Tag: Michael Millenson

Epic’s “Emmie” Chatbot Enhances the Patient Voice – For Their Real Customers

Sep 8, 2025

By MICHAEL MILLENSON

A Rock Health write-up of this year’s Epic Users Group Meeting captured the artificial intelligence vibe with a play on the names of three new AI chatbots rolled out by the country’s dominant electronic health record firm. “Epic Goes APE (Art, Penny and Emmie),” read the headline, using the first letters of the names of chatbots designed for, respectively, clinicians, revenue cycle managers and patients.

Emmie does positive things for patients – more on that in moment – but at its core the chatbot is a B2B play, designed to address the needs of the hospitals, medical groups and others whose fees have built the privately held EHR firm into an estimated $5 billion business.

Emmie is not an agent of patient autonomy. Its purpose is to help Epic customers (health systems and physician practices) provide more and better services to their customer, the patient, as long as that patient remains a customer.

That context is important. Yes, in a way it’s #PatientsUseAI, but that use is analogous to the AI algorithms deployed by Netflix. While you may marvel at their power of personalization, they’re never going to tell you that the best movie for your particular interest is harbored over at Hulu and, by the way, even if you’re watching tons of programs with medical themes, you’re still a couch potato.

I wasn’t present at the gathering at Epic’s Verona, Wisconsin headquarters, but news accounts and LinkedIn postings suggest that, unsurprisingly for this type of meeting, there was more drama than details. Much of what was unveiled and hinted at – the company said it’s working on more than 200 AI applications – will be rolled out over the course of 2026 and beyond.

Here’s Epic’s introduction of Emmie and Art from its LinkedIn account:

Informed by their chart and connected devices, Emmie is designed to support patients between visits. Whether it’s explaining test results in easy-to-understand terms, suggesting next steps, or guiding patients through open-ended conversations about their health, Emmie makes it easier for patients to stay on top of their health and walk into the exam room with a clear picture.

On the clinician side, Art is gathering data from Emmie to get the doctor the information they need before the visit even begins. Art is designed to reduce administrative burden, help doctors better understand their patients, and offer context-informed insights. This can take the form of generating pre-visit summaries, taking real-time notes, and even taking actions like placing orders or verifying prior authorization requirements.

That Rock Health analysis suggested that the real significance “may be less the function and more the channel,” since consumers are far more willing to share health data with their provider – in this case through Epic’s MyChart – than with a tech company (such as an AI vendor). “By capturing patient questions, decisions, and symptom-checking,” Rock Health noted, “Epic gains visibility into information consumers might hesitate to share with a generalist tech company. The EHR giant has already signaled that this data will feed back into [its] tools.”

Or as Epic did not say, “We empower our customers. We empower patients. We empower ourselves.” Good intentions alone do not get your product into use by more than half of all acute-care hospital beds in America, according to a KLAS estimate of market share.

At Healthcare IT Today, veteran tech journalist John Lynn sniffed out the actual schedule for all Emmie’s promised pro-patient wonders. According to Lynn (presumably from Epic itself),

proactive outreach and images is coming in March, 2026
active engagement in November, 2026
future screenings arrive sometime in 2026 (no month given).

Bill payment, scheduling abilities using SMS (texting) and a voice agent are all “coming in the future.”

As I commented on the Epic LinkedIn post, “How about patient-reported outcome measures [e.g. Proteus Consortium’s], whether from an app linked to Epic (i.e., like Twistle by Health Catalyst or others) or the patient’s own wearables?” I tagged Seth Hain, Epic’s senior vice president of research and development, who played a prominent role at the meeting, but got no reply. (To be fair, maybe he took some vacation time after an intense few weeks.)

In a recent STAT First Opinion that took up the topic of autonomy, I asserted that true informed consent means physicians should be obligated to inform patients what Epic’s Cosmos system says about the likely outcomes of treatment for individuals with their clinical profile. Those predictions come from a database drawing on a mind-boggling 15.7 billion patient encounters. But patients should be able to access that information about different hospitals’ results on their own.

At the Users Group Meeting, Epic announced a further refinement of Cosmos, with founder and chief executive officer Judy Faulkner proudly announcing that the company will be able to “predict the future” for patients. (For a deep dive into Cosmos, I recommend the posts of veteran medical informatics expert Mark Braustein.)

Faulkner did what any smart businessperson would do. She spoke about how her company’s product would enable an important segment of customers, clinicians and health systems, to provide better care. What those customers actually do (or don’t do) for their “customer,” the patient, with the Epic software? Evidently “Not my job.”

