Katherine Gerton @ Centene | Making sure you know other teams exist! | Data Science Hangout

Welcome to the Data Science Hangout. Hope everyone's having a great week. I'm Rachel. I lead our pro community at Posit. This is our open space to chat about data science leadership, questions you're facing and getting to hear about what's going on in the world of data across many different industries. And so we're here every single Thursday, same time, same place. So if you're watching this on YouTube, into the future, if you want to join us live, you can use the link in the details to add it to your calendar and join the party with us.

I'm so excited to have Catherine joining me as our co-host today. Catherine Girton is Manager of Data Analytics and Reporting at Centene. And Catherine, I'd love to kick things off with having you introduce yourself and share a little bit about your role and also something you like to do outside of work.

Yeah. Hi, I'm Catherine Girton. As Rachel said, I'm the Manager of Data Analytics and Reporting at Centene. In that role, I actually support our health insurance, Affordable Care Act marketplace products, supporting the risk adjustment mandate. When I am not working, I play volleyball and I do a lot of sewing. I decided I'm going to get actually good at sewing more than a rectangle and started making my own clothes.

I was really proud actually of a pair of jeans that I made that I embroidered with like outlines of different fruit. They don't currently fit, so they're not getting a lot of use, but they were really great when I made them.

Team structure at Centene

Something that we had talked a little bit about and it was shared in advance of the Hangout was about the power of embedded teams for kind of bridging the gap between a lot of cool data science the team's doing and then meaningful business insights. And I thought it might be helpful to just understand a little bit about the data analytics team structure at Centene and what that looks like.

So we have, it's kind of a hybrid between your traditional centralized team and your embedded teams within your business areas. We have a Centene Technologies corporate IT kind of data science, data analytics, where they support a lot of the tech stack, but also facilitate conversations with business teams that may not have their own analysts. And then we also have teams like the one that I'm a part of within the actuarial department where we are a data science team. We are involved in gathering the requirements, building the tools to support that business process, and then interfacing with the centralized teams that do control and maintain the tech stack for us.

Not having to know how computers work or how software is built and installed and maintained and all that is like great for me. I can spend my mental energy focusing on, okay, these are the data science problems that are facing my business unit and can really dig into those problems.

Risk adjustment explained

Yeah, so when the Affordable Care Act was put into place in about 2015, one of the things that was mandated there is insurance companies can no longer deny people coverage for pre-existing conditions. So one of the things that is in place in order to ensure that that happens is what's called risk adjustment. Throughout the year, all health insurance companies submit their claims data and their diagnosis data for their entire populations to these government-maintained servers. And then you get credit for the risk.

At the end of the year is when the risk adjustment takes place. So insurers who insure a riskier population actually get transfer dollars from the insurers who insure a less risky population. So it's entirely budget neutral. There's no government money involved in changing hands. They facilitate that transfer between those insurers but not anything else.

So if one insurer insures a bunch of like elderly people with multiple comorbidities and one insurer insures a bunch of healthy college students, the one who insures a bunch of healthy college students is going to owe money to that other insurer at the end of the year. I have a sister team who's involved in the risk adjustment and I am in charge of the audit of the risk adjustment. So once you've submitted all of this data, the government takes a representative sample of your population and comes back to you and say prove it. Prove that these members have the conditions that you say that they have.

Analytics work and Shiny apps

There's a huge part of this that's actually competitive intelligence analytics where you know because it is budget neutral and there's transfers between different insurers, you know how risky your own population is. But you don't really know a lot about the other insurer's population. So there's a lot of data gathering, data aggregation around that relative risk analytics. And then we also have a bunch of analytics involved with predicting which conditions are going to be easy to validate, which ones are going to be more difficult.

My team's been building a lot of Shiny apps and a lot of dashboards that we've been deploying onto our Posit Connect suite that track kind of like how the audit is going, where there might be points of intervention, if there's something that's like, hey we're not doing it as well as this, as we were last year, flagging that, sending emails. We've also made some data portals where some of our internal partners, where they might need something in a specific format like an Excel file or a CSV so they can do some additional work with it, where they can like upload data or download our results in a way that makes it really accessible for them to continue doing their work, but in more of a self-service way.

We're actually kind of shifting into that for this audit now, like maintaining what we've done in the past, updating it with some new tools, with some new Shiny development, as well as just, there's always lessons learned from last year, like oh we weren't tracking that but we really should have been and so we're tracking it this year.

