‘90% Accuracy Rate’: New Artificial Intelligence Can Predict Your Death

New deep learning algorithms, commonly referred to as artificial intelligence, are increasingly finding a home in the medical industry, where health care professionals are often overwhelmed. In palliative care centers, which include nursing and end of life care, an aging population has stretched medical and personnel resources to the limit, which just may mean this is the perfect time for AI to step in and help with both monitoring patients’ vital signs and making potentially life-saving prognoses.

The Food and Drug Administration (FDA) just approved the first algorithm that can monitor a patient in critical care and anticipate a life-threatening change in condition by up to 6 hours. The algorithm, called Wave Clinical Platform and developed by medical technology company ExcelMedical, is able to keep a 24/7 vigil over a patient, which is simply not possible for already overworked human care workers triaging in busy hospitals. The AI can also monitor patterns in vitals so as to determine whether, say, a sudden and simultaneous drop in oxygen saturation and a spike in blood pressure could be lethal.

Meanwhile, a research team at Stanford University is using an AI algorithm to predict patient mortality. A notoriously difficult task, predicting when a patient will die involves assessing a variety of factors from age and patient history to drug response and the illness itself. But it’s important to be as accurate as possible. Admitting a patient into end-of-life care too soon consumes valuable resources; admitting them too late risks leaving the patient cut off from their family.

Stanford’s AI has a 90% accuracy rate, which, according to some, is surprisingly and unsettlingly good. Created by feeding electronic health records (EHR) into a deep learning neural network, the AI is able to “sense” patient mortality within the next three to twelve months.

Jeremy Hsu of IEEE Spectrum described the model used as “an 18-layer Deep Neural Network that inputs the EHR data of a patient, and outputs the probability of death in the next 3-12 months.”

However, researchers admit they do not understand how the algorithm comes to its conclusion, a conundrum known as the “black box” problem. The Stanford team’s work is described in full detail in their paper, “Improving Palliative Care with Deep Learning.”

Ken Jung, a Stanford Medicine research scientist, and Anand Avati, whose team created the AI, hope the use of AI technology in medical environments could help with a number of problems, including palliative care and triage. Currently, while 80 percent of people say they want to die at home, only about 35 percent are able to. Employing new deep learning technology could enable better end of life care preparation and more personalized, consistent care once the patient is hospitalized.

That is, of course, if people can get past the stigma of having an advanced computer algorithm monitoring them when they’re ill.

Delivered by The Daily Sheeple

We encourage you to share and republish our reports, analyses, breaking news and videos (Click for details).

Contributed by Jake Anderson of theantimedia.org.

The “Anti” in our name does not mean we are against the media, we are simply against the current mainstream paradigm. The current media, influenced by the industrial complex, is a top-down authoritarian system of distribution—the opposite of what Anti-Media aims to be. At Anti-Media, we want to offer a new paradigm—a bottom-up approach for real and diverse reporting. We seek to establish a space where the people are the journalists and a venue where independent journalism moves forward on a larger and more truthful scale.