Automated detection of LASA medication errors

Look-alike/sound-alike (LASA) medications – also referred to as sound-alike, look-alike drugs (SALAD) (1) — have been a thorn in the side of healthcare professionals for as long as I’ve been a pharmacist.

Many solutions to the LASA problem have been proposed, including Tall Man Lettering (2), physical separation of look-alike drugs, printing of both brand and generic names on packaging and storage bins, use of colorful warning labels, and so on and so forth. The problem with all these solutions is that they involve humans. Working in acute care pharmacy has taught me over and over again that as long as humans are involved there will be errors.

Technologies can help. Automated carousel technology and robotics can help manage physical separation of the medications and eliminate visual bias generated by human eyes. Bar-code scanning can certainly aid in identifying medications correctly. Bar-codes don’t care that medications have similar names, they’re either right or wrong.

In a recent article by Rash-Foanio (3) et al. the authors use an algorithm to flag potential errors from LASA drugs when an order meets the following criteria:

1. a medication order is not justified by a diagnosis documented in the patient’s record
2. another medication whose orthographic similarity to the index drug exceeds a specified threshold exists
3. the latter drug has an indication that matches an active documented diagnosis.

In the study the authors perform a retrospective analysis to identify errors that involved cyclosporine and cycloserine. The algorithm wasn’t perfect. Sixteen orders involving unique patients were found. Additional chart review of the errors discovered that 5 (31%) identified by the algorithm did not involve a medication error, i.e. the intended medication was correct. However, the algorithm correctly identified 11 (69%) LASA errors.

While it may not catch all LASA errors, it seems that EHRs should give AI and some deep learning serious considerations for items like this. Preemptively catching greater than 50% of LASA errors is better than catching zero. (5)

——-

1. I came out of pharmacy school having learned the phrase “sound-alike, look-alike drugs” (SALAD). At some point it changed to look-alike/sound-alike (LASA). Not sure when, how, or why it changed, but them’s the breaks. Just go with it. Adapt or die, I suppose.
2. I’ve never been a fan of tall-man lettering, and it isn’t even clear that it works to reduce errors.
3. Rash-Foanio, Christine et al. “Automated Detection Of Look-Alike/Sound-Alike Medication Errors”. American Journal of Health-System Pharmacy7 (2017): 521-527. Web.
4. Kondrak, Grzegorz, and Bonnie Dorr. “Automatic Identification Of Confusable Drug Names”. Artificial Intelligence in Medicine1 (2006): 29-42. Web. 28 Apr. 2017.
5. Honestly, one of the simplest things we can do is force providers to select an indication when ordering LASA medications. That simple act has the potential to significantly reduce these errors.