Each member of staff was trained by way of a compulsory three-hour training session.
December 20, 2005 11:09 AM   Subscribe

Computerized physician system linked with increase in child mortality In an effort to reduce medical errors and mortality a children's hospital implemented a commercial physician order entry (CPOE) system.
After the system had gone live, analysis over an 18-month period revealed an unexpected increase in mortality from 2.80% to 6.57% (about one extra death per month)
It seems the big mistake was Changing the Systems of Work to suit the computer system, rather than fixing the technology to meet the needs of a specialist work area (intensive care).
posted by Lanark (13 comments total)
No real surprise: too many so-called "system architects" focus on re-training users to use Byzantine systems, rather than spend the time and effort to learn what the user really needs to do.
posted by orthogonality at 11:55 AM on December 20, 2005

My wife is a fourth-year medical student. For the past two years, she's worked in probably a dozen hospitals in CT for a month at a time. Since I'm a computer fella, she's described the systems to me, and a surprising number of them have very advanced and time-saving infrastructure. What I think would be more useful is a longitudinal study that compares doctor age to delta-mortality. I'd bet it's simply a case of older doctors having a harder time adapting to computer systems.

I'm not saying "old people are stupid" or anything. I just know that computers take a bit of dedicated brain-power to learn when you're not used to them, and considering the amount of medicine a doctor needs to know and continually learn and re-learn, I would not be at all surprised that that would end up decreasing doctor effectiveness.
posted by Plutor at 12:46 PM on December 20, 2005

it would be interesting to compare this with any studies with longer time-frames.
posted by stratastar at 1:34 PM on December 20, 2005

I dunno, you try something for a year and a half and get an extra dead kid each month, you don't go looking for longer term data. Parents are real funny about dead kids.

Back a few years ago, Healthcare Informaticists -- that's "doctors who want to see everybody use computers in the hospital" -- were of the opinion that systems that caused doctors to change how they worked were Bad. This sort of hard-liner would be in favor of a computerized drug resource because that's no different from looking up something in the Monthly Prescribing Reference, but against a lot of other things. I'm not willing to go that far because I think that there are places where computers can add the kind of efficiency that saves lives. Clearly this system did not, and I must wholeheartedly endorse the same conclusion as the study: "Although CPOE technology holds great promise as a tool to reduce human error during health care delivery, our unanticipated finding suggests that when implementing CPOE systems, institutions should continue to evaluate mortality effects, in addition to medication error rates, for children who are dependent on time-sensitive therapies." Blind faith in a computer is Ungood, and humans with clinical judgement should double check things.
posted by ilsa at 2:20 PM on December 20, 2005

Great post! This is especially fascinating considering that CPOE already has a reputation for being failure-prone during implementation (Cedars-Sinai was the poster boy I believe). I can't believe they installed it in 6 days. From what I understand,these systems dramatically change how clinicians sumbit data (to Plutor's point). At any rate, I've always believed that the reason healthcare IT hasn't delivered on its much-promised promise in the last 30 years is due to competition. That is, healthcare companies (i.e., hospitals, clinics, docs, etc.) do not want to share data, so there are limits to the value of networking. So maybe it's not just a matter of connectivity. This may suggest that the cost side of the cost-benefit equation doesn't just involve wasting money. Yikes.
posted by micropublishery at 2:42 PM on December 20, 2005

Concerning of course, but this retrospective study also screams post hoc ergo propter hoc. I'd like to see the data and find out more about the "mortality covariables" in their analysis. I mean they go about enrolling all these patients and get a massive N value, when in reality, if a CPOE played into the poor outcome of some patients, careful inspection of the medical record on a case-by-case basis would have been far more elucidating.
posted by drpynchon at 2:43 PM on December 20, 2005

...and CPOE is among the industry's hottest trends in recent years!
posted by micropublishery at 2:44 PM on December 20, 2005

Interesting post. I'm an industrial engineer retraining for a career in nursing. I can't help but think that Healthcare Informatics will do for Hospitals what MRP, MRPII and ERP did for manufacturing. It could get pretty ugly: ortho's restrictive byzantine systems, rogue spreadsheets and db's, users who can't spec requirements, IT guys who don't know the operations, outdated documentation, ugly user workarounds while waiting for enhancements, local customizations, managers fearful that automation will dismantle their kingdoms... This might not really be new ground - except for the part about people dying over the learning curve.
posted by klarck at 4:21 PM on December 20, 2005

I left medicine last year, having practiced geriatric medicine for 10 years. One of the primary reasons was an awful electronic medical record (EMR) system.

