Dr Andrew Wilner (Host): This is Transformational Pediatrics with Children's Mercy Kansas City. I'm your host, Dr. Andrew Wilner. My guest today is Dr. Mark Hoffman, Chief Research Information Officer. I invite you to listen in as Dr. Hoffman's helps us understand how data science can inform better healthcare decisions. Welcome, Dr. Hoffman.

Dr Mark Hoffman: Great. Thanks for including me today.

Host: Sure. Thanks for joining us. So, I want to learn all about data science. What is it and why is it important in pediatrics?

Dr Mark Hoffman: Yeah. So, data science, there's a number of definitions out there. But if you think about it, every activity in our lives, especially in the delivery of healthcare, generates massive amounts of information and data. Data science is the discipline of trying to make sense of that. So, sometimes we can develop predictive models, sometimes we can understand patterns of care using data science methods. And so, as a data scientist, we're focused on gaining insights from the vast amounts of information, especially in my area pertaining to healthcare.

Host: So, data is just information, right? A data point, the temperature, how long you were in the waiting room, what your temperature was. Is that right?

Dr Mark Hoffman: Yeah. And there's everything from high volume, low complexity data, for example, temperatures to low volume, high complexity data like genomic information. So at Children's Mercy, we're very active in developing diagnostic genomic sequencing tests to identify the causes of rare diseases in children, for example.

Host: Okay. So, let's talk about that. I thought we knew everything there was to know about the genome, that we could map it, and we got all that stuff. Is that not right?

Dr Mark Hoffman: There's been huge progress, but now we're really looking at individual variation and how that contributes to health. And there's still a significant number of rare diseases that the cause is still not recognized. And so, the Genome Center at Children's Mercy is leading many of the efforts in that area. Our genome center has sequenced probably 25,000 people. And we have, I think, five petabytes of genomic data that we manage from that process. So, you can imagine making sense of that is often a needle in a haystack type activity. And that's where data science comes in.

Host: Right. So that's what computers are good for, right? I think when it's a question of an autosomal dominant disease with a single gene, that's not too tough. But I was interviewing someone about autism the other day, and they said that they've identified 1000 genomic factors that all interact within the autistic spectrum and probably there's more. So trying to sort all that out, you either need a lot of dedicated people sitting with a spreadsheet or a machine with a lot of memory.

Dr Mark Hoffman: Yeah. And another area that we're really active in leading is epigenomics. So even if the DNA sequence is the same, the methylation of that DNA can vary from person to person. And so, there's also some really substantial effort in that area.

Host: So, the same gene, but it's got a tweak to it. It could be this way, it could be that way, methylated or not methylated.

Dr Mark Hoffman: Correct. How it's expressed and when it's expressed. We are also very active in gleaning insights from electronic health record data, and both specific to Children's Mercy, but also we work with National Electronic Health Record data sets to learn more about patterns of care.

Host: Okay. Now, I know that private information has been kind of a barrier to doing studies like that. Because of course, you want to share data, but if the data has a name on it, so then you gotta take the name off it, but then you don't want to forget who it was. Is that still a problem or has that been kind of solved?

Dr Mark Hoffman: I think for the resources we use that are fully de-identified, that's largely solved. All of the 18 HIPAA-sensitive fields are stripped from the resources we use. But the data that we have access to represents more than a hundred US-based healthcare facilities, and there's more than a hundred million patients represented in those data. So, you can learn a lot without needing those HIPAA identifiers. Our group is focused, for example, on how often does real world healthcare deviate from ideal practices and standards, and that type of data is ideal for that type of investigation.

Host: For example, like a screening, you know, when you're 60 years old, you should get a colonoscopy, and of course not for pediatrics, but that's what I'm more familiar with. How often does that actually happen? What would you do in pediatrics, for example? What's the standards of care that you're looking at?

Dr Mark Hoffman: We've published a few papers. One example is we looked at how often are opioids ordered for youth and young adults treated in the emergency room setting for migraines and it should be zero to 1-2%. We found that 22% of ED cases of youth and young adults presenting with migraines were administered an opioid. So, it can identify those really important variations between what should be happening and what is happen.

