Jump to Transcript

In this episode, Donovan & Matt talk with health services researcher Betsy White from Brown University about a unique new resource for researchers called the Long-Term Data Cooperative, a provider-led data sharing collaboratory that puts together nursing home EHR data from EHR vendors that can be linked to Medicare claims. This powerful tool is made available to researchers through an online application process. 

This episode references the RFA for the Real-World Data Scholars Program, which is now expired. However, Minding Memory listeners can find out about upcoming opportunities or future RFAs by emailing [email protected]. 

Betsy White Faculty Profile: https://vivo.brown.edu/display/ewhite14  

Long Term Care Data Cooperative: https://www.ltcdatacooperative.org/Pages/default.aspx  

Resources:

• You can subscribe to Minding Memory on Apple Podcasts, Spotify, Google Podcasts or wherever you listen to podcasts.

Transcript

Matt Davis:

Welcome to Minding Memory, a podcast devoted to exploring research on Alzheimer's disease and other related dementias. Here we'll discuss compelling research and talk with leaders in the field about how their work is improving the detection and treatment of dementia. I'm Matt Davis.

Donovan Maust:

And I'm Donovan Maust.

Matt Davis:

We're both researchers and associate professors at the University of Michigan. I'm a PhD with a background in data science.

Donovan Maust:

And I'm a geriatric psychiatrist, so I think a lot about the diagnosis and management of dementia.

Matt Davis:

I'll work to keep Donovan to the minimum use of medical jargon.

Donovan Maust:

And I'll make sure we talk about research with real world applications to patients and caregivers of individuals with dementia.

Matt Davis:

Thanks for joining us and let's get started.

Donovan Maust:

Our listeners might recall that Matt and I are doing this podcast as a part of CAPRA or the Center to Accelerate Population Research in Alzheimer's, which is funded by the National Institute on Aging. Just like the title says, one of the center's goals is to accelerate population research in dementia. So we hope that this podcast contributes to that by introducing researchers to new and important sources of data out there that can inform our understanding of the care people with dementia receive.

Today, we're going to talk about a particularly exciting new data source that is coming online for researchers called the Long-Term Care Data Cooperative, which can be found at ltcdatacooperative.org. In this episode, we'll speak with health services researcher who is part of the Long-Term Care Data Cooperative team to help us learn about this new resource.

Today, we're joined by Dr. Betsy White. Dr. White is a geriatric primary care nurse practitioner who completed her BSN, MSN and PhD in nursing all at the University of Pennsylvania. She has been at Brown University since 2018 where she's currently an assistant professor of health services policy and practice in the School of Public Health. During the COVID-19 pandemic, she assisted in the construction of large data systems of nursing home electronic health records to examine various aspects of COVID-19 management treatment and outcomes among nursing home residents and staff.

She's co-author on a number of publications using these data, including one she led, published in the New England Journal of Medicine that provided the first patient level data on the effectiveness of the mRNA vaccines to prevent infection in nursing home residents. This was a really big deal because this incredibly high-risk population was absent from the trials that secured approval of the mRNA vaccines.

She's here today to help us learn about the Long-Term Care Data Cooperative and its potential for dementia-related research. Betsy, welcome to the podcast. Thanks for joining us. So first, let's start with some nuts and bolts. What actually is the LTC Data Cooperative and what are the data that are in it?

Betsy White:

Yeah. So the Long-Term Care Data Cooperative or the LTC Data Cooperative is a provider-led data sharing collaboratory that puts together nursing home electronic health record data from three different EHR vendors and links that with Medicare claims. And this is an initiative that is co-run through the American Healthcare Association, Brown University and Exponent. So researchers can apply to use the data. They go onto a secure web environment and they can have access to this really novel EHR data that we haven't had previously for nursing homes. And this builds a lot on some of our work that we did during the pandemic with kind of an earlier version of an EHR data system.

Donovan Maust:

And for listeners who aren't familiar, could you say what is ACHA and what is Exponent?

Betsy White:

Sure. So AHCA is the American Health Care Association. It is one of the main industry groups representing nursing home providers, and then Exponent is our partner that's managing all the data infrastructure. So the company does a lot of work, but they have expertise in building secure data environments and managing all the technical aspects of this work.

Donovan Maust:

We'll probably get into this a little bit in subsequent questions, but is there a certain timeframe of data that are available? Does it go back to a certain point or how does that work?

