Transcript#

This transcript was generated automatically and may contain errors.

Hey there, welcome to the Paws at Data Science Hangout. I'm Libby Herron, and this is a recording of our weekly community call that happens every Thursday at 12pm US Eastern Time. If you are not joining us live, you miss out on the amazing chat that's going on. So find the link in the description where you can add our call to your calendar and come hang out with the most supportive, friendly, and funny data community you'll ever experience.

Can't wait to see you there. Today, I'm so excited to introduce our featured leader, Cori Campbell. She's a data scientist in the health data science realm. Cori, thank you so much for being here. I would love it if you could introduce yourself, tell us a little bit about what you do and something that you like to do for fun.

Great, thanks so much Libby and everybody on the team for having me to the hangout. I'm super excited to be here. Yeah, my name is Cori. I'm a health data scientist. My background is in quantitative epidemiology and health data science. For a large part of my career, I was in academia and research based in Europe and the UK and I was leveraging electronic health data and lots of administrative data sources to generate robust evidence with regards to health systems, analysis, drug safety, vaccine safety, those types of things.

And I am based now in Montreal in health data science consulting. So still in a very similar world kind of through a different lens, using a lot of the same methods and tools that we've talked about already that we've mentioned. And yeah, I would say for fun in the last year, I've been getting into enjoying the snow again, making that part of my, you know, self-care routine, getting back into skiing, all of that stuff. I love a good hike any time of the year, anytime I can get into nature and kind of feel that reset. So this is a beautiful place to be doing it. I'm super lucky.

Oh, it is extra beautiful. I have only seen pictures, but I'm very jealous. Someday I hope to go. Well, I heard tons of topics in there. I put some of them in the chat. So these are things that we can think about asking Corey about, including a PhD journey, EHR and admin data, academia to industry, drug safety, but also open source tooling for science and collaboration, obviously. And you are a globetrotter. You've been all over the world. So probably also collaborating across multiple time zones, multiple countries and multiple sites. These are all going to be fantastic things to ask Corey about.

I would love to start with the PhD question. Did you always know you were going to get a PhD? Was that always the plan? And what did you get your PhD in?

Oh, that's so funny. That's such a good question. I was just talking about this recently with some friends. My PhD was in, I guess, pharmaco epidemiology was really the main subject.

Wait, that's a long couple of words. Can you explain pharmaco epidemiology?

It's such a long word. And it's only a word I learned when I started a PhD in it, ironically.

It's kind of, I guess you could say epidemiology quantitative methods with regards to medicine use, medicine coverage, sometimes uncovering those mechanisms that are underpinning modes of action and that kind of thing. So I guess that's kind of a high level how I would describe it. And I kind of, I guess, not fell into a PhD, but I just had been really enjoying what I was studying on my master's. And it kind of just felt like the natural next progression. I did my master's in epidemiology in the Department of Population Health at the University of Oxford.

And then from there, I moved into the Department of Medicine for my PhD, which I did together in collaboration and co-sponsored by GSK. So that was awesome because I had the flavor of academic research and epidemiology and also the pharma side from them. And I feel like I was super lucky to learn so much on that process and work with great colleagues. But I would say that I was definitely guided by the team and my research interests. And I definitely didn't have a PhD on my radar from the beginning. I'm a huge proponent of PhDs, no matter what your motivation might be, but I wasn't necessarily always motivated to do one or thought that I would do one.

Well, that's the question that I thought was going to come up quickly was, do you still recommend a PhD in 2026 and the age of AI and everything? But also I wanted to make sure that I understood what epidemiology is, especially for anybody who's not in the science world. Epidemiology is the study of health and disease and populations of people, right?

Yep. Yep, exactly. It doesn't count how I describe it. Yeah.

Transitioning from academia to industry

Okay, great. Well, I have a question from Noor that came in on the Slido, and I would love it if Noor would unmute and ask that question live. And I see she's on camera too. Hey, Noor.

Hi, I'm Noor. I also came from the PhD land. And my question is, what are tips you would give for people transitioning from a more academic space or academic leading space to an industry space to not only successfully make that transition, but also thrive?

Oh, that's a great question. And I think it's still one that I'm in the process of answering, to be honest. I'm super, super green, super new in kind of the industry space. And I'm not sure if you made the same transition or if you'd like to. But I think that kind of seeing both sides of the fence less as one versus the other and looking more for the spaces where we're overlapping and opportunities to collaborate, I think those are a lot more abundant than a lot of people may realize. And that a lot of industry folks are super happy to be reached out to cold emailed and to look for collaborations in that regard, kind of in the similar style you would do in academia.

