Subscribe to The Podcast by KevinMD. Catch up on old episodes!

Join Shrujal Baxi, an oncologist passionate about AI, as we delve into how artificial intelligence is revolutionizing colorectal cancer detection while navigating the complexities of prior authorization. Discover how AI streamlines screening, reduces administrative burdens, and enhances patient outcomes in the face of health care policy barriers.

Shrujal Baxi is an oncologist.

She discusses the KevinMD article, “Removing workflow burdens and improving access for colorectal cancer screenings.”

Our presenting sponsor is Nuance, a Microsoft company.

Together, Microsoft and Nuance are leveraging their rich digital technology and advanced AI capabilities to tackle some of health care’s biggest challenges. AI-driven technology promises to revolutionize patient and provider experiences with clinical documentation that writes itself.

The Nuance Dragon Ambient eXperience, or DAX for short, is a voice-enabled solution that automatically captures patient encounters securely and accurately at the point of care. DAX Copilot combines proven conversational and ambient AI with the most advanced generative AI in a mobile application that integrates directly with your existing workflows.

Physicians who use DAX have reported a 50 percent decrease in documentation time and a 70 percent reduction in feelings of burnout, and 85 percent of patients say their physician is more personable and conversational.

Discover AI-powered clinical documentation that writes itself. Visit https://nuance.com/daxinaction to see a 12-minute DAX Copilot demo.

VISIT SPONSOR → https://nuance.com/daxinaction

SUBSCRIBE TO THE PODCAST → https://kevinmd.com/podcast

RECOMMENDED BY KEVINMD → https://kevinmd.com/recommended

GET CME FOR THIS EPISODE → https://earnc.me/h850pJ

Powered by CMEfy.

Transcript

Kevin Pho: Hi, and welcome to the show. Subscribe at KevinMD.com/podcast, and get CME for this episode by clicking on the CME link in the show notes. Today we welcome Shrujal Baxi. She’s an oncologist, and today’s KevinMD article is titled “Removing workflow burdens and improving access for colorectal cancer screenings.” Shrujal, welcome to the show.

Shrujal Baxi: Thanks, Kevin, I’m happy to be here.

Kevin Pho: So we’ll talk about your article in a bit. Let’s start by sharing your story and journey to where you are today.

Shrujal Baxi: Sure. As you stated, I’m an oncologist by training, but I really started out very interested in public health and health care delivery, and how we could improve patient outcomes by changing the way we actually deliver care. So my research when I was in academics was really focused on using data and technology to solve those types of problems. I transitioned out of academics about six years ago now, and I have spent time in industry really trying to find the space for technology to do just that: either make it easier for doctors to be doctors, so that they can do what they want to do, which is take care of patients, or find ways to bring solutions directly to patients so that they’re getting better outcomes. Now I work at a company called Iterative Health as the chief medical officer, and at Iterative Health we’re trying to see how computer vision can improve insights and efficiencies from endoscopies, which are predominantly performed by gastroenterologists.

Kevin Pho: Perfect. So tell me about your journey into industry from academics. I talk to a lot of physicians who make that transition for a number of reasons, but tell me about your particular journey.

Shrujal Baxi: It’s a really good question, and it was almost a serendipitous opportunity that presented itself in late 2017. In oncology we were really at the precipice of the game changer, which was immunotherapy, and I was trying really hard to develop large databases within the context of an academic center that could do the kind of real-time insight generation that I thought the space needed. It turned out that while I had the fundamentals from a clinical perspective, and knew what the data needed to look like, to really ingest it, process it and make it ready for research required a team of people that I couldn’t get access to in academics at the time.

I happened to be partnering as a consultant with a company called Flatiron Health that had done just that. They had figured out how to do tech-enabled large database development using EHR data. I actually joined Flatiron in 2018, thinking it was going to be a two-year internship to really understand the nuts and bolts of what it was. I got there and was just turned on by the speed and the pace and the cross-functional nature of what was happening. So in a really practical sense, I sort of stumbled upon the journey from academics to industry.

But when I look back and reflect on the kinds of grants I was writing to develop technology and implement it at the point of care, what was really interesting is that there are a lot of great ideas, but taking those ideas and actually commercializing them, or making them available to the population, requires a set of skills that most academic centers neither teach nor support. To do research to write a paper is very different from doing research to actually change care. And that was really what tech and industry allowed me to do, which was actually fulfill what I was trying to do in the first place.

