Interested in the intersection of healthcare, technology and business? You've come to the right place. The Health Care Blog is the leading online forum covering the business of healthcare and the new ideas that are changing the health care industry.
Electronic health records (EHRs) are a polarizing issue in health reform. In their current form, they are frustrating to many physicians and have failed to support cost improvements. The current round of federal intervention is proposed rulemaking pursuant to the 21st Century Cures Act calls for penalties for “information blocking” and for technology that physicians and patients could use “without special effort.”
The proposed rules are over one thousand pages of technical jargon that aims to govern how one machine communicates with another when the content of the communication is personal and very valuable information about an individual. Healthcare is a challenging and unique industry when it comes to interoperability. Hospitals spend lavishly on EHRs and pursue information blocking as a means to manipulate the physicians and patients who might otherwise bypass the hospital on the way to health reform. The result is a broken market where physicians and patients directly control trillions of dollars in spending but have virtually zero market power over the technology that hospitals and payers operate as information brokers.
What follows below are comments by Patient Privacy Rights on the proposed rule. The common thread of our comments is the need to treat patients and physicians, not the data brokers, as the real stakeholders.
Comments to the ONC Rule
Overview: 21st Century health care innovation, policy, and practice is increasingly dependent on personal information. This is obvious with respect to machine learning and risk adjustment, but personal information is now central to the competitive strategy for most of the health care economy, clinical as well as research. ONC’s drafting of this rule reflects the importance of competition to innovation and cost containment.
The Proposed Rule skillfully addresses the pro-competitive essentials but it leaves too much open to interpretation and delay by very wealthy and well-organized incumbents. The Patient Privacy Rights comments below endorse the structure and details of the Rule while pointing out ways to ensure that access to competitive services by clinicians on behalf of our patients must be “without special effort” on the part of either the clinician or the patient, ASAP.
We state clearly and emphatically that the Rule should be largely left intact in its spirit and in most of its details
Summary of Priority Goal: Clarify the scope and process of patient-directed interoperability
The common thread through almost all of PPR’s comments is to support and encourage patient-directed sharing via the mandated API as the foundation for meeting the pro-competitive goals of the 21st Century Cures Act “without special effort”. Patient-directed exchange inherently solves very difficult problems in patient matching, consent, and integration of sensitive information that cannot be shared under the HIPAA rules. Patient-directed exchange helps address the need for a patient-centered longitudinal patient record and provides a critical relief valve for both physicians who simply need “the data to follow the patient”. Patient-directed exchange also informs how we will implement TEFCA and various registries that can provide essential public health and health care innovation benefits.
Early versions of patient-directed sharing via API can make a visible and welcome impact for physicians and patients within 6 months of adoption of the final Regulation. That technical capability is already voluntarily enabled by some API Technology Suppliers and just needs to be mandated for adoption by API Data Providers. The timelines for standards development are long but when standards already exist for Dynamic Client Registration, Refresh Tokens, and User Managed Access, the adoption of these standards can begin immediately by new competitors and early adoption by CMS, VA, and other customers in the Federal Health Architecture can drive a competitive strategy.
Summary of Other Considerations:
21st Century health care innovation, policy, and practice is increasingly dependent on personal information and the rate of progress is increasingly limited by privacy and human dignity in how personal data is used. This is obvious with respect to machine learning and risk adjustment, but personal information is now central to the competitive strategy for most of the health care economy. Privacy now dominates the rate at which technology and policy can progress.
The cost and burden of interoperability at scale are both reduced if we approach the problems from the patient and clinician perspective rather than the institutional:
Patient matching is a non-issue when information is shared with patient consent and transparency. Modern-day automated bank transaction APIs are a good example. Once set-up by the customer, money can flow automatically and on-demand without further customer action. Email and text messages are used to notify of transactions. All transactions are logged and accessible to the customer online. The costs are lower with the API and transactions process faster.
HIPAA is a floor but Not Sufficient because it doesn’t cover the data originating in behavioral health practices on the sensitive end and data originating in consumer mobile devices and wearables which can also be quite sensitive. To avoid the limitations of HIPAA, we urge CMS to design interoperability on the basis of patient consent with full transparency to the patient. That also means patient notice and on-line accessible logs for all transactions including treatment, payment, and operations. HIPAA’s exclusion of T/P/O transparency is not justified with modern Open APIs and adds unacceptable security risks as we expand the scope and scale of interoperability.
Designation of Providers should be without special effort for both the patient and the providers using the Open API. That means accelerating and enforcing the need for providers to include voluntary digital contact addresses in their NPI and Physician Compare files. Patients can automatically link the digital contact info to their consent. Providers can use their digital credentials to automatically register their API client without special effort. It is easier and less burdensome to drive interoperability on the basis of the HIPAA patient right to designate recipients.
Competition for Authorization Services would be the ultimate cost and burden reduction for large-scale interoperability. The Open API, including FHIR, can be configured to allow the patient to specify the authorization server to the API Data Provider. (See User Managed Access standard in 2019 ISA). Current FHIR API practice forces patients to use a separate authorization server for each API Data Provider. Managing consent at a dozen or more patient portals requires undue effort on the part of patients. Allowing the patient to specify the authorization server would give patients market power to choose their consent service competitively and provide a competitive basis for health information network providers that want to serve the patient.
