Monday, April 13, 2009

Should Google Seek the Resignations of Those Responsible for This Healthcare IT Debacle?

Simply amazing. One of the richest and most "leading edge" IT companies in the world with almost unlimited resources and access to expertise commits one of the most fundamental biomedical information science (a.k.a. informatics) blunders, as in the taxonomy in my post here, at the level of "likely to cause patient harm in short term if uncorrected."

I have repeatedly written over at least the past ten years that applying the leadership and methodologies of business IT to clinical computing is both ill conceived and dangerous, as business computing and clinical computing are two very different computing subspecialties, the latter requiring quite specialized leadership and approaches.

I've written it at academic sites, in magazines, in newspapers, and other venues.

Yet, as we have observed at HC Renewal regarding other flavors of healthcare mismanagement and malfeasance, these words seem to suffer an anechoic fate.

Here we go again with another example of what appears to be gross mismanagement of clinical IT by business IT personnel and organizations. The following type of debacle is sooner or later going to kill patients and
must end, immediately:

Boston Globe
Electronic health records raise doubt
Google service's inaccuracies may hold wide lesson
["may?" - ed.]
By Lisa Wangsness, Globe Staff
April 13, 2009

WASHINGTON - When Dave deBronkart, a tech-savvy kidney cancer survivor, tried to transfer his medical records from Beth Israel Deaconess Medical Center to Google Health, a new free service that lets patients keep all their health records in one place and easily share them with new doctors,
he was stunned at what he found.

Google said his cancer had spread to either his brain or spine - a frightening diagnosis deBronkart had never gotten from his doctors - and listed an array of other conditions that he never had, as far as he knew, like chronic lung disease and aortic aneurysm. A warning announced his blood pressure medication required "immediate attention."

"I wondered, 'What are they talking about?' " said deBronkart, who is 59 and lives in Nashua.

DeBronkart eventually discovered the problem: Some of the information in his Google Health record was drawn from billing records, which sometimes reflect imprecise information plugged into codes required by insurers. Google Health and others in the fast-growing personal health record business say they are offering a revolutionary tool to help patients navigate a fragmented healthcare system, but some doctors fear that inaccurate information from billing data could lead to improper treatment.


(Addendum April 19: a first hand account of this problem is at e-patients.net here.)

What manner of amateurs made and approved the decision to
map semantically and often medically imprecise, and often deliberately overstated or misused billing codes to diagnoses, and then display the diagnostic terms to a user - ANY user, patient or "learned intermediary" - in an electronic health record?

Not to mention how poorly conceived and implemented many of the HIT billing systems themselves are, making billing data even less trustworthy...


Injecting humor into a most somber post,
Homer succinctly summarizes the situation.


It is common knowledge to any competent person in healthcare informatics that doing what was done by Google Health is prone to create exactly the kind of situation that occurred.

Insurance data, by contrast, is already computerized and far easier and cheaper to download. But it is also prone to inaccuracies, partly because of the clunky diagnostic coding language used for medical billing, or because doctors sometimes label a test with the disease they hope to rule out, medical technology specialists say.

One does not have to be much of a "specialist" to make this realization. Almost anyone who's ever practiced medicine could probably have told Google's designers, developers and programmers this. This raises a number of questions, which also do not require a specialist to raise:

  • What were the designers, implementers and management of this project thinking?
  • Who was leading the project?
  • What were there backgrounds?
  • Who made the decision to implement in this manner?

Danny Sands raises the obvious:

"The problem is this kind of information should never be used clinically, especially if you don't have starting or ending dates" attached to each problem, said deBronkart's primary care doctor, Daniel Z. Sands, who is also the director of medical informatics at Cisco Systems.

Indeed.

Personal health records, such as those offered by Google Health, are a promising tool for patients' empowerment - but inaccuracies could be "a huge problem," ["could be?" - ed.] said Dr. Paul Tang, the chief medical information officer for the Palo Alto Medical Foundation, who chairs a health technology panel for the National Quality Forum.

For example, he said, an inaccurate diagnosis of gastrointestinal bleeding on a heart attack patient's personal health record could stop an emergency room doctor from administering a life-saving drug.


And when such an event occurs and a patient is harmed or killed, who then is held accountable - and who is held harmless? (Oh wait ... we know the answer to that question thanks to Koppel and Kreda...)

This "billing data" issue and other EHR issues like it are not rocket science, they are Medical Informatics 101.

I've seen such issues before, such as at "AOL kerfuffle: information technology vs. information science, a distinction lost at industry's peril" and at "On Intel's and Walmart's prescription for Healthcare IT."

I summed the problem up like this at the post "A Biomedical Informatics Manifesto":

Biomedical Informatics as a specialty might as well be invisible. Amateurs** rule HIT.

(** Amateur in the sense that I am a radio amateur, not a telecommunications professional and would not deem myself appropriate to design and run a critical telecommunications project).

Perhaps, though, I should have added "amateurs rule HIT, and even worse are too often managed by incompetents."

I believe Google should conduct a top to bottom investigation of the management chain and the decision making process that led to such a fiasco, which can only further erode public confidence in electronic health records at a time of national distrust in Big Business and Big Medicine.

Those who made such design and implementation decisions without appropriate input from those who know better, or worse, those who might have overridden or ignored such counsel, should be dealt with appropriately. (If it were me, I'd ask for their resignation, but that's my opinion.)


Clinical medicine, Electronic Health Records and patients' well being are not an information technologists' learning lab.

