“… it’s not only transforming how we see the world but how we participate in it – from the factory floor to the meeting room.”

Satya Nadella

“This… is going to be far more pervasive and powerful than anything else. If one central company gains control of this, they will become more powerful than any government and be a god on Earth.”

Tim Sweeney

Those are quotes from a Forbes article back in 2022; BUT, they are not about artificial intelligence; no, I used ellipses to hide the fact that they are about the “METAVERSE.” I included them to make the point that we might just be a tad skeptical about AI changing everything about the world, especially given that nearly everyone talking about AI is selling you something.

What got me thinking about all this was being given the assignment to write an article addressing this question: “Can AI help vaccination programs move faster and further than the diseases they prevent?”

So I went to the Big Brains at STChealth to see how they are putting AI to use in their jobs… Not in some future world, but right now.

LOTS OF LITTLE WAYS AND ONE BIG WAY

We started by visiting with Lacey French, head of product, who seemed generally optimistic about the future of AI in her field, while finding ways to put Microsoft’s AI system, CoPilot, to work now. “We are always asking ourselves,” she began, “by asking how can AI help our team be more productive? For example, we’ve used it to create slide presentations. I’ve certainly had experiences where it created useless PowerPoint decks, but also others where it did provide a starting point to save time, such as for an internal training session where we were teaching a specific product methodology. It quickly outlined the fundamental principles that could then be built upon.”

 Lacey French

But Lacey lit up when talking about something more mundane… and perhaps eliminating the mundane is the short-term beauty of AI. She said, “It can be helpful with capturing key meeting takeaways by recording the meeting and transcribing it. The accuracy on who is speaking in a meeting has continued to improve; it’s a bit uncanny how well it can figure out who is saying what… a bit unnerving and awesome at the same time. Let’s say you miss an hour-long meeting; it will not just provide the transcript, but you ask for key highlights as well. So you can review an hour-long meeting in minutes.”

That notion of saving time by having AI sit in for you also came up when we spoke with John Mendez, STChealth’s Director of Software Engineering. He said, “I’m often part of long message threads. I’ll come back from a meeting and I might have 40 unread messages and I can ask ‘What have I missed?’ and it summarizes.”

John Mendez

And speaking of summarizing, John gave us what might be the best way to think of what AI can do for us right now…

“There’s this assumption that it’s going to replace people; no, it’s there to assist. It saves time on the task; it doesn’t do the task. It might save you five minutes on some task. But it you have ten tasks that day, and you save five minutes each…”

And unlike the endless verbiage from AI, John knew when to stop and simply shrug, thus offering us the opportunity to think for ourselves of the implications.

THE BIG WAY

John also offered us an understanding of how AI is helping coders code: “We’ve had a handful of developers trying it out. It’s in a code base and runs in the background and starts to understand how you work. It can predict what you’ll write next, learning as you go. It learns your style and takes away the mundane tasks. It will do the boilerplate code that you used to have to do.”

Here was something that seemed familiar and we asked if it was like those email programs which resemble an old married couple… continually finishing sentences. John replied,

“It’s better. It gives you options. And if you don’t like the options it offers, you can hit Escape and keep typing. Or, you can say what you want the code to do and it will write. It’s the coolest thing ever in the developer world.”

Even so, when speaking about the enthusiasm for AI among developers, John described how he recently asked two developers their opinions of coding with AI. “One shrugged and said, ‘It’s okay.’ No enthusiasm. The other lit up, saying, “It’s amazing. I can’t work without it.” Asked what was the difference between these two, John suggested, “It’s how the person thinks. Some people are annoyed when predictions keep popping up – it interrupts their thinking.”

OTHER BIG WAYS

We sought out another regular user of AI, Sawyer Koops, Director of Data Science, who leads efforts to develop the IIS of the future. “We’re still in early discovery,” he said of AI, but is already putting it to use saving bits of time by minimizing mundane tasks. “The most common use case is around de-duplication.” (That’s the time-consuming but important job of figuring out if two people in a database are really just the same person.) Sawyer continued: “Over many years we have developed a long series of business rules – you have to develop code for every situation you could encounter. You see a weird pattern and you code for it; it becomes a known-unknown. But with machine learning we can now deal with the unknown-unknowns and take some work off of users: We have a queue of people who may be in the system or not and we can send the people in that queue off to machine learning and it will come back and say, ‘Here’s what the model would say to do with those people.’ Then you can accept that answer or not.”

Sawyer Koops

So once again we see that AI would not replace human work, but rather, boost productivity. One fact to remember: AI is NOT taking over responsibility; that, as John explained, is still an entirely human endeavor: “I hear people say, ‘The AI got it wrong.’ No. It’s still up to you.”