Michael L. Millenson is president of Health Quality Advisors & a regular THCB Contributor. This first appeared in the “Patients use AI” Substack

As Shared Decision-Making Ails, AI May Save This Human Interaction

Aug 5, 2025

By MICHAEL MILLENSON

Shared decision-making between doctors and patients may be “the pinnacle of patient-centered care,” but three new medical journal articles suggest it’s encountering more problems than peaks. Yet counterintuitively, it may be artificial intelligence that rescues this intimately human interaction.

“Shared decision-making is at a crossroads,” declares a Perspective in the Journal of General Internal Medicine, “Saving Shared Decision-Making.” Unfortunately, its more-research-and-education recommendations for “advancing the science of SDM implementation,” seem more crossing guard than crisis management.

Even a cursory historical perspective shows that SDM is suffering from a failure to flourish. Back in 1982, a report by a presidential commission on ethics in medicine declared SDM “the appropriate ideal for patient-professional relationships” and called on doctors “to respect and enhance their patients’ capacities for wise exercise of their autonomy.”

Yet 43 years later, the Perspective authors – 18 members of the Agency for Healthcare Research and Quality Shared Decision-Making Learning Community – acknowledged that while some doctors respectfully ask patients, “What do you think you would like to do, given these options?” many others still believe that, “Let’s do this option, sound OK?” is a shared decision process.

That attitude reminded me of a tongue-in-cheek comment by comedian Stephen Colbert. “See what we can accomplish when we work together by you doing what I say?” he told a 2015 Colbert Nation audience. “It’s called a partnership.”

Cancer Communication Curtailed

In cancer, where patient-doctor interactions have the highest stakes, shared decision-making was named one of the central components of quality care in a 1999 report, Ensuring Quality Cancer Care, by the Institute of Medicine (now the National Academy of Medicine). Nonetheless, a review of SDM among cancer patients in the journal Psycho-Oncology found that for physicians, “making decisions and taking responsibility for the decisions remain an important part of the physicians’ professional identity.” The fear of losing this identity, the authors wrote, “tends to hinder the patient involvement and implementation of SDM.”

Not surprisingly, cancer patients who want to speak up feel as if they won’t be listened to or can’t really refuse whatever their oncologist considers clinically “optimal.” And, it turns out, oncologists are actually less open to SDM if a patient does speak up and resists the recommendations they feel are in the patient’s best interest.

“Health Care” vs. “Healthcare” Signals Change Greater Than Grammar

Apr 14, 2025

By MICHAEL MILLENSON

“The New Yorker House Style Joins The Internet Age” announced the magazine’s daily newsletter under the byline of Andrew Boynton, whose appropriately old-fashioned title was “Head of Copy.” Among the alterations Boynton acknowledged readers might feel “long overdue,” were “Internet” becoming “internet,” “Web site” consolidating to “website” and “cell phone” becoming “cellphone.” Other quirky spellings (teen-ager, per cent, etc.) were deliberately retained.

But what about “health care” vs. “healthcare”?

A New York Times interview described Boynton as “tight-lipped” about the style changes, which came as the publication celebrated its 100^th anniversary year. When I nonetheless sought to discover whether a descriptor central to a massive chunk of the U.S. economy was more like a cellphone or a “teen-ager,” the magazine graciously responded.

“’Health care’ is our style,” a spokesperson wrote me in an email. “There has not been any discussion of diverging from this.”

Not even a discussion? This was shocking news! But as I dug deeper, it seemed to me that the choice of the one-word versus two-word term often sent an underlying signal about the evolution of not just language, but of health care as both a profession and an industry.

Debating Evolution

Back in 2012, after I dived into the “health care vs. healthcare” debate for The Health Care Blog, my friend and colleague, the determinedly data-driven David Muhlestein, PhD, JD, accused me of ignoring language evolution by insisting on the “two words” usage. He eventually presented me with Google searches showing that the ratio of uses of the one-word to the two-word term ineluctably indicated “health care” was going the way of “Web site.”

When I solicited a 2025 update, Muhlestein obliged with a Google trends graph tracing relative usage since 2004.

Apart from a brief time that “health care” was more prevalent as discussion of the Affordable Care Act dominated the news, the preference for “healthcare” has steadily strengthened. “As of now, people use the one-word version more than twice as often as two words,” Muhlestein wrote in an email.

He added, “You can’t predict how language will evolve, you just have to go with what it is, and for the U.S., healthcare is definitely going to one word.”

Perhaps. But even a cursory qualitative analysis suggests a more nuanced picture than volume alone provides. After poking into the preferences of publications, corporations, the U.S. government and others, I decided that a 2022 April Fool’s column in Health Affairs actually provided a rough guide to understanding many usage decisions.

