“The Pen is Mightier than the Computer” — Medical Technology, Politics and the Database Problem

A brilliant post about medical record-keeping by Michael O’Connor:

When it comes to Computerized Health Information Technology (CHIT), also known as the Electronic Medical Record (EMR), the pen is mightier than the computer.
Why? Because regulators and billing professionals seek comprehensive documentation, and believe that more information generates a clearer, more useful picture of what is happening (and has happened to) a particular patient. Malpractice attorneys and quality experts lust for this level of detail, as it will afford them the opportunity to point out the myriad failings of the health care system, and serve as a perch from which they can direct the continuous improvement in the quality of care. Patients, anxious that critical elements of their medical story might be lost or unappreciated, are anxious that every caregiver have complete and total awareness of the details and trajectory of their medical history. While noble, this aspiration is part of the problem with CHIT, and perhaps the major obstacle to its being a solution to any problem in medicine.
The relentless quest for higher resolution of detail has driven a relentless increase in the detail provided. Unfortunately, the coding available is often a poor fit for the clinical information (a mild dilation of the aorta classifies out as an aortic aneurysm, the former something that bears minding over decades, the later a potentially life threatening medical problem that commands close follow-up). Worse, much of this coding is generated by administrators remote from the bedside, and who typically are deprived of the information required to code accurately. The imperative to code something, anything, invariably trumps accuracy, and little inaccuracies creep in to the documentation in droves at this point. Please note the shift in language from record to documentation in the last sentence. Only outsiders regard such documentation as containing useful information about a patient; you will likely never meet a healthcare provider who has this view. You will never hear ‘Could you please request Mr H’s medical and billing records from his hospitalization at memorial hospital?’ Not gonna happen. In fact, practitioners know that there is more noise than information in this documentation, which is why they do not and have never had any interest in it. It is almost certainly the case that the cost of improving the accuracy of this documentation far surpasses any benefit that might accrue to the patient. The fantasy that you can monitor the quality of health care from this perch, or improve it, is, well, a fantasy. This has not stopped major players from falling for this, hook, line, and sinker:

Read the whole thing.

I know little of medicine. However, it strikes me that O’Connor’s post is an excellent explanation of how information-gathering systems tend to fail unless they are designed with a strong idea of what information is needed, and with careful attention to the incentives created (intentionally and unintentionally) for system users. The general problem is that bureaucratic incentives tend to encourage collection of as much data as possible, regardless of accuracy or utility for practitioners, while databases tend to become decreasingly useful as their scope increases and errors increase. The extreme case is something like the government’s “no fly” list, which is heavily seeded with inaccurate data and does not provide much if any benefit for all the hassles it causes. Medical databases designed by bureaucrats rather than doctors are likely to have similar problems, and O’Connor says that medical practitioners now avoid the electronic system in favor of a “shadow” version of the traditional medical chart.

The top-down imposition of database-driven information systems on medical practitioners is a promised feature of the Obama administration’s health-care reform scheme. O’Connor’s post gives a hint of how destructive such politically driven “reform” might be.

16 thoughts on ““The Pen is Mightier than the Computer” — Medical Technology, Politics and the Database Problem”

  1. “You will never hear ‘Could you please request Mr H’s medical and billing records from his hospitalization at memorial hospital?'”

    If this is really true, then it’s a problem that needs to be addressed. In today’s society, it is very improbable that all of a patient’s healthcare will forever be done by the same physician. And it doesn’t seem either safe or efficient for each new physician starting care for a patient to begin with zero context, or with only the context available from the patient’s memory.

    In aviation, all significant maintenance on an aircraft is required to be logged, and mechanics do make use of these maintenance records.

    Creating the right incentives for doctors to provide records with meaningful detail may be difficult…frankly, I would have thought it would already have been considered as a serious professional responsibility…and establishing coding systems to meaningfully capture this information in machine-readable form (above the pure-text level, so that statistical analyses can be done) but is surely a worthwhile effort. I do think there’s probably a danger that coding standards intended primarily for billing may wind up playing way too big a role in the computerized medical-records systems.

  2. I’ve seen this problem is business systems. The core problem is the drive to create discrete labels for different degrees of a phenomenon that occurs on a smooth spectrum.

    The dilation of the aorta is a good example of this. The diameter of the aortas in the population varies in a smooth gradient. There is no standard diameter which represents a problem for everyone. Even if there was, the observation might simply be visual with no objective measuring device involved. Also, different diameters might represent a different degree of risk in different patients.