Kevin Pho: So how did you get the training to go into industry? Was it on the job? Did you take any courses for that transition?

Shrujal Baxi: I think by background I was always a data scientist before that had a name, because I was always thinking about what data looks like, and that was my first foray into tech. It was an easy transition because I was already doing outcomes-based research as an academic clinician, and now I was doing outcomes-based research in an industry environment. What that transition from my academic hat to the industry hat allowed me to do was still be a subject matter expert in oncology, which was natural, and work in the part of health technology where I was most comfortable, which was large database research.

What I learned, though, was all the other stuff: How do you run cross-functional teams, what does product-market fit look like, how do you generate research that is relevant throughout the drug development life cycle as opposed to only at the very end? So it was understanding the business side of it, as opposed to the academic side, where you get to ask the question that’s most meaningful to you. I learned how you ask questions, because asking the question is fun in and of itself, and how to do that quickly and effectively. I don’t have an MBA, I don’t have a business background, but I have had a crash course in everything from funding to financing, scaling, selling and building. Those are all on-the-job learnings that I’ve grown into over time.

Kevin Pho: So let’s talk about your KevinMD article. We talk about colorectal cancer screenings, and your article is titled “Removing workflow burdens and improving access for colorectal cancer screenings.” Tell us, how did this article come together?

Shrujal Baxi: I think the article really started as a question. Given where we are as a company, trying to advance technology and AI in endoscopy, seeing barriers put up for clinicians to actually be able to provide endoscopies to their patients hit really close to home, because we want to make it better, easier and faster and broaden access. And that decision, while I really appreciate what they were trying to accomplish, felt like trying to solve a problem with a sledgehammer, and we were trying to highlight that in that conversation.

But in the context of writing that paper, you have to take a step back and ask, what is our role as an AI company in the endoscopy or care delivery space? It allowed me to reflect a bit, and what I realized as I was writing it is that as a tech ecosystem in health care, we do need to do a better job of highlighting what we’re for, and where the value of what we’re doing is going to be generated. Sometimes the value is that it’s just the right thing for patients, and if it costs a little bit more, it’s worth it, because the patients are going to benefit and it’s a long-term return on investment. When you talk about colonoscopies as a screening or diagnostic tool, oftentimes that is where the return on investment might be.

But then we thought about what we’re also developing that helps providers with efficiency in care, or effectiveness in the care they’re delivering, such that it’s one less thing for them to worry about. And it actually turned out that what we’re building, polyp detection software, hopefully helps them feel a little bit less stressed when they’re doing their procedures, so that it’s just one less thing for them to be worrying about in this environment. So it’s a long answer to your question, but it started out as a gut reaction to what was happening in the payer space and became a bigger reflection on what our role is in gastroenterology and endoscopy.

Kevin Pho: So for those who didn’t get a chance to read your article, just give us the context of what the issue was and how your solution approached that.

Shrujal Baxi: Yeah, I’ll start by saying that our solution doesn’t actually solve the problem. It’s part of the broader conversation, right, which is that you want to make sure the right patients are getting the right procedures, to enable access for the right people at the right time and save the system money, because we know as a country we spend a lot of money on health care. The article talks about a decision that was made at the time by UnitedHealthcare to consider requiring pre-authorization for diagnostic colonoscopies. We work very closely with the GI community, and they were all very concerned that this type of administrative burden only added to what is already a series of choices being made in the space that are leading to burnout, right? We were putting forward the idea that the intent might be right, but the application across the board, writ large, was probably overkill. How do we get to a place where payers and providers can align on policies that benefit the payer by containing cost, but also benefit providers by not increasing their administrative burden, which is already quite intense and leading to record levels of burnout, which we talk about all the time?

Kevin Pho: So these are diagnostic colonoscopies. Can you contrast that with a screening colonoscopy?

Shrujal Baxi: Yes. Anything described as screening implies that you are doing a routine evaluation for something. So a screening colonoscopy is done to evaluate an average-risk person for early detection of colon cancer. That’s a screening colonoscopy. Most people will need a screening colonoscopy once every 10 years, and the recommendations say to start at 45. But if you are found to have something, if you have a reason to get a colonoscopy, some symptom, now it’s diagnostic. There is a purpose to it; you are looking for something.

Kevin Pho: So for instance, if someone had a fecal blood test that was positive and a subsequent colonoscopy was offered, that’s now considered a diagnostic colonoscopy, correct?