I worked for a big multispecialty group allied with a hospital. As best I could tell, the EMR system was chosen because (1) it spanned both outpatient and inpatient care, and (2) it could generate lots of different kinds of reports for the administrators to look at.

The EMR (LastWord by IDX) hindered my ability to practice medicine. It was simply a hole into which information was thrown, with no useful organizing tools. It didn't even have the capacity to search for text within a patient's record. For each patient visit, I had to re-assimilate data scattered throughout the record. The EMR did -nothing- to assist the onerous burden of documentation required by Medicare for billing.

Now, I'm honing my programming skills, and hope to become employable and to develop a front end for an EMR that will help providers care for patients. Wish me luck.
posted by neuron at 4:56 PM on December 20, 2005

Here are some notes from Why Things Bite Back--Technology and the Revenge of Unintended Consequences by Edward Tenner (1996)

quotes from the book with [my notes in square brackets]

Technological systems also multiply the opportunities for miscalculation.

There is a [improper?] hierarchy of medical evidence, with tests and imaging results at the top, the physician's direct visual and aural examination in the middle, and the patient's account of the illness at the bottom. [the EMR facilitates this hierarchy] An overreliance on tests can defeat medical common sense.

The shift from tool use to tool management enhances our power over the physical world while reducing immediacy of understanding.

The most fundamental revenge effects of contemporary medicine are systematic tendencies, not the dead ends and errors of therapeutics. The problem of today's medicine and main revenge effect of new therapies, is that contrary to our expectations of technology, the more advanced it becomes, the more it demands in vigilance and craftsmanship.

In medicine, the increased potential hazards of diagnostic and therapeutic equipment, complex procedures, and the possible interactions of drugs require an unusual degree of attention. The proof is the surprising frequency of errors in medical practice.

The great majority of avoidable medical errors go unnoticed.

Because the human body is a tightly-coupled system, in which treatments can make parts interact in unexpected ways, advanced medicine usually requires more rather than less human attention. More and more care becomes intensive, potential complications multiply, and deviations can be fatal. For physicians, new technology requires more rather than less craft. (Where do they get this craft?] For all health workers, it demands more rather than less vigilance.

Chronic illnesses run counter to most of the strengths of technological medicine.
Characteristics of chronic illness (from Anselm Strauss):
• long-term by nature
• uncertain prognosis
• demand relatively more symptomatic relief
• are multiple (tight coupling)
• socially disruptive for patients & require social services
• costly
• encourage unorthodox treatments

Citizens of the developed world have come to expect ever-higher standards of accuracy and protection.

One mark of newer technology is that while it is cheap in routine operation, it is expensive to correct and modify.

Getting experiments and machinery to work takes skills that don't appear in textbooks or manuals. So-called high-technology professionals learn these as artisans always have, by working with masters of the craft, and of course by trial and error.

Even as the cost of equipment goes down, the increased power and flexibility of hardware and software make them take more time to learn and to use.

More powerful hardware, even when it drops in price, does not run itself.

Remarkably little research has been devoted to the effects of computers on the quality of decision-making....There is growing evidence that software doesn't necessarily improve decision-making.

Computerization has helped reduce rather than promote the amount of time that these employees spend performing their highest and best work. Many highly paid people were spending a significant amount of their time performing what amounted to secretarial and clerical functions, usually working with computers but often not doing what they spent years at college and graduate school learning to do.

The growth of engineering as a profession has made a new type of error possible, as Henry Petroski has shown: overconfidence in the safety of a new design, the defects of which too often remain hidden until some new disaster occurs. But there is also a second type of error: failure to observe the repeated rituals that safe operation of advanced technology entails.

posted by neuron at 5:39 PM on December 20, 2005

UCSF has switched over to a new thingie to computerize the entire medical record. For this, 18 new computers were needed on my own hospital floor. They are called 'Computers on Wheels' (COWs for short).

