Does computerized order entry increase medication errors?

You might think so reading this article about 2 studies and an editorial in this months JAMA (Journal of the American Medical Association; these probably require subscription access).

The first interviewed house staff (interns and residents) at a teaching hospital about possible errors facilitated by the CPOE (computerized physician order entry) system in use there (which apparently has ~60% market share) and identified "22" ways in which the system could lead to errors.

It is crucial to note that no objective data on actual medication errors or harm to patient was collected and there was no control group. They just asked house staff about potential problems and then went back to ask how often they'd observed specific problems. More troubling, there is no control group of house staff who used non-computerized error entry, so no way of knowing if the many benefits of CPOE (no mistakes due to illegible handwriting, less delay in orders reaching pharmacy, computerized checking of dosage and allergies, etc) outweigh the problems identified. Such studies of CPOE have been done an result in an up to 80% decrease in medication errors.

I don't see this study as very informative. Any doc who has used a computerized order entry system knows they are far from ideal, but previous studies clearly show that CPOE does reduce medication error.

The one big advantage of CPOE is that they can be iteratively improved. In that sense, identifying sources of error is helpful in that the programs can be modified to improve or fix the problems.

A second study, systemically reviewed decision support software which, ideally, helps clinicians make decisions about diagnosis and treatment, and found they were overall disappointing. Not surprisingly, results were better when studies were conducted by the same individuals who designed the systems. Unlike CPOE, such decision support systems are not in wide use.

In general, I think decision support systems try to do too much. I think they would be most useful in suggesting alternative diagnosis for unusual or atypical patients, not in mananging more straightforward patients, which doctors are pretty good at already (ed really? At least we thing we are!).

The editorial is, to my mind, unnecessarily pessimistic about the role of technology in medical care. It brings up some good points, such as that new IT projects are usually initiated by some combination of non-care providers and senior faculty who don't have any sense of how most work is actually carried out and imposed on the actual care providers (e.g. nurses and house staff) without much input from the end users.

It unfortunately drifts into socio-babble:

Clinical work, especially in hospitals, is fundamentally interpretative,
interruptive, multitasking, collaborative, distributed, opportunistic, and
reactive.1, 13, 20, 24-25
In contrast, CPOE systems and decision support systems are based on a different
model of work: one that is objective, rationalized, linear, normative, localized
(in the clinician’s mind), solitary, and single-minded

Well now that we have that straight!

I think medicine, like all industries, will make big strides as younger people more comfortable with technology come into the system. Some peculiar features of the way health care is structured (fragmentation, transient nature of house staff, lack of focus on quality improvement, poor measurement of quality)have delayed the commitment to use of information technology compared to other industries.

Well, off to do some interprative, interruptive, multitasking, collaborative, opportunistic, and reactive work.