The draft rules for interoperability, CMS, ONC, TEFCA, USCDI are over a thousand pages. Most of the complexity stems from a design that avoids direct patient direction and transparency the way we expect banking and other automated services. This approach fragments the patient and physician experience and poses privacy and security risks that may never be solved. On the other hand, an interoperability design based on patient-designated sharing with clinicians that voluntarily post their digital contact info (personal, group, or institution) works across the full range of patient data (behavioral, HIPAA, patient-generated) and provides patients and family caregivers the transparency and accountability over health services that we need. Allowing patients to specify their authorization server further simplifies things by enabling competition for the authorization service – a digital concierge – that would give market power to individuals and deliver the pro-competitive benefits the Rule seeks.
Slide into Health in 2 Point 00 (or rather, Health 2.0 HIMSS Europe) with Jess and I today! On Episode 84, Jess asks me about the big news that CVS has now made it possible for employees to get reimbursed for Big Health’s Sleepio, an insomnia digital therapeutic, and about Atrium Health’s $10 million investment in an affordable housing plan, addressing the social determinants of health. Hear some of my key takeaways from the conference so far, too. –Matthew Holt
Health in 2 Point 00, Episode 84 | Health 2.0 HIMSS Europe, Part 2 - YouTube
Far more attention has been devoted to the ways in which
industry consolidation has driven up health costs than to proposals on how to
remedy the situation. But the introduction of Medicare for All and Medicare for
More bills—however dim their short-term prospects are—has changed the terms of
the debate. It is time to think about how we can eliminate the market power of health
systems without causing harmful dislocations in health care and the economy.
Before we get to that, here are the main facts about
consolidation: As a handful of health insurers have become dominant in many
markets, health systems have done likewise in order to maintain or improve
their negotiating positions. That has proved to be an effective strategy in
many cases. Even dominant health plans cannot do without the largest hospital
systems in their areas, especially when they employ many of the local
According to a Kaufman Hall report, 90 hospital and health system deals were publicly announced in 2018. This was a decline from the 115 deals unveiled in 2017, but the average size in the revenue of sellers hit a high of $409 million.
The biggest provider mergers are staggering in scale. In February 2019, for example, Catholic Health Initiatives and Dignity Health formed a new organization called CommonSpirit Health, which has 142 hospitals, 150,000 employees and nearly $30 billion in revenues. The union of Chicago-based Advocate Health Care and Wisconsin’s Aurora Health Care in April 2018 created a giant with 27 hospitals and $11 billion in revenues. A month later, Atrium Health (formerly Carolinas Healthcare System) joined with Wake Forest Baptist Health to form a system with 49 hospitals and combined revenues of $7.5 billion.
A substantial body of research shows that health systems use their market power to raise prices. For example, a 2016 study found that hospital prices in California grew by an average of 76% per admission between 2004 and 2013. Prices at hospitals belonging to large, multi-hospital systems grew substantially more (113%) than prices paid to all other California hospitals (70%). By the end of the period, average prices per admission at hospitals in the largest systems exceeded prices at other hospitals by about $4,000, or 25%.
Another study found that there were over 1,400 hospital mergers in the U.S. from 1998 to 2015. At the end of this period, nearly half of hospital markets were highly concentrated, dominated by one or two large health systems. Hospital admissions in these areas cost $2,000 more, on average, than admissions elsewhere, the authors noted.
A recent RAND study found that that commercial insurers paid hospitals an average of 241% of Medicare rates in 2017. One interpretation of this data, the researchers said, “is that hospitals, especially ‘must-have’ hospitals, have used their negotiating leverage to extract unreasonable price concessions from health plans.”
Health systems have also integrated vertically with physicians as their practice purchases have accelerated in recent years. Between those acquisitions and the increasing percentage of residency graduates who go to work for hospitals, hospitals now employ 35% of physicians directly or indirectly, according to the American Medical Association. Another survey by the Physicians Advocacy Institute and Avelere Health indicates that 44% of physicians were employed by hospitals in 2018, compared to 25% in 2012.
While hospitals offer a variety of explanations for employing
physicians, the real reasons boil down to two: they want to lock up the doctors
who refer patients to their facilities, and they want to make sure their
competitors don’t lock up those same doctors.
Until recently, hospitals had another incentive to buy practices: Under Medicare payment rules, physicians who worked for hospitals were considered part of hospital outpatient departments (HOPDs), and their services could be billed at a higher rate (including facility fees) than private practice physicians could charge Medicare. To even the playing field, the Centers for Medicare and Medicaid Services (CMS) has begun to implement “site-neutral payments,” but the final outcome remains in doubt.
Meanwhile, hospitals use their market power to negotiate higher commercial payment rates for their employed doctors. A study of claims data from commercial insurers found substantial differences between the prices negotiated by employed groups and private practices across the country. In two-thirds of the areas included in the study, physician prices increased as the result of practice purchases by hospitals. Another study found that physician prices rise nearly 14% when a hospital acquires a physician group.
Most important to hospitals is the downstream revenue generated by employed physicians. In 2016, the average net revenue that each employed doctor generated for her hospital was $1.56 million, up 7.7% from $1.45 million in 2013, according to physician search firm Merritt Hawkins. In 2019, the same company conducted a survey showing that independent and employed physicians generated an average of $2.38 million each for their affiliated hospitals.