And I think that especially leveraging those opportunities for exposure at conferences and those types of things is really invaluable. But that being said, I was doing the bulk of my PhD during COVID, so I actually didn't go to any conferences in person during my PhD. So I feel like that actually was kind of a blessing because I just had to leverage email, LinkedIn, all of these other types of avenues for networking. And I feel like especially in this day and age, it's super easy to make those connections virtually. So I think one of my biggest pieces of advice is never be afraid to send the email, send the DM, just send it. You'll always be surprised what comes of that.

So I think one of my biggest pieces of advice is never be afraid to send the email, send the DM, just send it. You'll always be surprised what comes of that.

Yeah, I think that that would be my first piece of advice. Does that answer your question?

Feel free to follow up with anything if there's anything more I could add.

I think that talking to people is always the answer. It's always my answer. But it's really nice to know that it feels like people in this space are open to cold DMs and cold emails. Often their emails are listed, right? Like you go to the university website, you go to the department website, and their emails are there, they're public, and they might take them a few weeks to get back to you, but they generally do. That's my experience when I was, you know, looking at all these things. Also, look for professors of practice who are working both in academia and industry. I find that they are so easy to talk to, so easy to sort of like approach and say, I just need some advice or some help. Thank you so much for the question. We have a question from Natasha, which is, what is an average day or week look like for you?

Day-to-day work and tools

Oh, that's another interesting one. I think that the pace of my day has kind of changed rapidly from academia to industry. I think that might be one of the major shifts that I've experienced so far. I had to kind of get used to now arriving to the office and being ready for my whole plan or my whole timetable to kind of shift majorly and for priorities to almost seem like they have changed overnight. And now I'm focused on project B, whereas I thought I would be coming in and focusing on project A and specific deliverables. And that's been an excellent lesson because I feel like in academia, because the timelines are kind of so much longer, it can be kind of easy to stay very planned and very regimented. And especially if your work isn't particularly interdisciplinary or you're not relying on a lot of other people or moving parts to deliver things, it can be really easy to work kind of according to your own schedule.

So the really necessary collaboration and reliance and teamwork that we have on each other in industry, I feel, has been super rewarding and a new skill that I've learned really quickly.

So on any given day, I'll come in and I usually have a couple tasks or deliverables that I know need to be delivered within the next few days, kind of trying to plan week by week. But there's always things that come up, always fun surprises and changing priorities. A colleague might ask you for help on something urgent. That's awesome. You know, we do have typical seasonal flows as well, similar to all industries. But yeah, it's super dynamic and exciting. I'm someone who likes to stay busy and feel challenged kind of all the time. So it's definitely been satisfying in that way.

Thank you. What about the context of tools? Like what tools do you find yourself working in day to day? Are you in R, Python, SAS, SPSS? And I'm curious what you use for collaboration as well.

Oh, well, that's another great question. Previously, when I've been doing a lot of disciplinary work, especially spanning different countries, we've relied really heavily on RStudio and GitHub for sharing our workflows. I feel like that's something that I really love and is still near and dear to my heart. I'd been undertaking federated analysis before in a previous role where we would share the workflow through a GitHub repo, and then each data site would pull that and run it locally. And we might have some little nuances or debugging to undertake kind of at each local site. But by and large, the methods and the processing pipeline was the same across all the sites. And that was so awesome. And it's really where I kind of fell in love with open source and transparent pipelines.

So I'm still trying to get managed things every opportunity I get. Still a lot of stuff day to day in R and RStudio. Trying to, yeah, make Positron my bread and butter IDE right now, which I'm loving. Loving learning how to navigate that. That's been awesome. I'm also new to SAS recently in the last year. So learning a new language has been awesome. And definitely feeling challenged in that regard. And then in my personal time, I try to keep up to date with, you know, novel methods and open source tools and languages and stuff. In my spare time, I'm currently trying to, for example, do some sports analytics stuff, develop some open source stuff there in Python. So it's really a mix. I feel like it's awesome to stay current and try to share things as much as you can. But always across, you know, all of the places I've been working, always sharing code and workflows with colleagues. And it's something that I love and always want to be doing.

