- Experienced data scientist with a passion for continued learning and research - Proven ability to manage multiple projects while meeting challenging deadlines - Excels while working in a fast-paced environment with high demand of services providedView the profile
About the talk
This video is part of the R/Medicine 2020 Virtual Conference.
so, For life. Okay so are next-to-last presenter is Corey Bridge from UW Health and he's going to be talking about using our to support covid response at the health system. Hello everyone. I mean like she said, my name is Corey Fritch. I'm from the other UW, which be UW Health in were based in Madison Wisconsin. Thank you, for braving, the Saturday afternoon, talk this late. So what I'm going to do is just over the next 10 minutes or so spend just a little bit of time, explain to what my team is as well as how we leveraged our studio server, profar, work with
covid-19, So kind of at a glance Madison hospitals are where are main functions are out of, so we have seven hospitals with one academic Medical Center, that would be the University Hospital in Madison, are we do it to Regional, Hospital's down in Illinois and as well as 77 clinics mainly around the Madison area. The team that I work on is called the applied data science team. So there is three of us three day, two scientist that work in a larger Department, called
Enterprise analytics. So we are a full-time analytics team, we basically support all business operations as well as clinical work. So we the real nice thing about my team as well as my department. It that might be a little different than some of the other presentations are speakers that were talking is we have direct access to the HR so we use epic. So we have access to not only go into any of the instances between prod and test. We also have access to any of the databases and data warehouses that are built off of that as well as I need the other liquor.
Employee databases and stuff like that. So our main focus on the plight of the science team is working on predictive models and getting those into production into operations at uwhealth. So, not only do we build our own custom models. We also work with researchers and others to translate and Implement those predictive models and get those into the into the workflows. As well as the other big thing that our department does, as we are consistently working on, kind of engaging and getting our other
analysts in Enterprise, Aunt, likes to use more advanced analytics as well as predictive models. so, I'm guessing similar to most data analyst both in health care as well as outside of healthcare, our work change dramatically in early, March going into mid-march and late March, so do the covid virus. Team joined a collaboration of other analyst as well as researchers clinicians and professors from UW-Madison. And we were tasked with there was an incident command team that was
stood up at UW Health as well with you W. Madison employee, young researchers and teachers and this model team was tasked with creating some Predictive Analytics around like modeling out what we thought was going to happen over the next week. Sat into next month's, both in Dane County, and South Central Wisconsin, so that we could try to make recommendations as well as provide data and analytics to That incident command around or I see you in an Icee bed capacity,
what we thought was going to happen in terms of spread of of the of covid to make sure that we were prepared as a hospital system if needed. So our city of server Pro is the main platform that our team uses prior to covid, as well as it became Immensely helpful during our work with covid-19. Some other visualization software is that like we use as a whole organization, Arceus is not our like organization-wide visualization, but what it did allow us to do, is it allowed us to stand up stuff
extremely quickly, and because, our team normally codes in arm python, right? Inside our server Pro, we were able to do our coding, some data Federation. We had a lot of data, we had to pull together from lot of different places. We built notebooks, get some cold reviewing, as well as Oliver. Versioning, we built a lot of different models in a lot of iteration of the model. So using Arceus everpro was hugely beneficial. So, it also allows all of our team members as well as two people that we were working with that normally don't work in our Su
Servpro to have the same packages inversions instead of everyone using their own on their laptop. So we could easily share codeshare notebooks and be able to run those no problem. I will say it's not widely used by everyone in Enterprise analytics yet, but hopefully soon that I like I said we're mentioning some people. Another thing that we're doing is trying to bring people into a coding in, in both arm python, which will be done through our server. All right, and what we did need it for was we had to provide a lot of answer. So like I said, we had a
incident command with the prep stuff for them at each week. So we started to use Arceus server Pro as a main solution for everything, we did from data collection through visualization. So our studio just kind of the actual using hours of coding language allowed us to bring in stuff from like Wisconsin Department of Health Services for through the apis. We start looking at data at the state county and census tract level. We really tried to drill down into different things to make sure that we were trying to really predict out and kind of see what transfer
happening. Even compared to just Statewide numbers that everyone was getting, so we created scripts and did some automation around getting that data in so that we could get pulled out and pushed out to an in command dashboard. That way, they have one spot that they were looking at data instead of trying to pull in metrics from hundreds of places. Are we developed three predictive models on that model? Team? One of them was a Seer model, so susceptible exposed infected and resistant.
Again there was a lot of iterations around this and it's been it's continue to be work on. I think our latest version was finished a couple weeks ago. Doing that coding in our, in our studio allowed us to updated your weekly. Put some iterations around it and get predictions both in and out of the model pretty quickly and easily. We also leverage our mark down quite a bit. Like I said, we were trying to be very effective and efficient on some of the the modeling that we are doing as well as the analytics around it. So
rmarkdown was really great when we wanted to just go export. Something is it as an HTML that we could get to any person if we needed to email it to him or if it was a quick look at something a really really allowed us to get everything out there. So, we also did a lot of the analysis of current past Active cases. kind of when we were sending stuff out, our connect was huge for us, a new piece to our stupid server Pro allow us to use our active directory to not only distribute everything but authenticate
to different groups and stakeholders And then we developed two pieces to our connect that were really beneficial to our stakeholders. The first one was an API with plumber that allowed for simulation of Sierra vent. So if someone wanted to kind of mess around with what they thought might happen, in the next couple weeks, they could do that as well as we develop the speedometer that allowed us to look at his capacity going to be over. What we are what we could handle in the next couple weeks, depending on like length of length of stay as well as how many covid deaths.
We had a four-hour capacity. and that is kind of a kind of kind of weird to drop yo 6 months of work into Like five slides, but that is kind of where we're at. Thank you. So I think we have time for one question. Are you looking at how Healthcare usage has changed since covid? In terms of analytics, it wasn't clear. So that So we that, that is our main. So I'll see how our Healthcare System was being used, was one of the things that we were in charge of looking at. So especially at our clinics, we had closed most of our clinics
at as did most people in Wisconsin. So we were starting to look at when can we safely reopen? And that was one of the driving things that we had to present each week. Great. Thank you so much. Corey, no problem. Thank you for having us.
Buy this talk
Interested in topic “Medicine, Health and MedTech”?
You might be interested in videos from this event
Buy this video
With ConferenceCast.tv, you get access to our library of the world's best conference talks.