HLTH_Taha Kass-Hout: this mp3 audio file was automatically transcribed by Sonix with the best speech-to-text algorithms. This transcript may contain errors.
Saul Marquez:
Hello! Saul Marquez here and welcome back to the HLTH Matters podcast, straight from the HLTH event in Las Vegas, Nevada. I want to welcome our next guest. His name is Taha Kass-Hout. He is a physician and also the director of Machine Learning and chief medical officer at Amazon Web Services and leads US’s Health AI strategy and efforts. He led teams at Amazon responsible for developing the science technology and scale for COVID-19 lab testing, including Amazon’s first FDA authorization for testing their associates, later offered to the public for at-home testing. A physician and Bioinformatician, Taha served two terms under President Obama, including the first chief health informatics officer at the FDA. During this time as a public servant, he pioneered the use of emerging technologies and the cloud, the CDC’s electronic disease surveillance, and established widely accessible global data sharing platforms, OpenFDA, which enabled the researchers and the public to search and analyze adverse events data and precision FDA, part of the Presidential Precision Medicine Initiative. Taha holds his Doctor of Medicine and Master of Science of Biostatistics degrees from the University of Texas, and he completed his clinical training in interventional cardiology at Harvard Medical School’s Beth Israel Deaconess Medical Center. I’m privileged to be with you here today, Taha, and thank you so much for joining us on the HLTH Matters podcast.
Taha Kass-Hout:
Thank you so much for having me today, Saul.
Saul Marquez:
It’s a pleasure. So let’s level set, talk to us about your role at AWS.
Taha Kass-Hout:
Yeah, so I’m Chief Medical Officer and Vice President of machine learning at AWS, Amazon Web Services, that’s our cloud business, where I work very closely with partners and customers on innovative solutions to healthcare challenges, which I’m really glad to talk about some of those today.
Saul Marquez:
Well, you’re in the right place and the right meeting to be focused on those things. At AWS, what’s your mission when it comes to health?
Taha Kass-Hout:
Yeah, so many folks know about AWS, but for those who don’t, so Amazon invented the cloud and that’s Amazon Web Services. It’s the widely available and adopted cloud platform with over 200 features and services available from data centers around the world. And at AWS, we’ve been inventing in healthcare since day one on helping healthcare and life science customers with breakthrough medical innovations on the cloud. Particularly our mission specific to healthcare customers is really to enable them with the right technology to be able to offer convenient, personalized, and high-quality care. And the way we go about doing that is we’re a customer-obsessed company, so, which means we listen to customers, and shame on us if we don’t help them ask the right questions given our breadth and depth and technology, machine learning, and scale, but also focus deeply on the end of mind, which is double down on use cases and challenges to help them achieve those.
Saul Marquez:
That’s fantastic, Taha. What are some of the challenges you hear from health customers to make sense of their health data?
Taha Kass-Hout:
So if you look more globally, the last ten years been great towards digitizing health records. There was massive digitization of health data in general. I mean, we went from something like 15% of medical record data, for example, in the digital form ten years ago to today, almost 98% of health data, whether that’s primary care, acute care, huge, huge value. Along with that, many other data became available in addition form. I mean medical devices that measure you are in digital form. You look at the way we sequence and analyze human genome or one’s proteome or cancer tissues are all in digital. We also have a propagation of new data. I mean, these like social data such as access to food and other factors, you have now more refined measurements from wearables and the likes, and all those are digital form. We’ve seen, since the pandemic, also like where people with remote monitoring receive care at home, with telemedicine, propagation of chatbots, and all that in digital form. People go to their local pharmacies and get their COVID-19 vaccine or flu vaccine, and all of those are in digital form. So as a result of that, there is, you know, patients now walk around with thousands of data points on them throughout the entire medical history, and so, making sense of that data is key. However, the majority of this data remains unstructured information, meaning that you have to extract and transfer this data before you can analyze or search it. So it is estimated about 97% of this data goes unused today, it’s almost all of it. When you think about the actual value, beyond what you code like for billing or beyond what you code in the medical record, I mean, the majority of the data, just if you look at that one lens, you know, the majority of your data is in medical notes, and the challenge to extract this information to many health systems, it’s a process that’s oftentimes very error prone and complex. Oftentimes these systems spend weeks or months just trying to structure what they can for, you know, just to get paid or to manage, the best of their ability, their patient population. So the cost of operation complexity becomes incredibly hard on these customers. And, but we believe that this highly contextual and multimodal health data is of great and immense value towards applications such as precision medicine operate more efficiently, be able to design better clinical trials, or manage patient population or subtypes of patients to really deliver the right diagnostics for them and the right line of therapeutics.