So I sit at Centene Corporate, but we also have subsidiary health plans and so certain, I have access to all of our health plan data, but not everyone who supports the audit execution at the lower levels does. So when they go in, they can just see like the California data or just see certain like pre-filtered sets of it that's really kind of tied to their login credentials. So I don't have to be like okay, who's asking for this? What health plan do they support? Filtering it on my end, we were able to build those things tied to their internal login credentials and then they can just do it.

Machine learning and upcoming work

We have been working really hard on turning some okay statistical models into some machine learning models where we can bring in some different data types and get a little bit more insight out of that. So we have a couple different models that are in the works around estimating final validation rates for these audits or estimating some of those like statewide competitive intelligence variables. And where it's something that we've done in the past on like an ad hoc basis with different, you know, hand-grabbing different Excel files and things like that. We're currently in the stage of like figuring out how to deploy those so that people can access the results.

The theme of a lot of the work that we're doing is not like purely self-service, but like mostly self-service where somebody who wants this can get it and knows how to get it.

For those applications, we're not really targeting whole organization reach. We have a couple specific partners during audit execution that we really target with those where they've, you know, they'll ask us for an ad hoc thing and that's fine. But if we, you know, get those same requests more than twice really, it's okay, let's put it together here. It's going to run every week. It's going to come straight to your inbox. And everyone's like, oh, this is great.

rhandsontable and data portals

The most complicated and business important app that I've built is really, it was a way for me to have to not email four people, get four different versions of an Excel sheet back and collapse it and then do analytics on it. I made a portal where they can go in everyone with their login credentials, edit basically the Excel file, but in the Shiny app through like an R hands on table format and save it in exactly the format that I wanted. And nobody can move anything and nobody can change data types and it's great.

For my team, business value is not measured in number of users or reach of that app. It's really the fact that this saved us so much time and so much money to do it in this way that like, that's the value of this app for my team.

So it lets you click into a data frame and make edits and save those edits back out. So those edits can then be re-redisplayed to the user. It's delightful. It's a little bit hard to learn if you're used to maybe like a GT or a DT data table, but it's basically magical. You can put this table in your Shiny app and people can feel if you want it to, you can even make it green. If you want people to really feel like it's Excel, you can, but you're basically putting this in your Shiny app and people are editing it, but you can say, this is a date data type. This is exactly the date format that it's coming in. This is going to be a number. It's going to have three decimal places. And when they hit save, that's exactly what you get back.

And I have that linked to our S3 storage solution. So every time somebody hits save, it's a separate record. Like we can go back, we can access historical records. It's made my life so much better. And I think, you know, building tools that make your own life better and then like spreading them to other people is still a really valuable way to do work.

And I think, you know, building tools that make your own life better and then like spreading them to other people is still a really valuable way to do work.

It doesn't do you any good to be like, well, have you tried doing it in R if the person doesn't even know what that is? So I think that's really helpful. I have no hate for Excel. Don't love it, but I have no hate for it.

Persistent data storage with S3

My team built a package that helps us like handle our internal S3 credentials. So that the app has the service account credentials. The app has access to S3 as part of like setting up our environment variables. So that was the first thing that we did that didn't make it so that the individual user needed to have the right S3 access to the bucket. And then I'll say invented, nobody invented this, but invented it for this app. We've got the submit button and we also have a save for later button where in terms of just like naming these files, it's like, okay, is this done? Save it for later.

Every time you want to save something, you can send it to S3 and it saves that portion of data that I would expect regardless of what's in it. And it's tagged with that like intermediate tag. And then when they hit submit, then it's tagged with that done. And what I actually have as part of that interaction is in that display, when you're going to say, okay, what rows do I need to edit? It brings that tag back in. So the user knows what's already been changed.

Day-to-day leadership balance

I would say I'm pretty, maybe like 60, 40 leadership to in, in the data. And I think a lot of that comes from the fact that I've been at Centene. I started there as like a data scientist one, and I'm now a manager of a data analytics team. My team is also pretty small. I have two direct reports and my sister team has three. And so together we're like a data science org of about eight. And so there's still a lot of need for like, hey, can you do this for me really quick?

In terms of like strategic leadership and direct management of my team, there's a lot involved in making sure that us as an embedded team stay up to date on what the tech stack development looks like and where those tools are going to live. And how can we make sure that we maintain access to the new tools and potentially transition processes, but also keep support for these processes that aren't ready to be transitioned yet.

I've personally been a lot more involved. I'm on the leadership council for our women in STEM initiative. And our user group, but I kind of, that's a given here on the R Hangout, but there's a lot of that community building and reaching out to people that I've kind of incorporated into some of my leadership work. A lot of things is like making sure, you know, that other teams exist, which in an organization at the size of Centene can be really difficult.

