I received an email from Jonathan Perlin, commenting on the yesterday’s post on CDS and “Stupid Questions”. It was based on his pioneering experience with CDS, has helped me to better understand ways in which an EHR can assure that CDS does not devolve to being “stupid questions.”
I am fatiguing of hearing “alert fatigue” as rationale against CDS. That said, the pre-flight checklist (which would be a great advance) is a little bit strained. That process is highly repetitive and predictable. The number of changing variables around a patient are much greater. Thus, from a management perspective (especially in VA), I tolerated some agitation about alert proliferation. The end result was that physicians, other clinicians and informatacists got together and developed much smarter alerts, such as nested lipid alerts that wouldn’t trigger three times for the patient with hyperlipidemia and diabetes and vascular disease. The sort of work Jim Walker is doing with anticoagulation blending best practices in risk communication with pharmacogenomics to deliver truly “personalized” decision support and, thus, medicine is a great harbinger of future direction
I take these ideas from Jonathan’s note.
- Contextual specificity is an important way to avoid alert fatigue. This goes all the way back to an early IOM meeting where the Daniel Masys reportedly said, “current [medical] practice depends on … autonomous individual practitioners for classes of problems that routinely exceed the bounds of unaided human cognition.”
- The recent work on checklists is valuable and illustrative in terms of the fundamental value of decision support and the difference between fewer good rules versus many rules. At the same time it draws attention away from the ability of EHRs to provide contextual specificity.
- Jonathan’s experience reinforces my perception that the development of CDS rules is a community activity with a social interplay between two distinct roles: advocates who finally decide on the rules to be put in place vs. users who feed back to the advocates using every social means conceivable, not all of which are always sociable. Primarily, though, their feedback is measured by observing their actual use of CDS.
I continue to wonder how we apply these big-institution lessons in the context of a small practice implementing EHR largely under its own steam. Contextual specificity depends on the availability of structured data and occasionally may drive documentation templates towards being more detailed. More important, without the healthy tension between the advocate and user roles progress may be limited to those cases where those in the user role can clearly foresee how a CDS rule directly ties to their ability to get paid more or participate in an accountable care situation. It is not that the users don’t want to do the best they can, it is just hard to create focus absent these externalities.