Betsy White:

Yeah. So it varies by providers. So different providers sign up for an EHR system at different points. So we have data from whenever a given nursing home provider started with that EHR vendor. So if they started in 2015, we have data from 2015 onwards that they just started within the last year or two. We have shorter period for those providers. So it varies across nursing homes.

Matt Davis:

When people say provider, it could mean different things. Are you talking about the facilities or insurance? What level of provider are we talking about?

Betsy White:

So when I'm saying provider, I'm specifically referring to nursing homes, so individual nursing homes. And so to date, we have about 2,200 nursing homes across the country. It's a nationally representative sample of nursing homes that have signed up with the Data Cooperative to have their data shared. We have the data from about just over 900 of those nursing homes that's actually integrated into the cooperative data system right now. It takes a little bit after the individual nursing home provider signs up for the cooperative. It takes a little bit to actually get their data integrated into the data stream.

Matt Davis:

So for our listeners, I'm sure they're going to check out your webpage, and when they do so, they're going to find that this grew out of the COVID pandemic. Could you tell us a little bit more about how that came to be?

Betsy White:

Sure. So starting in March 2020, so right at the outset of the pandemic, it quickly became obvious that there really wasn't good real real-time data available on the nursing homes to understand what was happening. It also was very quickly apparent that the people who live and work in nursing homes were going to be disproportionately impacted by this virus. So we were able to leverage pre-pandemic relationships that we had with various nursing home providers and the expertise that we have at Brown managing large data systems to pretty quickly build up data system of electronic health record data.

We actually collaborated with one nursing home provider, which at the time was the largest nursing home provider or nursing home organization in the country. It was Genesis Healthcare. And at the time, they had about just shy of about 350 nursing homes across multiple states. We were able to enter this collaboration with them where they were actually sharing their data on a nightly basis with us directly. So they use PointClickCare, which is one of the EHR vendors that's in the Data Cooperative. We were getting nightly downloads of this data.

So we were pretty early on able to look at resident-level characteristics and outcomes in pretty much real-time because most people that do nursing home outcomes research, the main data sources for those are Medicare claims and the minimum data set and wonderful research has been done with those data resources, but the key issue there is that you're using the newest data that you can get is usually at least two to three years old.

So we really had a major advantage at the outset of the pandemic when there was this novel virus and lots of questions and really the need to have real-time data to inform decision-making that we were able to have access to this electronic health record data system. And we were able to working very closely with our partners at Genesis, answer a number of questions using their data. So with that, here at Brown, we built the expertise of working with EHR data. That led into another predecessor of the Data Cooperative where we started getting electronic health record data, not just from Genesis, but from 11 other nursing home companies, all... Again, the head PointClickCare that the same EHR vendor.

And then that ultimately led into the Long-Term Care Data Cooperative, which is seen as being more of a public resource that researchers around the country can access and that is available for research, not just related to COVID. So it's supposed to be a longer term lens. So those are kind of the predecessors that ultimately led to the data coop.

Matt Davis:

It makes a little more sense now, when you said it was provider led or kind of led by the nursing homes. I was thinking to myself, why would they be interested in doing research? But it makes sense now when you talk about the history of the pandemic, and obviously they had skin in the game.

Betsy White:

Yeah. And I would say our work during the pandemic, working closely with our partners at Genesis was some of the most rewarding work that all of us on that team participated in because it was on the academic side, we're building and learning how to work with these new data and these complex new data. I mean, we were having weekly calls with their chief medical officer, their regional medical director, their technical team, and it was truly a collaborative experience, getting to understand these data, understand the real world context, learning how to apply them, and it was just a really valuable experience and we were able to produce a lot of high quality work from it.

Donovan Maust:

Essentially, it's like an incredible example of community-based participatory research. Just in this case, it was really the provider, the facility community, and in a tremendous time of need. So it must've been incredibly fulfilling. So, you touched on the fact that historically a lot of the work in this setting has focused on, say, Medicare claims data, Medicaid claims data and the MDS. So maybe for people who aren't as familiar with what's the added benefit or maybe the particularly high yield elements and pieces of information that you can get at using the EHR that you don't have through those other data sources?

Betsy White:

Yeah. So in addition to the benefit of having much more real-time data in terms of measures that you can... Or things that you can measure using EHR data that you can't from claims or minimum dataset, that would be things like daily medication administrations. So we have the medication administration records, so we see every single medication that was given to a patient. And that's really important for the pharmaco epidemiology world, which historically, again, have primarily used Medicare part D claims where you're sort of making the assumption that the person received the medication, but with having the actual medication administration record in the EHR, you can actually confirm that and you know on a patient day level that people received medication.