Sawyer ended with a big-picture view of his team’s AI efforts:

“We want it to be a force multiplier. We know that our clients are facing budget issues, including Covid-related cuts. Our goal is to help them do their jobs better and what we don’t want is some system where they have to hire more people just to figure out how to use it. We want to remove drudgery and let clients put those time savings into what they really want to do – improve public health.”

---

STATS OF THE MONTH

Next Big Thing: Improving Immunization Outcomes?

By Bill Davenhall, Geospatial Advocate

Last month  (June 28, 2024) the Advisory Committee on Immunization Practices (ACIP) issued a recommendation to the CDC to advise to all healthcare providers and patients over 75 years of age to receive the RSV vaccine, especially those having any chronic medically diagnosed conditions. In fact, they have suggested that people over 60 with any one or more chronic health conditions should also seek immunization. According to the US Census Bureau there are about 23 million people (see table below) in the US over 75 years of age in 2023, and that is expected to grow an average 24% by 2028. Fifty percent of these 75+ people are concentrated within fewer than 10 states.

Medical analysts at Medicare (CMS) also use data from their Chronic Conditions Warehouse called the (CCW) that reveals that the majority of these beneficiaries will also have one or more chronic disease – which make these beneficiaries  ideal candidates for many of the lifesaving vaccines that will be approved for use – especially the RSV. Analysis of chronic disease among older adults by the National Council on Aging (NCOA) reports that  95% of all adults age 60 and older have at least one chronic condition, while 79% have two or more.

The need to accelerate the use of all the data associated with immunizations, from the vaccine manufacturer to the medical and pharmaceutical  providers that provide the immunization, and the information system technologies must now strive to leverage the data they already have available to directly communicate the risks and benefits to the beneficiaries regardless of what geography they reside in directly. Would you agree?

Here are the data that are  immediately available today about the population over 65 most vulnerable to RSV:

#1.Medicare Claims data submitted for payment by health care providers to CMS by licensed health providers in every State seeking Medicare reimbursement for medically necessary care, and diagnostic tests and treatments. In the Medicare world there are two acceptable claim forms (aka UB-04 & CS-1500, used by hospitals,  physicians, and  other health providers), which collect all the information necessary for the medical information analysts to determine the presence of any one of the 10 most popular chronic conditions. Essentially a Medicare claim submitted to CMS can’t get paid otherwise! You can see the type of data that is routinely collected by Medicare on claims submitted to them in the hyperlinked reference above.

#2.Personal Health Records data – digital and electronically stored personal health record data exist almost everywhere – especially in hospitals and doctors’ offices who are now required to submit all such claims electronically. Any healthcare provider that has a Patient Portal system has an electronic record of their patients available, so they too can know immediately if  a patient over 65 has one or more chromic conditions, although there are still many “speed bumps” in how individual hospitals and physicians use their Electronic Health Records to populate Personal Health Records of a patient. Emergency facilities operated by large regional health systems have considerable experience delivering  information to patients through personal portals. It’s not uncommon however to have “too many” personal portals that are not interconnected!

#3.Immunization data – Robust immunization registration systems have existed for decades in most states. Almost every person immunized over the last decade has contributed data that would be useful in attempting to accelerate the delivery of a vaccine to any medically vulnerable  group, such as persons over 75 with more than one chronic condition. If fact, “…its nearly impossible for older adults to avoid becoming a chronic disease statistic,” according to the NCOA research. Accelerating the sharing of this type of information across all  administrative boundaries for use by patients and health seeking consumers would seem to be a useful service and greatly appreciated by those that receive immunizations in a variety of public health jurisdictions across the United States. 

#4.Population  data – The US Census Bureau has created a reliable data stream that offers access to social, economic, and age specific counts by a host of useful geographies, (such as Census Tracts, Congressional Districts, Health Referral Regions, Zip Codes, Counties) so  knowing where the population over 65 are living, geographically, and how many, is easily and immediately accessible.

Working with the four data streams mentioned above over several decades, the actual data these systems collect, manage, and distribute, would have few barriers to the engineering of a personalized notice system capable of advising such patients, like me,  of my need  to consider the RVS vaccine sooner rather than later.  As a “stakeholder” in my own health, I can attest that after receiving immunizations in three states within the last 10 years, only one health system has sent me a personal message about my need (as a multiple chronic disease patient) to consider any immunization. I would suggest that lack of data is not the problem or the capability of the IT that manages it, but lacks of beneficiary  input about what might be a more personal and useful system for managing my immunization outcomes 

Take a look at the table below and see where the 75+ population in your State will likely be in 2028! You might be in for a surprise!

I always appreciate a 2^nd opinion!