Early Detection with Routine EHR Data
Episode 575

Jeremy Orr, Chief Executive Officer at Medial EarlySign

Early Detection with Routine EHR Data

In this podcast, Dr. Jeremy Orr discusses how his company is leveraging the power of machine learning and unlocking the full value of data so clinicians can have all the information they need to give the best advice for their patients. He talks about finding patterns through sophisticated analysis of simple routine data test results and maximizing the potential of that data. Dr. Orr is going the extra mile and laying the framework for the time when clinicians’ actions will be augmented by machines. This interview is packed with learnings and insights, so please tune in!

Want to start your own podcast or have someone else manage yours professionally?

Don’t let technical busy work hold you back from sharing your genius!

Learn Now

Get The Latest In Your Inbox

SUBSCRIBE

Early Detection with Routine EHR Data

Episode 575

About Jeremy Orr MD, MPH

Jeremy has more than 20 years of clinical medical practice, population health, and healthcare IT experience. He served as EarlySign’s Chief Medical Officer before he was appointed CEO. A board-certified family physician, Dr. Orr was named a “Top 100 Physician” during his time with Kaiser Permanente, and then went on to launch a medical practice that became part of Centura Health. While an Assistant Professor at the University of Colorado, he was selected as Teacher of the Year by residents. Prior to joining Medial, Jeremy served as the CMO of Boston based clinical data analytics firm Humedica (later Optum Analytics) and CMO of Los Angeles based clinical decision support company Stanson Health. Jeremy earned his MD at University at Buffalo and his MPH at Tulane.

 

Early Detection with Routine EHR Data with Jeremy Orr, Chief Executive Officer at Medial EarlySign transcript powered by Sonix—easily convert your audio to text with Sonix.

Early Detection with Routine EHR Data with Jeremy Orr, Chief Executive Officer at Medial EarlySign was automatically transcribed by Sonix with the latest audio-to-text algorithms. This transcript may contain errors. Sonix is the best audio automated transcription service in 2020. Our automated transcription algorithms works with many of the popular audio file formats.

Saul Marquez:
Welcome back to the Outcomes Rocket, Saul Marquez here, and today I have the privilege of hosting Dr. Jeremy Orr. Jeremy has more than 20 years of clinical medical practice, population health and health care IT experience. He served as EarlySign’s Chief Medical Officer before his appointment as CEO. A practicing board certified family physician Dr. Orr was named a “Top 100 Physician” during his time with Kaiser Permanente, and then went on to launch a medical practice that became part of Centura Health. While an assistant professor at the University of Colorado, he was selected as Teacher of the Year by the residents. Prior to joining Medial, Jeremy served as the CMO of Boston based clinical data analytics firm Humedica (later Optum Analytics) and CMO of Los Angeles based clinical decision support company Stanson Health. Jeremy earned his MD at University at Buffalo and his MPH at Tulane. and today he’s making a huge impact with the work they’re doing at Medial EarlySign. And so today we’re going to be diving into the outstanding work that they do for payors, providers and really health care at large. And so with that introduction, Doctor, or I want to welcome you to the podcast. Thanks for being here.

Dr. Jeremy Orr:
Thanks for having me. Excited to have this conversation today.

Saul Marquez:
Likewise. And so, you know, I’ve been waiting quite some time for us to finally connect. And here we are. You’ve done some amazing work in the space. And today, the opportunity to drive change in this era of transformation is big and nobody could do it alone. So before we dive into the value you guys add at Medial EarlySign. Tell me a little bit more about what inspires your work in health care.

Dr. Jeremy Orr:
Sure. So essentially, it’s I wanted to maximize the impact I have. So we all have a limited time on this planet and we hope to do good and help to look back on our lives someday and feel like we made a big difference. And practicing medicine is extremely satisfying. And I think for some people, they really need that one to one interaction to feel like they’re having an impact. But for me, it was really a math problem. And I could see twenty five patients a day in clinical practice for my whole career and have a great impact on lives. And I and I have the greatest respect for clinical providers, but I’m not one of those who needs to have that immediate feedback from patients. And for me, it became a question of where do I allocate time and effort, resource, skill and I have the biggest impact and what we’re doing in advancing technology in health care, I can affect hundreds of thousands or millions of lives in a smaller, more modest way. But I think overall I have a more satisfying impact. So at the end of the day, that’s what works for me and that’s what motivates me. And that’s what gets me up in the morning. And I think it’s the same is true for everyone at Early Sign. Our stated mission is to have a massive impact on human health. Period. And we’re working on how to get there every day.

