Copies of 3898 electronically faxed prescriptions from multiple outpatient clinics to a single pharmacy chain operating in three states (31% Florida, 22% Arizona and the remainder Massachusetts) from a 4 week period were reviewed. The researchers were also aware of the EHR system being used to fax the 'scripts. A two member clinician panel reviewed the all the prescriptions for errors and, if any were found, a second two-member clinician panel confirmed them and identified the type of error.
Types of errors included omission of - or an unclear description - of the treatment duration, dose, frequency or route, as well as a mismatch between the directions, dose and written frequency. Examples included vague directions (like "take as directed"), conflicts (like "take 200 mg. twice a day" appearing with "take 400 twice a day for one day, then 200 mg twice a day"), and doses that were just plain dangerous (high levels of narcotics mixed with potentially toxic doses of acetaminophen)
48 copies were illegible, leaving 3850 available for study. 452 or 12% had errors and 165, or about 4% of the total, were classified as having a potential for patient harm. Nothing immediately life-threatening was found. Some EHRs appeared to have a higher rates of errors than others, with error rates ranging from approximately 5% to 37%. Only one system had zero prescriptions with a potential for patient harm.
The authors point to a considerable body of literature that document that these kinds of errors seem to be equally common among written and electronic prescriptions. What was surprising however, is that some EHRs seemed to have more than their fair share or errors.
The authors recommended that EHRs have "forcing" functions (nothing happens unless all the fields are completed), electronic decision support back-up ("are you sure you want to do this?") and calculators (that reconcile the number of pills prescribed with the instructions for use). As for the physicians, the authors recommend that docs carefully scrutinize the prescribing system when they select an EHR.
The DMCB is not surprised by the results of this robuststudy. This (and the other studies quoted by the authors) should make the EHR weenies think again about their insistence that computers will fix all that ails health care. As EHRs become more prevalent in health care settings, it may be appropriate to start directing quality (QI) analyses. Last but not least, physicians need to closely examine the "hold harmless" clauses in their EHR vendor contracts.
Posts
2 comments:
As an EHR "weenie," what this study describes seems to be a 'Model T' EHR. Firstly, nothing should be handwritten, and even keystrokes should be limited, to either the first few letters of a drug name or to an annotation comment box (that only clarifies an order). Everything else, i.e., route, dosage, frequency, indication, drug allergies, etc., should be from dropdown boxes (or a lookup box in the case of drug name, after the physician types in the first few letters). And, as the authors point out, forcing functions, calculators and, of course, the ubiquitous decision support function should be part and parcel of the EHR system. In other words, A.I. should be a driving force behind each and every field of a 21st century EHR – as two brains are better than one!
All good points but this study examined multiple EHRs that are currently in use. I recall on EHR that I relied on in the past and it was too easy to mix the number doses prescribed with the text "Sig" field. It seems so simple, but it appears we should take nothing for granted.....
Post a Comment