Open source and sharing code

Do you feel like that's a skill set that you had to learn? That openness to sharing and showing people something that feels a little bit vulnerable sometimes, like other people seeing your code, you can be like, oh, I don't want to look at that. Because I feel like when I came out of school, I had been teaching and tutoring so much that I was really used to just working out loud, showing everybody everything I was doing. I was used to live coding already, screen sharing. And I met others who really, really, really struggled with just showing people what they were doing. And I feel like it held them back and it helped me. What do you think about that? What was your journey there?

Oh, that's so interesting. It's awesome that you were kind of habituated to that right away. I think that's an excellent skill.

I got lucky. Everybody go tutor. If you're in school right now, go tutor other people. It'll teach you some skills. Yeah. Live coding, live debugging is, you're right, it's intimidating at first. But I actually, I love doing that with my colleagues. Any chance we get to debug something together? I really love to do it. I think I totally agree that it feels really vulnerable, especially if it's kind of a workflow and analysis you've been working pretty independently on. And then all of a sudden you feel like, oh, I have to open up this little world that I've just been programming, this whole system that I've programmed myself in.

I think especially these days where I've felt the biggest switch is with academic journals. I think the tide has kind of turned relatively quickly there where you are really often now asked to submit your analytical code or a repository alongside your manuscript. And that's been a big shift. I'm absolutely in support of it. I think it's all for the better. And it's, I think, changed the way that a lot of academic teams program their analyses from start to finish and thinking from the beginning, how can we make this accessible and legible to other people, reviewers, potential collaborators, that kind of thing. And it's always so, it feels so nice on the flip side when you learn to love that workflow, when someone uses something you've written before and they tell you how it's impacted their workflow or they give you tips or you know, submit a pull request or those kinds of things. It's so much fun and it really feels

like sharing. So you're right, there was kind of a barrier to entry there, but once I came around to it, I really learned to love it. Wonderful. Well, Isabella put a link in the chat as well to something that Cori mentioned earlier, which was Positron. If you are coming from RStudio and I'm sure Cori came from RStudio, loving RStudio and using it to Positron, I still love RStudio. I still use it as well. But if you are coming from RStudio to Positron and Positron looks unwieldy or you're worried about it or you have questions about the change in your experience, Isabella and I did a walkthrough with RLadies Melbourne and they have a video up already of the recording. So please, please go watch that.

Okay. Hubert had a question on Slido and I would love it if Hubert could ask that live. I'll stick it in the chat as well.

Sorry, I was slow with my unmuting over here.

I am so many. I have my two screens and I lost it. There we go.

Hi, Cori. My question is, what are some of your data challenges? Is it, you know, you've got a wide range of collaborators in many, you know, all over the world. Do you have, is it lack of kinds of data? Is it things buried in unstructured data? You know, I live with a lot of clinical notes and all the good stuff's in clinical notes. Yeah, stitching different data sources together. So what are some of the data challenges you face and how have you approached those?

Oh, that's really interesting to think about. And it's nice to hear from someone else who appreciates the challenge of clinical notes and getting useful information from that. They're so, they're so rich, but they're a challenge to work with for sure.

I think that in general, across my health data science experience, a lot of the big challenges that I've encountered kind of relate to how different health systems operate and how that information flows from general practices or doctors or hospitals into those records and how different linkages are undertaken, not just across countries, but even across different regions within a country, this can look so different place to place or hospital to hospital and understanding where things are coming from and why has been huge. And I think that especially for a lot of lab data or things that we would measure for patients or people kind of on a repeated basis on a regular basis, you know, the tools to undertake those measurements, the assays, those labs, the ways that things are measured can be so variable.

So just learning really how we can reliably identify our populations of interest or a study sample, given all of this nuance across data is massive. In my previous team in the Netherlands, we were harmonizing data for up to 10 countries for any given study. And that was so much fun. And we had this extra challenge about language and translating things also. And, you know, different things can be collected in some countries or some settings, not in others. In some countries, it comes from a general practitioner or a family doctor in other countries, things come from the hospital. So I think it really taught me to think high level about a health system and how information flows through different health systems.

The clinical notes piece is another beast that I haven't had to tackle majorly yet. I'm sure it's coming and I'm excited for it, but I'm always so interested to hear about the methods other people are using for that because they seem so diverse and interesting, but it seems challenging for sure.

Well, Hubert knows something about data harmonization and data modeling across many different types of data contributors. If anybody's interested, go watch Hubert's episode of the Data Science Hangout as well. It's wonderful to be asked a question by somebody who has empathy with what you do, isn't it, Cori? I can tell this person knows that this is a challenge.