Saul Marquez:
I think it’s really inspiring. And there’s that shift in physicians, not everybody, but I’ve been seeing a lot more of it. Physicians like you who said there’s an opportunity to do so much more than just the bedside and impact populations and communities. And so it’s great to hear that that was the catalyst for the great work that you’ve done and now the work that you guys are doing at Early Signs. So tell us a little bit about Early Sign. What exactly are you guys doing and how are you adding value to the ecosystem?

Dr. Jeremy Orr:
Yeah, it’s a great story. So Early Sign is you can consider us experts in clinical data as it relates to machine learning. So applying machine learning approaches to clinical data. Now, machine learning has become quite encompasses a lot of human activity these days, and it’s quite a heterogeneous group of activities. And so I’ll carve out our niche for you so you know where we operate. We’re just working in health care and the data substrate we use is pretty much structured electronic medical record data and with particular attention to labs, data medications, existing diagnoses and things like that. So we’re a little bit separate from lots of good companies doing good work and imaging that works a little bit more advanced. And that’s partly because that type of data is fully classified, unlike clinical data, which is very, very messy and requires a lot of prep work. So that’s the niche we operate. And we’ve done several interesting things that some of our initial work has been on advancing early detection in diseases where early detection intervention makes a big difference. So we’ve operated in the space of cancer screening, in chronic disease, progression of chronic disease, understanding complications. And then more recently, we’ve done work in COVID as well as influenza. So infectious diseases for us more recently, but taking it a little bit wider. We’re also now engaging with lots of different players in the health care system to figure out other ways to bring machine learning insights into care.

Dr. Jeremy Orr:
So a lot of our initial stuff in cancer screening, chronic disease was really next generation risk stratification. This particular group of patients is very high risk for having a cancer or a complication of diabetes and the next time period take care of them first. But where we’re trying to go is trying to actually move into more prescriptive types of analytics. And this is a much more challenging problem, but potentially a much more impactful approach where the machine, if you will, and survey the sum of all the clinical data for a particular physician, a group of physicians, and advise not just on who’s highest risk for what’s the next best action or intervention to take for individual patients. So this will really lead to the true augmentation of care. And we had early on believe that a physician will always be an intermediary in care, but we want to do everything possible to bring not just the highest risk patients, but the best advice in front of the physician at the right point, place and time and care so they can act on it. So we think someday it’s inevitable that clinicians actions will be augmented by machines like this and that we’re just laying the framework for that now.

Saul Marquez:
Well, I think it’s fantastic that you guys are daring to be bold and taking that step. A lot of companies are afraid and they’re afraid that they’re too early. And I think that it’s figuring out that right timing, but at the same time, not being afraid to lay down the framework, like you said. And so that prescriptive step taken just like people talk about monitoring and early warning scores. So what are you going to do with all that package so that you enable a clinician to do more? Maybe you could talk to us a little bit more about that and what exactly you guys are doing different and better than what’s available today.

Dr. Jeremy Orr:
Sure, sure. So many of the efforts to get machine learning and care, as I mentioned, they have focused on imaging or applying them to more sort of esoteric data, a patient source data, in some cases genomics, proteomics and so on. So one thing that’s different is we’re trying to present the broadest possible use case and really work on readily available very common elements. So some of our algorithms, for example, just the one blood test, age and gender to provide a very sophisticated risk profile for, say, a cancer. And behind that effort to act on routine data is a premise that routine data is not being nearly anywhere near fully utilized. Now that the current approach is actually pretty primitive. If you go to your primary care physician, you get a blood test and a chemistry panel and your doctor calls you your doc. And I did this in primary care or family medicine my whole career. If everything’s in the normal range and I’m very likely to say there’s nothing you have to worry about. So I’m getting more sophisticated. I might look at the last few tests and look at trends, but beyond that, we don’t do any more sophisticated analysis.

Dr. Jeremy Orr:
And that’s a very common and naturally human approach to interpreting labs. But the machine looks at it completely differently and the machine is trained on hundreds of thousands or millions of patients and those that are completely nonintuitive patterns. And sometimes, for example, in a blood count, there might be a relationship between the platelets and the size of the haemoglobin concentration. That’s mathematical. And there might be a subtle change there that precedes an obvious change outside the normal range. So often we will flag patients at high risk that have completely normal appearing labs to people. And that’s really the power of machine learning. It’s it’s not magic. It’s the pattern recognition over large numbers, but it’s stuff that people can’t do or don’t do. And so that is really unlocking the true and full value of all this data that we have just sitting around a very routine, inexpensive stuff for blood now cost about three dollars and really squeezing all the value out of it, putting it in front of the provider so they can act on.

