OPINION:
Here’s a comforting thought: Google’s Medical Brain can predict with up to 95 percent accuracy the day you will die.
Cue mad scientist laugh.
Is there nowhere Google’s now-near-omnipresent technological footprint won’t tread?
The company’s artificial intelligence breakthrough is being billed as a friendly hospital helper — a real boon to the warning systems already being used by medical professionals to better address patients’ needs.
Apparently, its beauty comes from its data collection system.
Medical professionals and technology whizzes have been trying for years to find A.I. ways to use all the data in patients’ records in order to provide better care. But for the most part, inputting and analyzing every single record was a too time-consuming, too costly affair that saw much of the information being tossed to the side. Enter Team Google.
Google’s Medical Brain, in testing, has been able to disseminate even doctors’ notes scribbled on old charts, or memos scratched into the margins of PDFs.
According to findings published in May in the journal Nature, Google’s model was then able to take all that data to predict with 86 percent accuracy on how long patients would have to remain in the hospital for treatment, and with 95 percent accuracy on whether they would die during their stays. Those percentages come in comparison to hospitals’ traditional computerized modeling which, on average, only rated 76 percent and 86 percent on ye olde accuracy scales for those two categories, respectively.
“These models outperformed traditional, clinically used predictive models in all cases,” said Google research scientist, Dr. Alvin Rajkomar, in Nature.
Wunderbar.
The Google researchers even gave a specific case of how they bested a certain city hospital’s own computerized predictions for a patient who visited with late-stage breast cancer. After the female patient received a radiology scan and visits from two doctors, the hospital’s internal Early Warning Rating gave her a 9.3 percent chance of dying during her hospital stay. Google, however, disseminated the same data and found this: the woman actually had a 19.9 percent chance of dying during her hospital stay. Who’s right?
Turns out, the woman died within 10 days of being admitted. Cheers and applause for Google — black mourning wear for the rest. Rating a human’s chances for death does remind of lab rats running around a maze, yes?
But that touches on a basic question that ought to guide all A.I. in the medical field, and it’s one that goes like this: How much is too much?
Insurance companies would no doubt have a field day with having an inside scoop on which treatments are worthwhile, which are likely to be wasted dollars. So, too, would cash-strapped hospitals, overworked nursing and medical staff — maybe even families considering the pros-cons of pulling plugs.
At its best, A.I. can offer medical professionals insight into treatments to better serve their patients, and in so doing, curb costs for all involved. At its worst? A.I. will be the sneaky way the medical powers-that-be decide and separate the haves from the have-nots, the gets, from the get-nots.
Will Medical Brain ultimately become a tool to bypass life-saving treatments deemed by bean-counters as too wasteful to try? It’s definitely too soon to tell.
But methinks Americans ought to read the warning labels on this one.
• Cheryl Chumley can be reached at cchumley@washingtontimes.com or on Twitter, @ckchumley.
Please read our comment policy before commenting.