Writer's block and journal articles

Okay. Well, update on the poll that I put, I put the results into the chat. I'm going to go ahead and stop that one. I asked, do you love being asked for help by coworkers? Because Cori mentioned part of her week sometimes is like being asked for help. And she was like, that's always fun, right? I think so too, but not everybody does. So we had about 17, 18% said, no, about 82, 83% said, yes, I love it. Count me in the yes. Nerd snipe me anytime. I have to stop myself from being nerd sniped because if that could be my whole job, I would love it. I see some people shaking their heads, yes, in the chat as well. Thank you, Hubert, for that great question.

We had an anonymous question. I'm so curious what you think about this one. It says, I have been told at work that I need to write more journal articles. I'm good at doing the analysis, but I get writer's block when writing anything up. Do you have any tips? Oh my gosh, I totally empathize with whoever sent this question. And I feel like it's a sentiment that's shared kind of widely across a lot of research jobs, research settings. I hear a lot of people say that they're allergic to writing, which I think is not uncommon.

It's a great question. I feel like on a high level, trying to think about the narrative or the story that your analysis tells can maybe help you find that motivation for writing sometimes. That's a piece of advice that I think I received in my undergrad and really stuck with me, is that find the way to connect your findings to the overall big picture and understanding what that is and how we can make it interesting and accessible to the people who are going to be reading our article, which really helps, I think, distill key messages and guide the flow.

In terms of practical tips, I guess everybody kind of finds what works for them. So it's maybe a lot of trial and error finding what works for you if it's about making a scaffold or an outline at first, and then kind of filling in the pieces as you go. Something that I learned to do, and I'm curious what other people feel like is their process, because I don't know if my process is shared necessarily, but I front load my methods. So that's the first thing I write with a research article. And normally they're huge, and they're long, and there's way too much detail. And then with my co-authors, we need to go through revisions, and they usually end up being cut by 50%. But it's almost like a guide for me as the analysis is ongoing and being iterated. Okay, why did we do this? Okay, now we're going to go change this sensitivity analysis or that one. And it kind of feels like it lowers the burden for the article, because then when I go to write everything else, it's like I have a decent chunk written already. That's been really helpful for me.

I love it. We have lots of people in the chat showing empathy for writing is hard. I love writing. I like editing better. But I don't like using LLMs to write. It's not a thing that I enjoy. I know some people do. There are lots of different apps nowadays that help you with dictation as well. Like I can talk someone's ear off. It has to be a person though. So I have to have my husband come in and sit with me and just sort of nod, even though he has no idea what I'm talking about. Hire your friends to get on Zoom with you and be like, can I just talk through this thing? And sometimes you'll find yourself just spilling out the entire thing with no blocks. Whereas if you had just sat there with a blank page, it would have been really hard.

Pharmaco-epidemiology timelines and decision-making

Oh my goodness. Well, thank you so much for that. Great question. Anonymous, ask her whoever you are. It was fantastic. We loved it. Mike K. Smith asked a question and it has an asterisk, which is rare for Mike. Mike, I'm going to read this for you. It says pharmaco epidemiology feels like it may take a while to build and test predictive features of the models. How do you make sure you get answers in time for decision-making? That's a good question. What is your horizon for just time horizon for decision-making for projects?

That's a really good question. And I almost feel like I've been a little lucky thus far because a lot of my pharmaco-epi work has been focused on defining disease populations and kind of unmet clinical needs. So maybe I've been a little lucky in that I haven't been tasked with major prediction challenges in the pharmaco-epi space. So I'm really curious to hear from other people and how they're tackling that.

I think maybe an underlying theme across these kind of behemoth pharmaco-epi analyses is really about understanding your data source and the strengths and weaknesses there and what might be missing and really how that's influencing the sample that's being selected and then thereby the generalizability of the results and how quickly you can deliver the product that you want. Because I think you can always come to the most accurate answer possible. The methods can always be more sophisticated. It just needs more time and probably more of that investment to understand the data you're working in. Whereas I feel like in a lot of industry settings, and maybe this is more common in industry pharma, is that we don't necessarily have the long time horizons that we'd have in academic research. So it's about understanding exactly the message that we'd like to deliver right now and what we might like to deliver in the next quarter, the near future, and working kind of incrementally in that way.

I think you can always come to the most accurate answer possible. The methods can always be more sophisticated. It just needs more time and probably more of that investment to understand the data you're working in.

But with that being said, I have enjoyed the work that I've done so far, which has been a lot of population at risk identification and that kind of thing. I'd love to undertake some prediction in the future.