Shrujal Baxi: That’s exactly right, because now you’re evaluating an abnormality.

Kevin Pho: So when a health payer adds an obstacle to ordering that, like prior authorization, how does that affect both the physician and the patient?

Shrujal Baxi: It’s a good question, and I’m sure there are people studying the impact of policies like this. I do think that for physicians, part of it is that it’s just yet another thing, right? I happened to train when we went from paper to the EHR, and that was probably the single biggest leap in care delivery in my career, but it was catastrophic in terms of the disruption it caused in workflow. So I think of every next thing that might seem small to an outsider looking in. Why are they complaining? They just have to fill out some forms. But if you can imagine that everything carries that next burden, and that next burden, it’s leading to this feeling amongst practitioners: Is it really worth it? It’s so hard. Why should being a good doctor be so hard?

I feel like that part of it for physicians can’t be underestimated, because of the number of physicians who are leaving practice because they just aren’t sure they want to keep doing that stuff. You rarely hear it’s because of the care or the patients. It’s this stuff that is burning people out, right? So do I think this would have broken them? I don’t know, right? But there might be a physician out there who says, “That’s it, I can’t anymore.” Those things are coming rapidly, and they’re just tired.

It also costs practices, right? Many GIs are still in private practices, or even large academic centers or large community-based practices, and you have to pay somebody to do those administrative tasks. So now you’ve added another layer of overhead, and one expensive part of our health care system is how much we have to pay just to get paid, right? I think that’s also very disheartening for physicians.

And then from a patient’s perspective, you have a symptom, your doctor’s telling you that you need to get this done, and now you’re going to wait because the insurance company denied it. Emotionally, as someone who’s getting a little bit older and thinks about my health, I take for granted that when I need something, it’s there and available. But it’s clear that these hurdles are just making it a little bit harder, and for the wrong patient they might actually dissuade them from seeking care. And that all matters, right?

Kevin Pho: So you’re in the tech space, and you mentioned AI. What are some innovative approaches to alleviating that administrative burden that you’re seeing in that space?

Shrujal Baxi: Right. So once we wrote that article, the interesting part was how we think about AI and its role to play in medicine, right? The one that I think is really sexy, that everyone likes to talk about, is that we’re going to see things a human can’t see: pathology, radiology, endoscopy. We’re going to pick up patterns faster or in a more standardized way. A lot of what computer vision does is directly pass information that the AI can generate to the physician, to either give them confidence in what they’re doing, or provide insights that a tired radiologist or a tired gastroenterologist at the end of the day might not pick up, that they would have picked up at the first read of the day, right? All of that is really sexy. It’s really hard, though, because that work requires an incredible amount of validation, because you are now putting that AI into the physician’s psyche and their delivery of insights, knowledge and care, right?

Kevin Pho: So as it relates to colorectal cancer screening, how exactly would that work?

Shrujal Baxi: Oh, OK, that’s a great question. For those who’ve never had a colonoscopy: When a physician does a colonoscopy, you’re in this room, and they have a camera at the end of a long tube, the endoscope. As it’s inserted into the colon, the camera projects a view onto a screen, almost like you’re playing a video game. The physician is going to advance the endoscope all the way to the end of the colon and take a look around for abnormalities, polyps, maybe some areas that are colored differently, things like that. They’re going to look along the way for the abnormalities, and the things that look suspicious they’re actually going to biopsy and take out, so that you have pathologic confirmation of what they saw.

The way that AI works in this space is that it’s running right along with the physician. So as you’re looking at the computer screen, the AI is going to generate a square on the screen that says, hey, I think there’s a polyp up here, hey, I think there’s a polyp here, there’s a polyp here. And the physician can say, “You know what? Thanks, AI, I don’t think so,” and make their choice. It’s not making the decision. It’s aiding the physician, to make sure they don’t miss something, or maybe it picks up something and the physician says, “I need to look a little closer here, I didn’t see that,” right? It’s just computer-assisted detection. It’s not diagnosing anything.

And the goal of it is really to make sure that when you and I do our once-every-10-years colonoscopy, we get the full look, so that we feel confident as patients that someone looked at all the things that might be abnormal. There are clinicians out there who are very good at this, and their benefit from the AI might be a little less than for someone who’s new and just coming out, or really tired because they’ve had the flu for three weeks. You could just see how this kind of technology will level everybody up to a standard. So that’s how our technology works. It’s not fully diagnostic, but it’s helping the physician feel more confident and be better, and helping the patient feel that they had a good-quality exam where everything was looked at, regardless of whether they got the best gastroenterologist in the world or someone who’s in the middle of the pack.

