Despite all the investment in digital health and billions of dollars that have gone into it—the apps, the ingenuity that's gone into electronic health records—we’re not yet at a point of being able to understand how to use these systems. One of my colleagues at the Brigham, best-selling author Atul Guwande, wrote an article in the New Yorker recently about why doctors hate their computers. And if you read through this article, there are a number of workflow-related issues that are problematic in and after the implementation of electronic health records. He does a very eloquent job of articulating what some of those problems are, and some of them come down to the fact that you have fragmented workflows that have been adapted digitally, some of them are just associated with the fact that the computer feels like it's in the way between the patient and the provider relationship.
Atul wasn't the only person to complain about the fact that there are opportunities for improvement. Another, more recent, article came out, better research, called Death by 1000 Clicks. And I round every Friday across the entire system at partners, speaking with physicians and nurses and pharmacists, and everybody else that's part of the healthcare team, and I'll sit in the waiting room and talk to patients on a regular basis, trying to understand better what is working and not working for them, and the one thing I do hear from the clinician side is there's too many clicks. So this, kind of, resonated very well, and this article goes through a series of different topics, which I won't go into in broad detail here, but basically, kind of, outline some of the problems that exist with these electronic health records.
I just want to frame a way of looking at this and make the case that electronic health records are just the tip of the spear. There's a number of things that surround the electronic health record that are contributors, whether it's the fact that the data that's being mined from it is to be used for AI is a data hungry things. The fact that researchers need to get more information to many necessarily be useful for actual clinical care. In the U.S. we enjoy an enormous administrative burden, especially from the billing side of the equation, and that gets embedded into electronic health record and requires many more clicks than is necessary.
Several years ago when I would hop on a plane, I would be able to get to the gate about 10 to 15 minutes before my flight left, I could just make it and get on for that flight. These days, I can't do that anymore. Why? Because I'm going through security, I'm partially undressing myself as I go through the _____ [00:02:50] things, you know, I have to take my electronics out of my pocket, and so forth, and so that experience, compared to how it was before, is certainly not the same. And the same is true for the electronic health record, the experience of being able to access, have passwords that are expiring, have different ways that you gave…you know, kind of, re-authenticate, and so forth, caused the ability to, kind of, just go into the system, get what you need, and get out more complicated and more difficult.
One would assume that in a system that now has mostly digitalized healthcare, as in the U.S., that the interoperability between the systems would be solved. But, in fact, that's not the case. And it comes in all kinds of different ways and forms, and we will get into it a little bit, but I'm just trying to make the case that there's a number of barriers to usability that plague EHRs in and of themselves that cause things like burnout.
So, when it comes to the next topic, which would be The Application of Artificial Intelligence in Healthcare, I think it's safe to say that there are some practical applications. You'll notice that you'll see algorithms right now that are being put into papers and research articles that show, for example, that a dermatologist is equal to a computer, in terms of being able to interpret certain kinds of moles for melanoma. You'll see that images of the back of the retina is processed by AI and can as easily identify retinopathy in patients that have diabetes and, in fact, do so in a way that is independent of gender and ethnicity, in other ways that sometimes can hide a condition or a disease. And, you will also see that there are algorithms that show that you can diagnose and interpret a mammogram better than a pathologist that are looking at these things. So, it's just important to be aware of the fact that there are a series of different tools that exist out there that are augmenting today what the clinician can do. They're not replacing what the clinician can do, but they're certainly augmenting that.
In my experience, in terms of things that actually work and things that have actually saved lives, if you take information that comes from a diagnostic imaging report and you're able to extract some of that data from it, you can actually automate some of the ways to go find patients that fall through the cracks, in terms of what's going on with somebody who comes in with a trauma, who gets a pan-scan, CAT scan, they find some other incidental findings that need to be followed up, but then they don't get followed up. And these are patients that could sometimes have very serious conditions like an abdominal aortic aneurism. So we built a whole series of different things that would identify loss to follow-up patients with certain key conditions, we would contact them, in some cases they would have gotten their condition solved elsewhere, and in some cases they were still running around with basically a time-bomb in their belly. We'd invite those patients back to the hospital, perform the procedure, and these are all patients who would have succumbed to that illness if we hadn't gotten to them.
So there are some ways that actual, you know, approach of taken AI and natural language process, therefore, can actually help and, in fact, save lives. That's an important component. I'll give you another example, very practical, of a lady who comes in to your office, who's 86 years old, who feels a little bit dizzy and you're not quite sure what's going on so you go ahead and get a CAT scan. And it's on a Thursday, and you just want to make sure everything's looking okay. That film is an ambulatory film and gets read in the privatization of an ambulatory film. Sometime on Friday afternoon the film shows the patient has a stroke or, even closer to that, the patient has a stroke immediately after they've been examined. Well, in the way that it normally, kind of, first in and first out, that film may not actually be read until it just gets read based on all the work that's being done.
So, what we've done is we've gone through the process where AI actually looks at those CT scans instantly as soon as they come out of the scanner. And when AI establishes a fact that actually as a potential 96% likelihood that this patient has a stroke, that film just gets put at the very top of the reading list. So just reorganizes the stack. To the radiologist, and their workflow, it's nothing different than them grabbing the next film that's there. But to the patient, it becomes incredibly important to go from a four-day down to several minute recognition of that, so then we can do something about it.
I also wanted to mention that the new currency these days it to data. So a patient walks into the hospital, various different components of examinations are being performed, and that outcome, that secondary use of that data, becomes incredibly useful to biotech, it becomes incredibly useful to pharma, researchers, and a number of other people. And that currency, the same kind of importance of the currency of data when it comes to, you know, when it comes to all the other industries that are using that data to sell us things, and so forth, is becoming just as important when it comes to AI.
So, in terms of patient experience, I think the key thing is quality and cost was a key value-driver before expense has been added, and I'm just making the case that adding digital innovation is a very important component that people want access to care, they want it online, they want to know what the cost of the care is before they engage in it, they would like to have it to be personalized, of course, they want a caring team around them. The caring team can be just like an example of a cardiologist, the patient feels very cared for, even if the cardiologist isn't reaching out to the patient, so long as somebody is, even if it's the care navigator. They expect ultramodern technology and all this digital information to be shared amongst other hospitals, and they want to be guided. The hospital is an incredibly intimidating place, when you're not use to going through it, it becomes less intimidating when you're used to it, and I think patients have enough to deal with, with their illness that they don't need the burden of having to figure out how to get there and who to reach out to, and there are ways we can help with that.
Last few slides. New models of care, in terms of virtual, a number of different clinics that can put everything online, you click it, walk, call in, just make an Omni-general view of being able to access care services. A new technology, in terms of consumer relation, management, transactional systems that are missing in our midst, so imagine if you pick up the phone and this is, kind of, a thing that pops up immediately after you pick up the phone, to find out that the patient likes to be called Jenny, she's scared of hospitals, don't book her on a Tuesday because she never shows up on a Tuesday, so offer her appointments on another day, and then at the bottom shows what the experience has been in the past few years, in terms of how she's experienced the healthcare system. Is there anything we can do for her?
[This article has been reprinted, with permission, from Rotman Management, the magazine of the University of Toronto's Rotman School of Management]