Saul Marquez:
You know, you’re calling out something really important is that we’re already taking so much data and not doing what we can with it. We’re not maximizing the potential of that data. We’re not serving the people as well as we could. And so tell me a little bit about how you’ve already done this, doctor or and some examples of how you’ve improved outcomes or made business better for the people you serve.

Dr. Jeremy Orr:
Sure. So we have our algorithms live now at five health systems. And again, we use cases for early detection. And we’ve worked closely with those health systems to get them into workflow, whether it be electronic medical workflow, management workflow, and then also work closely with their questions about how to talk to patients when they when they speak with them about them being high risk as detected by a machine learning algorithm and then defining the follow up actions. And in some cases, it’s a screening test, like a colonoscopy for colorectal cancer or a low dose C.T. scan for risk of lung cancer, or in the case of diabetes, an earlier intervention into a prevention program or a better control program. And so you hinted at something earlier that’s very important in our field. There is a lot of concerned that we may be too early with this, I don’t think we’re too early, I just think it’s hard to drive adoption and do it well and actually get the outcomes which which is which is all that matters at the end of the day. And there are many sort of more technology focused companies who provide the technical solution, but then may not go the extra mile to work closely with the clinicians to understand the workflow and the communication with the patients to operationalize it effectively.

Dr. Jeremy Orr:
So one of our early lessons has been that adoption takes a lot of effort to get Right. and we’ve learned a lot along the way, implementing at these now five health systems and with our partners. And we’ve actually gone so far as to work with behavioral scientists. There’s people that work at Nudge that their job is to optimize patient messaging to get better outcomes. And we worked with them on what is the best way to tell patients about their risk with these algorithms. And we came up with some language that we think is the right amount of activating. And so that’s that kind of work, that kind of detail and nuance that’s going to lead to better outcomes and it has to be operationalized effectively. So that’s what I’m sharing with you, some of our progress, but also some of our learnings along the way. And we’ll continue to put in the effort because we think it’s worth it. You think it’s going to have an impact on lives that’s worth that effort?

Saul Marquez:
Yeah, that’s a really great call. It’s that adoption, that operationalization making these things know, putting them into the workflow, writing them in the policies, making them useful. And it sounds like you guys are doing some fascinating work. I mean, employing behavioral scientists to figure this out. I mean, that’s a step in the right direction shows the commitment that you guys have to hey, we don’t just have some technology here. We’re trying to figure out how to make this work for everybody, right?

Dr. Jeremy Orr:
Yeah, it takes a commitment and it takes time and effort and it takes the right partners. You know, we’re very fortunate to have some partners.

Dr. Jeremy Orr:
Two of our development partners in the US are Kaiser and then in Israel, we work with a health system called McAvey, a very high performing HMO. And these guys are their culture is a fantastic fit for this because they understand their transformation and their docs know that it’ll be worth the effort up front. So we’re lucky to have partners like that. Now we have to get this process right, because when we take it to the rest of the country in the world, we have to make it as easy as possible. And one of our things that early sign is try to make I made easy, bring pre trained algorithms, bring studies that prove their validation. We’re going to need to have that kind of evidence for docs to adopt new technology, bring examples of how we communicate to patients, bring workflow to make it easy. So that’s sort of next phase. We’re in the middle of refining that now.

Saul Marquez:
Now that’s exciting. And so there’s so many places to play Right. the domains you guys planned, GI, kidney, lung. I mean, can you give us an example of one of those areas and how you’ve made a difference already?

Dr. Jeremy Orr:
Yeah, certainly. So the work that’s most advanced is in lower GI. Ok, so this is an algorithm that looks at routine labs and age and gender is basically the example I used before. And we’ve operationalize this to scan the data at health systems and flag the highest risk patients. And we work closely with the health system to understand what’s your colorectal cancer screening program look like, what your colonoscopy Pasley look like. Are you concerned about missing high risk patients? Are you concerned about patients who are due for screening colonoscopy, for example, who refused it for a number of reasons? And then we work with them to find the high risk patients and no one on the noncompliant patients give them extra ammunition to convince that patient. And it’s pretty remarkable the conversion rates we’ve seen among previously noncompliant patients. They’re coming in for scopes. Now, among those, we played positive at a clip of about 70 to 75 percent. And that’s got a couple different health systems. And if you’ve worked in public health at all, you know, it’s pretty remarkable for a patient who said no previously to come over at rates anywhere over 50 percent. So we’re very proud of that work. And then I’ll get some additional. Thank you. And then also providing an additional safety net. So the data’s there. As you said, all these labs were already done. We know age and gender and virtually your entire patient population. You can scan the whole population on a periodic basis and make sure we’re not missing any of the highest risk patients. So kind of low cost, no effort safety. That’s very appealing to many organizations. And it also helps them elevate the value of data they already have.