It’s Money That Changes Everything (Or Doesn’t) For Surgeons

Jan 30, 2025

By MICHAEL MILLENSON

“Money changes everything,” Cyndi Lauper famously sang about love to a pulsating rock ‘n’ roll beat. So, too, when it comes to financial incentives for surgeons, two new studies suggest, although “How much money?” and “What do I have to do?” are the keys to unlocking monetary motivation.

The first study, a JAMA research letter, examined the impact of a new Medicare billing code for abdominal hernia repair that paid surgeons more if the hernia measured at least 3 centimeters in size. Previously, “size was not linked to hernia reimbursement,” noted University of Michigan researchers.

Surprise! The percentage of patients said to have smaller, lower-payment hernias dropped from 60% to 49% in just one year. Were “small hernia” patients being denied care? Nope. Were surgeons perhaps more precise in measuring hernia size? Maybe. Or possibly, wrote the researchers in careful academic language, “the coding change may have induced surgeons to overestimate hernia size.” Ambiguous tasks, they added, “can be conducive to perceptive [cq] bias and potentially even dishonest behavior, perhaps more so with financial incentives at play.”

This being an academic publication, two footnotes informed us that dangling money in front of our eyes can cause people to “see what you want to see” and come up with an “elastic justification” for truth.

If a simple coding change can apparently boost the number of large-hernia patients by 18% in just one year, what about a payment incentive meant to induce more urologists to follow the medical evidence on low-risk prostate cancer and adopt “active surveillance” (formerly known as “watchful waiting”), rather putting patients through a painful and expensive regimen of biopsies and surgery?

A second study, also in Michigan, involved commercial and Medicare-age members of the state’s Blue Cross and Blue Shield plan. However, after three years and more than 15,000 patients, “the payment incentive was not associated with increased surveillance use among patients with low-risk disease,” researchers concluded in a JAMA Network Open article.

My Totally Wrong, Expert Predictions for Health Care 2025

Jan 5, 2025

By MICHAEL MILLENSON

January

In a blistering commentary, the American Medical Association’s flagship journal, JAMA, condemns the corrosive effect on patient care of the profit-seeking practices of health insurers. Separately, the organization announces that it’s selling the 13 journals in its JAMA Network to a private equity firm for $375 million “in order to enhance our mission of promoting the betterment of public health.”

February

Quickly following up on a campaign pledge to slash the federal budget, the Trump administration announces a radical consolidation of various entities at the Department of Health and Human Services. The new organization will be known as the Agency and Bureau for Children, Drugs, Explosives, Firearms, Families and Food (ABCDEFFF). Reflecting the new president’s strong personal preferences, “alcohol” will no longer be permitted in any agency name.

March

Bipartisan legislation demanding transparency from Pharmacy Benefit Managers dies in committee after industry executives explain that secret rebates to PBMs are like secret political action committee contributions to politicians: they allow you to loudly proclaim you’re an “advocate” for those supposedly paying you while actually serving the interests of those who are really paying you.

April

Pfizer announces that its once-a-day pill version of the wildly successful GLP-1 agonist weight loss drugs will shortly be submitted for government approval, and also that the company is moving its headquarters from New York to Louisiana, a state with a 40 percent obesity rate. Coincidentally, Louisiana is also the home state of Republican senators Cassidy and Kennedy, senior members of the Senate committees overseeing health care and all federal appropriations.

May

The new private equity owners of the JAMA Network say that all staff except one editor at each journal will be replaced by ChatGPT. A source at the private equity firm tells the Wall Street Journal that OpenAI won out over Gemini “because our CEO is a Leo” and over Claude “because nobody likes the French.”

June

Controversial right-wing firebrand Rep. Marjorie Taylor Greene, long the subject of rumors that she’s had cosmetic surgery, is diagnosed with a serious infection after an unspecified procedure. The House quickly schedules its first hearing on medical error in over two decades, but then cancels when the American Hospital Association points out the official term for what the Georgia Republican contracted was a “healthcare-associated infection,” so it’s entirely possible she accidentally brought the infection with her to the pristine hospital. Meanwhile, with House leadership telling Members they were free to vote their conscience, a resolution to send Greene a “Get Well” card passes unanimously after deletion of the word, “Soon.”

