JA: Hello, and it’s my pleasure to be able to speak with again, a HITLAB Symposium and Columbia University Digital Health Strategy speaker, Kelly Zou from Viartis. Kelly, welcome to the program. Those in the audience that don’t know who you are, we’d love to get a 60 second intro about what you’re currently doing.

KZ: Thank you, Jerry. I guess this is not my first time being here and I’m really enjoying sort of this interactive forum, and to talk about important topics in healthcare, particularly in my space, that will be sort of data science. I am the Head of Global Medical Analytics and the Real World Evidence at Viartis. Viartis was formed about two years ago through a merger between parts of Pfizer, and also my own. So, I oversee in terms of medical side of analytics, and within my organization, you have various functions that includes real world evidence, as well as more focusing on outcomes research side. So thank you for having me again. I also want to note that the opinion expressed in my conversation from now really represents my own, may not necessarily represent my company’s.

JA: No, of course not. Thank you so much, Kelly. And let’s – well let’s dive straight into it and talk about health equity, and the importance of equity within healthcare related to real world data and evidence that you that you work with on a day to day basis.

KZ: Yeah, so first of all for those of you who have heard about definition of real world evidence, really this came from the 21^st Century Cures Act, that was 2016. So subsequently, the US in particular, according to the regulatory sort of definition if you look at the FDA’s website, such data are really related to patient health status, or the delivery of healthcare routinely collected from variety of sources. So the presenters before me talked so much about different kinds of data, for example, electronic medical records claims throughout through billing registries and other sources such as digital tech that other speakers talked about. So these are type of data, as you can see, because those data are collected from patients, we really need to take a patient-centric view. In addition, data eventually are being used to gather insights, to generate evidence, whereas in terms of that particular evidence generation process is really to think about a framework of benefits and risks in terms of eventually the products derived from the analysis of such data. So when you put all these elements together in terms of patient-generated patient perspectives, as well as potential benefits and risks in that framework, we really need to care about variety type of patients, patients who are representative of the future use of our products, and such patients really we need to think about critically, the health equity aspect.

JA: Yeah, no, absolutely, you’re absolutely right, Kelly. I think that there’s definitely a lot to think about when it comes to merging those two pieces together. And as you mentioned, with some of the folks that had been on earlier talking about various different things, one of the things that did come up was breakthroughs with AI and digital health, but people still, the wider public still not having access to those kinds of basic needs, if you like. And there’s a lot of talk around social terms of health and how that plays a critical role in identifying that sort of data gap that you work with. But how can that data be complementary to such a wide spectrum of patients in your view?

KZ: Yeah, thank you, Jerry. You’ve mentioned about two important sort of concepts that we hear quite often these days. The first one is really social determinants of health, in short, SDOH. And if you look at government’s definition, so there’s a website by the US government, which is called Healthy People 2030. So we’re not just looking at now, we’re looking forward. In terms of this forward thinking process in terms of healthy people, we really think about what kind of conditions people have. When I say conditions, I do not necessarily just mean health conditions. It’s the conditions of the environment. The definition of the SDOH is the conditions in the environments where people are born, live, learn, work, play, worship, age, and all kinds. It’s a range of these elements. And these variety of elements can be sort of categorized in terms of different domains. When you think about this kind of data really ultimately related to access to care? Well, the first domain is really economic stability. Second one is education, access and quality. Third one is healthcare, access and quality. The next one is neighborhood and built environment. Lastly, social and community context. So we focus a lot about healthcare access and quality. That’s what we do in healthcare. However, patients as our ultimate people and also consumers, they really face with a whole set of decision making factors and these other dimensions really come to play. So ideally, we want to know a lot about patients. And in reality, there are several challenges, right? In order to harness that kind of data, the first question is really holistically, do you have this kind of a sort of a linked connected data, which is in the US, it’s quite siloed in terms of looking at various databases, getting all that different aspects is not trivial. The second thing is, even if you are able to link electronic medical records with patients’ billing, maybe some economic background information, you may still not get the patients who are not in the system who may not have the insurance. So those patients, how do we really get access to their information? That’s when communities and also community health centers are so important in terms of potential collaboration opportunities. And finally, after you link those data, how would you ensure the privacy under HIPAA? That’s where ensuring that when you link all that information, you don’t have the risks associated with really getting access to all that massive amount of data. So I think I’m just talking in terms of the data space. But when you also mentioned another element, which is artificial intelligence, that’s how you can use such kind of information. A lot of information is really structured data, as I mentioned, but the vast amount of data are really unstructured. Early on, another speaker talked about text languages talking about natural language processing. When you think about text images and also potentially conversational way of interacting with information through large language models, that’s where it’s important that you want to think about potential biases and also the representativeness and trustworthiness of whatever way, however way you analyze in terms of the algorithms, in terms of the transparencies, and that comes with a lot of the considerations about new technologies. I also want to mention that just this past Monday, I was very fortunate to attend the very first inaugural face-to-face workshop held at Columbia University. So this was actually the large language model workshop co-hosted by the American Statistical Association. It’s a section on text analysis, the local New York budget chapter, as well as Columbia University. So we really had a chance to discuss many facets and elements about data, about algorithms, about the benefits and also risks.

JA: No, that’s lovely. How was the turnout, by the way?

KZ: It was great. It was roomful, it was packed. And we even had a lingering sort of a face- to-face sort of interactions. That’s also one aspect that people also really miss in terms of data, it’s not just something dry, but also has so much context underneath. That’s why the social determinants of health also talk about the contexts, right? Not just say, you look at the face value, what is really the kind of conditions that people are facing with?

JA: No, absolutely, absolutely. And I know we are really at the top of our segment, but as a last question, I think that the audience would love to know if you can sort of explain how better access to telehealth and e-health can be beneficial in with our buzzword that we always love to use around patient-centricity.

KZ: Yes, first of all, I think there’s a reset in terms of sort of digital health post-pandemic. During the pandemic, because the telehealth gave so much potential that we could interact. But on the flip side, since we are using Zoom, there’s also, as we know it, the Zoom fatigue and also the bedside doctors, the interaction in the office, that’s also really that human side. I don’t want to say just the patient side, but it’s also the human side of trustworthiness, right? And I think post-pandemic, there are several values. For example, decentralized clinical trials, where we talked about social determinants of health and some patients, such as elderly, such as those who have some cognitive impairment, mobility issues, and then this kind of a potential can really open up in terms of enabling not only patients, but also caregivers to access to our healthcare system. And also some of the recent reforms in terms of being able to think about out- of-pocket expenses, whether that’s being waived or whether that will really be helpful in terms of patient-centric perspective to really let them be able to access to the care sooner rather than wait until when it’s too late. So I think that’s really, really timely.

JA: No, that’s really, really wonderful. Kelly, thank you so much again for your time to spend with us today again. It’s really lovely to have you back on the program, and no doubt we’ll have you back on again.