So medication administrations, that's one really important one. Vital signs data. So for example, if you wanted to do a study where you're looking at trends and beta blockers and how that impacts pulse rates or blood pressure levels or what have you, we have that granularity of data. Things like lab data varies a little bit across facilities. So not every single nursing home in the collaborative has their lab data integrated into their electronic health system, but still a number of them do. So we have already used those types of data to look at questions like we had a doctoral student here, Becca Forness, who did a really nice paper looking at the impact of chronic kidney disease stage on mortality in people with COVID.

So, lab data, that's a really unique piece. We have all the medical orders. So, somebody wanted to look at things like nutrition or fluid consistency, look at particularly in a dementia population. Look at prevalence or outcomes related to sickened liquids or food consistency, that sort of thing. That can all be captured in the orders data, advanced directives. Advanced directive status can be captured from medical orders. That's something that is really missing from the current minimum dataset.

It was in the earlier version of minimum dataset. It was in MBS version two. There were a lot of validation issues with that measure, so it's not in the current version of MDS, and that's always been a key area of weakness in the MSs, but using the EHR data. We have information on advanced directives. So those are just some examples. And I would say those are the data elements that we use the most. And the ones that tend to be the most complimentary to what we can measure out of minimum dataset and claims data.

Matt Davis:

It's a good reminder of just how limited claims can be. I recall working with people that are new investigators and stuff with claims, and they're building their data set. And it's not until they start working with the data, they realize like, "Oh, you don't have BMI," and things that you just assume that you would have that you really do need to go into the clinical notes and EHR types of data sources to get that stuff.

Betsy White:

Right. And it's important to always understand the data inputs. So Medicare claims, we use them for clinical evaluations, but they're there for billing. So we make a lot of extractions when we're using claims data in terms of what's happening on the clinical side, but it's not the actual clinical documentation. So that's sort of the advantage of the EHR data system. And same thing with minimum dataset. The MDS has a number of wonderful measures in it, but it's also somewhat limited. So for example, immunizations. So you can determine pneumococcal and influenza vaccination out of the MDS, but MDS doesn't currently capture COVID vaccinations or shingles vaccinations or other things like that. So there's just the opportunity to get into a lot of nuance about the clinical environment that previously wasn't available with our existing data sources.

Matt Davis:

It's incredible. I mean, the little bit of work that I've done with EHR data, it's a mess usually. And I'm not sure why, but it's such a huge thing just to pull off a single research project with EHR data. I can't imagine trying to put something like this together in terms of the infrastructure and informatics kind of part of it and the programming and stuff. It must be an enormous task.

Betsy White:

Yeah. And this is where I'll sort of cheerlead Exponent for a moment. Because recall, this is a data system that pulls from three different EHR vendors. So just right there building something. So for example, each of those vendors has some version of a medication administration record, but they're all a little bit different in terms of how they capture different fields or what have you. So what Exponent's role in this is they have built what's called a core data model where basically they take a medication administration record from those three different vendors and build it into basically a standardized form that is then analyzable.

So that's really the key value of the cooperative is that it is able to integrate data from different EHR vendors that are all formatted and programmed a little bit differently, but they're able to build it into a common data structure that you can actually analyze.

Donovan Maust:

So that totally anticipated my next question, which was to what extent is there a standardization of data? And then is there something like a data dictionary out there for researchers to really understand these are the variables, these are the kinds of information that are available in this kind of format?

Betsy White:

Sure. So we have a technical user guide. If you go on the Long-Term Care Data Cooperative website and you go to the data page, there's a link there. It was just released within the last week or so. It's the initial version. In terms of the specific elements that are in the cooperative, it's going to evolve a little bit over time. So that's going to continuously be updated, same as nursing home compare other things. Different measures, different data will be added over time. But yes, the technical user guide is available on the website and it provides a background on the core data model, the narrative context of the various different data elements that are available. And then it also has the more granular data dictionaries to understand what the specific fields are.

Donovan Maust:

And then, so you had mentioned that some say some EHRs, some chains will say vary in whether laboratory data are available or not. So does the dictionary also give a sense for this is available for 100% of participating facilities versus two thirds of them have it? Is that part of what's in there?

Betsy White:

That's a great question. I don't think that specific information is in there right now, partly because it's still something that we're learning still. We have a number of plans for validation work that's going to be done on this. And so there's definitely a discovery process as you get to work with these data. I mean, we had our experience working with the Genesis data. I mean, it was a very steep learning curve and to get up and going with these data.

