Pleasure to be here today. I am going to move away from research now to, to talk about the, not only the practical applications of some of this, but even commercial applications of some of this in an area that I pay close attention to. First, SMART and FHIR have both been mentioned but I, I want to make it absolutely clear to everyone what the implications of those 2 technologies are to the practical use of some of the tools and applications that have been alluded to earlier. Everyone here has a smartphone in their pocket, or in their purse, or somewhere on their possession I'm sure. And we all take it for granted that whenever we wish we can download apps on to that phone and that with our permission, hopefully with our permission, those apps can access whatever data is available on that phone. Well, the practical ramifications of FHIR and SMART
are that you can do the same thing with an electronic health record system. You should be able to download apps that a physician or some other care provider, or even a patient, feels would be of value to them, gives the apps permission and they can access whatever data they need to operate. And indeed, that's not only happening, it's happening commercially particularly the United States, where ,I don't have time to go into what the federal government did, but the federal government has essentially mandated that this stuff should happen, it's the law. So all of the major EMR vendors, and many of the smaller vendors now support what I just described. We're going to look at Cerner.
I would be happy to look at Epic, but they won't let you show their stuff. It's kinda a strange company. And I'm going to illustrate some of this using three commercially available SMART on FHIR apps. First is DoseMe, which is Australian, comes from Brisbane. Second is Rimidi. for full disclosure I'm an advisor to the company and on their board. And they're based in Atlanta and the third is Suki, which is in Silicon Valley.
Before we do that, I'd like to outline what I feel, are some of the key success factors that are going to be necessary for this technology to to do what it has the potential to do. The first is we need a predictable data model. These apps are designed to work across EMRs. I can't overemphasize to you the importance of that to a commercial enterprise. You have one product where,
it's not quite that that true, but close to one product, and it can work with all these EMRs. That has enormous ramifications for the cost of operating the company secondly, and this is was alluded to earlier, I think by Sean, but I'm not, I think it was Sean, and I cannot overemphasize important to this. The tools, no matter how good they are, to be successful must be integrated into the work flow and process of the intended user, particularly. if it's a physician user and it can't be duplicate data entry. We've spent 50 years actually demonstrating that if you don't do this you can develop wonderful tools and nobody uses them. The 3rd thing is one of the objectives of, of many of these technologies is to provide clinical decision support.
To help physicians, other care providers, patients, make good decisions, something that they demonstrably don't do as often as we would like them to. The key here is not just to provide the support, but to provide it appropriately. If it doesn't fit this patient, or it doesn't fit the current clinical status of the patient, that quickly generates something that has got a name alert fatigue and it's been shown that providers turn off and don't pay any attention to the alerts even when they are appropriate.
And finally it's becoming increasingly important that we consider it and indeed apply Artificial Intelligence and natural language processing into these clinical tools to make them smarter. Something that you've already heard about, I think some of the points I'm going to make, have already been made, but I'll, I'll remake them, I guess. Well, FHIR provides the predictable data model. I haven't gotten any detail, but, The word "resources", the last of the 4 words in the FHIR acronym refers to the units of that data model which are highly predictable, although not totally prescriptive, descriptions of patients, providers, clinical observations, conditions, medications, and a host of other things.
And an app developer can count on these resources being quite similar, no matter what the underlying EMR is. That's incredibly important. The 2nd is integration of FHIR apps into the workflow and the process. And here's an example of that. This is DoseMe. As I said, an Australian FHIR app running within Cerner's power chart. Here on the left, you see that it's initiated by clicking on a menu item that's in line with all the other menu items that the provider is using to do charting.
So it looks to the provider like part of the EMR, and even looks more like part of the EMR when you consider that the area where they app operates is the same area where charting and clinical review is already done, so not only are these apps integrated into the EHR, they appear to be the EHR. The major difference is the user interface is, often quite a bit better than the EMR. Uand if I had enough time, I would show you some apps whose primary purpose is that. 3rd, and this was alluded to again by Sean I think it was, these apps can get the data they require at launch, using a facility built within SMART so that there is no duplicate data entry. In fact, some of them just come up with the results the provider wants there's no need to enter any data at all. The 3rd is providing critical success factors, providing appropriate clinical support.
So, here, using Rimidi is an example of that, leveraging a technology from the same group in Boston that provided SMART. And it's called clinical decision support or CDS hooks, and this allows the embedding of clinical logic into the process, so that the clinical decision support is provided at the appropriate time for the appropriate patient. Here we see that the whole paradigm is event driven so clinical decisions support is evoked when the chart is open, when orders are written, places when it might be appropriate to provide advice. So, I, I emphasize it not only needs to be appropriate for the patient, but for the point in the patient's care.