Yeah. Everybody does different sorts of tasks with their data, right? I think that this is a big difference, that time horizon, and also the pace is a difference between academia and industry as well, right? Where you have different goals. Because I heard you talking sort of around a business question in a way. I know that health and epidemiology, that doesn't go to one-to-one, but that's what I hear from people coming from academia to industry when they're not in health spaces as well, is that they go from a rigorous question that has to do with the data, the subject, nothing else, to I'm answering a business question and I'm on a timeline and the rigor is not quite as important. Have you found that shift in the health space as well or no?

Definitely. I think understanding how the priorities are shifting across the two environments is a major learning curve for sure. And that's something that I've heard over and over again across people who are moving between those spaces. And I think that maybe the longer you spend in both spaces, it's easier to see how they kind of go hand in hand. But there's definitely definitely different, I guess, priorities with regards to analytical timelines and objectives and that kind of thing across both spaces. And yeah, it's so interesting to being in health specifically, maybe that makes them feel like more one side of the fence versus the other, where if you come from a different academic discipline, maybe they tend to align better. But it's always really interesting to think from, I think, a pharmaceutical perspective in general, because a lot of that common thread, I think, is addressing unmet clinical needs, addressing unmet patient needs, understanding really comprehensively what's our disease population of interest. Those types of themes, I think, span across the spaces, which is really nice to keep in mind those principles.

Awesome. Thank you. Mike, thank you for that great question. I'm sorry that you weren't here to answer it or ask it yourself.

Biking data and personal tracking

Well we have a fun question in the Slido which is, I see your road bike in the background. Do you use your biking data for anything? Do you track your biking data? Oh that's that's super interesting. I don't have any biking data for myself because I don't have a watch or like a smart watch or anything like this and I know that it's kind of unusual these days. I would be really interested to start to play around with that data and it it did kind of motivate me a little bit to think about like a smart watch or a smart device. I've heard different preferences from different people with regards to maybe like the watch they choose and how granular their health data is. I have a couple friends who have whoop watches I think and if I understand correctly the data you get from your whoop watch is like super detailed and granular and you can really kind of do analysis on your physiology. So that might be something something coming for me in the future. I think for now I'm I'm happy to stay a little blissfully ignorant with regards to not tracking that stuff and not not uploading to Strava or anything but but it would be really interesting to just start to dig in into my my own data. My friends, I love seeing my friends stuff on Strava. I just I'm not putting stuff myself you know.

There is nothing wrong with that although the wealth of data that you can get about yourself is sometimes how how would I say surprising? I'm like oh wow I am allowing my phone to track a lot. Some of it's really fun. I have several years of my own sleep data because I used to use the sleep app that would record sounds while sleeping. Ostensibly record your snoring. It never caught me snoring. It caught my husband snoring all the time so that was a little messy data thing where I'm like oh it says I snored all of these days right but I generally don't snore and I would listen to it and it would definitely be my husband. So if you are recording your own data know that there's some mess in there too probably for you to clean.

would love to know a little bit about the ways that you've used Shiny because I love Shiny and I know that you've used Markdown outputs. I'm curious if you use Quarto or Markdown or any other things. Do you have any recent Shiny work that you could talk about or any past work that was really fun? Oh I love this question. Yeah I've actually been like lucky to use it with colleagues across a few different examples. I haven't yet started in Quarto. I did have a colleague who he'd moved all of his like teaching over from Markdown to Quarto so I feel like that's kind of an ex-challenge that I'm excited for.

But in regards to how I've used it in the past a really cool use case that we had at one of my previous jobs actually relates again back to the kind of harmonizing international data across different countries and health systems. We were mapping all of the data from the different hospitals and the sites onto a common data model and so an intermediate that a brilliant colleague had built was kind of a Markdown score sheet or data quality check about how well the data at a given site was formatted right now to be mapped onto the CDM, kind of what changes would need to happen and the different coverage and things that were missing that were key that we wouldn't really expect to be missing like very very kind of essential parameters like date of birth or postcode or these kinds of things.

So that was I think the one of the first use cases I'd seen of like Shiny or Markdown output where I was like wow this is having such an impact and it's so important to our workflow and I think since then trying to find small ways to try to work that in it's such an invaluable tool to I think share results with people who maybe don't necessarily come from a data science or programming or quants background and to give them something that they can kind of play with and see what they need and find what they need I love.

it's such an invaluable tool to I think share results with people who maybe don't necessarily come from a data science or programming or quants background and to give them something that they can kind of play with and see what they need and find what they need I love.

