Search the KHIT Blog

Friday, November 21, 2014

A damning Health IT quote from Joe Flower

"A Medical Economics survey published last February showed that over 70 percent of physicians would not buy their current information system, if they had the chance to make that choice again, because they hate the way it works. Nearly 70 percent have seen no improvement in care coordination with the hospitals. And 45 percent believe it has actually hurt patient care."
From "Why You Should Ditch Your IT System."

So, three survey findings. The first one we can take at face value. The second and third, though, would be stronger with some underlying data having firm prior operational definitions, i.e., some scientific metrics going to adverse process "care coordination" and HIT proximate causes that "actually hurt patient care." Absent those, we're left mostly with partisan shouting-past-each-other points of contention. See, e.g., my prior post "An Epic battle: Did the EHR kill Dallas Ebola patient zero? On the double-edged sword of Health IT."

I am reminded of the saying "the plural of 'anecdote' is not 'data'."

Nonetheless, Joe's post is highly worthy of consideration. He's one of the most astute and charitable healthcare observers I've ever met. Read the entire post carefully.
So you spent millions to billions of dollars on information systems over the past few years, right?

How’s that working out for you?

For a large percentage of you, whether or not you admit it, not so well. What you bought needs some serious tweaks, re-engineering, re-thinking, re-vamping.

For an even larger percentage, maybe most of you, the best advice is: Junk it. Throw it out and start over.

Poorly designed and poorly implemented information systems are worse than useless, worse than a waste of those millions and billions of dollars. As we go through rapid, serious changes in health care, poor information systems will strangle your every strategy, hobble your clinicians, kill patients and actually threaten the viability of your organization.

A lot of health care executives dismiss the complaints about the new systems as the carping of stubborn technophobic doctors and nurses who should just get with the program. If you are tempted to do that, you need to take a step back. You need to get real. The complaints and concerns are too widespread, too deep and indeed too frightening for that kind of blithe denial. And they are not just coming from disgruntled docs...

In a recent Black Book survey, 98 percent of 13,650 registered nurses polled said nurses in their facility were never asked to help design the system; it was just imposed on them. As a result, 85 percent say the system is flawed and gets in their way, 94 percent feel that it has not improved communication among the care team, and 90 percent feel that it has damaged communication with the patient...

It’s Time
It’s not like this is new. It’s not like we didn’t see this coming. I have been writing about and advocating for the digitization of health care for 30 years, as have many others. I have to tell you: We knew back then that interconnectivity through industry standards — and smart user interfaces that assist clinicians in their normal workflow rather than hinder them — were hard problems that needed solving...
Yeah. "How's that working out for you?" Maybe Joe might inquire of Kaiser Permanente, users of the much-maligned Epic platform? See my Feb 6th, 2014 post "Meaningful Use 2013 review, ONC Working Group Stage 3 draft report, and discussion of KP's book "Connected for Health."

See also my May 28th, 2014 post on Joe, "Joe Flower on the Health tech revolution."

"It's not like we didn't see this coming." I return to my 2011 post "Use Case" and the observations of medical economist J.D. Kleinke, written nearly a decade ago.
HIT market failure. The underlying cause of Joe’s death is health information technology (HIT) market failure. If the state of U.S. medical technology is one of our great national treasures, then the state of U.S. HIT is one of our great national disgraces. We spend $1.6 trillion a year on health care—far more than we do on personal financial services—and yet we have a twenty-first-century financial information infrastructure and a nineteenth-century health information infrastructure. Given what is at stake, health care should be the most IT-enabled of all our industries, not one of the least. Nonetheless, the “technologies” used to collect, manage, and distribute most of our medical information remain the pen, paper, telephone, fax, and Post-It note.

Meanwhile, thousands of small organizations chew around the edges of the problem, spending hundreds of millions of dollars per year on proprietary clinical IT products that barely work and do not talk to each other. Health care organizations do not relish the problem, most vilify it, many are spending vast sums on proprietary products that do not coalesce into a systemwide solution, and the investment community has poured nearly a half-trillion dollars into failed HIT ventures that once claimed to be that solution. Nonetheless, no single health care organization or HIT venture has attained anything close to the critical mass necessary to effect such a fix.

This is the textbook definition of a market failure. All but the most zealous free-market ideologues recognize that some markets simply do not work. Indeed, reasoned free-market champions often deconstruct specific market failures to elucidate normal market functioning. The most obvious examples of such failures (such as public transit and the arts) are subsidized by society at large because such subsidies yield benefits to the public that outweigh their costs. Economists refer to these net benefits as “positive externalities,” defined as effects that cannot be captured through the economic equation of direct cost and benefit.

