Use of machine learning to optimize referral for early palliative care: Are prognostic predictions enough?

In the article that accompanies this editorial, He et al used a machine learning (ML) algorithm to identify patients with advanced cancer who were receiving non–curative-intent treatment and at risk of death within 1 year to allocate early palliative care services at least 6 months before death as a means to increase use of early palliative care in the context of limited resources. While ML prognostic models are one promising strategy for triaging palliative care services, initiation of palliative care based exclusively on estimated survival and in the months before death likely excludes patients who would benefit from early palliative care initiated at the time of diagnosis, regardless of their estimated survival.

Gary E Weissman, Joseph A Greer, Jennifer S Temel