Beyond the Hype: AI Won't Fix Healthcare Burnout Until We Do This

AI was supposed to fix burnout in healthcare, wasn’t it? 

So why are the latest results in terms of documentation time and time to action lackluster? Why hasn’t AI saved the day after all?

I’ll tell you. 

According to Dr. Díaz here, AI is not proving to be healthcare’s knight in futuristic armor because it has yet to be applied to an appropriately curated group of end-users.

I am not saying that careful thought and consideration has not been applied to date, but rather that the traditional targets for health information technology (HIT) are not what’s needed at this unique moment in history.

For so long we have used terms like laggards (even Luddites) and early-adopters to sort through potential candidates for new HIT.  What I have noticed is that this approach tends to favor the Maddie Make-dos & Persnickety Petes, and I’ll explain not only who these characters are, but why they’re the wrong ones to pick.

Maddie Make-do, as the name would imply, makes do with whatever technology they have at their disposal. They are efficient with using dictation tools, scribes, or even brute-force typed entry. Maddie Make-do is hellbent on leaving work at the office and protecting their homelife; their notes might not be literary masterpieces, but the necessary information is there. If you introduce Maddie Make-do to an AI scribe, they’ll take it in stride. 

Persnickety Pete, on the other hand, almost always brings their work home with them. They need the time to pore over notes and make sure everything is documented just so. Persnickety Pete will not clamor for new HIT, but based on their EHR usage metrics they’ll appear to be most in need of an intervention. 

Though they’re at opposite ends of the healthcare technology spectrum, both Maddie Make-do and Persnickety Pete will show no appreciable difference in their user metrics like pajama times nor efficiency metrics when new HIT is applied because they are, for better or worse, fixed in their ways. A new tool such as AI ambient scribes will not change that. 

At this point, you might be thinking that my suggestion is merely to focus on everybody else. The answer to that is, not exactly. My method is to take the remaining clinicians and create a curated list of potential AI tool recipients using meaningful metrics. These metrics are, of course, meaningful in my opinion. I base that on years of experience and observations. Were I to have the hard data to prove my theory I’d be publishing this in an academic journal instead of opining here on LinkedIn… But I believe my time will come!

So here they are, Dr. Díaz’s meaningful metrics for AI candidate selection:

• Schedule template variability
• High clinician inbox volume compared to back office pool
• High chart prep time compared to that of assigned medical assistant

What you’re going to notice here is that operations play a heavy role in AI tool success. This is why, before assigning AI scribes, I partnered with our operations leaders across the state to ensure these issues were assessed and addressed prior to deployment. This approach will not only increase our chances of seeing significant change in clinician efficiency, it is the right thing to do to ease operational burdens in our clinics. 

In other words, don’t drop an AI tool on a busy clinician whose workday just went sideways because they were expected to inject an arthritic shoulder between annual wellness visits, and expect efficiency scores to improve. Don’t scratch your head wondering why AI didn’t alleviate pajama time when your clinicians are bombarded with inbox messages they should have never seen in the first place. Be aware that if humans are not setting our busy clinicians up for success through pre-visit planning, they’ll be highly suspicious of AI tools that promise to do the same. 

The right candidate with optimal clinic operations will demonstrate the immense potential of AI tools in healthcare. Do not be deterred by the current reports of insignificant findings or even outright failures in AI’s ability to support overburdened clinicians. Seek instead to tweak the deployment of AI in the clinical setting, partner with operational leaders, and you’re going to see success after success until even Persnickety Pete starts to rethink his ways!