William A. Hyman
Professor Emeritus, Biomedical Engineering
Texas A&M University, w-hyman@tamu.edu
Read other articles by this author
EHRs have the capacity, to varying degrees, to manipulate data that they have and to present that manipulated data to the user in a way that is presumably believed to be helpful. Integrated Clinical Decision Support (CDS) is an example of such data manipulation in that the CDS uses patient attributes contained in the EHR to generate reminders, suggestion or advice to the user. But CDS is not the only way in which an EHR can use data to control and alter what is presented. This was made clear in a discussion by Dr. Michael Chen, writing on the blog KevinMD.com.
Dr. Chen addressed the issue of EHR presented patient medication lists, which he notes might carry a tacit assumption of actually being correct. Moreover computerized physician (or provider) order entry (CPOE), which is required as a core measure under Meaningful Use, would presumably help populate such a medication list. It turns out that at least one EHR calculates when the use of a particular drug should have concluded based on the order date, the dosage, the frequency and the amount ordered. If this calculation concludes that all of the drug should have been by now consumed, it deletes that drug from the current medication list. While there may be documentation somewhere that explains that this is what happens, the EHR itself is apparently silent on the subject. The problem is that for any number of reasons the patient might still be taking the drug because they did not, in some way, follow dosing instructions. And at least some drugs may have been prescribed on as as-needed basis but this calculation assumes that as-needed use was at the maximum permissible dose and frequency. Dr. Chen suggests that the code creator could do the math, and write the code, but did not understand the realities of drug prescribing and drug taking.
The automated appearance of an EHR generated list may also fall under an observation by the FDA that electronically presented information can, at least for some, give an appearance of veracity that is actually undeserved. This observation occurred in the context of Medical Device Data Systems (MDDS) which in general receive, move and store medical device data. (While sounding something like at least some aspects of an EHR, an MDDS is something else, which can, using circular reasoning, be demonstrated by the fact that an MDDS is FDA regulated while an EHR is not.) Interestingly, Dr. Chen reported the opposite effect, noting that normally he doesn’t trust the EHR because he has been jaded by bad design. This may be an attribute of bad design that I have not previously considered–a design so bad that you know the product has to be used with extreme caution.
A physician I discussed the medication list issue with pointed out that the reliability of such lists, however generated, has always been a problem. However true this may be, an EHR that generates a list based on unconfirmed and arguably incorrect assumptions is not using the health IT revolution to solve the problem but is instead making it worse. This physician also noted that his EHR had a required medication reconciliation check box, which he and others regularly check even when they might be aware at some level that reconciliation is a tricky business and that their knowledge is in fact incomplete. This is one example of the fantasy of data “validation” in which users are asked to “confirm” data about which in general they have no independent knowledge. In some cases such validation comes closer to confirming that the information is not obviously wrong than it does to confirming that it is correct. In this regard it is good to remember that the provision of check box confirmation may be to more to protect the EHR vendor than the patient or the user, by allowing them to say, after the EHR does something strange: well, you accepted it.