I'd love to do more in my spare time. I had one friend in her spare time she had built like an interactive Shiny for all of their household expenses like their bills and their rent every month and she and her husband were kind of managing their finances that way which I loved so I think a small personal project like that too is something I'd love to do and kind of a low barrier to entry introduction to those things as well but yeah I and if I oh yeah do you have any cool are you are you doing a lot of Shiny Libby in your in your free time? In my free time not anymore although I did I do love doing Shiny apps I have done Shiny apps for my job but I only ever got to work on really boring ones or ones that I didn't agree with right where like I didn't get to build it I just had to maintain it and I didn't agree with the premise of it or the way that was designed or whatever but I really like the idea of building a Shiny app for yourself especially if nobody cares about it or uses it that's my favorite thing I haven't I have one that I've shown people on the data science lab before that just makes an art piece that's it has absolutely no use case.

Mike says Libby has free time yeah not a lot of it but I love the idea of tracking finances I love personal finance that's like a niche of mine. I shared a poll that said going back to a previous topic that said do you or do you currently struggle with sharing your code or sharing your screen while coding and I wanted to get everybody an update on that we had 30 votes about 50% of people said yes I struggle with this type of anxiety and only 13% said no I never struggled with this so you are not alone if you have struggled with this or if you are struggling with it but the most hopeful thing is that about 38 39 percent said not anymore but I did in the beginning so maybe there's hope that you can practice your way through it and it'll be okay oh yeah it's changing people are voting as I'm talking about it now.

um okay let's head back over to our questions in Slido and I see one from Zach that has an asterisk so I will read it and it says how do you see AI affecting health data science what a good question that's a really great question. I think I feel like we're only just now releasing how it might have a major change for maybe health system processes I know there's so much incredible literature out there like really quantifying robustly all the different use cases or utilities we're integrating it but maybe now we're really understanding how how we can do that.

AI in health systems

I think the paper that comes to mind is and maybe I'm sure a lot of people people in the hangout are familiar with it a lot more than I am but it's I think it was they'd used an LLM to triage emergency admissions and kind of weighing that up against kind of human triaging and the findings were just remarkable and so I guess thinking about how we can use these tools alongside healthcare practitioners and and help to integrate them into workflows so yeah everybody is benefiting the patients the healthcare providers or practitioners you know there's less on their plate they're under less stress in any given moment I can't imagine what it's like to be a doctor working in an emergency room especially on a weekend when when the waiting rooms overflowing and so just having a super robust tool or something to add to your arsenal that can really take that load off take that burden off I think that is the way that I'm really excited about AI impacting I guess you know health systems research and health systems in general.

yeah I guess and then I guess there's all of you know maybe the ways a lot of us are more familiar with using LLMs for writing or helping us code boilerplate stuff all of those use cases there's so many so I guess it's about like the scale of the impact as well but if I think high-level I think there's some really exciting opportunities there.

awesome yeah I saw Luke in the chat pop up at the mention of an emergency department I saw that we had several epi folks in our chat which is awesome and lots and lots of like empathy with a lot of the the topics that we're talking about which I love and also a lot of people saying that they build Shiny apps for fun for stress relief you know to do all kinds of stuff and also I put another poll in the chat I can't stop with the polls they're fun Cory has so many like poll worthy questions and topics today about whether or not anybody tracks any of their own data so go answer that poll we are at like 11 or 12 votes here but a lot of people for the most part saying that they like tracking the data but they don't analyze it themselves this is a perfect opportunity for a fun Shiny app y'all that has no consequences that you can go play with.

Administrative data across health systems

all right let's take a look over it I lost my Slido one second everybody I had a question for you about in the beginning you talked about administrative data and I was wondering what administrative data is and what that looks like and whether or not you are seeing differences in that administrative data across all these different places that you work because you have worked in Europe you've worked specifically in the UK also in the US but you're in Canada right now do you see a difference between the administrative data and all of those places because we had Vitas who is from the central statistics office in Ireland talking all about Ireland specifically and we had some questions I remember that day about how it compared to the US and he was like I don't know but you might have some comparison right.

