Tag: Eric Topol

Can we trust ChatGPT to get the basics right?

Apr 12, 2023

by MATTHEW HOLT

Eric Topol has a piece in his excellent newsletter Ground Truth‘s today about AI in medicine. He refers to the paper he and colleagues wrote in Nature about Generalist Medical Artificial Intelligence (the medical version of GAI). It’s more on the latest in LLM (Large Language Models). They differ from previous AI which was essentially focused on one problem, and in medicine that mostly meant radiology. Now, you can feed different types of information in and get lots of different answers.

Eric & colleagues concluded their paper with this statement: “Ultimately, GMAI promises unprecedented possibilities for healthcare, supporting clinicians amid a range of essential tasks, overcoming communication barriers, making high-quality care more widely accessible, and reducing the administrative burden on clinicians to allow them to spend more time with patients.” But he does note that “there are striking liabilities and challenges that have to be dealt with. The “hallucinations” (aka fabrications or BS) are a major issue, along with bias, misinformation, lack of validation in prospective clinical trials, privacy and security and deep concerns about regulatory issues.”

What he’s saying is that there are unexplained errors in LLMs and therefore we need a human in the loop to make sure the AI isn’t getting stuff wrong. I myself had a striking example of this on a topic that was purely simple calculation about a well published set of facts. I asked ChatGPT (3 not 4) about the historical performance of the stock market. Apparently ChatGPT can pass the medical exams to become a doctor. But had it responded with the same level of accuracy about a clinical issue I would be extremely concerned!

The brief video of my use of ChatGPT for stock market “research” is below:

Dr. Topol’s comment on LongCOVID and the heart is misleading/lacking context

Jul 18, 2022• 1

By ANISH KOKA

It’s been a while but Anish Koka, a one time regular writer on THCB and occasional THCB Gang member, is back publishing up a storm on his Substack channel. You may recall that his political and clinical views don’t always mesh with some of the wooly liberals we feature on THCB (cough, cough, me), but we are delighted to be back publishing some of his pieces–starting with a look at a tweet from one of America’s most prominent cardiologists.–Matthew Holt

Given Twitter’s commitment to the truth in Medicine, I thought I would try to give them a hand by analyzing a semi-viral tweet about COVID and the heart.

Earlier this year (April 2022), the most influential cardiologist in the world tweeted about a study on the long term cardiac effects of COVID (LongCOVID).

Medical trainees who trained in the early 2000s like I did know Dr. Topol as an absolute legend in the field of Cardiology. He was responsible for seminal work in Cardiology in the 1980’s on the use of clot busting drugs for patients having heart attacks, and became head of cardiology for the famed Cleveland Clinic at the age of 36! (I vaguely recall feeling like I was starting to understand Cardiology at the age of 36.) He’s since moved on to do many other things, and is a potent voice that may have been instrumental in the FDA delaying approval of the mrna vaccines until after the 2020 election.

Nonetheless, this paper that he is giving his significant stamp of approval to has significant issues. As far as I can tell individuals with LongCOVID were recruited by advertising in LongCOVID support groups. No independent assessment carried out as far as I can tell clinically. If you say you have it—> you’re in.

Eric Topol: The Patient Will See You Now …

Sep 23, 2014• 2

JANE SARASOHN-KAHN

Moore’s Law is coming to medicine. And it will look and feel a lot like Uber: with rich technology underpinning, consumer-service oriented and friendly, and shaking up the professionals at the front line of the business (from taxi drivers to physicians).

Eric Topol, physician and editor-in-chief at Medscape, told a standing-room-only audience at the kickoff of the 8th annual Health 2.0 Conference that the democratization of health care is coming based on consumers’ use of eight drivers: sensors, labs, imaging, physical exams, access to medical records, transparency of costs, and digital pills.

Breaking: Health 2.0 Fall Conference Lineup Is Out!

Jun 25, 2014• 1

By SOPHIE PARK

Tech giants storming the digital health landscape will be center stage at Health 2.0’s 8th Annual Fall Conference in Santa Clara, CA. An impressive line-up of health and tech executives headline three full days of live demos and innovative sessions. Highlights include keynotes from visionary physicians Eric Topol and Patrick Soon-Shiong as well as Samsung Electronics President, Young Sohn in conversation with Health 2.0 CEO, Indu Subaiya. Leaders from Intel, Humana, IBM Watson, Qualcomm Life, Merck, athenahealth, eClinicalWorks and the Office of the National Coordinator for Health Information Technology (ONC) will showcase and discuss their latest technologies and initiatives on the main conference stage this fall. As always, Health 2.0 features over 150 live demos of new technology, 250+ speakers, 50+ sessions, more networking, and deals-done than anywhere else in health technology.

The main stage will feature the following panels:

Smarter Care Delivery: Amplifying the Patient Voice: Matthew Holt, Co-Chairman of Health 2.0, sparks the discussion on how new technology platforms, payors, and providers are working together for enhanced patient care delivery and engagement.