Oh, I didn't know that this team was doing this kind of thing with membership. I have a need to do something with membership. Let me make sure that I know who to reach out to. And that's something that we've been really working hard on in our next internal, our user group meeting. And we're trying to do it a little bit in everyone moving forward is someone just highlighting their team. And this is what we're doing. This is how we're using the Centene tools. Maybe it's not always R, but it's just, this is how we're using our data. This is the data science that we're kind of doing to keep those relationships built.

Tracking R packages

So we have in our bi-weekly team meeting, after we go around and we say what everyone's been working on, what do they need any help with, we actually have a little slide that's called R spotlight, and there's a web scraper that just runs, and it lists all like, oh, here's what's new in this package from R Weekly. It kind of like scrapes that and just plops it into our team slides for us, and so when me or the other manager have a little time before the meeting, we'll go through that and kind of pick out the ones that we think we should individually highlight. And then we also have within our data science community a couple R devoted channels, so if somebody's asking a question about a package or about something that they heard, we can keep an eye on that. Oh, okay, yeah, there's now been four questions about this package. Let's learn a little bit more about it.

I think I kind of outsourced knowing everything that's going on in R to the point that where it does filter to the top of my consciousness, that's something that I actually, you know, should learn about.

Managing requests and project prioritization

Our primary method for tracking that is actually we use GitLab and we have issues and epics and milestones related to a lot of that and then can use the comments of, like, God asked for this again, something like that to prioritize. We have been having to do a lot more with value analytics in terms of saying, like, to prioritize our work because we've started getting more and more as we've integrated with more teams, taken on more responsibilities.

What's the priority for us and what's divorcing the this is valuable to me and my team and would save us time versus this is what's valuable to the organization and this is the amount of money that we will save or not spend or whatever if we implement this is something that does require a little bit of tracking.

Moving from individual contributor to manager

I feel great about it. I think it's something that fits within like my personal evaluation of my own skill set. So like I was ready for that transition. But I think one of the things that really helped me with that is like in my individual contributor role, there were certain projects that I was wholly responsible for already. And so I knew a little bit about managing that project, what was involved in terms of talking to the people.

As I got promoted in my pipeline, I actually had a really great leadership mentor talking me through how he was managing his team, setting me up with like, okay, now that you're taking over this responsibility, these are my primary contacts. So somebody really set me up for success in that transition into the people leadership.

Something that I've had to balance with myself is part of that of like, there are certain things that I am very good at and I have a lot of experience with within Centene. But now if I choose to spend my time doing that, that's time that I'm taking away from something else. So balancing that, like I do need to train people to do this task because that's no longer my primary responsibility is something that I've been navigating and trying to get better at.

Building relationships across the organization

I think it starts small. Like, when you're really close to a data project and you, there's one piece that you don't know, it's like, shooting the email to the person that you saw last access that database, because you can check those things. You can say, like, hey, I saw you were using this data, like, what's going on with it? And so those kinds of small emails, and it's also getting over that hump of, like, self-doubt of, like, oh, am I emailing the right person? Are they going to, you know, hate me for derailing their day, for emailing the wrong person? And just, like, send email, because the worst thing that happens is that they don't respond to you.

And then as I moved through into a leadership role, it's getting plugged in with other team leaders, where they, okay, like, now I know the name of your team, but do you have somebody on your team who knows about this or who can help with this? And that's also something that, like, the women in STEM group has helped with, because we have women in STEM across the entire organization. I've been exposed to so many more people and so many more incredible women at Centene through that other organization that now I know there's a new team, our member insights team, brand new team, but I'm like, their manager immediately got plugged in with women in STEM. And so now I'm connected to her on, like, both pieces.

Career advice

I think it's something that even coming into this meeting, where I'm like, I'm great. And then I'm also like, oh, my goodness, what if people find out I'm not great? That sort of imposter syndrome of just like, let people believe in you until you can believe in yourself, where it's like, no one's giving you opportunities that they don't think that you can do. Because everybody's looking out for themselves.

So even like, I've given that advice, I give myself that advice, because it is like, let other people believe in you. And I think that that's something that makes the most difference. And it comes even back around to like, if you think you're doing okay, like send that email to the stranger and hope that they can help you with, like with your data. So it's just like, yeah, you probably have a lot of, you probably have the opportunities that you have, because you earned them.

Let people believe in you until you can believe in yourself, where it's like, no one's giving you opportunities that they don't think that you can do.