oh that's super interesting I yeah I guess I do I use the term administrative data as a super broad umbrella but I guess it means something kind of specific for me actually which is really data related to health systems any kind of part of a health system that's collected routinely and kind of passively but isn't collected necessarily for the purpose of research I think in a lot of health systems it's collected for the purpose of billing and and payment or reimbursement and I think that this was something that blew my mind on my PhD in the UK and and my work thereafter because I was like we have like population level data it's so granular and it's not even collected for us to do research on it but it facilitates such incredible research so that that really inspired me.

and then of course that that comes with the caveat that because it's not collected for research it often requires so much cleaning and harmonization and processing I I feel like the the typical split I hear is that we spend 90% of our time cleaning and 10% of our time actually running our analysis and I feel like I definitely empathize with that and share that sentiment.

and yeah in terms of geographical variation it's in my experience it's been it's been pretty broad and it's so interesting to think about how that reflects the health system design in general so I guess the example that I'm most familiar with is the UK where it's you know public system and it's free at the point of access or accessible kind of at the point of care and I think that that makes kind of the UK so exceptional in the way that they're able to link data across primary and secondary care sources and link to external sources like census data or Office for National Statistics records death records cancer registry all that stuff it really is such an incredible place to be doing health research.

and I I really enjoyed digging into that data but then if I think about other health systems I think especially a lot of insurance based systems maybe the purpose of collecting that data and the payment mechanism looks a little different and that also has implications for how to clean how to process and what kind of population coverage we have and all of these things so it it really can impact your research question too and the data source that you're choosing kind of the health system design and and how the system works.

and I think it's really exciting for common data models it's a really exciting time I'm thinking of the OMOP common data model I I don't know if anybody in the chat or anyone here is familiar with that but I'll thinking it will be oh exactly yeah for sure yeah yeah yeah so thinking about how you know such different health systems in different countries are all mapping onto the same common data model it's it's an incredible challenge but it's been so much success so it's really awesome to see that.

oh awesome and Dan just added Odyssey which is there's a link there about data data standardization this is a really big topic I think that I before meeting so many people on the hangout who work in data modeling in data harmonization I don't think I really truly understood that the challenges that go along with modeling data not statistical modeling not machine learning modeling but modeling data absolutely something that I have massive respect for.

Instrumental variables and causal inference

I had a question while you were describing that and I'm gonna ask it really quickly so we can get to a career advice question it sounds like you were talking about lots and lots of variables that were not meant to be variables right like these were things that we do collect and we can look at but they're not meant to be health data which brings up instrumental variables and I thought maybe you could talk a little bit about what an instrumental variable is and some of the ones that you see in your work because if anybody hasn't learned about them they're fascinating.

sure yeah and you know Libby I might love your input here because I haven't been doing instrumental variable analysis for a long time we're thinking oh I'm sure there are lots of people in the chat who are who can give me a lot of pointers but how I'm kind of describing the methodology or the framework but I think that the way that I think about it most frequently is kind of in regards to how we emulate a trial or simulate a randomized trial using this administrative data there's a methodology called target trial emulation which I was using quite frequently to kind of design my observational studies previously when I was working in academic research.

and I think that thinking about instrumental variables as kind of the instrument for our exposure I guess in that regard is is really is really the context in which I think about it so just thinking about kind of the mechanisms of exposure and the different exposure groups that we're interested in and instrumental variables analysis I think I think some people are using it in target trials I've actually more commonly used inverse treatment probability weighting which is kind of just another way to ensure that we can make causal inferences basically with our study design it's super super convoluted and really smart folks are developing those methods but really confusing I think the first time I learned about instrumental variables I was like oh no I don't even think I can explain this everybody should go Google instrumental variables though if you're interested in statistics causal analysis hmm complex statistical analyses but I wanted to end in our last eight minutes with two things one being

a shout out to Phil Boucher who works at Posit who met you at a conference which means that you got to meet us which means you're here with us you talked earlier about tips for people coming from academia to industry right and I always say go where the people are I wanted to ask you like how did that happen what how many conferences do you go to every year do you go in person do you attend virtually how did you bump into Phil and just the importance of talking to people and how it can lead to things like this yeah it's a great question and I think I totally agree with the message you know go where the people are if something's interesting you something looks cool really follow it and and you know don't have any qualms about reaching out so I think