Saul Marquez:
Now, that’s great, you know, making sense of the data, cleaning it up, making that data fuel that could actually be used for the engine of intelligence is the goal.
Taha Kass-Hout:
Absolutely, and our customers, I mean, they’re really excited about how can they deliver convenient, personalized, and high-quality care, and a lot of that, you have to think about really a data-driven approach to this. And these are, the area we’ve been making a lot of material investment in purpose-built health services, almost all enabled by machine learning to tackle all these Lego pieces from every type of information. How can you structure it? How can you piece it together to really bring a 360 view of a patient’s entire medical view and aggregate all this data so you can now do analytics and machine learning on the population as a whole? To answer many, many of these questions, I’m happy to talk to you today also about some of those tools we’re building.
Saul Marquez:
No, absolutely, definitely want to hear more. Can you share with the audience a little bit about what tools and resources AWS offers for healthcare and life sciences more broadly?
Taha Kass-Hout:
Absolutely, so I’ll start from the top. In 2020, we introduced Amazon HealthLake. It’s a hyper-eligible managed service for customers to store, transform and analyze both their structured and unstructured information, health information, at petabyte scale and in the cloud, and today we announced two additional features of the service. The first one is Amazon HealthLake Imaging, to be able to now analyze, to store access, and analyze medical imaging in the cloud at petabyte scale, and Amazon HealthLake Analytics to be able to, to enable you to analyze multimodal health data and query that data across whether at the patient level or population level, also at petabyte scale. So really excited about these services and the partners that are using these services to build the next generation of medical imaging applications, analytics in the cloud, for example, or to now derive population analytics on the entire population so they can tailor treatments and diagnostics and be able to intervene early in the course of a patient’s journey and be able to personalize that journey for them. We also have now specific tools for the, in a form of managed services to do specific jobs. So Amazon Comprehend Medical, as you know, like the majority of medical record data are in the form of patient notes. You have nurse triage notes, radiology reports, surgical reports, all those are narratives, documents upon documents. Within those documents are answers to a lot of features that you can extract and have better predictions. So Amazon Comprehend Medical automatically structures this massive amount of information in ways that now, you can, it’s ready for analysis interpretation, also maps it to ontology. I mean healthcare is domain-specific, highly contextual. So for a medical code or condition, there’s a code for it, like an ICD ten or ICD nine for a medication, there’s …, for a concept, there’s …, and there’s like 370 thousands of those concepts, over 70,000 ICD codes and whatnot. So the ability for a service that uses machine stateless service, that uses machine learning to be able to extract all this information, map it to the right ontology, also maintain the relationship in the text, such as family history or negation, someone was ruled out or was referred, doesn’t mean that they have the condition or maybe the family member has it. This condition is associated with this medication, this medication associated with these, you know, dosages and routines and that sort of thing, here’s a follow-up. So imagine that you’re taking that bulk of unstructured data and now be able to process through the service, and now all this information is available to you with confidence core about for full visibility into how the machine did. The other service we have is Amazon Transcribe Medical is a way to add ASR capability to your voice application, whether it’s a chatbot, a transcription or a telemedicine conversation, or in-person conversation. So be able to also analyze the entire medical context in that conversation and be able to structure also for analytic purposes. With a lot of payers and many health customers, we also see the propagation of a plethora of document-based, like your lab reports and scanned PDF or the whole claims and scanned PDF, and we have also tools specific, purpose-built for, to be able to extract that information, whether that’s a narrative with us in a form or a table, maintain the accuracy, and have full visibility into also the accuracy how the machine did, and that’s Amazon …, it’s also a managed service. All these services are HIPAA-eligible and GDPR compliant, and so, and a whole sleuth of other, you know, regulatory aspects because security jobs zero for us and patient data is very, very private, so each one of these services maintain that posture. So when you use them, you can be assured that patient data remained encrypted and secure at all time. And then we give access to the data via many modern ways, how you can interact with it with these microservices via application programming interfaces or APIs are also secure. But also you can start building on top, which is healthy flywheel with partners that work very closely with us in addressing a lot of the customer’s challenges in healthcare and life sciences.
