Not so elementary, my dear Watson
In just the last few hours, I’ve seen a huge wave of pushback and doubt about Watson, the IBM supercomputer, being used for clinical decision support.
Yesterday, I covered a “healthcare leadership exchange” at IBM’s new Healthcare Innovation Lab in downtown Chicago. I posted some of my observations on the EMR and HIPAA blog, and made the case for diagnostic decision support.
I also wrote a story for InformationWeek, but that hasn’t run. Instead of posting my story, InformationWeek healthcare editor Paul Cerrato wrote a column about Watson already being “beaten in the medical diagnostics race” by Isabel Healthcare, a diagnostic decision support tool that’s been available for years. I have to admit, he’s right. I first interviewed Isabel founder Jason Maude probably in 2002 or so, and I first blogged about the company in 2005. I mentioned Isabel in a 2007 post that, interestingly, also alluded to the work of Don Berwick and Larry Weed.
Cerrato mentioned Jerome Groopman’s 2007 book, “How Doctors Think,” which discussed, in part, how IT could help doctors avoid many types of cognitive errors. “[D]octors tend to lean toward diagnoses that are most available to them in their day-to-day routine,” Cerrato wrote (emphasis in original). That’s exactly what Weed has said for decades, and exactly what Atul Gawande talked about in his groundbreaking book, “Complications.” Computers should not make decisions for physicians, but rather should help them reach the right conclusions, particularly when they see rare cases.
Wouldn’t you know, “e-Patient” Dave deBronkart commented on my EMR and HIPAA post to say he just finished reading Groopman’s book. He tweeted a link to my post, which a few of his 6,500 other Twitter followers noticed. They also noticed EMR and HIPAA grand poobah John Lynn’s comment that the example in yesterday’s Watson demo, a 29-year-old pregnant woman being prescribed doxycyline was “pretty weak.” (He’s right, by the way.) Aurelia Cotta, who blogs about issues such as infertility and adoption, started this thread that also got South Carolina nurse Sunny Perkins Stokes interested:
[blackbirdpie url=”http://twitter.com/#!/AureliaCotta/status/76775042503028737″]
[blackbirdpie url=”http://twitter.com/#!/AureliaCotta/status/76775279565090816″]
[blackbirdpie url=”http://twitter.com/#!/sunnystill/status/76776486119555072″]
[blackbirdpie url=”http://twitter.com/#!/AureliaCotta/status/76782462893699072″]
[blackbirdpie url=”http://twitter.com/#!/AureliaCotta/status/76782822198743040″]
[blackbirdpie url=”http://twitter.com/#!/AureliaCotta/status/76783250663682048″]
[blackbirdpie url=”http://twitter.com/#!/sunnystill/status/76783933106307072″]
[blackbirdpie url=”http://twitter.com/#!/AureliaCotta/status/76784835267534848″]
[blackbirdpie url=”http://twitter.com/#!/AureliaCotta/status/76785122996789248″]
Well, there’s a reason why I call myself a “healthcare” reporter and not a “medical” reporter. I don’t know the science, and I do occasionally get myself in trouble when I start talking about things like whether doxycycline is contraindicated during pregnancy. (To my credit, I did attribute the statement to IBM’s chief medical scientist, Dr. Marty Kohn.)
As I was reading the above tweets and contemplating this blog post, I came across a link to some tongue-in-cheek pushback against Watson in healthcare. An anonymous radiologist who blogs about PACS as “Dr. Dalai” compared Watson to HAL, the diabolical mainframe in “2001: A Space Odyssey.” Dr. Dalai wrote: “Watch out, boys and girls, Watson is headed to a hospital near you, and he (it?) may challenge you as much as he did Ken Jennings.” Jennings, of course, is the Jeopardy! champion whom Watson beat earlier this year.
