Can AI Predict Your Death?
How artificial intelligence could make people more comfortable at the end of life and save money in the process
As a physician who trained to preserve life, I find discussions with people around death can be extremely difficult, regardless if the end of life is near or distant. During my residency training, I was seeing an older person with leg pain. I helped him with some topical medication and other pain solutions. Things stayed stable for a while, but a few months later the pain worsened. Imaging showed a mass growing on the femur. He had metastatic cancer. He deteriorated quickly, and the family was kind in letting me be there when he passed.
If I’d had to guess just months prior how long this man was going to live, 10 years would have seemed a reasonable guess. But survival time—the predicted time to death—is something that clinicians have been getting wrong for a long time.
Since survival time is so hard to predict, it leads to patients often not receiving the care they desire. Approximately 70% of Americans facing end-of-life care would like to die at home. Between 2003 and 2017, home deaths increased from 23.8% to 30.7%. While this shows movement toward making death at home more of a priority, it is far from the overall preferences of patients.