A separate example is we use data to evaluate how often are sickle cell patients tested for A1c. And again, when you think about it, they're both related to hemoglobin. And if you have a mutation in your hemoglobin gene for sickle cell, that makes the A1c test less reliable. We found that 11% of sickle cell patients had an A1c test order. So, doing data science with these massive data sets can help us identify things that can then lead to quality improvement initiatives.

Host: So for the most part, the data sets come from the EMR?

Dr Mark Hoffman: In those examples, yeah. We contribute data in. And then through our relationship with our EHR vendor, we're able to query the entire data sets.

Host: So, you guys have like a giant room there at Children's Mercy all filled with computers, or is this done in the cloud or how does that work?

Dr Mark Hoffman: Most of that large scale work is now done using cloud-based systems. So, my team recently led an initiative to move our genome centers bioformatics platforms to a completely cloud-based strategy. Likewise, the electronic health record data resources is also accessed through a cloud platform.

Host: Great. So now, you can work 24/7 from home.

Dr Mark Hoffman: Yeah. That's the hazard of having that option. We do have a really world-class data center as well. There's still a number of types of work that that's still beneficial.

Host: I'm going to ask you a tricky question now. Who pays for all this?

Dr Mark Hoffman: So, again, I'm focused on the research perspective. But, in my case, much of our work is funded through grants.

Host: Okay. So, like other kinds of medical research. Because, you know, you can't really-- I don't know if you can, can you bill somebody, say, "Okay, we're going to look at your gene. What's the code for that?" That sounds sort of a whole new kind of area of investigation. I'm not sure if the economics of it have caught up.

Dr Mark Hoffman: There's been a lot of progress towards making sure that diagnostic genomic analysis is billable. Also, for many of our large initiatives, we have philanthropic support for much of the work in the genome center.

Host: That's great. Okay. So, where do you think this is going? What would be your sort of vision? What are you going to be doing five years from now if all works out the way you want it to?

Dr Mark Hoffman: I personally see us still at a foundational stage right now. So, we're building all of the foundational pieces to be able to really benefit from data science. More and more each year, there's a need for, a lot of work in the ethics of artificial intelligence. And so, I'm also involved in initiatives related to that. For example, if you develop a predictive algorithm and the population used to train that algorithm was not very diverse, then you're really not serving your entire community. So, we're also very active in making sure that those data resources represent a more equitable population than may often be the case. And so, just bringing those topics to the forefront so that when we see these algorithms applied towards improving and accelerating care, that it really does so in a way that benefits everybody.

Host: So, I'm an Associate Professor of Neurology and I often advise medical students and residents about their careers. Do you see medical physicians playing a role in data science?

Dr Mark Hoffman: Absolutely. I believe it's a team process. So, we need subject matter experts with that clinical knowledge. We need the technical experts who can write the code to do the data queries. Neither one of those groups can do this work alone. I worry when I attend conferences and I hear data scientists telling physicians, "Oh, you have to learn how to code in Python." And my belief is you really don't. You need to be at the table and bring your unique knowledge to the data science process. So, I really believe it's a team effort.

Host: Okay. Well, we're just about out of time. Is there anything you'd like to add?

Dr Mark Hoffman: I really appreciate the chance to talk with you. I think we're really at a tipping point where data science is already starting to influence the trajectory of healthcare, and we're, excited to be a big part of that at Children's Mercy.

Host: Well, Dr. Mark Hoffman, I want to thank you very much for joining us on Transformational Pediatrics.

Dr Mark Hoffman: Thank you.

Host: For more information on research at Children's Mercy Kansas City, visit childrensmercy.org/research. To refer your patient or for more information, please visit childrensmercy.org to get connected with one of our providers. This has been Transformational Pediatrics with Children's Mercy, Kansas. Please remember to subscribe, rate and review this podcast and all the other Children's Mercy podcasts. I'm Dr. Andrew Wilner. Thanks for listening.