Chapters Transcript Video Update on Hypertension Studies at Sentara Dr. John Brush describes Sentara’s research study focused on hypertension and community efforts to improve outreach. Good morning everybody. um, I'm John Brush. I'm chief research officer at Centera Health, and it's my pleasure this morning to give cardiology grand rounds. Um, I wanted to share with you some of the research, some hypertension research that we're doing currently here at Centera. You've seen some of this, but I wanted to just review, um, some of the studies that we've been able to, uh, do and we're currently doing here at Centera. So thank you all for uh for uh attending this morning and uh let's go ahead and get started. So I, I don't have any financial conflicts of interest or uh relationships with industry to disclose relating to this, uh, talk. So I keep saying this research makes us better clinical trials expose us and our patients to cutting edge therapies. Uh, research helps our reputation. It helps us recruit, uh, top notch, uh, cardiologists to come to our, our area and practice. And we're in the knowledge business, uh, medical, uh, taking care of patients is, uh, utilizing knowledge for the care of patients and doing research is the process of generating new and generalizable knowledge. So why do I do research? Well, I'm, I'm hoping that perhaps maybe I can lead by example, but also I'm hoping that we can build research capacity here at Centerra and we can do that one project at a time. We're building, uh, and I think I'll be able to show you that we're building the capacity to do big data analytics and also the capacity to compete for and manage federally funded research grants. So uh many years ago, what seems like a previous lifetime, I did a fair amount of hypertension research when I was at National Institute of Health and also at Boston University. It ended up resulting in a number of New England Journal papers and jack papers and what have you about uh micro microvascular angina in patients with hypertension. Uh, endothelial dysfunction among, uh, hypertensive patients both in the periphery and in the coronary vasculature as well as diastolic dysfunction. One thing I learned from that, uh, research is that NIH style research is very, very intensive research is very hard to translate outside of NIH. I found myself doing the impossible, trying to do this, and so, uh, when I figured I was trying to do the impossible, I decided to come here and join a private practice group back in. 1992, but it's kind of ironic and, and it's kind of fun that I'm be able to loop back and do hypertension research after all these years. So some background on hypertension, it's the most prevalent cardiovascular risk factor. It affects almost half of Americans. There are persistent and troubling racial disparities related to hypertension. The control rate for hypertension is in the like the 40 or 50% rate. We're failing over half the time. It's unacceptably low. When I went to medical school about 45 years ago, uh, 95% of patients with hypertension were classified as having essential hypertension. In other words, the cause was unknown. Uh, now 45 years later, it's still 95% of patients who have hypertension have it because of an unknown cause. It is inherited if you have hypertension, your relatives are about 4 times more likely to also have hypertension, and yet genetics doesn't explain it. There are genetic abnormalities like Little's uh syndrome and uh and Gordon's syndrome, uh, those are monogenetic abnormalities, but those are extremely rare. And genome wide association studies have revealed snips, uh, single nucleotide nucleotide polymorphisms in hypertensive patients, but the sum total of all those snips only explains about 3% of blood pressure variation. And so the genetics now with, with, uh, the human genome project, the genetics really don't explain what appears to be inherited. So here at Centerra we've got two approaches to hypertension. One is sort of an aerial view using big data, sort of a high tech approach, and one is more of a boots on the ground approach, sort of a high touch. So we've got a high tech approach and a high touch approach. So let's talk about the high tech approach. We're doing data science and data science can really be defined as the art of making your data useful. It not, you not only have to have useful data, data that's accessible and, and useful for purpose, but also you have to create use cases where the use case, your, your, your intention. Is it sort of conforms with the data you have. You can't do research with data that you don't have. And so matching up the use case with the useful data is sort of a key and hypertension is an ideal use case. Blood pressure is a number. Computers keep track of numbers really well and so doing hypertensive research is is a very good use case. So what we did to make our data useful is we organized it into a common data model. So each night, uh, EPI uh transfers uh data from the production part of Epic into a static SQL