What the government can
What all of this shows is that much of the growth in health spending can be attributed to industry consolidation. However, the Federal Trade Commission (FTC) has tried to stop very few hospital mergers; and, as health economist Paul Ginsburg has pointed out, federal antitrust policy doesn’t directly address hospital acquisitions of physician practices. Even if the FTC were to suddenly take an interest in healthcare mergers and the courts were more disposed to rule against not-for-profit entities, consolidation has gone too far for antitrust regulators to have much effect. After all, the FTC is not going to break up corporations that include thousands of hospitals and hundreds of thousands of physicians.
The solution, therefore, lies elsewhere. My argument, in
brief, is that the Medicare for All and Medicare expansion proposals create a
new space for countering industry consolidation, should anything come of these
Obviously, if the U.S. adopted a single payer system, the government could set hospital fee-for- service payments or give each hospital a global budget. But if Medicare for All meant that hospitals would get paid at Medicare rates, both conservative and liberal experts say, many hospitals would be seriously damaged. The adverse impact would vary, depending on how much of a particular facility’s revenues came from private insurance. But in the aggregate, a Stanford University study estimated, the average hospital would see a net decline of 16% in revenue and a negative margin of 9%. As many as 1.5 million jobs could be lost as a result.
A Democratic Congress under a Democratic President is more likely to pass a bill like Medicare for America, which grew out of proposals from the Center for American Progress and Yale health policy expert Jacob Hacker. In brief, this legislation would achieve universal coverage by enrolling the uninsured, people who buy individual insurance and those on Medicare, Medicaid and CHIP. Large companies could continue to provide insurance to their workers, or they could let their employees enroll in Medicare and pay 8% of payroll to the Medicare Trust Fund. Employees could also opt out of their company’s insurance plan and enroll in Medicare on their own. Over time, it’s possible that this model would morph into Medicare for All.
During the transitional period, states could curtail health systems’ market power by adopting “all-payer” models similar to those in Maryland and West Virginia. Under Maryland’s law, every insurer, including Medicare, Medicaid, and private health plans, pays uniform hospital rates negotiated between the state and the hospitals. In Ginsburg’s view, it would be impractical for other states to replicate this model, which Maryland introduced 40 years ago, because commercial rates are now so much higher than Medicare and Medicaid rates. A more feasible approach, he said, would be to emulate West Virginia, which sets only commercial insurance payments to hospitals. But in either case, an all-payer system would eliminate the ability of dominant health systems to extract higher rates from private payers.
All-payer systems are not a panacea. Maryland’s hospitals, for instance, raised the volume of services to compensate for lower payment rates under the all-payer law. As a result, the state introduced global budgets for hospitals, fully implementing them in 2014 after a phase-in period. Hospital cost growth dropped in the ensuing years, but ambulatory and post-acute-care costs grew more rapidly than before. This is prima facie evidence that it’s impossible to restructure just one part of the healthcare system.
To prevent hospitals from using their market power to obtain
higher rates for their employed physicians, the government could simply
prohibit them from employing doctors. This would not only curtail spending
growth, but would also allow more physicians to form group practices and ACOs in
which they could be incentivized to pursue value-based care. The incentive of hospital-employed
doctors to emphasize value-based care will always be limited, because hospitals’
business model is based on filling beds, not emptying them.
The legal basis for mandatory divestment could be derived from state “corporate practice of medicine” laws that bar corporations from employing physicians. Found in many states, these measures were enacted to avoid conflicts of interest between physicians’ duty to provide the best care for their patients and their employers’ dictates. Most states with such laws allow hospitals to hire doctors, however, since they’re also in the business of medicine.
The sole exception is California. That state’s corporate
practice of medicine law prohibits any non-professional organization except for
a public hospital, a narcotics treatment program or a nonprofit medical research
firm from directly employing physicians. Unfortunately, the California
corporate practice of medicine law has not had the intended effect. Instead of
hiring doctors, private hospitals and health systems simply lease their
services from “foundations” that stand in for professional corporations.
But the states could enact stronger laws that prohibit
hospitals from directly or indirectly employing doctors. It’s unclear whether
most hospitals would be worse off economically if their medical staffs were
independent rather than employed. Considering the losses that hospitals incur
on their owned practices, some hospitals would benefit financially from
divesting them. The hospitals’ main concern would be to prevent competitors
from controlling their referring doctors. If no health system could employ
physicians, that wouldn’t be a problem.
States versus the
Considering the variability of states’ responses to the
Affordable Care Act, it’s not likely that all or most of them would enact
all-payer laws or corporate practice of medicine statutes that applied to
hospitals. Realistically, only the federal government could make these things
happen. Perhaps all-payer laws in some states—at first just for commercial
plans–could help pave the way for a single-payer system under the gradualist
scenario of Medicare for America. And when enough people joined Medicare, maybe
Congress could pass a national corporate practice of medicine law. These
changes might occur in a number of different ways. But, while individual states
could serve as laboratories to test and adjust these policies, ultimately the
federal government would have to implement them nationwide.