Matt Davis:

One thing that I think that's worth pointing out is sometimes when you think about using EHR data, you have what I would call catchment issues. So you might have an individual's data from one facility, but you don't know where they're getting care in different places. But this population in theory, right, you have all their EHR data, I assume. So you really do have a sort of complete picture of a population.

Betsy White:

So it does vary to some extent. So if a patient moves from one nursing home to another nursing home where... I mean they have to move to another nursing home that's signed up with the cooperative. Now, that's the advantage of having it linked with claims. And that came out of some of the challenges that we had with our COVID work because we had visibility on what was happening to the patient in the nursing home. But when we wanted to measure things like hospitalization or mortality, we were a little bit limited because oftentimes the nursing home would document that the person was transferred to the hospital.

We would have that information and usually they would document if the person died outside of the facility. But we didn't have anything to... Because those specific data were not linked with claims. We didn't have a way to validate how many hospitalizations or deaths we might be missing. So that's the real value with this upgraded version with the Data Cooperative is it is linked to claims. So while you might lose some information if a patient goes from one nursing home to a different nursing home that's not in the cooperative, you'll still have their claims data. So you can measure things like hospitalizations. I mean anything that you would be able to capture in the claims data that happens to the person externally to the nursing home that has the EHR data.

Matt Davis:

Quick question about just working with different stakeholders. I think that's really an interesting aspect of this, and those of us that do research have our goals and the way we see things. Could you talk a little more about what it's like to work with the facilities? I mean such a diverse group of people, and how much... Are the incentives, and the goals, and objectives pretty well aligned?

Betsy White:

Yeah. So the approved uses for data in the data coop are for public health surveillance, comparative effectiveness studies, and then for trials. And we actually have a provider. So we have a review committee that consists of academics and providers, and we also have a public comment period where providers can provide input on any particular application that comes into the Data Cooperative. So any researcher that wants to use the data, there's an application process they have to go through. And part of the review process seeks provider input.

It's actually a really important lesson. And part of what we're hoping to gain with this is getting better collaboration between researchers and providers. And again, speaking to our experience during the pandemic, that collaboration was so valuable because the clinicians that were in the nursing homes and dealing with the daily operations, they understand the policy environment, they understood changing in guidance that was coming from CMS and CDC, they understand how the data in real time gets inputted into the EHR data system and what the meaning of those data are, and then they're also able to just provide a real world gut check on what's happening there.

So it was very informative for us more on the academic side to have that real world collaboration. So yes, I think sometimes in academia we talk a certain language or we ask certain questions that if you put it in front of a clinical audience that's actually working full-time in the nursing home, there's not always great alignment. So part of the application process, we actually make the researchers write a real or a plain language abstract where they basically have to describe their research in a way that conveys both the value of it and the clarity of their research for the providers who are part of the review committee. So it's an added incentive of this initiative that we're trying to improve collaboration between researchers and nursing home providers.

Donovan Maust:

When you sort of described the evolution of the cooperative, it sounds like it started with, say, some relatively large nursing home chains. To what extent would you say if there's now over like 2,000 participating facilities. To what extent is the membership comprised of facilities in chains? How do you think about the representativeness of the facilities that are participating versus those who aren't participating?

Betsy White:

Yeah. That's a great question. Oh, I'm sorry. The numbers are actually updated. We're upwards of about 2,400 nursing homes now that are enrolled. And those 2,400 nursing homes come from 220 nursing home organizations, so we have a variety of some of the larger chains that are signed up, smaller chains, and then we have a number of just single operated nonprofits. So there's quite a variation in terms of the diversity of nursing homes that are engaged. I think we have representation in all 50 states. There might be one or two states that we don't, but for the most part it's a nationally representative sample. Yes, there's pretty good variation in the provider population.

Donovan Maust:

I could imagine. You mentioned that for researchers they need to present to a committee and there's an open comment period. I could imagine a situation where, I don't know, maybe two thirds of participating facilities like it and are on board, and there's a subset of people who are not on board. Is it an all or nothing kind of thing in order to get the go ahead for a project?

Betsy White:

No. It is a majority decision. So they are allowed to be... It doesn't have to be unanimous consent to go forward with a project. We have been working also with researchers. This is kind of a new thing. Academics are used to writing to like an NIH study section, and there's a certain way that you write when you're spending a proposal like that. So we have been working with some of the initial researchers that have been applying to the cooperative to help them learn how to write in plain language because a skillset that many academics need, and to make their application appropriate before it even goes before the review committee.

