennsylvania.

More articles on AI:
Microsoft genomics director: The future of healthcare depends on AI
New Google partnership brings self-driving car tech to hospital rooms
Dartmouth-Hitchcock granted $1.5M to build AI tech for cancer care

© Copyright ASC COMMUNICATIONS 2020. Interested in LINKING to or REPRINTING this content? View our policies by clicking here.