In the current environment, there is no political will to make such radical changes. Congress would have the impetus and the popular support to move in this direction only if the country were transitioning toward a single payer system. But I believe that that day is coming, and when it arrives, it would be far better to eliminate the market power of hospitals than to reduce their revenues to the point where many of them could no longer function properly or would be forced to close. The healthcare system needs to be restructured, not destroyed.
Ken Terry is a veteran healthcare journalist and the author of Rx for Health Care Reform (Vanderbilt University Press, 2007). This article is adapted from a forthcoming book.
There I was, my 10th-grade science fair. My mother made
sure I had a tie that fit properly and a shirt that was perfectly pressed. I stood among my peers
with our cardboard presentation displays highlighting what we did to make it to
this point. I was a little nervous but also extremely proud of myself and
excited to see the looks on the judge’s faces when they saw what my project was
of Enzymes on DNA”
Boom. Oh, I wasn’t doing something that many people had seen
already — I was working inside an NIH facility with a brilliant scientist
mentor/coach, to get this done. The memories of taking multiple modes of
transportation after school throughout the week for what seemed like forever
wore me down enough to make sure that I knew this was going to be worth it. And
then after the judges were introduced to all of our concepts and families
poured throughout the gymnasium to see what we all came up with — now was the
moment of truth.
Sweaty palms and teenage anxiety wouldn’t deter me. First place goes to….oh ok, yeah of
course, they deserved that. They worked really hard I’m sure. Second place goes to….oh wow, I didn’t make
second place? At least, I’ll get something. After a third place winner was
announced and the applause faded. I looked, stunned, over at my mother in the
audience whose face was covered in tears. I was ready for the night to be over.
Did I not wear the right tie? Did I seem
too confident? Not confident enough? The questions would consume me until
later that evening when my science teacher told me that the judges thought I cheated or didn’t actually do any
of the work.
He was very blunt about it. Prejudice was something that I
wasn’t extremely familiar with — maybe my mother did a great job of shielding
me from what she could. But that night, it was different. After my teacher told
my mother to get me out of the school and into a public one (my mother quickly
found out that private, faith-based schools weren’t the pillar of perfection
she thought) where I could shine, my life would never be the same again.
the real world, André. Now what?
Naiveté was lost and I decided to work harder on things
that I had control over.
around figuring out how things work took me from childhood (wondering why I couldn’t make a robot from
a stuffed animal and radio parts) all the way to working at NASA in high school
(yep, the public school I transferred to) with one of the country’s leading
astrophysicists. This ultimately led me to enroll into the amazing aerospace
engineering program at the University of Maryland College Park, feeling like my
calling was solidified in helping get more people into space. Thanks again, Mae Jemison and Bill
However, during my second year at UMD, I took an elective
course on epidemiology (the study of disease) and had a particular interest in
a lecture where I learned about resistant tuberculosis impacting underserved
communities. Ravenous learning mode
engaged. Within weeks of discussions with my professor Dr. Donna Howard and
learning about the world of public health, I was sold. The discipline around
preventing large groups of people from getting sick and ensuring their ongoing
health just made so much sense to me. And so it began — my fascination with
public health would never go away and within weeks I switched my major.
to 2016. After several years of building
a name for myself, working on projects around digital health innovation,
starting a consultancy and growing a massive network of
innovators/change-makers — the opportunity of a lifetime came my way where my
alma mater, the University of Maryland School of Public Health, chose me to be
the Spring commencement address speaker. Joining the ranks of the U.S. Surgeon
General and other distinguished speakers in the School’s history gave me
pause — but then I knew what time it was. It
was time to redefine how I showed up in the world.
Me: 2016 Spring Commencement, University of MD School of Public Health
Days after the commencement (my 9 minute speech) I reflected. A lot. I thought about the opportunities
that were springing up in the health innovation landscape around data science,
technology, venture capital, design and more. The startups that were coming
into view to tackle mental health, chronic disease, healthy cities, and access
to quality care. I thought about all the connections I’ve been fortunate enough
to make over the years as well as the valuable career coaching I’d done with
friends and colleagues. Finally, I reflected on the kinds of talented
individuals that would need to come together in order to create long-lasting
impact for the future of health. It was also time for me to reflect on my own
beginnings in this promising landscape as a person of color who often times was
“the only” in many rooms and conferences. If we really wanted to see change
happen in this industry — that actually delivered value-based care to those who needed it and still propelled
innovation —we would need to close the representation gaps in the workforce.
Over the course of my life and career, there have been 3 consistent
approaches to the future of health
people to resources, opportunities & information that change their careers
I think there’s a reason why these have been constants for
the past 12+ years — I feel the most
energized and effective when all of these areas converge.
It didn’t take long to come to the realization that I
wanted to build something that allowed scalable opportunities for those who
want to build the future of health — especially those who reflected the
diversity of our society. Honestly, it makes absolutely no sense to see the
teams of the next promising/shiny/highly publicized new startup tackling public
health or healthcare, that is devoid of diversity — especially in leadership
ranks. For the world of public health and healthcare, it is critical to have
diverse talent representation and inclusive workplaces to create real, valuable solutions that actually
meet the needs of our population.