Matt Davis:

So nursing home research is new to me. You said there's about around 2,000, 2,500 participating facilities. How many facilities are there in the United States roughly? Do we know?

Betsy White:

So there's about 15,000 nursing homes across the US. At any given point, there's about 1.1 million nursing home residents in those nursing homes. So in those 2,400 nursing homes, we have data on just over 530,000 unique residents, and that's across the years of the data that are available. So pretty significant numbers to do some robust work with.

Matt Davis:

So for our listeners who are thinking about their own research questions listening to this podcast, could you talk a little bit about what the process is like of getting access to the data?

Betsy White:

Sure. So on the website, there's an application process. There's a tab for researchers. You can go on there and it explains the process for application. Basically, step one is you just upload a specific AIMS page, it gets a preliminary review. So we have people that are applying either while they're in the process of writing a proposal before they're actually funded, and then we have people that are applying to use the data after they've been funded. So if you're still kind of in the preliminary stage, it's a lower lift application, we just take a lot of the basic information about the proposal and AIMS page, do an initial screening.

So that's the preliminary review. If people need a letter of support in order to submit with their proposal, they can indicate that and we can provide that after a preliminary review with the review committee. Once somebody is funded and they're going to submit a full application, then we take more information about specific data elements. They're looking to use the IRB submission a little bit more about that. And that's then when it goes in front of the full provider review committee. They also have to submit this plain language abstract, which again, we work with them a little bit on that to make sure that it's appropriate.

Matt Davis:

So there is a cost associated with using the data, I assume?

Betsy White:

Currently not. Now, it varies. If there's particular specifications that a researcher might need that requires us to do above and beyond what the standard data product would be. There might be some added costs. But for the most part, there's not a cost to the researcher to use the data.

Matt Davis:

Do you have students doing projects using the data?

Betsy White:

So with the Data Cooperative, the first researchers for this project specifically are going to be using... Are just starting to use the data this spring. But we've had multiple dissertations written on the in house electronic health record data we have here at Brown. We've had some wonderful doctoral students doing all kinds of work, everything from prescribing projects to vaccine effectiveness, to looking at chronic kidney disease as a risk, looking at TAMIFLU prophylaxis, patterns for influenza. So we've had some really interesting projects being done by our doctoral students here at Brown.

Donovan Maust:

So is there anything that we haven't asked you about that you think would be really important for listeners to know that you want to add here at the end?

Betsy White:

Yes, I would like to plug... So we have a new initiative, it's called the Real World Data Scholars, and it is on the website. It's also available through the IMPACT Collaboratory. So basically what this is we're looking for a couple of primarily early career investigators. So that's doctoral students, postdocs, junior faculty. We really want to get more people oriented on these data, and we want them to help us with some of the validation work that we need to do on the data. So it's called the Real World Data Scholars Program. It is going to be a mentored one year, $50,000 training grant where investigators from around the country can apply.

They will be mentored by someone who is one of the more senior investigators with the Data Cooperative and they will do some of the validation work. And in the process of doing that, we'll also get oriented to the data and hopefully we will be able to use it for their own future work. So the Real World Data Scholars, the application is online now and applications are due August 4th, I believe. And then it's a one-year training grant. Projects would be starting, I'm sorry, by October 1st.

Donovan Maust:

And there will probably be maybe a couple annual cycles of this.

Betsy White:

We're starting with the first year and seeing how it goes, and then we'll kind of make decisions after that. But yes, I mean, we're definitely looking to expand the network of people who have expertise on using this type of data.

Donovan Maust:

Great. Well, thank you so much, Betsy. This was super informative and just sounds like such an incredible new data source to have available to the community. So hopefully we'll send some people your way to check it out.

Betsy White:

Great. Thanks for having me.

Matt Davis:

If you enjoyed our discussion today, please consider subscribing to our podcast. Other episodes can be found on Apple Podcasts, Spotify, and SoundCloud, as well as directly from us at capra.med.umich.edu where a full transcript of this episode is also available. On our website, you'll also find links to our seminar series and data products we've created for dementia research. Music and engineering for this podcast was provided by Dan Langa. More information available at www.danlanga.com. Minding Memory is part of the Michigan Medicine Podcast Network. Find more shows at uofmhealth.org/podcast. Support for this podcast comes from the National Institute on Aging at the National Institutes of Health, as well as the Institute for Healthcare Policy and Innovation at the University of Michigan. The views expressed in this podcast do not necessarily represent the views of the NIH or the University of Michigan. Thanks for joining us and we'll be back soon.