Saul Marquez:
Thank you, Taha, it’s certainly a broad list of capabilities that AWS, Amazon Web Services, offers health providers and people in healthcare, whether it’s life sciences or providers, payers. You know, you guys did some exciting launches today, so congratulations on those. Can you share any partners or customers who are already using these products and how they’re using them?
Taha Kass-Hout:
Absolutely, so today we announced Amazon Health Imaging and that’s to be able to store access and analyze medical image data at petabyte scale in the cloud, and a whole sleuth of customers, including, for example, University of Maryland and Radboud University Medical Center, who are excited about Amazon Health Imaging to build the next generation of medical imaging intelligence in the cloud and be able to derive analytics and machine learning on that data for all sorts of purposes, whether that’s diagnostic personalization and also for improving radiologists’ workflow and clinician workflow. The other partners include GE Healthcare, who’s evaluating Amazon HealthLake for streaming their viewers and … application as well as look at optimized medical storage in the cloud. Intel Arad building next-generation PAX systems on Amazon health-like imaging to ensure a scalable architecture for their customers with subzero retrieval and that clinical workflow, ensuring high availability and low latency. We have NVIDIA, a great partner with its MONAI, a connector that worked with AWS to provide that connector to Amazon HealthLake, and MONAI is an open framework, medical AI society and suite of tools to help analysts and data scientists, and researchers be able to build, deploy, and train medical AI applications for a whole sleuth of applications such as early diagnostics, image recognition, pattern recognition. And imagine, now those inferences can become available in real-time in Amazon HealthLake Imaging. We have partners like Arcturus, who’s building really innovative AI technologies in the clinical workflow for radiologists to identify anomalies, for example, CT scans, that decrease the cycle of radiologists’ operational, sort of overburdened and be able really to help them fine-tune and zoom in on the areas of interest. And they’re also building their AI platform on Amazon HealthLake to enable them with high availability and also viewer performance for their medical AI platform, and many others that you know, are really, we’re really excited to see them, part of the flywheel that now we’re creating for healthcare customers and whether radiologists or in a clinical workflow of a clinician, to be able to really start now reinventing and reimagine medical image viewing as well as analytic applications. The other services analyze is Amazon HealthLake Analytics to be able to, to enable you to query and analyze multimodal health data at scale. And while we announced it today, Rush University Medical Center, a leading healthcare institution here in the US, is already using Amazon Health Tech Analytics to be able to bring together medical records data, social determinants of health data, as well as cardiometabolic information from their population at risk. And as a result of that, we will then use the service to create advanced analytics and machine learning to personalize care for those patients, identify risks early, and also improve health equity and close gaps in care for the population they serve primarily here in West Chicago region.
Saul Marquez:
That’s fantastic, Taha, and very exciting. Congrats on the launch of these new features and the partners that you’ve already gotten on board on the flywheel, as you call it. So really, like the one-on-one of it is the data lake helps customers store, it helps them analyze and helps them build their solutions, their AI and ML solutions in a way that’s simple and seamless.
Taha Kass-Hout:
Absolutely, and index every point in a patient entire medical journey. So now you have a 300, truly 360 view of someone’s entire medical history, then be able then to study entire populations at scale and do these advanced analytics and machine learning in just a few clicks. We’re on a mission making machine learning as boring as possible. Part of our responsibility, the way we’re both responsible technologies and responsible AI should be available to everybody, not just software developers and engineers, but also data scientists, business analysts, physicians, and clinicians that you really you use no code or low code depending on your skill set to be able to really start taking the to the next level and see really what you can do with all this massive amount information, but just with few clicks and then be able to learn from the data and manage, better manage your population, have designed better clinical trials, operate more efficiently, and many more.