So I asked myself: how can I help bring more talented
people of color, women and the LGBTQ community into the tremendous
opportunities that are happening around chronic disease prevention, access to
care, food sustainability, healthy cities, mental health, and more? How can I
make it easier for these talented professionals who want to lend their skills
to create solutions with some of the most forward-thinking startups and
companies shaping our well-being? How can I make access and acceleration
to these possibilities part of an ecosystem approach?
Diversity and inclusion is a hot topic to discuss these
days, however in the health/healthcare landscape, it absolutely cannot just stop there. We’re not dealing strictly
with products here — these are actual lives at stake. How we hire, retain and
advance the talented individuals from underrepresented communities (women,
people of color, LGBTQ, disabled) is the only way we will see widespread,
At Onboard Health, we’re extremely dedicated to bringing together a diverse
workforce and equipping them with opportunities that not only open the door for
roles at companies — we’re equipping the mission-driven with resources,
coaching, and access to partnerships to build a healthier future.
For instance, we’re supporting the…
scientist who cares about the impact of food security can have in Detroit
storyteller/content strategist can have in building the brand of a mental
health tech startup
engineer can have in solidifying the framework of a company building new
solutions for chronic disease
nurse providing ongoing insights and advisory services to a human-centered
design company working on a population health project
And that’s just a few examples. Have a conference you want
to make sure is representative of our society? Yep, we’re building a community
for that. We’re making sure companies who have a lens on the future of health
(bonus points for upstream solutions and community-based care) can easily share their vision
of impact with a whole new generation of doers and thinkers who have the skills
as well as experiences, they need.
We’re gearing up for some amazing things this summer for
our talent community as well as how we work with startups, larger organizations
and academic institutions to build the future together.
Sound like you want to be a part of this? Let’s talk (but more
importantly, let’s do).
André Blackman is the Founder and CEO of Onboard Health where he and his team are dedicated to building a diverse ecosystem of talent and companies to build the future of health. This post originally appeared on Medium here.
Today on Health in 2 Point 00, Jess and I are in Helsinki for Health 2.0 HIMSS Europe. In Episode 83, Jess asks me about Roche cheating on mySugr—Roche announced a new partnership with digital diabetes provider GlucoMe, about the new $100 million hospital venture fund in Iowa coming from UnityPoint Health, and about Infermedica’s recent $3.65 million raise for their cool symptom checker complete with an AI chatbot. Stay tuned for more updates from the conference. —Matthew Holt
Health in 2 Point 00, Episode 83 | Health 2.0 HIMSS Europe - YouTube
I’ve been talking in recent posts about how our typical methods of testing AI systems are inadequate and potentially unsafe. In particular, I’ve complainedthat all of the headline-grabbing papers so far only do controlled experiments, so we don’t how the AI systems will perform on real patients.
Today I am going to highlight a piece of work that has not received much attention, but actually went “all the way” and tested an AI system in clinical practice, assessing clinical outcomes. They did an actual clinical trial!
Big news … so why haven’t you heard about it?
The Great Wall of the West
Tragically, this paper has been mostly ignored. 89 tweets*, which when you compare it to many other papers with hundreds or thousands of tweets and news articles is pretty sad. There is an obvious reason why though; the article I will be talking about today comes from China (there are a few US co-authors too, not sure what the relative contributions were, but the study was performed in China).
China is interesting. They appear to be rapidly becoming the world leader in applied AI, including in medicine, but we rarely hear anything about what is happening there in the media. When I go to conferences and talk to people working in China, they always tell me about numerous companies applying mature AI products to patients, but in the media we mostly see headline grabbing news stories about Western research projects that are still years away from clinical practice.
This shouldn’t be unexpected. Western journalists have very little access to China**, and Chinese medical AI companies have no need to solicit Western media coverage. They already have access to a large market, expertise, data, funding, and strong support both from medical governance and from the government more broadly. They don’t need us. But for us in the West, this means that our view of medical AI is narrow, like a frog looking at the sky from the bottom of a well^.
What would be really cool is if anyone who knows details about medical AI in China wanted to get in touch and let me know what is actually happening over there. I’d love to do a blog post highlighting leading companies and projects that actually have products working with real patients in real clinics. The same goes for AI teams in Africa, India, Southeast Asia and anywhere else that doesn’t get news coverage or exposure.
It bills itself as a prospective randomised control trial. Others have claimed to do AI clinical trials before, but have all (to the best of my knowledge) fallen short.
This one lives up to the billing.
An AI team/company/startup called Shanghai Wision AI Co. produced a system that detects polyps (little tumours) in the bowel wall during a colonscopy. They did performance testing previously which showed a per image AUC of 0.984 in a retrospective experiment, and a variety of other promising results. But the defining characteristic of a clinical trial (in my opinion) is; “how does it change patient outcomes in practice?” In this case, does using the AI system translate into diagnosing more cancer, and does it lead to more unnecessary biopsies?
In the paper, they use the system in actual clinical practice. The endoscopist did their normal colonoscopy, but the AI was watching them work in real time. If it saw a polyp it would beep, and the endoscopist could then turn to look at a different screen which shows a floating rectangle overlying the video to highlight the polyp.
The two screens the endoscopist has available. They only look at the right image (the AI augmented view) if the AI alerts them it has seen something important.
