from Bing / Dall-E image generator
Opportunity: Opt out of alerts?
We have a new tool from our EHR vendor to allow physicians/APP’s to “opt out” of future similar interruptive alerts about medications.
For example: “Magnesium in IV for pregnant patient” or
“Ibuprofen interaction with __ drug”
Previously, our physicians/APP’s would either take action (cancel the prescription) or override the alert and state that benefit outweighs risk, EVERY TIME.
Now, our colleagues can FILTER or SUPPRESS these warnings specifically or categorically. As an informatics working group, we have some choices on how to configure this.
Won’t you come along and be a back-seat driver for our conversation?
Q1: Shall we turn this on at all?
I mean, we are doing SUCH A GOOD JOB ALREADY (we are top/worst quartile in “overridden alerts” by our users, among our EHR vendor customers).
A: Yes, we are going to turn this on.
Q2: The pop-up alert allows the prescriber to choose:
Filter for ALL patients or Filter for THIS patient. When we suppress this alert, shall we turn it off FOREVER? Or bring it back every 6 months? Or longer, or shorter? On the one hand, we should respect physician decisions, right? And never show it again? On the other hand, who knows how fast that physician is moving when selecting to “suppress everything — I hate all of you.”
A: Well… after debate, we will leave it at 6 months. Not too burdensome to ask a prescriber twice a year whether they still want this to be filtered. That is 363 days a year we stop asking. And, intermittent re-inforcement is more sustained learning (showing alerts sometimes prompts better retention than every-time alerts that become noise and aggravation).
Q3: What about that choice for THIS patient?
Should the alert be suppressed FOREVER if a prescriber chooses THIS patient? Obviously, the incremental benefit to the prescriber is tiny (I have thousands of patients, so if I choose THIS PATIENT ONLY, I MUST have thought about THIS patient in more detail and we should honor that choice with a FOREVER suppression). For example, “ibuprofen in hypertension” alert. Perhaps THIS patient has tried dozens of other treatments and the ONLY help for migraines is ibuprofen, and yes, I know it affects hypertension, and the risk is worth it. Why remind me again? And, because the benefit is small to the prescriber, the THIS PATIENT setting should be infrequently used, and ONLY with thought.
A: We are investigating if we can make THIS PATIENT suppression setting different from the ALL PATIENT suppression setting.
Q4: What if there are particularly HIGH RISK drug interactions?
Can/should we remove the ability to suppress those interactions? How paternalistic should we be? EG: Paxlovid and lovastatin: can we allow this to be suppressed?
A: TBD. We are unsure of the granularity of the tools, and we will investigate this further. This will end up being a consensus decision of our committee, backed up by proposing these major decisions to our clinical leaders for approval. It is always a good idea for operational committees like EMO (comprised of practicing clinicians / informaticists / pharmacists / IT analysts / clinical nurse informaticists) to work together to build consensus. Many of these decisions are NOT slam-dunk, as you are starting to see.
Q5: There is an option to “require a reason” to suppress.
What shall we put there? Should we give several “best guess” reasons, for example:
- Low risk interaction: benefit exceeds risk
- My literature review disagrees with this alert
- Only show this to students and residents. I am an expert
- This is stupid, remove this for everyone (GROSS: Getting Rid of Stupid Stuff)
A: Although the last option is a real project (look it up: GROSS is a clever and memorable acronym invented by other informaticists), this is too many choices. We went with ONE choice “Clinical Judgement – Suppress for 6 months”, because A) No one wants to spend more time on this alert, there will not be much signal here to detect, and B) Suppress for 6 months RESPECTS THE USER by telling them what to expect, and when this will re-appear, to reduce the frustration of “I thought I got rid of this.”
from Bing / Dall-E image generator
Q6: Shall we require a narrative explanation after the choice “Clinical Judgement” override?
Maybe they’ll tell us something useful that we did not expect!
A: NO. NO. NO. In the past, our required clinical fields have often been perceived as Nuisance. Requiring completion results in answers like: “.” and ” “, or even “jkljlkj”. The least number of keystrokes to make the stop sign go away. Or it results in hate mail “This is stupid. You are stupid. ” Burned out physicians/APP’s can easily pass the burnout to our IT and informatics team members. Leaving it optional leads to better signal/noise and more thoughtful comments from the rare commenter.
Q7: Whose job is it to review these exceptions and comments? And how often shall we review?
A: We plan to bring these exceptions back to the EMO committee to look at, perhaps 1 to 2 times year and see if it is performing as expected. Also, if we can spot patterns of the most ‘overridden’ alerts, we can look at removing the alert completely, looking to reduce our need for overrides in the first place.
Q8: Shall we give a choice called GROSS (get rid of stupid stuff) in the suppression reasons?
Why don’t we leverage the wisdom of the crowd to finding the lowest value alerts?
A: Unfortunately, doing so may empower our “go fast” physician/APP phenotype: “I hate all alerts, so I’ll choose the one least likely to show up again”. We worry about these “over-suppressers” who over-estimate their ability to catch important interactions and defeat the systematic protection of the EHR. We do NOT plan to offer GROSS as a choice, but instead will look at “Clinical Judgement” frequency (if a LOT of prescribers choose to suppress, there is signal there, don’t rely on individual prescribers to accurately choose GROSS). A subtle point, but there you go.
Q9: Who do we report these decisions to?
Shall we announce it in the newsletter? Shall we get approval from our busy CMO (chief medical officer) council?
A: Machiavelli teaches us, whenever you make a bad change, do it all at once. Whenever you make a good change (this is a good change; ability to suppress alerts customized to the individual prescriber), TAKE CREDIT FOR EACH CHANGE SEPARATELY. Yes, we will announce this in the newsletter as a separate news item every time there is an improvement. As for authority, we have agreed with our CMO’s to bring, on a quarterly frequency, a summary of the changes the EMO committee is making, to ensure they are overseeing the changes. This will come as an FYI, no decisions necessary, as we have established trust that we are not making crazy decisions, based on past presentations this year.
CMIO’s take? This is a taste of the discussion and decision-making at EMO. Maybe you have an interesting or different take on this informatics work? Let me know!