Hello, everyone. So very appreciative to be here with my brilliant critical care colleagues. I am in the Department of Family and Community Medicine. I do a lot of community and then also clinical research kind of across the severity spectrum for COVID. And so these are my disclosures. I have a lot of different COVID related funding, again, from community surveillance, household transmission, the Sari Prep, and then also some long COVID funding. So I'm really kind of talking about the impact and then also the beyond, kind of how we can take some of these lessons that we've learned in other aspects of understanding COVID in the pandemic and apply it to Sari Prep. Big picture, right, when we think about using real-world data, when we think about trying to understand emerging threats or even existing threats, we really think about these gaps and limitations with collecting and using data. So these are kind of some constant issues that we continually run into when we think about clinical data. Timeliness, we can collect it quickly, but can we use it quickly? Data interoperability, so we have multi-sites, we have different data collection, we have different data capture forms. We also have different sort of terminology and understandings in our own institutions of the way things are kind of working and how we're reporting. Relevance, I think we learned this hard lesson during COVID, things that were really relevant if we weren't sort of agile and able to really adapt. We were out of sync with the pandemic in our research. And then continuity of care, which being in family medicine is something that I think about a lot, right? So folks come into our hospitals, they come into our ICUs, but do they get their follow-up care or their pre-care somewhere else? And so really kind of thinking about trying to link some of these data sets and really trying to learn from what we were able to do with community surveillance. So I'm working on some large community surveillance cohorts with essential workers and kids. We've learned very quickly to do kind of constant QC, QA, merge data from multiple sites, very large data sets to do weekly analysis. That has allowed us to, for example, we were able to do real-world vaccine effectiveness on COVID vaccines within four months of the vaccines being available. So we had a system we were constantly monitoring and we can actually do power analyses every single week and decide when we have enough to sort of go forward with an analysis that we're confident in and that we can trust. And I have a couple of citations here just to kind of demonstrate. We were able to publish adult VE in March of 2021 when the vaccine came out in December and kids VE in December of, time doesn't work in my brain anymore, December, 2021. Wow, that doesn't, that feels like much longer ago when the vaccines for kids came out in over the summer. So again, trying to take these lessons and these tool sets and kind of this approach and apply it to clinical data. It seems kind of simple, but it's incredibly complicated, right? It's simple on the surface, but when you start to go deeper, it can kind of unravel a little bit. And so what we've done is we've leveraged the data infrastructure that we created for this community data to create some flexible and standardized systems to manage the sort of the multiple streams of clinical data that's coming into SariPrep. We have some very, again, on the surface, maybe simple goals, but in practicality, some complex goals. We do, our goal is to do weekly QA and QC to identify missing data, but also data that just doesn't make sense. So we're always testing for missing. Is it really missing? Should it be missing? But also sort of internal and external validity, kind of thinking about it the way we would validate maybe a survey. This makes sense in the context of this REDCap form, but does it make sense in the context of clinical care? Because if anybody that's dealt with EMR data knows, sometimes it just like, it just doesn't quite make sense. We are working on merging data that's being collected outside of the study EDC. It's a multi-site study. We have a standard electronic data capture system, but institutions use them differently, which means there's always a level of merging discordant data. There's always a level of merging old CRFs and new CRFs and slightly different sort of homegrown CRFs. Being able to generate descriptive tables regularly to understand patient characteristics. This is something that a lot of the community studies are able to, it's difficult to do in an EHR, like a clinical cohort that isn't an EHR only cohort. So our EHR cohorts that are funded by CDC and such, they're able to do this, but they sometimes lack the context and the depth and sophistication that studies like SARI PrEP have. We have the biologics. We also have a human sort of going through and doing some interpretation of what's really happening and entering that data. And then to create emerged analysis ready data sets monthly that can be used to answer sort of existing questions. We have questions in the protocol. We have questions in the protocol paper, but there will be emerging questions. There are going to be new treatments. There are going to be new symptoms. There are going to be new variants. And so to be able to work towards being able to respond and really quickly using clinical data to analyze those. So just some kind of nuts and bolts stuff. Not everybody cares about this. So I apologize if you really don't think about data or enjoy data, but for the few of you that do, we've got code that was developed and we spent a lot of time kind of developing that code to be able to standardize and clean variables. So they load properly into all statistical analysis programs. That seems so small, but it just is so important. It allows for really easy kind of frictionless kind of plug and go data analysis from a variety of folks. We check variables again to ensure they make sense. So yes, red cap didn't flag it as out of range, but like, does it actually make sense for something that we should be seeing? And then we build missing data reports so that we can examine missingness across and within sites, but really quickly. So the idea is while that patient is still in the hospital, if something's missing because it's on the EMR, we can still go get it. We can also get it quickly again while we're still in that context. We also merge data. So again, we're merging this data across multiple systems. We've talked a little bit about, Dr. Savransky talked a little bit about the individual patients, but really we're thinking about the higher level kind of more urgent data, the data that will allow us to answer really quick questions if there are new variants or we're seeing a new spike or a new peak. So the core data, the administrative, the demographic and sort of characterization of the illness and then the outcomes is really what we prioritize to make sure that we're ready to be able to merge and analyze data really quickly. It allows us to do things like create a descriptive table. This is just a snapshot of a descriptive table that includes data from last week. So we were able to get the data from all the sites, merge it and put out kind of a descriptive table. And so this doesn't include all of the participants because we're sort of working through things iteratively. So it's about half of the participants in the cohort, but this does include the 40 people, I believe that were enrolled since the beginning of 2023. And so just we can kind of see some big picture things that would help us make decisions about analysis or help us understand the nature of the pandemic, like number of comorbidities. Again, I'm sort of a, I used to be before the pandemic kind of more of a disparities researcher. So I'm always interested in sort of who we're treating, if age and race, ethnicity distribution, I will eventually get back to that work. And then, so looking also at the type of comorbid disease, we know that as the different variants have progressed, they've impacted different individuals, right? And so we'll be able to better kind of track who's being really impacted, which can filter back to care and planning practices. And then incorporating community data. So George mentioned this a little bit, kind of going through and incorporating what was happening in the community. I have much less sophisticated graphs. These are really much more descriptive, but what this is, is in contrast to George's sort of stress, this is admission. So this is community cases compared to admission rates. So admission is the orange and community is the blue. And then this is to the ICU. And what's interesting for me as somebody that lives in both the clinical and the community world is we had really good community surveillance when sites were doing, when communities were doing testing. But as we moved to the antigen testing, our community numbers are gonna be less and less reliable. So there's a potential opportunity to actually kind of flip that. We've been always thinking about like, what's happening in the community so we can predict what's happening in the emergency rooms. But there might be a, there might come a time and that might be soon where we're actually using what we're seeing in the emergency rooms and the ICUs to predict what's happening in the community. And so as antigen test says, not everyone's reporting their antigen results and things. So again, our ability to sort of understand and link will deteriorate because of the sort of the lack of community testing data, especially in some communities where they've just stopped it and folks are kind of on their own. And our ability to clinically characterize cases with detailed histories and links to sequencing and links to the biologics, I think can really help understand sort of the community. So again, instead of having surveillance informed, community surveillance informed clinical care, I think we have the possibility to go the other way around. And so I wanna again, thank everybody. This work takes a village. I have a whole team of analysts that do a lot of this really smart kind of coding and things with me. And then all of my brilliant physician colleagues that are working on the project. So thank you. Thank you.

real-world data

COVID-19

clinical data

community surveillance cohorts

multidisciplinary approach