< + > Key Ideas on How to Scale AI with UC San Diego Health and Notable

One of the most important topics at the HIMSS conference was the focus on AI in actual practice versus AI in theory.  Plus, a key part of this conversation was does the AI solution scale to the problem or does it just work as a pilot.

In the session “Build, Deploy, Transform: UC San Diego Health’s AI Playbook” presented by Karandeep Singh, MD, MMSc, Chief Health AI Officer at UC San Diego Health, Jeffrey Pan, Director at UC San Diego Health, and Aaron Neinstein, MD, Chief Medical Officer at Notable, we got a heavy dose of what’s really working at scale with AI.  Plus, they share some key insights into the mindset shift that’s needed to really benefit from the healthcare AI solutions out there.

Here are some of the key insights and perspectives they shared during the session along with some additional commentary.

I love 2 elements of this session.  The idea of reimagining is a really good one.  I know when I first implemented an EHR we replicated the previous workflow.  Then, we went back and redesigned the workflow based on what the EHR could do that we couldn’t have even thought about previously.  I think AI is forcing us to go straight to reimagining workflows because of what’s possible now.

I also love that they approached this as a system wide project.  We’re going to have a lot of shadow AI projects that were done by departments and that can lead to all sorts of drama later.

These stats are a great example of being able to reimagine a process.  When a human is auditing something it’s expensive and time consuming.  The right AI technology can make reviewing everything trivial.

I know many healthcare organizations that still live in this scarcity mindset.  To be fair, it’s a hard mindset shift to make.  The above audit example is a good one to reinforce this concept as well.  When we’re human constrained, scarcity is real.  With technology, we can often change from a world of scarcity to abundance.  That shift in thinking is a powerful one that I’ve seen play out across healthcare organizations.  Can we reach out to every patient?  Can we audit every claim?  Can we follow up on every low balance?  etc etc etc.  All of thse are a challenge with humans, but are possible with technology.

The beauty of healthcare is that we’re all impacted by what happens here.  One thing I love about this story is that the AI technology can adapt to the person regardless of age and tech skill.  However, at its core is ensuring that the patient has the best experience possible.

I’ve often told people that if you don’t want to reach me, send a letter.  I barely check my mailbox.  However, I check my email, texts, phone calls all day every day.  Plus, mailing is expensive.  The AI should consider a patient’s preference, but the opportunity to be able to do outreach to 100% of patients is a powerful one.  Obviously, that’s proven out in the amazing no-show and cancellation rates that they described.  When it comes to surgery prep, this translates to a massive ROI.  Plus, many healthcare organizations have capacity issues and long wait lists.  Improving no-shows and cancellations can help get patients seen sooner.

Another major theme at the HIMSS conference was orchestration.  It was great to see this highlighted in this session as well.  Orchestration really has become key.  Plus, deep integration with Epic is powerful as well since most providers aren’t going to visit another system.

What a powerful idea.  How much time has your organization spent thinking about “should we automate this workflow?”

I find that this shift almost happens naturally once you start implementing AI.  In fact, it’s why every healthcare organization should start implementing AI.  The act of doing something expands the minds of your users and enlivens their creativity for what’s possible.  Of course, the real challenge today is that every few months AI is progressing so that what wasn’t possible 3 months ago is now possible.  That’s the most exciting and also challenging parts of what’s happening with AI right now.

Where are you at in this framework of healthcare AI agent adoption?  Have you started linking agents?  Have you seen the power of linked agents?  What seems clear to me is that you don’t generally jump to linked agents.  You have to start with specific agents and then can link them over time.

I think that this is one of the things that holds many organizations up.  They have fear of the “what if it goes wrong.”  It’s an important question to ask and test.  I love this session’s comment about it not needing to solve everything.  Escalating something it can’t solve can still be a win for your organization without putting anything at risk.

This reminds me of the reality that we often want to compare AI against perfection.  That’s always been a mistake.  We should be comparing AI against the alternative.  The reality is that humans aren’t perfect either.  However, we allow much more grace for humans than AI.  That often holds organizations back from benefiting from the improvement that AI can provide.  Don’t throw out the benefits of AI because it can’t solve everything for you now.

Trust really is the key to any AI implementation.  If they don’t trust it, they won’t use it.  This is a nice framework for building trust in an AI solution at your organization.

What did you think of this session on implementing AI at scale with UC San Diego Health and Notable?  Has your experience implementing AI been similar to there’s?  What else would you add to the conversation?  Let us know on social media.