The positive externalities of an HIT system approaching the functionality of our consumer finance IT system include reduction of medical errors like the one that killed Joe Wilson; elimination of tens of thousands of redundant and expensive tests, procedures, and medications, many of which are not only wasteful but harmful; and the coordination and consistency of medical care in ways only promised by the theoretical version of managed care. These public health benefits are well beyond the reach of a health care system characterized by the complexities of medicine and conflicts of multiple parties working at economic cross-purposes. They are trapped outside the economic equation, positive externalities of a stubbornly fee-for-service health care system that inadvertently rewards inefficiency, redundancy, excessive treatment, and rework...
Back to Joe:
"[For] maybe most of you, the best advice is: Junk it (your EHR). Throw it out and start over."
Interesting. Maybe a small shop ambulatory practice will have that luxury, but I don't see larger acute care institutions going there, absent crushing exigent circumstances.

Joe in the comments (and my reply):

I subseqently went on with more.
“Meaningful Use,” btw, is a red herring whipping boy. The typical ambulatory EHR houses close to 4,000 variables in its RDBMS, with hundreds or more having to be accessed, viewed, updated, transmitted, synthesized etc during a conventional pt visit (and, yeah, to your point, a lot of them just to get paid). The subset numerator/denominator MU vars comprise less than 2 dozen (including the CQMs), most of which can be handed at the sub-MD support staff level.

Just to be clear: I’m no big MU fan, btw. I mocked the bozo-ness of it even while working in the MU program. See my Clinic Monkey (a “Survey Monkey” riff)

The fact that it may take 5-8 clicks to get to a MU data target is stupid (e.g., in eClinicalWorks there are 4 different workflows for doing CPOE, ranging from 5 to 8 clicks). EVERY MU criterion should be macro-accessible in 1 click (or Dragon voice command). It’s not much to have to learn. I’d have insisted on macro capability within every ONC certified system as a condition of cert. Windows no longer has a macro utility, but there are 3rd party vendors of inexpensive macro utilities. Still, no one wanted the added expense and hassle of going 3rd party macro. And, in fairness, even optimizing MU workflows might still result in an additional $4,000 a year in labor cost (even if it only added minute or less to a note; I could show you the math).

apropos of all of the foregoing, let's fast-forward up to late 2014. Where might we be headed? Jerome Carter, MD, at EHR Science:
A Question for EHR Vendors: What Is a “Real” Computer?

This might seem like a silly question with an obvious answer, but is it really? The solution to any problem grows out of the environment in which it appears and from the mindset in which it was conceived. In 1970, the answer to this question would have been a mainframe system. By 1981, after the Apple II and a few other microcomputers had been around for a few years, the answer for most people at that time would still have been mainframes (or maybe minicomputers as well)  because microcomputers were still considered to be toys. When IBM released the IBM PC AT in 1982, microcomputers began to be taken seriously as computers—that is, computers that could be used for real business applications.  The arrival of reliable local area networking technology cemented the status of PCs as real business computers.

Initially, local area networks (LANs) were used to share printers, disk storage, and applications.  However, as servers became more powerful and disk storage more dense and affordable, database management systems and sophisticated client/server software appeared...

Personal computers had to mature for a few years before they could tackle real problems, and smartphones and tablets are following the same pattern. Among other things, mobile computers add new user interface options and portability to the computing mix in ways that no other computing platform can match.   Solutions to clinical information management problems must now embrace mobile computing capabilities: touch-based interfaces, multi-media data management, communications functionality, and location/ movement awareness. Without question, the iPad Air 2 is a real computer, and it and other tablets with similar specs can be used to solve real problems.

Most current EHR software was designed well before tablets and smartphones existed, and many were born before the Internet really caught on. These EHR systems were designed back when LANs were state-of-the-art computing platforms, the cloud did not exist, Wi-Fi was painfully slow, and pointing was done with a mouse. The computing platform and development tools dictated how developers approached clinical information management problems...

Looking at clinical care and its computing needs, I see requirements that are distinct when compared to standard business computing.  Clinical data are varied and numerous. Clinical work consists of interacting with patients to obtain information, consulting information sources  (e.g., chart, guidelines, articles, other clinicians), making decisions, recording information, and moving on. Support for clinical work requires large, searchable data stores, fast networks, sophisticated communications functionality, and portable computers capable of displaying text, pictures, sound and video.  Tablets and smartphones are the first computers to meet all of these requirements.

Writing for mobile means stepping back from web and client/server applications and being willing to see a problem purely from the standpoint of mobile computing; that is, adopting a “mobile first” attitude.