database called Clarity. In clarity there are about 20,000 tables and it's uh some of those tables are referred to by some IT people as junk drawers. It's just the data is strewn into these uh these tables. Uh, clarity wasn't designed for research. Uh, it was designed to create a static repository for epic data. OMO is a uh is a common data model that was designed by researchers for research. It has 38 tables as opposed to 20,000 tables. When you go to find something in OMO, it's much easier to find it, and much of the, uh, sort of the sorting and, and, uh, and harmonizing and structuring it goes into the process of the extract transform and load into OMAP. So we, uh, we invested a fair amount to get our data into OMOP. We hired an external company called Refactor Health. And they pick up the clarity data in our enterprise data platform. They have a business associate's agreement so that they can work within our data platform. They run, um, they run their data, our clarity data through the refactor software which uses actually AI to convert the clarity data into OMO concepts. And then the OMO concepts are then um organized, put in a file that are are given back to us and our epic uh um our IT ingestion team puts that into a schema back in the enterprise data platform. A schema is a is a isolated collection of data that can be sort of walled off in uh in the Azure cloud so that we can offer um access to that data. So uh we can offer uh Windows virtual desktop access to that data and with that data we have SQL Server and our studio so the data can be analyzed right where it is. No data in any part of this process, no data leaves our firewall. Everything happens within our firewall, so that provides a very, very safe and secure way to analyze big data. So we put together a team. Uh, we're a little bit limited on, on, on our, uh our local analytical uh capacity and our statistical capacity, but fortunately we were able to put a team together with uh with uh people that I know at Yale. And we were able to get funding from the Batton Foundation to support some of this, this teamwork so you can see our Yale team, uh, uh, Yan Liu, an epidemiologist Harlan Krumholz, uh, two analysts, Yin Tian Liu and, uh, Chung Soo Kim, our Centerra team, uh, John Burroughs, a cloud architect, and Mark Anderson, software engineer. And then we've got the refractor engineers that that have helped us making sure that the data is where it's supposed to be and and it's in analyzable condition. So, uh, our first study was a serial cross-sectional study. This data is an enormous amount of data. It's 1.7 terabytes of data that sits in the Azure cloud, and to analyze this, we, we sliced it into two year segments to do a serial cross-sectional study. This is quite similar to what the Nahanes study does. The Nahanes is the National Health and nutritional examination study. Which they sample 5000 nationally representative patients annually and they weight that data and extrapolate that data to give us an idea of what's happening with hypertension across the country. So we, we decided that we were gonna do a similar a similar study where we were gonna look at two year segments in a serial cross sectional study. Unlike uh Nahanes, which samples 5000 nationally representative patients in our big data we're looking at over a 376,000 patients uh over uh over 12 years. So rather than roughly 10,000 patients per two year segment, we're looking at upwards of. 600,000 patients per second. We're not sampling. We're looking directly at the entire population in our database and so you can see at the top you can see that the numbers of patients that we're looking at with each two year cycle and it's it's absolutely representative of who is in our in our database and so it's roughly 25% are African Americans, similar to our population. Uh, in that we serve, um, so there's no extrapolation and, and, and waiting that's necessary when we're looking directly at real world data. So we looked at our age adjusted hypertension prevalence rates and we looked at it using various operational definitions. When you're looking at data retrospectively, you can use a number of different operational definitions to define how you define hypertension in your database. Many studies use just the coded diagnosis, but there, there's a lack, there's a delay in coding and also there's a lack of coding sometimes in hypertension, and so you miss patients. You're doing it that way. The column to the very right, uh, in, in during a two year period of time, a patient had to have at least two blood pressure elevations greater than 140/90 or be on first line medications. And so this is similar to a guidelines recommendation for hypertension. The next one over is a random blood pressure during a two year period of time or on first line antihypertensive medications. That's quite similar to how Nahanes does it. They go and they randomly sample patients and if you've got hypertension on that particular day, uh, you're considered to have hypertension. So