Saul Marquez:
That’s really insightful, Jeremy. And I just think about employers, too. I mean, how useful this could be to them. It sounds like you guys are mainly working with the providers, but is there an employer play here in the go to market, maybe down the horizon?

Dr. Jeremy Orr:
Yeah, absolutely. So I think the initial work in the provider space was really critical for us to understand how to implement this stuff and how to get it working, both technically and clinical workflow and Some results, and we’ve published now 14 papers based on these experiences, so that’s important to sort of build credibility and momentum. But you’re absolutely right. We need to come at this from every possible angle. And we’re working with many different aspects of players in the health care system, both with life sciences companies. Now we’re working with payers. We would love to start to do more work with employers. That’s an area we haven’t really gone yet. We’re working with diagnostics companies, working with commercial labs. And really, I think when we fully mature as a company, what we would like to be is an inside engine. That is really our expertise is on clinical machine learning. And there’s many different pathways to get our insights to patient care. But we drive the innovation from the consumer and clinical data standpoint, and then we work with our partners to deliver it in various ways.

Saul Marquez:
So interesting and valuable. So in the work that you’ve done and you kind of hinted at this earlier, Right. the adoption piece, anything else that sticks out to you as a setback that you guys have learned so much from?

Dr. Jeremy Orr:
So another piece that’s going to be critical for driving this forward is explainability. And machine learning often gets knocked as being a black box. You know, it’s how can you know the algorithms so complex? The computer operates in its own way. And you can understand humans can’t completely understand why a patient landed on a high risk list. Well, we’ve actually added to our products a feature we call Butterfly, which goes partway towards explaining. It’s not, at our current understanding, fully possible to explain every bit of why, but we go most of the way there. And so some of the factors that land a patient on a high risk list and that, again, will power adoption because the clinicians, first of all, will have more faith in it because it aligns with the clinical instincts. And in most cases, there’s always some successes. But most of the reasons will align with the clinical instinct. And then it also powers a better conversation with the patient. So it is called the patients at your high risk for having chronic kidney disease related to your diabetes. We’re going to do X, Y and Z, and the patient says Y. It’s not a very satisfying answer to say the computer said. So it’s going to have to go back to what they’re comfortable with, which is saying, well, we see this in your labs and we see this trend in disease progression. So those are the kind of conversations we’re trying to empower.

Saul Marquez:
That’s excellent, Jeremy. And, you know, I’m just really impressed with the thoughtfulness that you and the team have put into just the approach and the soft side Right. like the human side of it. You know, usually you don’t see that much. And I want to recognize you guys for that work early on.

Dr. Jeremy Orr:
Thanks for that. We have a saying in a few of our slides. The talk is great, but it’s about the people and and we always have to remind ourselves of that.

Saul Marquez:
That’s so awesome. And, you know, as I say, it’s so interesting that you said that the conversations we have here on the podcast, you know, I mean, high level people like yourself and these amazing folks, it does the best at what they do always say, don’t have that shiny penny syndrome Right.. Look at your people, focus on your people. And you guys are doing just that. And so what would you say you’re most excited about today?

Dr. Jeremy Orr:
Well, I’m excited that we’re on the bleeding edge of a new frontier. And everyone we talk to in the health care ecosystem, all these parties I previously mentioned, they know deep inside five to 10 years, clinical machine learning is going to be incorporated into care in a wide variety of ways. It already is making inroads into image, interpretation and many administrative functions. And we’re trying to move closer to the common use cases, and especially in primary care and screening and infectious disease. We’ve done quite a bit of work in covid and Influenza Two, which is exciting, but it’s just the beginning and all these things we’re figuring out how to talk to patients, explain ability, how to implement, how to prove the value over time. We’re really we’re creating a paradigm for the future for us. And there’s other companies doing really good work here. I’m sure we’ll make inroads as well. And from this foundation, we can expand into other data substrates, we can expand into other clinical areas and we can expand, as I said a bit ago, beyond this very comfortable use case of risk stratification. It’s something clinicians are very comfortable with. That’s why we started there into where it can really make a massive impact, which is more prescriptive advice, and then bringing sort of this computer intelligence to augment the clinical judgment of practicing physicians. And that’s when the care will really advance. And it’s also, I think when DOCS will start to appreciate this really makes me a lot better. And I owe it to my patients to do this. Well, we think that that is on the near horizon, that that sentiment and everything will change after that.