Kevin Pho: Now, where are we in terms of the development cycle, and in terms of outcomes data for AI in a technology like this?

Shrujal Baxi: I think that’s a really good question. One of the things that we’ve focused a lot on is making sure we don’t create a lot of noise. We don’t want to create the alarm fatigue phenomenon with this tool that we’ve seen with other helpful technology, and we are doubling down on making sure that we’re not only sensitive, but specific enough in our presentation of what we think is a polyp or an abnormality, such that physicians don’t ignore us.

Outcomes data takes a little bit of time, right? So we have surrogates. A surrogate would be true histology, making sure that the things we’re pulling out are actually premalignant, or adenomas, and not the hyperplastic, never-going-to-become-malignant things. Physicians in general are going to pull out everything they see; if physicians knew what was bad and what wasn’t, they would never cut out everything. So we want to make sure that we are not increasing the burden on physicians for biopsies and things like that. We are currently measuring the true histology rate of our work by different factors, to really give insights to clinicians that this is actually a good thing: It doesn’t increase the duration of your colonoscopy, and it doesn’t create unnecessary work for you. So we’re building some of that out.

But the true outcomes data that matters is really whether AI-assisted colonoscopies decrease the rates of colon cancer. We are lucky, because we ran the largest randomized clinical trial in the U.S. at the time to actually get our polyp detection tool to market, and now we’re going back to see whether we can collect outcomes data, because screening data needs time to mature. I’m sure you’ve spoken to plenty of people about the constant debate in the public health space about what the right frequency is and what the true benefit is. We’re currently working on collecting some of that outcomes data to do the true follow-up from our randomized controlled trial, because sometimes it’s hard, given a lot of other factors, socioeconomic and family risk factors, to really get a clean answer. And it may be that we’re not mature enough yet, because it takes years to see the benefit of screening protocols.

Kevin Pho: So you mentioned that the ultimate goal, of course, is to maximize that sensitivity and specificity. How close are you to your goal?

Shrujal Baxi: We’re actually already on the market with a commercialized device that’s getting good reviews from the physicians who’ve had the opportunity to use it. So we think it’s good, and we’re getting a really good signal, but because we’re an AI company, we always want to be better. So we continue to iterate on what we can do to make our device help physicians better, and we’re taking feedback from clinicians. We all just came back from the American College of Gastroenterology annual meeting in Vancouver, and we heard clearly what physicians are looking for. They’re telling us that there are certain kinds of polyps that are really challenging for them, and asking whether there’s something our AI can do to help. That’s the kind of feedback that we look to continually evaluate, build models against and then push out, because we imagine that this is an iterative process, no pun intended.

Kevin Pho: We’re talking to Shrujal Baxi. She’s an oncologist. Today’s KevinMD article is titled “Removing workflow burdens and improving access for colorectal cancer screenings.” Shrujal, tell us some of the take-home messages that you want to leave with the KevinMD audience.

Shrujal Baxi: I guess I would have take-home messages directed to the different stakeholders in the space, right? So first, to the physicians who listen to your podcast: I think technology can help us. It can help us deliver better care, and it can give us confidence in the care we’re delivering. But it is good to be skeptical, to ask questions, and to push companies like mine and others to always do better. We are all in it to give better care. I think one of the challenges of being a physician in this space is threading the needle between what we can do, what we should do and when we can do it. It’s not that we’re not going to solve all the problems eventually, but what is the natural sequencing, such that as we’re putting out iterations of products, we’re actually solving pain points in the day-to-day workflow of physicians? So to those who speak to people like me, keep the ideas and the pain points flowing, because that’s how we know what to build.

To patients, I was going to say that I think AI is the next level of care that can help guide you and give you confidence in the care that you’re getting. So while there is sometimes a distrust that’s built into the system, I think AI that helps physicians be better physicians is exactly where I want to see it positioned. As an oncologist, I don’t want anybody but a human talking to me about my conditions, but if AI can help my physician understand the data faster, then I want AI to be part of the conversation.

Kevin Pho: Shrujal, thank you so much for sharing your time and insight, and thanks again for coming on the show.

Shrujal Baxi: Thank you, Kevin.