A new study published in JAMIA gave us a good look at how EHR design – particularly choice architecture – can significantly influence provider behavior.
University of California, San Francisco researchers investigated several workflows at a UCSF medical center where the existing choice architecture was potentially nudging providers toward waste or misuse.
The first workflow involved ordering “free phenytoin” levels, a costly send-out test that often results in delayed care for patients, despite a readily available “total phenytoin level” being sufficient in most cases.
- The researchers hypothesized that the EHR alphabetically presenting “free phenytoin” before “total phenytoin” to providers searching for “phenytoin level” was influencing them to order the more costly and time-consuming test.
- They then replaced the alphabetical structure with an order panel presented to any provider searching for “phenytoin” that gave an explanation of the circumstances when each test is appropriate.
- They simultaneously nudged providers toward the “total phenytoin” test that is “almost always correct” by making it the default selection. The intervention improved the rate of correct test orders from 92% to 100%.
Another workflow the researchers examined was the prescription of benzodiazepines for procedural anxiety. The EHR originally set the default quantity for benzodiazepines to the same quantity needed for patients routinely taking the medications for a chronic disease, therefore nudging providers to overprescribe the pills.
- The researchers created a new order called “Lorazepam (Ativan) tablet 0.5 mg for imaging/procedure” specifically for imaging patients with a default quantity of two tablets with zero refills.
- The new order included a default comment “for anxiety (prior to imaging study or procedure)” to nudge providers toward the appropriate quantity. This intervention was also successful.
Despite the success at UCSF, the authors emphasized that organizations must balance the potential benefits of any EHR improvements against their implementation costs. The phenytoin nudge consumed six hours of implementation time and the Lorazepam nudge took almost 3x that long, which might make other investments more worthwhile depending on the org.
Although choice architecture is the name of the game for pretty much every product and design team, it’s doubly important when the choices directly impact people’s health. This study was great at wrapping numbers around how this plays out in a medical setting, and it was also interesting to see the cost-benefit analysis that still takes place when deciding whether to implement a solution that clearly improves outcomes.