we looked at it in a number of different ways and depending upon how you define it, the prevalence, uh, you, you, you develop a more sensitive or more specific definition and the prevalence can change. So we looked at it in a number of different ways. So this slide shows the hypertension prevalence rates by race, ethnicity, and by operational definition, and you can see that African Americans, no matter what type of definition we use, African Americans have about a 12% higher prevalence rate of hypertension compared to other races and ethnicities. This is in our data. This is looking at our data. This is not looking at some extrapolated, uh, abstract Hanes data. This is our data. Uh, we had to use something called a generalized estimation equation method to analyze this data because, uh, we had to account for the possible, uh, fact that patients would go some patients went from, uh, segment to uh to uh time frame to time frame and so there, uh, there were correlations between observations from, from different time points. One of the advantages of working with the Yale team is we have access to absolutely first rate statistical uh collaboration. So this shows the age adjusted blood pressure control rates by operational definition. At best our control rates are about 70 73%. So, um, you can see that we've got room to improve control rate a failure rate of 30% just seems, um. It it just seems unacceptable. This is a disorder that the reason why we're treating it is to reduce someone's long term risk of having strokes, heart attacks, heart failure, and kidney failure. And so we need to think of ways that we can do better. This is our own data looking at holding a mirror to what we are doing here at Centera. These are the age adjusted control rates by race and ethnicity, and you can see that again African Americans have a lower control rate than other races and ethnicity, and this is despite the fact that African Americans were more likely to have a coded diagnosis of hypertension compared to other races. So we're not missing it, but we're just, it's harder to treat an African American, so we need to, we need to work harder on that. We need to work harder to uh solve these racial disparities. So this was published in the Journal of the American Heart Association last May, and this really kind of put us on the map. It showed us what we can, we are capable of doing here at Centera through Big Data analytics. So a question came up, did COVID affect risk factor monitoring and control at Centerra? Well, we were very easily able to look at this because of our, our data, and what we found is that in the second quarter of 2020 when COVID was really raging now. Our um our outpatient visits had a dramatic fall and our tele telehealth visits had a dramatic rise. Well, we knew that we, we were all living through that. But the surprising thing is that it all came back to normal very, very quickly. Within a quarter it came, it reverted back to normal, uh, in terms of the outpatient visits. There was a persistent, uh, higher rate of tele visits afterwards, but, uh, it came down dramatically after just a single quarter. Well, did that affect. Um, our ability to keep track of, of risk factors and treat patients' risk factors ends up, uh, surprisingly, no, we, we thought that we were gonna see a drop off in blood pressure control rates, A1C control rates, uh, lipid control rates. We really didn't see that. We saw a very slight and very transient drop off in the control rate of hypertension. It affected all races equally. We, uh, we published this as a research letter in, in Jack, um, last year. So next question about this racial disparity business with hypertension. Could racial disparities and hypertension be related to social determinants of health and social vulnerability? Just some background, neighborhood level social vulnerability is associated with hypertension prevalence and severity which may stem from neighborhood attributes or individual characteristics of the people who live in those neighborhoods or maybe even both factors. It's, it's uh it's something that still needs to be untangled. The association between neighborhood level social vulnerability and cardiovascular complications in hypertensive patients has not been studied. Understanding this association may help target interventions to patients with the highest risk of developing cardiovascular complications. After all, that's the whole reason why we treat hypertension. It doesn't make anybody feel better because they're generally asymptomatic to begin with. And so the whole point is to reduce their long-term cardiovascular risk. And so how do social determinants of health relate to long-term cardiovascular risk? So we were able to take the addresses of our patients in our database, geocode them to something called a PIPS code. A PIPS code represents a census track. A census tract is established by the Census Bureau every 10 years when the census is done, and the census track represents an area. That includes about 4000 people, so