Phil and I actually cross paths virtually about the pharmaverse packages actually in our because I'm really interested in those especially thinking about regular regulatory grade evidence and that kind of thing and developing open source tools together with pharma something I'm really passionate about and I'd like to move more into the space actually so we connected with regards to that and then you know it happens so often that if someone looks like they're they're doing stuff that is interesting to you there's probably so much more they're doing that you're gonna find interesting as well so once you start talking you know you find so many shared areas of interest and it kind of can so easily snowball from there so I'm super thankful that we connected and I got to meet you guys and yeah trying to hang out um and I think it's it's so it's so important it's it's such a opportunity especially again with all of the ways we have virtually now post COVID to connect and how that's become a lot more normalized I was really unfortunate that a lot of people had to stop you know those in-person meetings and conferences and things that we that are such a great opportunity but at the same time we became so creative about thinking about how we how we do that I would love to be going to more conferences always I always would love more I really love that side of things meeting people in person and chatting and getting excited about somebody else's research or findings at a poster or or a talk is just one of the best things for me so always would love more and I think that if I'm not mistaken R has some really exciting ones coming up I'm thinking of R/Pharma I'm sure you there are a lot more on the docket that I I'm not aware of but yeah R/Pharma something I'm really excited for coming up I love it R/Pharma and R/Medicine are both wonderful it doesn't matter if you're in pharma or medicine there are so many great talks so many great topics and amazing workshops that are accessible affordable and with Posit Conf you can attend virtually as well in person is always amazing I went in 2023 in Chicago it was amazing well we have five minutes left I'm wondering if you have a piece of career advice you would be willing to share with us something that you wish you knew before you got started or that you like to tell other people to help them along their journey oh oh that's a great question I think I think that I I've always found it so helpful to always just ask the people around me you know supervisors or our colleagues or everybody about their journeys and what they found worked for them and kind of take all those answers and understand what works for you and what applies to your journey and remembering there really is no one way or one one normal route or there's no one way to cut a cake you know everybody's path looks so different and everybody's timeline looks so different I think remembering that especially maybe because I came from an academic background where our our steps and our timeline can look really regimented and at the beginning I thought oh you know I guess it's there's there's one way to do something and if you leave academia you know you can't come back where it's really you're either in or you're out and then now that I'm kind of on the other side I see there's so many opportunities for collaboration and continuing to stay in both worlds or multiple worlds and just finding ways to be creative about building the career that you want and what works for you it's so much about seeing what's possible and how you can make kind of your ideal scenario a reality because I think lots of things are possible and we think outside of the box you can always find a way to get there

there's no one way to cut a cake you know everybody's path looks so different and everybody's timeline looks so different

that is the best advice I have heard in a really long time because I am the type of person who if I don't see something illustrated or scaffolded for me especially growing up I didn't realize it was an option so a lot of the women in my life especially growing up in the 80s and early 90s I didn't see a lot of people who were working in science working in math I didn't have math teachers who were women or science teachers who are women now my friends are women who are science teachers and math teachers and when I got to college I had two women who were my math teachers who completely turned my life around and I went oh my gosh I'm actually really good at math and I didn't think that I was before representation is so so important in that way but yes I love the advice of like think outside the box I will add to Corey's advice and say if you struggle like me with seeing the gray points between A and Z and all the options in between like she said go talk to other people and ask them what they do what they think and options because they will come up with more creative like oh yeah you could just do XYZ and you're like oh man I didn't think that like one time I this is this is like two months ago my ice maker is broken and I ran out of ice and I couldn't fit a bag of ice in my freezer and I texted my friend and said what do I do I'm so frustrated it's so hot and we don't have any ice and she's like just make ice in an ice tray I was like wait what she's like just put water in an ice tray and put it in the freezer you have one of those right and I stood there in my kitchen I was hot all over I was so embarrassed I was like I didn't think of that I truly didn't I'm a brilliant person I think I'm very smart had no idea ask your friends very vulnerably what the gray area is that you don't see

okay everybody Corey we've reached the top of the hour I'm so glad that you could join us today talk about health data science I hope you had a good time oh it was so awesome thanks so much everybody and for the privilege of joining the hangout it's it's a pleasure we loved having you everybody we hope that you had a good time I hope that you'll join us again at the data science hangout next Thursday and also on Tuesday at the data science lab Isabella and I will be there talking about technical topics as we always do we cannot wait to see you also join the discord server go hop on go find us if you want a way to just always be at the hangout we are always there all of us are hanging out having fun and we enjoy this time with you so much if you would like to share the chat save the chat there's three dots at the top of your chat if you are on the browser it might be at the bottom where it might not work I'm so sorry browser based zoom is not ideal but if you're in the app you can save it and Isabella just put the meeting saved chat right there so you can download it hopefully as a txt file thank you everybody we love hanging out with you and we will see you next week bye