It is my impression thus far that EHR products have not been written by clinicians for clinicians.

They've been written by programmers for administrators who make purchasing decisions.

Once the sale is made, the product deployed, and a patient database created ... everybody's stuck with the product, like a bad marriage. Why?

It's too difficult to leave one EHR product for another. Nobody's happy, except for the financial beneficiary, and it's too expensive to divorce.

Data migration cost + new product cost = prohibitive cost

POSSIBLE SOLUTION:
Standardized data constructs. If data migration cost can be eliminated, and clinicians can change EHR products at will ... guess what?

The EHR industry will become very interested in making products that practicing clinicians will actually want - and at much more competitive prices!

Even as empirical evidence is presented to prove that current EHR and CDS technology does NOT improve patient care, the ability to accept that fact is difficult for technology advocates to swallow. It seems the prejudice that assumes this next great idea simply must produce universally positive results precisely because it's such a great idea is practically impregnable.

Even as the damning evidence is presented, excuses which hold out hope are always immediately provided. The transition is in its early phase - implying that if we just hold the course, things will get better. Greater attention should be given to monitoring the implementation so that the potential improvements can be realized - reasoning that improvements are certainly achievable (after all, they must be), if we just try a little harder. To quote the study directly, "'In the absence of governmental impetus and standards, current adoption patterns may have fostered incomplete implementation and use of less effective technologies,' they write". I would argue that it is precisely the government impetus itself which has spurred adoption of technically immature solutions yielding these results.

This study is actually mild compared to another on this site which indicated that in clinical tests, medical errors were missed 50% of the time - and 30% of the time when such errors would have proven fatal. Hardly the trend we are repeatedly told will happen as we transition to EHRs and CDS.

Let me make clear, once again, that I believe the inevitable march to computerization of medicine is indeed, inevitable, (if for no other reason that the political and economic momentum is irresistible at this point), and that it will eventually be a very good thing. But it is not all good, and it is certainly not a panacea. It has high costs in both economic and professional terms, downsides, problems (many) and it will not yield its most attractive (and therefore highly touted) potential positive result, that is, improvement in patient care, for some time - especially with the technology that is being bought and installed during our current 2011 "blue light" special, funded by the American taxpayer. (I'm not against government incentives - I'm insisting we get our money's worth.) So in that sense, I also agree that we continue to move our country down the path towards EHR/CDS implementation. But I believe the path we should take is not the one we're currently on.

Why are these problems just now coming to light? Why are we so surprised? Because this study focuses on CPOE, the weak point of this whole migration to computerization. All the hoopla has been about the benefits to be realized after the data is entered into the system. Little attention has been paid (by the advocates, that is) to the relatively slow, inconvenient, even tedious and sometimes downright aggravating process which gets the data there in the first place, all of which must be carried out by providers at the point of care - it's CPOE after all. Because this part of the process is largely excused, ignored and even belittled by EHR advocates, those who get to input data will continue to do so grudgingly and with what seems like (to the EHR advocates ) ineptitude.

The fact is, we have not yet developed systems or even standards to develop systems which accurately mirror the information they are required to record in the patient care setting, and therefore we do not, for the most part, have training regimens appropriate to achieve excellence in data quality. Furthermore, the data required is more than the provider needs to support their own ongoing care of the patient, yet the value of the additional data now collected is unreimbursed by those who benefit from it. The economics of the process are out of whack.

When it comes to CDS, there is a curious assumption that automated processes to reason and catch errors are usually better than human judgment. Yet we've forgotten that we are just now, literally, just now, getting to the point where a computer can perform as well as a normal human being at something as common and ordinary (to us humans, anyway) as playing Jeopardy (I refer to this week's articles about IBM's Watson computer).

While algorithms may excel when dealing with elementary processes like filling in IRS 1040's or traffic tickets, what we're attempting to do in medicine is to replace the learned and practiced judgment of some of the brightest and most highly educated people on the planet, dealing with the complexities of the entire human anatomy and physiology from the cranium to the (ingrown?) toenail, spanning development and disease literally from before the cradle to beyond the grave, encompassing everything which might go wrong (and what should go right) therein. Add to that the continually changing, on-label/off-label, side-effect ridden, medical contra-indication laden, drug interaction prone world of prescription drugs! It is no surprise whatsoever to me that an experienced medical doctor, left to his or her own naked, technology stripped abilities, can practice better medicine than an algorithm driven decision support tool that probably runs on Windows 7. We're not talking about Watson anymore. For crying out loud - Microsoft Word still can't even do a 100% accurate spell and grammar check. (I ran this comment through it - any mistakes are therefore not my fault).

To give an analogy: Teddy Roosevelt sponsored the largest technological project up to that time in history when he put his considerable political shoulder to the completion of the Panama canal. Initially, he and all the advocates of that project were so focused on the grand benefits to be realized that they were practically obsessed in a rush to "make the dirt fly" as soon as possible. It wasn't until the urgent project nearly failed, at great expense and even loss of life, that the new project manager took a step back and realized the whole endeavor had to be rethought, retooled, and approached from a completely different angle. Only then did they succeed in creating the 20th century's first great technological miracle.

They eventually realized that the project wasn't just bigger than other canal projects, and it wasn't just more expensive. They eventually realized that it was UNLIKE any other canal project ever attempted. Applying only what they had learned before would not suffice. They had to gain an appreciation for the truly unique challenges that this project presented, and match their techniques and solutions accordingly. I believe the current rush to create a nation-wide medical information exchange system is perilously similar right now to the "make the dirt fly" phase.

The analogy fits in many ways. We are often reminded that other countries have surpassed us in the digitization of healthcare. It is a sort of national embarrassment. But we forget that our system is still better than theirs, not in every way, but in many many ways, in no small part because of the decentralization and non-standardization of practice - factors which make it harder to computerize at a national level, but which provide greater variety and innovation in care. Our healthcare system is therefore unique, and in trying to match their systems only to fit a technological model, we may be sacrificing the benefits of our current, admittedly messy (in terms of data and standardization) healthcare delivery model. This thing is trickier than politicians realize.

I hope I won't be mistaken yet again as a naysayer just because I don't believe that a strategy of using current flawed technology designed for smaller, more localized, non-interoperating applications, and applying it all over the place a grander scale as fast as possible will work.

Apologies to all for the length. I'm no good at bumper stickers.

Patrick J. Casey, Consultant