    The best way to deal with problem is to use “fuzzy logic” in which each labeled phenomena has a numerical weight attached to it. So an entry for a mild aortic dilation would be “aortic dilation 1” while a dangerous one would be “aortic dilation 10”. This helps but does not eliminate the problem.

    The real problem with detailed documentation is that the time it takes increase exponentially with the level of detail. Let’s say you have a system with 100,000 weighted labels. To access that, you set up a menu system of 10 menus, 10 menus deep. So to find each individual label you have to click through ten levels of menus, then click to add a weight. So each recorded label takes 11 clicks. With ten entries that’s 110 clicks assuming no mistakes or backtracks. Now multiply that for every patient.

    People forget that it takes time and resource to generate and process information. The more information a system captures, the slower the system works. So, the attempt to improve care could harm it if the information slows to the point where people don’t get treatment in a timely fashion.

  3. “You will never hear ‘Could you please request Mr H’s medical and billing records from his hospitalization at memorial hospital?’”

    Uh, that’s not true, I mean about the medical part. Doctor’s and nurses want the medical records from prior hospitalizations, if it’s pertinent to the case at hand, and ask for it pretty routinely. It’s a pretty standard part of taking a medical history, actually, in sicker patients in hospitals, especially transfers. You definitely want the record.

  4. Okay, I see now reading through. My specialty is different and in my specialty, it is extremely useful for me to see things in an EMR. I work in a lab and look at patient biopsies and excisions – so, instead of trying to page a busy doctor with a question regarding the biopsy, I find the EMR much more efficient. But, in a specialty where you regularly chart, the EMR may be a pain. I still don’t understand – the doctors I work with say almost the exact opposite, in fact my critical care-cardiologist brother would love to have the data and says he’d love a more unified electronic charting system. Just goes to show you, one size fits all doesn’t work in medicine.

  5. Strike me that the med record has at least 4 different functions:

    1)for the doc’s own use, to remind him of stuff when he sees the patient later
    2)for use by *other* docs who may see the patient later
    3)for use by medical researchers doing studies of disease spread, drug effectiveness & side-effects, success rate of various treatment strategies
    4)for billing and audit purposes

    ..different applications may require different formats: for instance, scrawled text notes are probably fine for (1) and neatly-formatted text notes for (2), whereas (3) and (4) can benefit greatly from a more structured approach.

    (Park Street, does the above seem right?)

    Re the “shadow” charts being maintained outside the formal system, this kind of thing is fairly common in businesses that suffer from elaborate and overdesigned/misdesigned information systems…in many manufacturing environments, people are using simple Excel spreadsheets to do tracking that they don’t feel they can rely on the ERP systems (usually bought & installed at costs well over $1MM) to do things in the way they want it done. Indeed, some LEAN consultants have celebrated companies that threw out their computer-based information systems completely and went with Gantt charts on the wall!…one such company *is* using the Internet in its scheduling, but only to broadcast the pictures of the wall chart so that they can be seen throughout the company!

  6. David Foster,

    First off, I’m kind of embarrassed because I posted before I read the original link! I think the writer of that piece sure knows more about this than I do, although, I am still confused because I’ve heard such different things about EMRs. The one constant, though, is that documentation is a major pain.

    “Anyway, I think you are absolutely correct on those categories – EMR means different things to different people.
    Indeed, some LEAN consultants have celebrated companies that threw out their computer-based information systems completely and went with Gantt charts on the wall!…one such company *is* using the Internet in its scheduling, but only to broadcast the pictures of the wall chart so that they can be seen throughout the company!” Now, this just makes me laugh, because it seems that there is always some new fad in my business that is supposed to make everything magically efficient and wonderful. Our LEAN meetings seem, well…..I won’t say much more.

    I think the problem with EMRs is that you purchase a big system, the user uses it, finds something they would like to tweak, and the tweaking doesn’t happen. You need to tweak stuff in practice, and the big commercial systems just don’t allow it, it seems.

    Also, the old paper charts had a lot of low-value junk in them too – you just flipped to the best note-writer, and read his or her notes. You looked for the high value stuff and ignored a lot of the poorly written notes. Some of the problems may have to do with the poor writing skills of docs, in general, too, so that even if you free text, it’s a poorly written note. Witness my comments :)

    I wonder if the real complaint is not EMR, but the kind of mindless documentation that is part of an overly bureaucratized system? EMRs and the reg have risen together.

  7. I have spent 30-some years reading medical records as an expert witness in med-mal cases, both for defense and plaintiff. The recent adoption of EMR, especially for nurses, has generated tons of useless chaff because the risk managers got hold of it and pounded all real information out of it. Now you get a set of perfect practices logged in mind-numbing detail and there is no human input. I always read nurses notes first to see what was going on when nobody else was looking. That source of real info is now gone from hospital charts.