Also of concern to me, this is the type of data our government seems to be touting for use in Comparative Effectiveness Research. (It is also of concern to me in this regard that our new Secretary of HHS was the former Kansas commissioner for insurance from 1994 to 2002, and such billing data is likely where the majority of her experience with medical datasets resides.)

Finally, like the financial schemes of the past decade, I can only wonder when the computational House of Cards that is being built in healthcare as a result of the quasi-religious
Syndrome of Inappropriate Overconfidence in Computing, and worship of its priests, the IT Whiz Kids and consultants to whom domain expertise is optional, will come crashing down.

-- SS

Addendum:

A physician correspondent who wishes to remain anonymous writes (emphases mine):

[The Boston Globe article] could not have come at a better time.

Just today, a spouse had his "home grown" PHR for his wife who was hospitalized with multiple medical problems, including advanced metastic breast cancer and complex vascular disease. He has an elaborate PHR with history, treatments, allergies, medication lists, etc.

It was so impressive that when this 80ish year old patient was admitted, he gave the medication list from the computer to the physicians and nurses. It appeared so reliable that not one health care professional bothered to question it or reconcile it with the labels on the bottles (everyone is so busy nowadays clicking and scrolling the computer silos for information).

As it turns out,
he left out a decimal point on a dose of a potent medication that should have been 2.5 mg. The computer printed a legible list (with other errors too) stating the dose of this med was 25 mg per day (10 times too much). It was ordered that way by the doctors. It got to the pharmacy, but somewhere in this complex chain, a non physician non nurse individual got the dose to the patient correctly as 2.5 mg [fortunately, the error was caught, this time. What about next time? - ed.]

Being a detective with an eye for detail and a stickler for accuracy, I happened to notice the error when the spouse was showing off his PHR to me.

Again, this is one case with potentially dangerous consequences of a pervasive error generated in the PHR by flawed data entry. It was not a Google or Health Vault device, but I cannot believe that these companies have garbage filters on their devices to prevent the "garbage in, garbage out" syndrome.
Good medical care is being subverted by these experimental devices.

Upon scratching the surface of PHR, EMR, and CPOE devices' functional impact on the administration of medical care, the dangers are widespread. This toxicity is covered up from scrutiny by the "non-disclosures" and "hold harmless" contractual obligations described in the Koppel and Kreda report
.

One wonders how many incidents like this happen every day and are being concealed by the HIT industry and the pundits profiting handsomely from selling defective HIT devices. I am quite concerned that nobody really knows. This is not science.

On a final sobering note, as the "hold harmless" and "defects gag" clauses are purged from HIT contracting, which they will most certainly be, I would suggest the many amateurs in HIT obtain some very solid liability insurance covering patient harm related to their systems and their advice.

For they may just find themselves as defendants answering questions on the witness stand in front of a hungry plaintiff's attorney, a jury of average citizens, bereaved relatives of patients who were harmed via IT misadventure, and questions composed by people of my background. These questions will place the true nature of their expertise and qualifications to be tooling around with medical care under severe scrutiny.

That will likely not be very pretty.

-- SS

April 22 addendum:

In comment #15 to this post
Matthew Holt issued this filled with absolutes ad hominem comment ...

Seriously, MedInformaticsMD, you are so pissed off with everyone in IT [everyone? - ed.] that you're now part of the problem [problem of vendors creating bad IT? - ed.] Do you seriously think that the people at BIDMC, Google and everyone else in health IT (even Cerner) just dont give a shit? [I cannot read minds. I can only see results - ed.] Or do you think that they might be trying to figure out how to solve these problems [of course they're trying to solve problems, but good intentions without requisite ability and expertise are inadequate in healthcare - ed.], and perhaps could you some constructive help. Rather than a barrage of attacks on anything they try to do. [Anything? You mean, such as in this post praising Google in areas where they do leverage their expertise properly? -ed.]

Perhaps my direct Chairman of Medicine-after-patient-mishap tone in offering the most constructive of criticism - i.e., don't embark on medical projects in which you are over your head, find people who do know the domain and let them lead, don't release anything in medicine without appropriate, rigorous premarket trials - upset him. In addition to the inserts above, in the comments section I replied:

I'm sorry you feel that way.

I'm not sure what "problem" you're referring to, but if it's harming patients due to badly implemented HIT, I'm certainly not part of that problem.

As just one example, my website on HIT difficulties serves as a resource read internationally on how to best avoid HIT errors, has been online for a decade, and is quoted in one of the newest and best books on HIT, specifically "Medical Informatics 20/20."

Did anyone at BIDMC, Google or Cerner ever read it? Did you ever read it? If not, why not? It is in fact the first link that comes up on a google search on "healthcare IT failure", for example. Do they take it seriously? If not, why not?

I believe they were negligent on this project. This suggests they need to give a bit more of a s--- about their work, expecially since real, live patients are involved and the mistake made was so fundamental.

Finally, see my post "A Software Engineer's Eloquence on Health IT" for what I consider an attitude of someone who really does give a s--- about such matters.

Finally, I see no links to my decade-old academic website on HIT difficulties over at Matthew's blog. One wonders why. It may have to do with a tension between the statement that "
the Health Care Blog (THCB) has acquired a reputation as one of the most respected independent voices in the healthcare industry" and the post "A Shout out to our sponsors."

Healthcare Renewal has no sponsors and does not take advertising. We report, you decide.

-- SS