And this is real, this is from Rimidi here. A physician is provided with a CDS hooks recommendation based on data from a continuous glucose monitoring device in the patient's home. So, interesting stuff going on here, all of a sudden data from the patient, pr from devices in the patient's own is part of the EHR workflow in process. Something we've talked about for years but now it's real. And this alerts based on the percent of the time, the patient is in glucose range.
And here is an alert provided based on a guideline, in this case, the American Diabetes Association guideline for diabetic care, And it's recommending that the physician order some stuff, because the patient's most recent hemoglobin A1c, a test that measures the control of diabetes over the previous 90 days is higher than it ought to be. And finally, at least from Rimidi, bills all of this into the EHR's order flow. So if the physician clicks the button to order what was just recommended, they're taken directly in this case, this is part of Cerner's order flow, into the. order flow of the EMR, making this as seamless and built into the workflow process as possible. So, we've looked at DoseMe, and we looked at Rimidi now we're going to look at Suki. As the talk, I was asked to talk about AI powered SMART on FHIR apps.
Suki is to health care, or to charting what Siri and Google assistant and Alexa are to everything we do every day. So here the physician says Suki create a clinic note. It creates a note based on what's going on in this patient and because it's learned providers behavior when presented with certain clinical situations.
Provider goes on to say, insert my normal review of systems and it does. And then the provider can say, but I want you to change part of that to something else and it does then, so the provider can switch from giving commands to charting seamlessly and Suki can understand what's going on. It's also used for what I choose to call intuitive data retrieval. So here the providers say, show me Mrs Ramirez's medications.
And it does and provider notices this, let's say they are not familiar with Mrs Ramirez, notices that 2 of them are diabetes medication. So it says, Send me a graph of the hemoglobin A1c over time. And it does. So, how does this work?
I am not a PhD computer scientist, like Michael and other people here. So I'm going to give you a fairly high level view, but that's probably appropriate. So and what technologies it use, well, this is a FHIR. Or I wouldn't be talking about it and it pulls the data. It needs
from the chart using FHIR and SMART. Then it uses natural language processing. 2 specific technologies that the company tells me they use a transfer learning where it applies knowledge gain from elsewhere and training the neural network against the kind of healthcare data sets that the app is going to see. It understands context and intent. This is critically important and you'll see applications of AI and medicine that don't understand context and then they do dumb things.
That, like, chatGPT is is is capable of doing. So, it understands the action that's conveyed, what it is that the provider wants to do. And it has through another technology, the ability to extract structured information from text, keep in mind, most of charting is still text. Talk about that in a bit and then it can push the note to the patient's chart using FHIR, or where it has to the API is provided by the vendor then the provider reviews, makes any changes they want, and signs the note. And they're done.
This is this, this is not the I made 1 change to this presentation, and I didn't get in here. So, let me just talk to it. Pretend like you're looking at a slide from the American Academy of family physicians, the largest physician group in the United States, and a major advocate for digital health. In fact, the AAFP operates an incubator for digital health companies and Suki is their most successful example. They did a study, and they offered a 130 odd family positions, the opportunity to try Suki and about half of them decided to buy it, to adopt it. And the academy surveyed those 60 odd positions
and found that the amount of time they were spending doing documentation was reduced by 72%. I want to make sure that sinks in - 72%. Now, why is that important? Well, obviously, physician time is valuable. Secondly, if you survey physicians about electronic health records, it doesn't matter whether they like their record, or they hate it. Their number 1 complaints going to be "It just takes too much time."
So, what about the future? Where are we headed. There's the slide, I put it in the wrong place. Apologies to David and Marianne. You did give me my, I just I did
this last night. I realized, you know how your brain works, I realized I forgot a slide that it will be in here. So, but I've already covered all that. So, the National Health Service in England.
Actually commissioned Eric Topol the famous physician researcher, futurist, digital health guru, to look at the impact of digital technologies on workforce, genomics, digital medicine, AI and robotics, and there is a link if you want to read the report, it's really interesting to read, at the bottom. And this is an illustration from the report on the likely impact of these technologies over time, with the darker colors. meaning greater impact. We're going to focus for a second, on 1 of the areas that they looked at. The impact of automated image processing using AI, something we've already heard about this morning and they actually did a study and projected that the widespread use of that technology would free up all of the time of 500 radiologists in the UK.