In my epilepsy unit I have the advantage of 24 hour video recording. I can prove that the introduction of these god-double-damned COWs reduced the amount of time nurses spend with my patients by about 60%.

Other things the COWs do is make it impossible to navigate a bed past them in the hallway (which was designed as a hallway, not as a computer work area), and cause the pharmacy drug-toting delivery robot to become stuck and beep loudly (even at 4 AM; I know because my patients EEGs switch state abruptly, activating a computerized brainwave-change detector, and when I review the event, it is an arousal caused by a beeping, stuck drug-delivery robot.)

Unfortunately, no one gives a fuck what I can prove, because I'm a doctor, not a businessman, and "the medical record must be computerized."
posted by ikkyu2 at 6:12 PM on December 20, 2005

These are some selected notes from a lecture I used to give to medical residents on care of hospitalized elderly patients .

[Do electronic medical records help with any of these things? I sincerely doubt that any do.]

(ADL = Activities of Daily Living such as bathing, dressing, toileting, eating, etc.)

Recognize the patient who is at increased risk for complications. Explicity look for indicators: dementia or confusion, poor mobility, needs assistance with ADLs, depression, multiple chronic diseases, multiple medications, extreme age.

Recognize (explicitly look for) risk factors for malnutrition: social isolation, low income, inability to drive or shop, neurologic disease, liver disease, chronic infection.

Maintain a problem list throughout the hospital stay. Include the above risks .

Don't restrict physical activity unnecessarily. Expect five days of recovery per day in bed.

On daily rounds:
Ask the staff about the patient's mobility, intake, elimination, cognition, medication use, behavior, and sleep.
Ask the patient about these things, plus mood.
Assess attention. Look at pressure points on skin. Look at all devices attached to the patient.

In the daily chart note:
Document mobility and function.
List all devices attached to the patient; consider their necessity.
List all medications; consider their efficacy, toxic effects, drug-drug and drug-disease interactions.
Maintain a problem list. Look back at yesterday's problem list. Tell what you're thinking. Give yourself (and others) reminders about what to expect. ("If test is negative, consider ___.")

In daily orders:
Stop unneeded medications and devices.
Try nonpharmacologic treatments. Avoid prescribing meds for minor complaints.
Avoid sedatives, anticholinergic medications, new drugs.
If you start one drug, try to stop another--keep the med list from growing.
When prescribing, consider potential drug-drug and drug-disease interactions.
Look up any drug you're not very familiar with before prescribing.
Dosing: start low, increase gradually ("start low, go slow"). In high-risk patients, meds have a narrow therapeutic window.

Regarding physical restraints:
Restraints are proven to -not- improve safety. Restraints increase agitation.

What to do if:
* malnourished or poor intake:
consult dietician, ask food preferences, assist feeding,
supplement calories, protein, and vitamins, evaluate mouth and
swallowing (dry mouth from meds?), reduce meds
* not mobile or is falling:
assist ambulation, order PT, unrestrain, disconnect from tubes,
stop sedatives, lower bed and rails, evaluate for skin breakdown,
evaluate for delirium, check postural blood pressure, move closer to
nurses' station
• incontinent
mobilize, stop sedatives, check for overflow, evaluate further
• diarrhea
check for impaction, if on tube feed add fiber, replace fluid loss,
stop diuretic, avoid loperamide
• agitated/confused
Assume that it is multifactoral. Causes include medications, infection,
pain, hypoxia, heart failure, uremia, electrolytes, urinary retention,
hypotension, med withdrawal. Rarely is from stroke unless focal signs
are present. Agitation/confusion is facilitated by sleep deprivation,
sensory deprivation, sensory overload, physical restraint. Zeitgebers.
Quiet room, open curtain, no television, appropriate lighting. Get patient
out of room BID. Stop all medications not absolutely necessary to sustain

I'll stop now.
posted by neuron at 6:22 PM on December 20, 2005

A longer article with comments, based on the full Pediatrics article.
posted by Lanark at 11:01 AM on December 21, 2005

« Older The Conyers Files   |   Paper Forest Newer »

This thread has been archived and is closed to new comments