Consumer Tech and Wearables: Powering Healthy Lifestyles: Bringing together the most innovative wearables that are pushing individualized medicine into the future, Indu Subaiya, CEO of Health 2.0, leads this session focused on how consumers are experiencing new lifestyles centered around technology. Don’t miss the live fashion show featuring all the latest trends in digital health wearables!

Buy, Sell, Exchange: New Markets for Consumers, Employers, and Providers: Nearly a year after ACA implementation, this session will dive into the new ways benefits are being offered to consumers, how employers are buying care directly, and what new technologies are enabling change in direct care provision.

Data Analytics: From Discovery to Personalized Care: This panel focuses on how data analytics and powerful visualizations are pushing forward clinical research. Highlights will include genomics, non-invasive diagnosis tools, and integrated data collection are uncovering new discoveries, promoting personalized medicine and new care protocols.

Returning crowd favorites include 3 CEOs … (and a President!), The Unmentionables hosted by Alexandra Drane, The Frontier of Health 2.0 hosted by David Ewing Duncan, and Launch! with ten brand new companies unveiling their products for the very first time! Many more sessions and panels can be found on the Health 2.0 online agenda.Continue reading…

Massively Open Online Medicine

May 6, 2013• 6

By David Harlow

The new darling of the online educational community is Massively Open Online Courses (MOOCs). The example which figures most prominently in the popular imagination is the Khan Academy, though its founder says otherwise, noting that MOOCs are merely online transplantations of traditional courses, while Khan Academy offers something different.

Others would take issue with his conclusion, or characterization. A “connectivist” MOOCis based on four principles:

Aggregation. The whole point of a connectivist MOOC is to provide a starting point for a massive amount of content to be produced in different places online, which is later aggregated as a newsletter or a web page accessible to participants on a regular basis. This is in contrast to traditional courses, where the content is prepared ahead of time.
Remixing, that is, associating materials created within the course with each other and with materials elsewhere.
Re-purposing of aggregated and remixed materials to suit the goals of each participant.
Feeding forward, sharing of re-purposed ideas and content with other participants and the rest of the world.

Sounds great, but is it working? Can it work? A piece in the current issue of The Washington Monthly took a look and concluded:

Given the current 90 percent dropout rate in most MOOCs, an 8-point gap in completion rates between traditional and online courses offered by community colleges, the 6.5 percent graduation rate even at the respected Western Governors University, and the ambiguity of many other higher education reform ideas, there’s good reason to think that an unbound future might not be so great.

The best American innovations in education were the Land-Grant College Act of 1862, which helped create a system of public universities, and the GI Bill of 1944, which ensured that an entire generation had the money to attend college. This widespread access to the college experience enabled people from working-class backgrounds to advance en masse into professional jobs that required reasoning and logic and extensive knowledge of the world. The question is whether or not we will continue this trend or simply give up and say that a few online classes and specialized training are good enough for the majority of Americans.

In other words: Democratization of higher education – good; MOOCs – not so much.

Why is this relevant to you, gentle reader?

The question is whether the promise of MOOCs, or their inability to deliver, will characterize MOOM — Eric Topol’s neologism, “Massively Open Online Medicine,” used in his HIMSS 2013 keynote.

In health care, a perfect implementation of big data and data analytics, combined with open access for clinicians and patients, would yield a success in MOOM along the lines of a connectivist MOOC.

The Ghost of Steve Jobs and Your Bottom Line

Mar 19, 2013• 12

By JOE FLOWER

The progeny of the iPhone and the iPad will change the shape of your institution — and your balance sheet.

One of the more striking images, to me, out of the online spew in the last few months was from the inauguration. It was a wide view of an inaugural ball. There was the president waltzing with the first lady, and a crowd of several hundred watching them. What was striking about that image was that the several hundred people held several hundred small glowing rectangles in their hands. Practically every member of the crowd was carrying a smartphone and was photographing or videotaping the moment.

The scene was commonplace in its moment, remarkable only in the perspective of history — but such a short history. We could not have imagined so many people carrying smartphones at Obama’s first inaugural only four years ago. Four years before that, we could not have imagined any. The iPhone had not been invented.

There had been attempts at smartphones before the iPhone, and devices like tablets before the iPad. But the rampant success of iOS devices did far more than establish two profitable niche. It changed our relationship with the world.

Through a Scanner Darkly: Three Health Care Trends for 2013

Dec 25, 2012• 17

By DAVID SHAYWITZ, MD

As we anticipate a new year characterized by unprecedented interest in healthcare innovation, pay particular attention to the following three emerging tensions in the space.

Tension 1: Preventive Health vs Excessive Medicalization

A core tenet of medicine is that it’s better to prevent a disease (or at least catch it early) than to treat it after it has firmly taken hold. This is the rationale for both our interest in screening exams (such as mammography) as well as the focus on risk factor reduction (e.g. treating high blood pressure and high cholesterol to prevent heart attacks).

