I mentioned before that all this fancy schmancy policy requires glue together these systems to talk to each other, so that whether you go to the doctor's office or the hospital, people have your information. So, it's been interesting to watch that, what sounds like a technical issue, interoperability has become a policy, initiative, and imperative. In fact, the Macro Legislation talks about promoting information, the actual regulations being used by the ONC talking about promoting interoperability. So, what is this issue of interoperability? So, let's start with a book, it was published in 1992 called ''Computer-Based Patient Record.'' So, if you go back to equinology and get a sense of this, like in the ear and the Hospital Information Systems, people are beginning to think about, the record is a real integrated thing. They're thinking about who needs to communicate with this information infrastructure? They came up with all these different sources of data or clients of data. I won't read through the list, but I will ask you to think about who or what is missing from this picture by this August Institute of Medicine 1992, we'll get to who's missing in a minute. So, over several years, I gave students in a course of mine this picture, or a less pretty version of this, and I told them to go out on the web and look for products that connected any two people or any two agents or stakeholders in this picture, and I loved the pixel that came out of this. So, what I got was a system of pairwise beginning. We found where they found, an application or product for every possible combination of entities. There are 13 entities on this screen, you know that the number of pairwise connections is on the order of thirteen squared over two, so when you go to 14, 15, 20, you're talking about a lot of different products. We have, at Johns Hopkins, a 150 different products just inside Johns Hopkins. If we needed a connection for every pair of systems, we have 150 squared divided by two, which is a very big number. So, that's not great, and yet the vision that the 1992 people had was limited. So, here's a vision that's much more current, that shows the mix between institutional connectivity, which is what you saw in the previous slide and consumer conductivity, "Oh yeah, the consumer." Look back at that slide and you'll see no patient, no consumer. So, this is Chris Gibbons and colleagues that came up with this scheme that in the future only 5 to 10 percent of data will be impatient, and as you go on, you'll see that the vast majority of data will be consumer oriented. So, that scheme that we just saw, the spaghetti scheme of connecting institutions just won't work. But wait, there's more because here are the people who were left out, the patient, the instrumented home, the wearables, the retail sites that I mentioned before, Internet of Things in the upper left, the built environment itself. When you have smartcars running around and you have highways that are instrumented, they're not just tracking the cars, they're tracking you. So, there's a lot of data going on and a lot of data being mixed up, and how do you make sense of all this stuff? We tend to think about the cloud, right? I'd said the Internet of Things tend to think, "Okay. I will put all the data from these places on the cloud." That's just not a useful metaphor because it kicks the can down the street to be dealt with later on. I'm here to tell you that it's a bit more complicated, and let's see how it's complicated. So, the stack. No surprise that if you want interoperability, you have to think about the stack, and in fact interoperability requires connection at each level. There are names for different connections at each level, and the're even standards that are used at each level. So, we'll just kind of go through the top and the bottom now, and we'll go into more detail in later sessions. So, let's just think about the top of the stack. So, if you're trying to connect two organizations, what is it that connects? How do you get interoperability between organizations? I think one answer is that the way one organizations talks to another is through a contract, a data use agreement, a business agreement, something legal of that nature. So, policies is what's shared at the world level, we saw what the WHO had to share, contracts, and agreements, and that is at the organization level, there were also use cases. So, if I'm talking about caring for patient, what am I talking about? When I gave you that little prevent illness, treat illness, manage death sequence, I was actually giving you a use case for clinical care. If two organizations agree in the use case, they can start talking about how to interact. At the bottom level of the stack, you start getting into functional standards, but what do we mean by a prescription? What do we mean by an admission? These sound straightforward, but in fact unless the two entities, the two organizations don't agree what they mean by those functions, then they can't share information. At the bottom of the stack, we get two things that sound more informaticsy and more IT. So, let's go to the bottom, we talk about transport protocols, right? That's the Internet. The Internet is a technology level protocol for sharing information at the technology level. As you go up, you need to have data standards, so both sides know what you mean when you say S-O-D-I-U-M, oh, you mean sodium. Higher up is exchanging data, making sure that I can send and receive, and in the highest level there you have business processes Which are a bit more detailed in those functions like admission and prescription that I mentioned before. I already mentioned that the ONC cares about interoperability at the policy level, and there's a whole big book about this, from 2016 is still their vision and strategy guide. You can see that the goals that they have for interoperability, the top three of them have nothing to do with IT whatsoever, right? So, we have advancing person-centered Health, and we talked about person-centered versus patient-centered versus provider centered, right? We talked about transforming healthcare delivery, that's the advance payment models and such, and then fostering research, and that's that learning health system cycle we talked about in the first module. Then finally, they get to enhancing the nation's health IT infrastructure, and we go right back to the 1992 picture, we realize this is the bottom of the stack called the national health information infrastructure. So, if you want interoperability at the world, I'm not going to read this, but these are things that the ONC lists as things that they need to be worrying about, some of them have IT buried in them like electronic case reporting, others the electronic is not necessarily part of the use case like Send a Summary of Care, you could send it by mail, doesn't have to be electronic. Then the components that they have to support all that those higher levels, and I think you can see that we begin going further and further down the stack, I haven't shown the pictures for you. I think by now you're getting a flavor of how this all works. Now, interoperability is something that policymakers loves to talk about, it's something that informaticians loved designing for, I'll simply close out brief sojourn through interoperability with pointing out that there are at least these issues that just don't seem to be solved. It's not clear who makes money out of interoperability. Unless you have a really effective and efficient alternative payment environment like an HMO, or like an ACO or whatever, where there is real value in sharing data. Many hospitals and systems don't want to share their data with you, because you're a competitor and I don't want you making money over my data, so that's a real problem. The second is, if I have a shared record and I see no allergies listed, does that really mean you have no allergies or might it be that somewhere tucked away in some office somewhere and some doctor didn't elicit that you have an allergy, but never made it through to the whole system? So, what point do we trust that the absence of information means no? Nobody has that. The third issue is this notion of fine-grained information. It's great to have a code for sodium, but when I send the sodium level from point a to point b, the reason why I got it, the reason why I might trust it, the reason why I might not trust it, the contents for that lab value, those could be lost, and we don't have a great way of communicating all that extra information along with the core information. So, the bottom is that we don't have a complete set of standards, the joke goes. Here's a standard to combine all the other ones, you already now have now n plus one standards, and my path where I started from needing to integrate standards together. So, interoperability is a great example of a dielectric and informatics where we need it, we have a sense of what it is, we have a sense of how it would work, and it's really really tough to get it to work and it's going to be a source of permanent employment to some of you, and a source of exacerbation to others, and a source of delight in others of you when you actually get the information altogether.
