Artificial intelligence is quietly moving into one of the most sensitive areas of medicine: end-of-life decisions. New research suggests AI could potentially predict a patient’s wishes regarding life-sustaining treatment more accurately than family members, and even encourage proactive planning before crises arise. But this raises complex ethical questions about the role of algorithms in deeply personal choices.
The Rise of Predictive AI in Death Care
Two recent studies highlight how AI is being tested in end-of-life care. One European study found that an AI “patient preference predictor” accurately guessed end-of-life wishes (like whether a patient would want CPR) 71% of the time, outperforming both medical professionals and even the patient’s partners. The other, conducted at BJC HealthCare hospitals in St. Louis, showed that sharing AI-generated mortality risk predictions with doctors led to a significant increase in end-of-life planning, including more use of palliative and hospice care.
The key takeaway: AI isn’t about making decisions, but about prompting conversations and ensuring patients express their wishes before it’s too late. The St. Louis hospitals saw a 33% drop in 30-day mortality rates after implementing the AI-assisted approach.
Human Oversight Remains Crucial
Researchers are quick to emphasize that AI isn’t meant to replace human judgment. The European study stressed the need for “co-reasoning” between AI and medical professionals, while the St. Louis group trained clinicians to have better “goals of care” discussions. The lesson from past failures – like the 1990s APACHE III algorithm – is that blunt, unguided predictions can be traumatizing, especially for patients already under stress.
The modern approach: AI alerts are first reviewed by a second clinician before being shared, avoiding “alert fatigue” and ensuring the information is presented with sensitivity.
The Next Frontier: “Moral” AI?
Some researchers are even exploring the possibility of an AI surrogate that doesn’t just predict preferences but also considers a patient’s values, relationships, and cultural worldview. This concept, proposed by a University of Washington researcher, aims to create an AI that is not only accurate but also “morally adequate” in its representation of the patient.
The big question: Can an algorithm truly capture the nuances of human values, or will relying on AI create a dangerous detachment from the emotional and spiritual aspects of death?
Real-World Concerns and Economic Implications
The growing use of AI in medicine raises concerns about overreliance on technology. Critics warn that predictions are not static, and patients may seek conflicting information from other AI sources. Meanwhile, hospitals are also examining the economic impact, including potential savings from reduced hospital stays and intensive care unit usage.
The bottom line: AI in end-of-life care is not a distant future scenario but a rapidly evolving reality. The challenge lies in ensuring that technology serves human values, not the other way around.