    There are extremely powerful applications of EMR, for example in trauma centers and ICU where decision support systems can write default orders that are based on lab and biometric data. For example, Intermountain Health Systems about 15 years ago did a study that accidentally showed how powerful EMR can be when used correctly. This book probably has the story:


    The story is that they were trying to decide to buy an extracorporal membrane oxygenator (ECMO) for patients with lung failure. These are not end stage emphysema patients but young trauma victims that get what is called ARDS (Adult respiratory distress syndrome) as a consequence of massive trauma and shock. The mortality rate at the Mass General (considered the best respiratory care in the country at the time) was 85%.

    Putting a patient on ECMO cost $100,000 for the first day. That is the setup cost. Every day after is less but the minimum is high. They decided to see how good a job they could do without ECMO by standardizing care. That is a scare term in the present healthcare debate but for a respiratory ICU it was a goal. They set up a decision support system that was a computer program based on a series of “if-then” decision points. It wasn’t even fuzzy logic, just a fairly simple algorithm. They came up with consensus guidelines from a committee and allowed individual modification by signed participants. You could modify the orders for fluids, drugs, respirator, etc., but you had to sign each order. Every few days, they met again to assess progress. If changes got better results, they changed the algorithm.

    By two weeks, they had established a standard protocol and the computer was writing the orders, making changes hour by hour as needed. By one month, the mortality rate of ARDS had dropped to 45%, the best in the world. That was without ECMO, the purpose of the whole study.

    Most medicine, the scientific part anyway, consists of algorithms. A medical center in MInnesota reported an experience with a “learning system” that predicted the probability of positive coronary angiograms based on pre-exam data from Physician Assistants in rural clinics. Initially, there had been too many referrals to the city and Medicare was complaining. They set up this program, once again algorithm based, and cut the false positive rate to about 10%.

    A lot of this is going to be human nature at work. Something like the difference between Microsoft and Apple. I was a programmer before I went to medical school in 1961.

  8. Michael Kennedy,’

    I think there is a qualitative difference between a focused collection of data intended to clarify a specific phenomena and a shotgun collection of every considerable piece of random information that floats through a system.

    Rational ignorance is the goal. We have a finite ability to process information so we have to choose what we pay attention to and what we ignore.

    The problem with the imposed standards like the EMR is that (1) it has so many interest wanting to collect information and (2) no clear focus for the information collected. This will end up with a system swamped with imprecise information. Bad data is worse than no data at all.

  9. Oh, I agree. I was pointing out that there are good, even great, applications of EMR. The usual application is junk. When I joined the faculty of UC Irvine about 15 years ago, the medical center pharmacy did not even have the simple drug conflict and dosing error software that every corner drugstore had. They had a primitive EMR system and complained that surgeons didn’t use it. Only 17% of surgery residents entered their orders on the computer while 80-some % of medical residents did so. I pointed out that the practice patterns of surgeons and internists explained that. A surgeon treats with a knife and an internist with a pen. I suggested that they incorporate a system of default orders, something like the Intermountain example. If a surgeon could set up standard orders for pre-op admission and post-op care, it would be a huge labor saver and would prevent errors. It could even be linked to the lab and other admission data like body weight, sex, age etc, all information that may adjust doses of drugs. If they did it right, they could get the surgeons using the system 85% of the time. All they needed was to make it user friendly and to contribute real value.

    That got nowhere. I moved on to another medical school a year or so later.

  10. “Oh, I agree. I was pointing out that there are good, even great, applications of EMR.”

    I agree with this Michael Kennedy. EMR is like all technology – it’s use can be productive, counterproductive, or kind of ‘null’. Depends on how you use it. The problem in big teaching hospitals is that you get a one sized fits all approach, that, guess what, doesn’t fit all. Fantastic comment.

  11. Well EMR system has some advantages and disadvantages like all new technologies have as time progresses these systems developed very well. I agree with the point of view of Michael O’Conno that EMR system lacks many parts which helps doctors. But i believe EMR system is the future of medical field.

  12. As the family member of patients who have spent a fair amount of time in hospital, I can tell you that the vast majority of attending docs and nurses don’t seem to read much of, well, anything. They are standing their holding a 6″ file and still ask you “what medications do you take” – after you gave the previous nurse, earlier that same day, a photo copy of the entire list.

    I blame the HMOs who don’t seem to have a way to pay for people to actually take the time to read the file and think about it – instead docs get paid for “procedures” and no one (at least under 60) takes the time to think.

Comments are closed.