July

Following through on years of promises to reveal a “really great” replacement for the Affordable Care Act, President Trump on July 4 announces the “100-100-100” Make America Healthy Again plan. In keeping with the GOP’s advocacy for “skinny” plans with low premiums that encourage “consumers” to “comparison shop,” the plan will cover 100 percent of any medical bill for up to $100 a day for a premium of just $100 a month. Separately, Elon Musk tells a meeting of health insurance executives the plan can also replace both Medicare and Medicaid, enabling the federal government to cut spending by almost as much as the market capitalization of Tesla.

August

Before Congress recesses, a coalition of progressive organizations issues a press release declaring that all basic health services, whether provided by government agencies or the private sector, should be “available to the entire population according to its needs.” Shortly afterwards, the coalition is forced to make an embarrassing retraction after ChatGPT alerts the lone editor of JAMA that the coalition accidentally re-released a section of the report of the Committee on the Costs of Medical Care, formed in 1927.

September

The Business Roundtable says its members are committed to improving the quality of health care for all employees because “quality health care is good business.” An 85-year-old freelancer for The New York Times notes that this was the exact title of a September, 1997 policy paper by a Roundtable task force in which an executive for Sears, which at the time operated over 3,500 stores, declares, “We believe that quality health care is lower-cost health care.” Sears currently has about a dozen stores.

October

Medicare Advantage plans step up their advertising expenditures after public opinion polls show that nobody anymore believes the portrayal of happy and healthy seniors playing pickleball instead of writing tear-soaked letters pleading for approval of hip surgery. The trade associations for hospitals, drug and device companies and PBMs call on Congress to provide greater oversight of greedy insurers. The editor of JAMA resigns after ChatGPT writes an editorial extolling the merits of MA plans run by for-profit companies.

November

The National Rural Health Association says that in the spirit of the Thanksgiving holiday, its members will accept live turkeys in partial payment of the medical debts that now affect 99.99 percent of all Americans after passage of the administration’s “100-100-100” Make America Healthy Again plan. A KFF survey explains that the number is not 100 percent because Congress retained conventional health insurance for itself and top federal officials and because America’s billionaires had opted for self-pay.

December

A Washington Post editorial declares, “The bottom line is that if we want to contain spending, we will have to make critical choices about how care is delivered, to whom, and under what conditions.” Different chatbots differ on where that quote originally came from, but agree that if any humans believe the American public is ready to make critical choices, they’re hallucinating.

Michael L. Millenson is president of Health Quality Advisors & a regular THCB Contributor

THCB Gang Episode 148, Monday December 16

Dec 16, 2024

Joining Matthew Holt on #THCBGang on Monday December 16 at 1pm PST 4pm EST are patient safety expert Michael Millenson, physician, entrepreneur and technologist Shantanu Nundy; and Digital Health and Emerging Med-Tech Practice Co-Founder at Marsh & McLennan, Beracah Stortvedt.

You can see the video below live (and later archived) & if you’d rather listen than watch, the audio is preserved as a weekly podcast available on our iTunes & Spotify channels.

THCB Gang Episode 147, Thursday December 5

Dec 5, 2024

Joining Matthew Holt (@boltyboy) on #THCBGang on Thursday December 5 at 1pm PST 4pm EST are patient safety expert Michael Millenson, patient advocate & entrepreneur Robin Farmanfarmaian; futurist Jeff Goldsmith; and employer & care consultant Brian Klepper.

You can see the video below live (and later archived) & if you’d rather listen than watch, the audio is preserved as a weekly podcast available on our iTunes & Spotify channels.

Medicare’s Hidden Information Hurts People & Policy

Nov 12, 2024

By MICHAEL MILLENSON

Open enrollment season for Medicare, which began Oct. 15 and ends Dec. 7, triggers a deluge of information about various options. Since I’m a health care consultant and researcher as well as a Medicare beneficiary, I’ve looked critically at what we’re told and what we’re not. Unfortunately, information crucial both for the individual and for the broader policy goal of moving toward a “value-based” care system is often difficult to find or not available at all.

The most glaring example involves Medicare Advantage, the increasingly popular insurer-run plans that are an alternative to traditional fee-for-service Medicare. Plans receive a quality grade from one to five stars from the Centers for Medicare & Medicaid Services. Those grades are designed to incentivize providing the highest quality care for the money — the very definition of “value.” A high grade triggers both a boost in payment from Medicare and a boost in enrollment. Not surprisingly, almost three-quarters of people chose a plan with a 4-, 4.5- or 5-star rating, according to CMS.

Those ratings, however, should come with a large asterisk attached. It’s not just that the methodology can be controversial, particularly when a lower grade is meted out. It’s that the star ratings aren’t anchored in geography, as one would naturally expect; i.e., the rating is for the plan offered in my area. What is colloquially called a “five-star plan” is actually a plan that’s part of a five-star Medicare contract — and those two typically are not the same thing.