They measure how often the endoscopist agrees with the AI system, but if they left it there it would just be another example of performance testing gussied up with the veneer of prospective cohort selection (which some audacious researchers have claimed makes it a clinical trial by itself).
But this team took the step that actually elevates the work from an experiment into a clinical trial: they removed the polyps!
These are polyps that the AI noticed, which the endoscopist had not seen (although the endoscopist could still overrule the system, which would be recorded as a false alarm). In this study, they performed invasive medical procedures on patients because of the output of an AI system.
Before you get all “of course they did, it was China, something something safety standards”**, I want to be clear – they did exactly what needs to be done to show that a system is safe. After you do your performance testing and get promising results, you need to actually test what happens in practice. This is right. This is good.
A colonoscopy AI in a clinic doesn’t just make a visual decision. It (with the endoscopist) decides who needs a biopsy. If your testing doesn’t include actually doing the biopsies, then medical AI safety, you’re doing it wrong.
So, what did they find?
Unsurprisingly, they did a lot more biopsies. They removed almost double as many polyps in the AI group (500 vs 270 in the “normal colonoscopy” group, in roughly the same number of procedures). This number in itself isn’t that interesting, the exciting part is that they specifically found a lot more adenomas when the removed lesions were examined under a microscope (adenomas are the polyps at risk of turning into cancer). They found 1.89 times more polyps overall, but also found 1.72 times more adenomas. This seems like a huge increase in the number of potential cancers.
But the fact they find adenomas doesn’t mean the patients will be better off. The team recognised this, and also analysed what type of adenomas they found.
As you would expect, the AI mostly found small “diminutive” adenomas. Humans are unlikely to miss the big, dangly ones (dangly is the technical term, but some people call these lesions “pedunculated”). The AI couldn’t add much in this group of lesions, detection rates are already nearly 100%.
We also know the smaller lesions that the AI system founds have a lower risk of cancer than big ones (more cells = more risk), but the team acknowledges this. They say “further studies should address the role of CADe on decreasing interval cancer, which is the main goal of any screening colonoscopy.” This, again, is sensible.
But it will take years to run those experiments, so the question to ask right now is “when is it safe enough to use?”
Medical AI safety: doing it right
While we don’t have data on the primary endpoints of interest (interval cancer rate, cancer mortality), we do have safety data. They recorded both the false alarm rate (when the endoscopist over-ruled the AI system and said “I’m not biopsy-ing that”) and the complication rate (the risk of biopsy is that you might puncture the bowel).
Amazingly, the false alarm rate was tiny. Despite a per image false alarm rate of around 5% reported previously, they somehow end up with one false alarm per 13 colonoscopies in practice! I don’t know exactly how they achieved this (presumably they chose a pro-specificity threshold and did some “only if it is on multiple frames” heuristic magic), but it sounds amazing.
The complication rate was also tiny, it was zero! With almost 500 biopsies in the 500-ish cases with CAD and no complications, we are fairly safe to assume the risk is unlikely to be much higher than normal.
One other major issue I have with most medical AI papers, which I haven’t written much about because I am actively researching the effect, is that you need to carefully investigate what cases the system gets wrong. Just because an AI system is as good as a human (gets as many or fewer cases wrong), it doesn’t mean they are the same cases. What if the human errors are benign and never lead to harm, but the AI errors are life-threatening^^? The only way to understand the error distribution is to actually look at the images (something I have said quite a bit).
You can probably see how useful this is for a clinician. If the system decides it sees a polyp, the user could immediately look for drug capsules and bubbles, the equivalent of a radiologist considering imaging artefacts as a cause of unusual visual findings. I feel this sort of error analysis is waaaaay more clinically useful than any interpretability technique (although in practice it is fairly common to have no appreciable reason for the errors).
So given this safety data, and additional analysis, what do I think?
If they had only doing performance testing, even if I had ten times the guts^ I wouldn’t be willing to use this AI system on patients. There are tons of ways the results in performance testing can be misleading.
But they did a clinical trial, with an appropriate surrogate endpoint. They even had a very even handed discussion of possible failings of the study design, and delightfully refer to the possibility of user bias due to the lack of blinding as the endoscopists showing too much “competitive spirit” when they are being watched, which has apparently been reported in previous research.
Regarding blinding, obviously it was impossible here. They have a machine that goes ping for the AI arm. Who would run a blinded AI experiment …
well, this team would!
They have recently posted an abstract for a double blind study with sham AI! That is a serious commitment to scientific exploration, and something that I am greatly looking forward to seeing more detail on if/when they publish a full report.
In summary, I do think they have done the work needed to show a reasonable level of safety and efficacy, enough to justify clinical use. We still need to see if it works with long term outcomes, and if it works in other populations (which they also acknowledge), but in the mean time their own sign-off sounds about right:
So there you go. An AI clinical trial that (in my view) provides enough evidence to justify clinical use.
Such a thing is as rare as a Phoenix feather^.
* Even Eric Topol missed it initially, despite managing to tweet about pretty much every decent paper that comes out. He had eyes, but could not see Mt Tai^.
** I don’t mean to sugar coat it, as well as a lack of access there is clearly an element of sneering disinterest if not overt racism at play. Almost every news story I have seen about Chinese AI is focused on the negative aspects of their medical/legal/political systems, highlighting concerns around lack of privacy, regulation, security and so on. It is fine to raise those issues, and many of them are serious at least from my long-distance view, but to only focus on those is obviously biased when there is a ton of cool work happening over there.