Saul Marquez:
Fantastic, one more question for you, Taha. You know, it’s the cloud, you know that we’ve been talking about it. It’s, you know, it’s here and people need to start adapting it. What’s the role of the cloud and machine learning to help healthcare companies with their digital transformation?
Taha Kass-Hout:
Yeah, absolutely. I mean, it’s a massive I mean, still day for, day one for us. You know, it’s a massive, massive opportunity, especially for healthcare. I mean, as I mentioned, healthcare generates a massive amount of information. If you just look at medical imaging, I mean, five and one half billion studies and procedures done every year globally, and the size of the medical image has doubled just in the last decade. And then you look at institutions locally storing multiple copies of the same data over and over in their clinical and research systems, which increase in cost or level is unsustainable. Not to mention also the, a lot of this data is not structured so data scientists’ research spends weeks and months just trying to structure the information to derive any insights in order to drive patient care and population management and disease management, using advanced analytics and machine learning. So the cloud is immense, is immense, and Amazon, as I mentioned, invented the cloud, and that’s where our mission to provide you with the most secure, best storage scalability as well as privacy, while also giving you built on the most robust infrastructure around the world, but also give you purpose-built services to be able now to understand these medical contexts, understand the complexity of this data, help you bring all that data together in a cohesive way. So now you can really focus on just taking care of patients and drive insights from that data. There’s also a lot of advancements that we’ve seen over the last decade with advancements in deep learning and machine learning, where the dust has settled, for example, in natural language processing and understanding, pattern recognition as well as predictions. And we’re making all those capabilities purpose-built so that it integrates with your clinical workflow or integrates with your population health analytics, or be able to integrate with your clinical viewers so that way you can really streamline this information from around the globe and securely, and be able to protect patient privacy, and also the application programming interface is access to this data. Today, the world’s health data, you can access it via modern APIs or microservices, I mean, it’s really stored locally. So kind of, you know, bring that to the modern age about how systems can be able to talk to each other, how researchers can securely share that information across the enterprise, bring this data together in a way that now you can enable these machine learning, advanced analytics, and be able to stitch the information together, but also how you can do that at scale. I mean, it’s, this information, we’re talking a massive amount of information, it’s highly contextual and multimodal, and this is really where the cloud is really just impressive and definitely the right tool for the, to do the job from scalability and offer those deep insights. And now we have, providers can share data security among each other, patients can have unfettered access to their data through these modern APIs, through third-party vendors and applications, truly move the needle on advanced diagnostics and advanced therapeutics, personalized medicine, and try to really tailor clinical pathways and care for these patients as well as prevention, and in the mix, also, make sure that you improve in health equity for population at large.
Saul Marquez:
That’s fantastic data, Taha. Well, no doubt there, the innovation that you guys are creating is making a difference. So I thank you so much for sharing some of the most recent innovations with the data Lake Analytics and the data Lake imaging, thank you for that. And folks, everything that we discussed today with Taha is available in the show notes. So make sure you check out those show notes. Check out the entire list of services that Amazon Web Services provides on there. So Taha, what closing thought would you give the listeners before we conclude?
Taha Kass-Hout:
Oh, absolutely. I mean, if we look at healthcare today, it’s an unusual business. It’s $8 trillion globally, yet 40% of world population doesn’t have access to care. I’m not talking about high-quality care, any care. And if the last decade was about digitization, I really believe the next decade is going to be about how can we make sense of this data, and I believe there’s a great role for cloud, AI, machine learning, and structuring this information in a way that we can start closing gaps in care, where we can have more applications like precision medicines, and intelligent chatbots, and ways how we can truly close gaps in care, and improve health equity and patient outcomes for everyone around the world.
Saul Marquez:
Thank you so much, Taha, I agree. And this type of technological advancement will allow for bridging those inequities in health, and we’re excited that you guys are partnered to do this. Taha, thanks for joining me today.
Taha Kass-Hout:
Thank you so much for having me. I really appreciate it.
Sonix has many features that you’d love including automatic transcription software, world-class support, secure transcription and file storage, advanced search, and easily transcribe your Zoom meetings. Try Sonix for free today.