Well, this stuff, this report's a few years old, this stuff has moved from speculation to and reality, and in fact, 2 days ago, 1 of the front page stories in the New York Times was about the success in using the technology to detect breast cancer and the concerns radiologist starting to have about whether they're going to have a job in the future. Now, we're going to look at something else that they focused on which is the use of speech recognition and natural language processing. Now this study was released in 2019. Just 4 years ago, and there's an illustration of how fast things are moving, The study then, this is Eric Topol speaking, said as digital technologies become more prevalent, there is a risk that a deluge of automatically transmitted data will overwhelm health professionals.
Well, and I will stop and say, and if you look today, at the patients who drive most health care costs in the U. S. and in Australia and the rest of the advanced industrialized world, patients with a variety of chronic diseases are seen by the healthcare system with great frequency, their charts are already unwieldy and not, for all practical purposes, usable. So the application of API to generate patient summaries, they provide a clinical useful solution to this problem and I actually think this is 1 of the most important potential applications of the technology. First to save physician time and secondly,
to make things sure things aren't missed. Well, this is 2023 and lo, and behold, there is in New York, not far from where I live, a new startup company called Abstracted Health. spun out of Cornell Tech, the new technology Institute built in the middle of the East River on Roosevelt island. This company is less than a year old, and it it focuses on, currently is focusing on discharge summary, something that physicians manually write. I can tell you, I'm a physician and physicians hate doing this. When I was in training they wouldn't give me my paycheck until I had done my discharge summaries. The only way they could get us to do it.
And we already know that physicians complain about spending too much time doing charting anyway. And there are statistics that suggest that for one, each hour patient care. physicians spend 2 hours in her electronic health record. So could you automate the production of a discharge summary, which tends to contain a history of the present illness, a summary of the daily course of care And follow ups, and the most complicated part of that is the middle part, the daily narrative? So the company says they use encoder decoder sequence to sequence transformer models. to get sentences that summarize each day of the care and put it in chronologic order.
Well, that's all cool. Does it work? Well, it currently, besides this company is not a year old yet, there summaries as evaluated by physicians - currently 62% of them meet the physician's definition of sufficient quality to be clinically useful. And are they going to get these slides? So Quick quiz, uh, it's going to be an easy quiz. I'm going to give you the answer, but we have time to to do it.
Which is which? I can tell you, if you read these carefully, it's less than obvious which is which. One of these is the discharge summary written by Abstractive Health and the other is a discharge summary written by a physician. In fact, 1 on the left is done by the computer and was evaluated by a panel of physicians on a scale of 10 has 8 on these 4 criteria - quality, readability, factuality and completeness. This is the 1 done by the physician and it ranks slightly better, but not that much better. And in fact, what I've highlighted here are the major, it's hard to know why the physician ranked higher because, there's only 1 place where this is noted in there.
Well, I guess that's right the physician's rank higher, because it's only 1 place. I got that backwards and the main, criticism On both sides was that there were things mentioned And there was no detail given, not that there was an error, just didn't go into enough detail. Something that I assure you the computer can get better at. So, how do they do this? Well, I want to thank the CEO of the company who's worked with me to get this in a language that almost anybody should be able to understand. So, at the heart is not GPT3. but a similar tool developed by Facebook called BERT.
There are 2 reasons why they use this and not GPT3, is BERT was designed to summarize text. If you use Facebook, the news summary you can see are done by BERT. Secondly, it's open source, which means the company can operate it, which is important if you're talking about putting protected health information into the thing, you can't just have it anywhere. And the open AI people who they did talk to about using GPT3 refused to make it open source and refused to sign what in the U. S we call a business associates agreement, which you have to have if you're using protected health information. So, for practical reasons, they're using Facebook's BERT and I can't help but point out it was trained on CNN, that's how it learned English, And it was trained on health care data and that's how it learned the specifics of health care data. This is how this stuff's actually done.
And then uses something called Constrained Beam Search. Michael might know what that is, I really don't, but I know what it does. It tells the model that there are certain words, it must be included in the output.
Think about it. That's pretty important. Then, down here, it uses google's BERT, which is sort of the Swiss army knife of natural language processing. It's good at a lot of things. And in this case, it's good at figuring out what's important and identifying follow on care, I'm meant to point out there are 3 parts of the discharge. I mean, the last part is follow on care and that you can't forget that that has to be.
In the summer, and of course, then there's the most important technology at all. FHIR. That's how it gets the data from the EHR, and that's how it puts the summary back in the EMR, this thing is, in fact, a FHIR app. This is just all the technology sits behind it, the physician just clicks on the app and they get the discharge summary. Quick commercial for my book, if you want more, if you want, my book has about 40 odd examples of of real world applications of FHIR and FHIR apps. If you're interested in that sort of thing.
And if you have an, listen, if you have an academic affiliation, it's free. Can't beat that, just go to Springer link and your institution. You can download the book for nothing. Thank you very much.
2023-04-07