At first glance, I thought Dr. Dalai was yet another whiny physician clinging to the status quo. But he hit on the real issue: application of knowledge. Quoting from an interview with one of Watson’s programmers, Dr. Dalai noted that the supercomputer is being loaded with all kinds of medical reference material in preparation for “learning” human physiology and ultimately gathering experience in medicine. “This isn’t fair! If I could just take a text book, stick it up my, ummmm, brain, and have it instantly memorized, I would be whiz, too!” he wrote.
Yeah, isn’t that the whole point of clinical decision support?
Nicely done – you’re quick. :)
Dr. Dalai, concerned about Watson replacing human docs, clearly didn’t see your post, nor my comment, both of which talked about Watson as the all-powerful information server, not the decider. I’m SURE little old IBM wouldn’t want to be anywhere NEAR liability for diagnostic decisions. Can you just imagine how a Snidely Whiplash tort lawyer would salivate at the idea of THOSE deep pockets?
Dave, funny you should mention that. After the demo, someone in the audience asked Kohn about liability. His response: “It is a concern, particularly for IBM’s attorneys.” He said that IBM makes it clear that Watson doesn’t make diagnoses or make decisions for doctors. “We are an information service,” Kohn said.
I’m glad I saved my notes. :)
I think that accuracy of sources are important whether or not the reader is a human or a computer. I’m sure Watson will be able to take into account if patient’s are allergic and cost-benefit decisions. What will be more tricky will be the less tangible decisions doctors have to make. Teasing the real symptoms and concerns from patients is something a computer will find very hard. However, advanced CDSS is something that will undoubtable help doctors a lot in the future and prevent many errors. We’re a long way from getting rid of doctors but they might find their role changes from being repository of all knowledge to being a trusted communicator and carer.
Resources in patient care (that the physicians depends upon) have their own accountability/liability factor and while IBM’s Watson may not get labelled to be ” a decider”, it will be “an influencer”.
Isabel, while (relatively) new and less known on this side of the pond, has been in wide use across Europe for a long time – I recall a UK physician saying, 6 years ago, how the word had become a verb, as in , “Did we Isabel that patient in ward 7?”
Decision Support – in any setting, not limited to healthcare – depends on dialog and exchange of thoughts, as scenarios are evaluated, analyzed and sorted in terms of probabilities. When a decision is reached/executed, all other possibilities are cut-off, killed (the root of the word “Decide” comes from Latin cide, which means to kill, as in homicide.)
Sorry for the verbosity in this comment but since your post ended with “whole point of decision support”, I wanted to highlight how human decision making is asymptotic at best.
[…] I decided I’d just go with a small quote from a comment that Chris Paton made over on Neil Versel’s Meaningful Healthcare IT News. Here it is: We’re a long way from […]
Great ending to this post. What a great illustration of the point of CDSS. It is unfair and that’s why it’s so important that we find a way to use it effectively.
I agree the clinicians will always be involved and while the example given is poor it does not detract from the value proposition of the technology. I was quoted in this piece relative to the X-Prize link. What I found sad was a tweet from @HFDigest
[…] At Meaningful Health IT News geek jounro Neil Versel examines pushback this week to IBM’s recent effort to sell its advanced supercomputer technology as the future of clinical decision support in “Not So Elementary, My Dear Watson.” […]
[…] IT leads off the latest edition, hosted by Matthew Holt and John Irvine at The Health Care Blog. My post on pushback against IBM’s Watson makes the cut. Check it out. Related Articles:Health Wonk […]
Actually, I’m JEALOUS of Watson. I wish I could learn at the speed of light.
Despite my acerbic tone, I am incredibly impressed with the project, although I have some reservations about Nuance providing speech recognition. They haven’t perfected that in the Radiology field as yet, and Heaven knows what Watson will or won’t be able to understand.
I’m surprised you didn’t pick up on my little Twilight Zone reference….
DD
Dr. Dalai,
No, I didn’t pick up on the Twilight Zone reference. I’m no spring chicken, but I was born in 1970. I should be proud of myself for remembering HAL, from a movie made in 1968.
When I was in college, we used to joke about learning by osmosis. Just put the book under your pillow and the information would seep into your brain while you slept. Somehow, it never worked. :)
NV