Mobile first requires a willingness to rethink past approaches. At the top of the list is use of cloud capabilities. Like mobile computers, the cloud is a new way of doing things. Building mobile applications that link to cloud storage and use APIs to interact with other applications is a new way of delivering functionality.    There is no reason to have local terminology services if they can be obtained via a cloud application. The same is true of workflow engines or another service that supports clinical work. Mobile first also means not taking a client/server app and putting a mobile face on it.  That will not work any better than putting a browser interface on a standard desktop app. It might work to some extent, but the original design limitations will show through...
How many EHR vendors will bite the bullet and start serious mobile-first projects? Few, I imagine, because if the past is prologue, most will cling to the prevailing wisdom that mobile devices are not real computers. And we know how that story ends…

OK, where will we get the requisite top gun HIT programming / UX design talent seemingly in short supply? From the current issue of The New Yorker:
Want to hire a coding superstar? Call the agent.


Not long ago, Stephen Bradley, a New York tech entrepreneur, was looking to expand his company, AuthorBee, which aggregates tweets and Instagram posts and puts them together in story form. Instead of following people, readers can follow their interests—“Breaking Bad,” for example, or the New England Patriots. Bradley is not a stereotypical startup founder, a hoodie-wearing college dropout; he’s been working in tech and media for decades. To launch AuthorBee, he raised three-quarters of a million dollars from angel investors and hired programmers in Pakistan and Bangladesh to build a prototype. Now he wanted to build a bigger, better version of the site, so he had to find someone to write the code that would form AuthorBee’s DNA. The guys in Pakistan and Bangladesh were O.K., but the cultural differences and the language barriers slowed things down. He needed “one really good developer” with a mastery of all the coding languages and frameworks that AuthorBee uses: Python, Django, Angular, JavaScript, the Twitter A.P.I. The search for programming talent was the part of building a startup that Bradley most dreaded. “It is a nightmare,” he told me. “And I’m as plugged in as you can be to the New York tech scene.”

He put up a job posting on the Web site AngelList, and was immediately flooded with calls from headhunters and e-mails from offshore companies wanting to set up a “short online telephonic meeting.” “I could have had two hundred résumés on my desk,” Bradley said. But he knew that the people behind those résumés weren’t the ones he was looking for. His dream developer might be buried in there somewhere, but Bradley had come to think that developers were like social media itself: “Ninety-nine per cent of them suck.” He added, “The entire problem is wading through the noise.”...

The world is being rebuilt in code. Hiring computer engineers used to be the province of tech companies, but, these days, every business—from fashion to finance—is a tech company. City governments have apps, and the actress Jessica Alba is the co-founder of a startup worth almost a billion dollars. All of these enterprises need programmers. The venture capitalist Marc Andreessen told New York recently, “Our companies are dying for talent. They’re like lying on the beach gasping because they can’t get enough talented people in for these jobs.”

The computer science taught in colleges still focusses more on theory than on commercial application; the business of teaching practical coding skills has the whiff of trade school. So-called coding “boot camps,” such as General Assembly, founded in 2010, are trying to fill the gap, teaching crash courses in how to design Web sites and write code. But Jake Schwartz, the co-founder and C.E.O. of General Assembly, told me, “There’s simply not enough senior people in the system.”...
Well worth your time. Ponder the ramifications for Health IT.

More Joe Flower:
Zane Burke, the president of Cerner (which just bought Siemens), made a strong point about interoperability in a recent interview with H&HN Daily’s Matthew Weinstock:

“We all owe it to the country … to really perform true interoperability and create openness…. You need platforms that don’t just open up your APIs [application programming interfaces] but actually create ecosystems for other players … to perform well. We can’t use the operating system at the EHR lever as a competitive advantage. It just can’t be that way.
Yeah. See my posts on "Interoperababble" See also "Interoperability solution? HL7® FHIR® -- We ® Family."

More to come...


  1. RE "A Medical Economics survey published last February showed that over 70 percent of physicians would not buy their current information system, if they had the chance to make that choice again, because they hate the way it works." <= I strongly suspect that if given the chance to change, they wouldn't like their replacement software any better.. IMHO there are no extraordinary systems to be had out there, so changing systems is just a way for providers to reward HCIT vendors for that by keeping their sales pipelines filled.

  2. This is a great read. Everything mentioned are things that I have been sharing for a very long time. The applications are out of touch with practices. You cannot have IT people design an EMR system. Development has to be driven by the users (providers). Our company supports over 150 medical practices, mostly on eClinicalWorks. If it were up to them, they would all go back to paper.