^ I have (unrelatedly) been reading a lot of Chinese fiction recently, so my idiom game is on point. The header image is a reference too – essentially “Crouching Tigers and Hidden Dragons” describes experts who are undiscovered, which was too relevant to not include in the context of this paper.
^^ I have (unpublished) real world examples of AI with good performance making life-threatening errors, this isn’t just hypothetical
Luke Oakden-Rayner is a radiologist (medical specialist) in South Australia, undertaking a Ph.D in Medicine with the School of Public Health at the University of Adelaide. This post originally appeared on his blog here.
Today on THCB Spotlight, Chris Gervais, Chief Technology Officer of Kyruus, tells us about what Kyruus is doing to improve patient access and help health systems match patients to the right providers. Health systems often don’t know enough about their providers, and Kyruus is working to empower health systems to use that data in a computable way in order to coordinate patient demand with physician supply.
THCB Spotlight | Chris Gervais, CTO of Kyruus - YouTube
Office of the National Coordinator (ONC) and the Centers for Medicare and
Medicaid (CMS) have proposed final rules on
interoperability, data blocking, and other activities as part of implementing
the 21st Century Cures Act. In this series, we will explore ideas
behind the rules, why they are necessary and the expected impact. Given that
these are complex and controversial topics are open to interpretation, we
invite readers to respond
with their own ideas, corrections and opinions.
Interventions to Address Market Failures
Many of the rules proposed
by CMS and ONC are evidence-based interventions aimed at critical problems that
market forces have failed to address. One example of market failure is the long-standing inability for health care
providers and insurance companies to find a way to exchange patient data. Each
has critical data the other needs and would benefit from sharing. And, as CMS
noted, health plans are in a “unique position to provide enrollees a complete
picture of their clams and encounter data.” Despite that, technical and
financial issues, as well as a general air of distrust from decades of haggling
over reimbursement, have prevented robust data exchange. Remarkably, this happens
in integrated delivery systems which, in theory, provide tight alignment between
payers and providers in a unified organization.
With so much attention
focused on requirements for health IT companies like EHR vendors and providers,
it is easy to miss the huge impact that the new rules is likely to have for
payers. But make no mistake, if implemented as proposed, these rules will have
a profound impact on the patient’s ability to gather and direct the use of
their personal health information (PHI). They will also lead to reduced
fragmentation and more complete data sets for payers and providers alike.
Overview of Proposed CMS Rules on Information
Sharing and Interoperability
The proposed CMS rules
affect payers, providers, and patients stating that they:
Require payers to make
patient health information available electronically through a standardized,
open application programming interface (API)
Promote data exchange
between payers and participation in health information exchange networks
Require payers to provide
additional resources on EHR, privacy, and security
Require providers to comply
with new electronic notification requirements
Require states to better
coordinate care for Medicare-Medicaid dually eligible beneficiaries by
submitting buy-in data to CMS daily
Publicly disclose when
providers inappropriately restrict the flow of information to other health care providers and payers
These rules apply to:
Health care providers
State Medicaid and
Children’s Health Insurance Program (CHIP) agencies
Insurers that offer qualified health plans (QHPs)
Medicare Advantage plans
Medicaid and CHIP managed
While, the broader commercial market, employer-sponsored health
insurance, and stand-alone dental plans are currently exempted from these rules,
the hope is that some will still adopt these new
Data Exchange Requirements for Payers CMS has proposed substantial data exchange requirements that define both the types of information to be shared and, where appropriate, the technical approach and standards to be followed. One key requirement is to implement and maintain an open API that allows third-party applications (some with approval from the patient) to easily retrieve a variety of information as shown in the table below:
Other key data management provisions include:
Payers must be able to exchange data elements
outlined in the United States Core Data for Interoperability (USCDI) standards.
Payers must incorporate received
data into their own records.
When a patient (member) requests it,
the payer must (1) accept data from a patient’s prior health plan for up to five
years, (2) send data to other health plans for up to five years, (3)
send data to a recipient designated
by the patient for up to five years.
The proposed rules for exchanging data should lead to reduced fragmentation and more complete datasets for payers, providers and patients.
Importantly, the rules also specify response times where
Claims, encounter, and clinical
data must be available through the API no later than one business day after a
claim is processed or the data is received by the payer.
Provider directory data must be
updated within 30 business days of changes to the directory.
No specific timeframe for
submitting pharmacy directory or formulary information.
A key issue will be the payer’s dependence on providers sharing
data with them in a timely manner so the payer can meet these requirements. CMS
is urging payers to consider whether their contracts with providers should
include timing standards regarding the submission of claims and encounter data.
API Standards for Payers
CMS and ONC have been moving in
tandem to address interoperability and information blocking. It’s no surprise CMS
will require payers to comply with a separate ONC proposed rule to use APIs to
meet certain technical standards and address standardized content and
vocabulary for data available through the API. They also address behaviors that
can limit interoperability or lead to information blocking. A good example is
the requirement to deliver clinical data which mandates USCDI be available via
a standard FHIR API. Other requirements specify (among other things) that:
The API must be publicly
accessible on a payer’s website and accompanied by documentation on technical
aspects (such as API syntax, function names, and various other parameters).
