More certain predictions could help patients make better end-of-life decisions
A new algorithm developed at Stanford could make it easier for seriously ill patients to estimate how long they have left to live, helping them make critical end-of-life decisions.
“Analyzing data from hundreds of thousands of anonymized medical records, the model predicts which patients are likely to die in the next 3 to 12 months,” said Lloyd Minor, dean of the Stanford University School of Medicine, in a column for the Wall Street Journal.
The program, which uses a type of artificial intelligence called deep learning, was initially tested using medical data of patients who had already passed away. In nine out of 10 test cases, the algorithm correctly predicted their life expectancy.
According to Minor, the ability to predict how much time a seriously ill patient has left with some degree of certainty would provide “enormous value” for them and their families. “It’s a question that informs many others,” he said. “How long should a dying patient ‘fight’ a terminal illness, and when should that person focus instead on minimizing suffering? Is a day at home more valuable than a week in the hospital?”
Minor noted that late entry into palliative care — care that focuses on minimizing suffering — could mean extended periods of hospitalization and aggressive treatment with limited chances of improvement. He also suggested a significant “financial burden” from pursuing expensive, curative treatments too late: Medicare spending on beneficiaries who died at some point during the year more than doubled between 2000 and 2014.
“On the other hand, those who enter palliative care too early risk missing out on treatments that could improve their condition and extend their life,” he added. Such early availers of palliative care might also be taking a valuable opportunity away from those that need it most: citing data from the National Palliative Care Registry in the US, he said less than half of patients that need palliative care actually receive it.
But Minor clarified that any AI program to determine life expectancy isn’t going to take decisions away from human beings. In the end, mentally competent patients would still have the chance to decide whether they’d rather spend time with family, avoid severe pain, or explore every chance to beat their disease.
“I don’t pretend AI is a panacea. An algorithm isn’t going to make decisions for doctors or patients,” he said. “But it can help inform their choices by providing them with insight they’ve never had before.”