For instance, one large insurer contract that I tracked included at least 17 plans scattered across the country. It defies common sense to believe that care quality is identical among plans in, say, Rhode Island, Mississippi, Illinois, Colorado, and California just because they all share the same government contract number.

If you’re wondering who benefits from this not-very-transparent transparency, some insurers have been known to improve the rating of a low-performing plan with a small number of members by merging it into a contract with more members and a higher rating.

In 2024, nearly 33 million people, or 54% of Medicare beneficiaries, were enrolled in an MA plan, according to KFF (formerly the Kaiser Family Foundation). KFF expects that number to increase to nearly 36 million in 2025. It’s a long-accepted truism that “All health care is local.” Medicare beneficiaries deserve local plan information.

THCB Gang Episode 143, Friday November 8

Nov 8, 2024

Joining Matthew Holt (@boltyboy) on #THCBGang on Friday November 8 are THCB regular writer and ponderer of odd juxtapositions Kim Bellard (@kimbbellard); Principal of Worksite Health Advisors Brian Klepper (@bklepper1); patient safety expert and all around wit Michael Millenson (@mlmillenson); and digital health investment banker Steven Wardell (@StevenWardell). There may well be a discussion about an election.

You can see the video below live (and later archived) & if you’d rather listen than watch, the audio is preserved as a weekly podcast available on our iTunes & Spotify channels.

Why We Still Kill Patients (And What We Need to Stop Doing It)

Sep 11, 2024

By MICHAEL MILLENSON

This article is adapted from a talk given Sept. 7 at the 11^th Annual World Patient Safety, Science & Technology Summit in Irvine, California, sponsored by the Patient Safety Movement Foundation. World Patient Safety Day is Sept. 17, with a series of events in Washington, D.C. from Sept. 15-17 sponsored by Patients for Patient Safety (US). An agenda and registration, which is free, can be found here.

Since I started researching and writing about patient safety, one question has continually haunted me: given the grievous toll of death and injury from preventable medical harm that has been documented in the medical literature for at least 50 years, why have so many good and caring people – friends, family, colleagues – done so little to stop it?

To frame that question with brutal candor: Why do we still kill patients? And how do we change that? The answer, I believe, lies in addressing three key factors: Invisibility, inertia and income.”

When it comes to invisibility, we’ve all heard innumerable times the analogy with airline safety; i.e., plane crashes occur in public view, but the toll taken by medical error occurs in private. That’s true and important, but there are other factors that promote invisibility that we in the patient safety movement need to address.

For instance, while I’m not a physician, I can say with certainty that every patient harmed in the hospital had a diagnosis (right or wrong), and often more than one. Yet disease groups such as the American Heart Association and American Cancer Society have been uninvolved in efforts to eliminate the preventable harm that’s afflicting their presumed constituents.

Why have we let these influential groups sit on the sidelines rather than make them integral partners in raising public and policy visibility? For instance, there are a number of Congressional caucuses – bipartisan groups of legislators – focusing on cancer. While much attention is paid to the Biden administration’s cancer moonshot, what about the safety of cancer patients treated today, while we wait for an elusive cure?

In a similar vein about missed opportunities for visibility, the stories told by patient advocates about the harm a loved one has suffered are always powerful. However, the specific hospital where the harm took place is typically not mentioned, perhaps for legal reasons, perhaps because it’s become a habit. The effect, however, is to dilute the visibility of the danger. The public is not confronted with the uncomfortable reality that my reputable hospital and doctor in a nice, middle-class area could cause me the same awful harm.

Finally, one time-tested way to hide a problem is to use obscure language to describe it. Back in 1978, RAND Corporation published a paper provocatively entitled, “Iatrogenesis: Just What the Doctor Ordered.” It concluded: “In terms of volume alone, we are awash in iatrogenesis.”

That would have been a compelling soundbite decades before the 1999 To Err is Human report if everyone in America studied ancient Greek. “Iatrogenesis” is a Greek term meaning “the production of disease by the manner, diagnosis or treatment of a physician.” In short, patient harm is “what the doctor ordered.” Although there was plain English in the paper, the technical focus allowed the stunning prevalence of patient harm to remain publicly invisible.

Of course, today we don’t need to use a foreign language to hide unpleasantness. We can use jargon and euphemism. We have “healthcare-acquired conditions” and “healthcare-associated infections.” At least the Greek term acknowledged causality and responsibility.

The invisibility of the scope and causes of patient harm leads inevitably to inertia and complacency.