Saul Marquez:
Well, Doctor Orr you certainly made me excited about. I’m sure all the listeners that are tuning into this as well, hey, you guys might be the early signs of what’s to come. And I definitely think that it’s an exciting future for all of us in leadership seats, but also mental patients that will need this care. This has been great. I can’t thank you enough if you can, Doctor, or why don’t you leave us with a closing thought and then the best place for the listeners to get in touch with you or somebody on your team to continue the conversation.

Saul Marquez:
Yeah, thanks Saul for having us. I’m a big fan of the podcast, and it’s an honor to be here today. Two quick closing thoughts. You know, the COVID time. We’re still in the midst of it. But during this time, routine care was dramatically diminished. Now, some of that starting to come back. But I think there’s still a lot of suspicion on the part of patients, whether it’s safe to go to their health system. And we know there’s some good published research out there that a lot of cancer screening was delayed, a lot of chronic disease care was delayed. And we will see excess cancer burden and complications from this if it’s not taken care of properly. And what we’re doing now is we launched a campaign called Back to Care. And basically we have a set of algorithms that the health system could run and they can understand who to put at the front of the queue and make special outreach to on cancer screening backlogs, on backlog for diabetes care, chronic kidney disease care and even flu season is really just two months away. And everyone’s obsessed with covid now. They should be, but the vaccination season and influenza will be upon us and even prioritizing the highest risk patients for complications from flu. And COVID, we have some work there, too. So I think we can’t neglect this population. These will be the silent victims of covid if we don’t pay attention to them very soon. So that’s one another thought is we’re also, as you said in your question, engaging with other parts of the health care ecosystem that we just launched a program called our A.I. Accelerator. And it’s a way for life sciences companies, payers in particular, but also health care provider organizations to augment and accelerate their own efforts. So if they have a project that could use machine learning on clinical data, we would love to hear from you and see how we can help you do that faster and get results much more quickly. And we’re all about accelerating time to results and working with partners to do that. And partnership is really critical, as you heard several times here, for us to do that. Well, so we’re going to learn from you and hopefully we can help you, too. So we would love to have conversations in that realm as well.

Saul Marquez:
Well, that’s fantastic, Jeremy. And folks, here’s two really great ways that you can engage with Dr. Oz and his team early sign dotcom. You know, how are you going to deal with the massive influx of patients after this? There’s a way to do that and they’re thinking about it. And then secondly, if you’re thinking about implementing AI machine learning to your processes in your life sciences, your payer, here’s another opportunity for you to interact. They’re looking to help. And I think they’re very clear about their intention. So take Jeremy out on his invitation. And so, Jeremy, I know early sign dotcom is where they go to learn more. Is there a specific call out for those two programs or can they find those on the website?

Dr. Jeremy Orr:
Sure. I’m happy that people email me terribly early sign dot com for a small company. We handle a lot of things ourselves and also solutions. That early sign dot com is another way outstanding.

Saul Marquez:
Well, this has been fantastic. Can’t thank you enough for coming on and sharing what you guys are seeing and certainly looking forward to staying in touch.

Dr. Jeremy Orr:
Thanks. I’ll keep up the good work. Really enjoy the podcast.

Saul Marquez:
Thank you.

Automatically convert your audio files to text with Sonix. Sonix is the best online, automated transcription service.

Sonix uses cutting-edge artificial intelligence to convert your mp3 files to text.

Rapid advancements in speech-to-text technology has made transcription a whole lot easier. Automated transcription can quickly transcribe your skype calls. All of your remote meetings will be better indexed with a Sonix transcript. Better audio means a higher transcript accuracy rate. Are you a radio station? Better transcribe your radio shows with Sonix. Sonix takes transcription to a whole new level. Do you have a podcast? Here’s how to automatically transcribe your podcasts with Sonix. Get the most out of your audio content with Sonix. Automated transcription is much more accurate if you upload high quality audio. Here’s how to capture high quality audio.

Sonix uses cutting-edge artificial intelligence to convert your mp3 files to text.

Sonix is the best online audio transcription software in 2020—it’s fast, easy, and affordable.

If you are looking for a great way to convert your audio to text, try Sonix today.

Things You’ll Learn

  • Adoption takes a lot of effort to get right.
  • A successful company takes commitment, time, effort, and the right partners.
  • When you take your technology to the country and the world, you have to make it as easy as possible.

 

References
https://earlysign.com/