The problem, however, is that intervention itself carries a risk, which is sometimes well-characterized (e.g. in the case of a low-dose aspirin for some patients with a history of heart disease) but more often incompletely understood.

As both Eric Topol and Nassim Taleb have argued, there’s a powerful tendency to underestimate the risk associated with interventions. Topol, for example, has highlighted the potential risk of using statins to treat patients who have never had heart disease (i.e. primary prevention), a danger he worries may exceed the “relatively small benefit that can be derived.” (Other cardiologists disagree – see this piece by colleague Matt Herper).

In his new book Antifragile, Taleb focuses extensively on iatrogenics, arguing “we should not take risks with near-healthy people” though he adds “we should take a lot, a lot more, with those deemed in danger.”

Both Topol and Taleb are right that we tend to underestimate iatrogenicity in general, and often fail to factor in the small but real possibility of potential harm.

At the same time, I also worry about external experts deciding categorically what sort of risk is or isn’t “worth it” for an individual patient – a particular problem in oncology, where it now seems fashionable to declare the possibility of a few more months of life a marginal or insignificant benefit.

Even less dramatically, a treatment benefit that some might view as trivial (for hemorrhoids, say) might be life-altering for others. For these sufferers, a theoretical risk that some (like Taleb) find prohibitive might be worth the likelihood of symptom relief. Ideally, this decision would ultimately belong to patients, not experts asserting to act on patients’ behalf.

First, Do Net Harm?

Oct 12, 2012• 14

By Eric Topol, MD

Recently, the US Preventative Services Task Force reiterated its recommendation that women not undergo routine screening for ovarian cancer. This was remarkable, not simply because it was a recommendation against screening, but because the task force was making the recommendation again, and this time even stronger.

The motivation for the recommendation was simple: a review of years’ worth of data indicates that most women are more likely to suffer harm because of false alarms than they are to benefit from early detection. These screenings are a hallmark of population medicine—an archetypal form of medicine that does not attempt to distinguish one individual from another. Moving beyond the ritualistic screening procedures could help reduce the toll of at least $765 billion of wasted health care costs per year.

We already know the common changes in the DNA sequence that identify people who have higher risk of developing ovarian, breast or prostate cancer and most other types of cancer. Consumers can now readily obtain this information via personal genomic companies like 23andMe or Pathway Genomics. But we need to do much more DNA sequencing to find the less common yet even more important variations—those which carry the highest risk of a particular cancer. Such research would be easy to accomplish if it were given top priority and it would likely lead to precision screening. Only a small fraction of individuals would need to have any medical screening. What’s more, it will protect hundreds of thousands of Americans from being unnecessarily harmed each year.

Medicine Unplugged

Jun 21, 2012• 15

By Eric Topol, MD

Just as the little mobile wireless devices radically transformed our day-to-day lives, so will such devices have a seismic impact on the future of health care. It’s already taking off at a pace that parallels the explosion of another unanticipated digital force — social networks.

Take your electrocardiogram on your smartphone and send it to your doctor. Or to pre-empt the need for a consult, opt for the computer-read version with a rapid text response. Having trouble with your vision? Get the $2 add-on to your smartphone and get your eyes refracted with a text to get your new eyeglasses or contact lenses made. Have a suspicious skin lesion that might be cancer? Just take a picture with your smartphone and you can get a quick text back in minutes with a determination of whether you need to get a biopsy or not. Does your child have an ear infection? Just get the scope attachment to your smartphone and get a 10x magnified high-resolution view of your child’s eardrums and send them for automatic detection of whether antibiotics will be needed. Worried about glaucoma? You can get the contact lens with an embedded chip that continuously measures eye pressure and transmits the data to your phone. These are just a few examples of the innovative smartphone software and hardware — apps and “adds” technology — that have been developed and will soon be available for broad use.

Eric Topol: Too Clever by Three-Quarters

Feb 22, 2012• 4

By Matthew Holt

Eric Topol was once a lowly (well not that lowly) cardiology professor at the University of Michigan, but he’s now without question the leading renaissance man in health care technology. Virtually every week sees him on some big stage disgnosing his own heart murmur with an iPhone app or showing off how his sleep brain waves and his genome interact or don’t.

His new book, The Creative Destruction of Medicine is a tour de force romp through basically every type of cool new medical technology. He covers the Cloud/Web/Wireless/Sensor phenomenon from both a social, transactional and diagnostic point of view–leaning heavily on his connection to the West Wireless Health Institute which he helped persuade Gary & Mary West to fund. He’s the creator of a new medical school program at Scripps focusing on the genomics and proteomics revolution, and the book covers in great detail the evolution of the human genome project and its impact on disease discovery (coming eventually) and matching patients to the right drug (available more or less now). Finally he was of course the head of Cardiology at the Cleveland Clinic where he not only was heavily involved in the testing of tPA (the drug that built Genetech) but also in unveiling the problems with Vioxx not limited to the drug itself but also concerning Merck’s behavior at the time. (Remember Dodgeball?)