Payers cannot require a reader to
pay a fee to access the documents, receive a copy via email, or agree to
receive future communications before making the documentation available.
Payers can deny or discontinue a
third party’s connection to their API if the payer determines—using objective,
verifiable criteria —that the connection threatens the security of protected
health information (PHI).
make non-standardized data available through their APIs but are required to ensure that their API documentation
provides enough information to developers to handle this information.
Economic Impact on Payers
In general, the rules proposed by CMS and ONC are subject to a
Regulatory Impact Analysis (RIA) to estimate the costs and benefits of specific
rules. Interestingly, CMS suggests that promoting data exchange between payers
and participating in a trusted health information exchange may qualify as
“quality improvement activities” for purposes of an insurer’s medical loss
ratio. This is an important consideration for payers since these costs
could be counted against the requirement to spend 80 or 85
percent of premium revenue on claims and quality improvement.
This is Getting Real – Real Fast
CMS has proposed specific time lines and actions for payers to
meet the new requirements as illustrated below:
It seems likely payers will
object to the January and July 2020 deadlines and that CMS and ONC will
accommodate some delay, given the current timelines.
Data Must Flow for the Benefit of
An overarching theme of the proposed rules is that patient data
should flow freely and at the direction of the patient unless there is a
compelling, common-sense exception (seven
of which are spelled out in detail). The proposed rules for
payers reflect this theme and directly address the long-standing failure of
market forces to encourage robust information sharing. They also hold the real
promise of benefiting patients, health care providers and payers by enabling better
care at a lower cost.
Dave Levin, MD is co-founder and Chief Medical Officer for Sansoro Health where he focuses on bringing true interoperability to health care. You can follow him @DaveLevinMD or email Dave.Levin@SansoroHealth.com
Nikki Kent, SVP of Operations at Sansoro Health, is an accomplished health care executive having specialized in Operations, Human Capital and Sales for Payer and Provider organizations.
You’re running late and many things didn’t go right today. You knock on the door and enter the exam room with an apology. If you’re like me, you have a few papers and an iPad or a laptop in your hand. You sit down and open the patient’s chart in your device or perhaps on the big desktop, eyes not exactly locked on the patient.
Only after getting to where you need to be in the computer do you really look the patient in the eyes. Your body language has been one of hurry and distraction. Now you try to repair the damage of that, so you try to show you’re settling down now, at least for a few moments. You might sigh, move your arms in a gesture of relaxation and say something to get the history taking underway.
So far, you’re failing. I do that often, too.
Here’s what we all know we need to do, but often don’t; we should follow these ABCs:
A – Attention:
Clear your mind. It doesn’t matter what happened in the other room with the other patient, or on the phone with the insurance company or the smug specialist or ER doc who pointed out the diagnosis you missed. Open the door (I always knock first) and immediately look at the patient. Make eye contact and observe them. Pay attention to how they look, what they are signaling. The computer can wait; a few moments of focused attention will usually save you time in the end. After all, red or teary eyes, a leg cast, a big bruise or change in grooming can make the visit go in a direction you wouldn’t have expected from he listed chief complaint. How many times have we heard a patient comment about another doctor: He didn’t pay attention to me. Do we always do that ourselves if we’re rushed or preoccupied?
B – Behavior:
Behave like a doctor. I keep saying that. But the clinical encounter is like a dance, where either one of us can lead, and we lead a little too often. Behave in a way that signals respect, interest and both confidence and humility. Behave like someone who serves, guides and helps the patient heal. Behave in a way that behooves a doctor. You have paid attention to the patient. What did you see? What does he or she need, or need you to be like, in this moment?
C – Connection:
The goal of contemplating how a good clinical encounter should begin is to establish connection. Learning about someone, counseling someone, treating someone, comforting someone all require having a connection with that person. They tell you that strangers you meet like you better if you invite them to talk about themselves. Making connections with patients requires showing genuine interest, inviting disclosure and reciprocating just enough to show that you are a real person, but not so much that you seem too fallible or self absorbed. It is better to talk about your interests than about yourself. Sharing about pets, children and hobbies that don’t portray you as uppety is safest.
In the fast paced, high pressure day to day work we do, I sometimes catch myself not engaging quite enough with my patients. Even after forty years of doing this, I need to remind myself to start every patient encounter off in a way that sets the stage for making clinical and interpersonal progress. My demeanor builds relationship equity over time so that if I sometimes don’t live up to my ambition and miss one of my ABCs, my patients are a little more likely to overlook it.
Hans Duvefelt is a Swedish-born rural Family Physician in Maine. This post originally appeared on his blog, A Country Doctor Writes, here.
Today on Health in 2 Point 00, I’m back (despite Jess’s attempt to replace me). In Episode 82, Jess asks me about Talkspace’s $50 million raise, Heal getting flack for adding telehealth to their house call service, and Apple acquiring Tueo Health last year—and we’re just now hearing about it. Jess also gets riled up by Pokemon Sleep and Pillo’s $11 million raise. —Matthew Holt
Health in 2 Point 00, Episode 82 | Talkspace, Heal & Apple - YouTube