it's a fairly, it's a large neighborhood. It's sort of an extended neighborhood. Each one of them has about 4000 people, so a dense area is going to have a very small area, a country area, a not so dense pop populated area is gonna be very, very much bigger. The CDC publicly reports the social vulnerability index by census track and so based on patients' home census tracks and based on the social vulnerability rankings, which you can get them they're ranked as a percentile on a 0 to 1 scale, we were able to apply those uh social vulnerability index rankings to each one of our patients in our database. So this shows the social vulnerability index. um, it's, it's uh derived by the American Community Survey at the CDC. And it, uh, it has, um, you can look at the overall social vulnerability index, but you, it's also consists of four different themes socioeconomic status, household characteristics, racial and ethnic minority status, and housing type and transportation and you can see there are the things that uh that go into the calculation of the social vulnerability index. So and this this slide shows the uh uh heat map of the social vulnerability index by census track in our area. The dark blue are are are are more socially disadvantaged areas and the light, uh, the light sort of yellow or beige is the socially advantaged areas and so you can see that there is a broad array of social vulnerability or social advantage within our area. So to predict outcomes, uh, what we needed to do is a different type of study rather than serial, uh, study that we, we performed the cross serial cross sectional study, we needed to create a longitudinal cohort. So to do this we had to define a longitudinal cohort within our data. The way we did that is we uh we were looking at a data that extended over about a 14 year period of time. So we thought that any patient that we had seen for at least 5 years would be a candidate for the longitudinal cohort. That gives us adequate length of time to observe the patient, but also we wanted to make sure that we had seen the patient with adequate intensity over the time period that we had seen them. So we came up with something called the rate of years with visits and so um you can see down uh down below where um patient A had been seen over 5 years and was seen every one of those 5 years. So the rate of years with visits was 1. Patient B over 6 years was seen 4 times and so the rate of visit is 4/6. And the last patient had been seen over 12 years, but it had only been seen twice and so relatively a small amount, relatively low intensity of follow up and so what we did is we decided that the rate of years with visits of at least 50%. Would give us an intensity and so we defined a longitudinal cohort out of out of over 2 million people, almost 3 million people who had at least 2 all type visits. uh, we found about 1.5 million with at least 5 years of follow up and out of that, um, we found a little over a million who had um an adequately intensive follow up. We remove patients where if the follow up was just simply a lab visit or if it was an unmapped visit if we didn't know whether it was an inpatient, outpatient or ED visit then we uh remove those patients and so we came up with a longitudinal cohort of 981,000 patients. Um, you could, uh, compare this to the Framingham study. The Framingham study looked at 5000, uh, the initial cohort was like 5000 patients. We're looking at 981,000 patients over time. So, uh, among the longitudinal cohort cohort patients we defined a longitudinal hypertension cohort and so, uh, to do that for after the first, uh, start date of the patient's observation period, we look for either a coded diagnosis in EPIC, uh, uh, evidence of being prescribed a first line anti-hypertensive medication or, uh, two blood pressure measurements greater than either 40 140 over. Uh, either systolic greater than 140 or diastolic greater than 90, those two had to be at least 30 days apart and within 10 years. So that was our definition of hypertension according to the longitudinal uh for longitudinal hypertension cohort definition. Um And this, this, uh, slide shows the index date. The index date is the earliest of those three events after the first visit date. And so that was the index date for our, our longitudinal uh hypertensive cohort. What we, I, I analyzed, we analyzed the association between social vulnerability index quartiles, the most advantage to the, uh, being quartile one, the least advan the most disadvantage being quartile 4. We looked at the association between the SVI cortiles with the incidence of the composite endpoint. Our composite endpoint was a combination of new myocardial infarction, heart failure, or stroke, and we analyzed this using Kaplan-Myer curves and sequential Cox proportional hazard models which sequentially added demographics, BMI and baseline BMI and blood pressure and other comorbidities. And we uh we we'll demonstrate these hazard ratios by the SBI quartile with quartile one as the reference. So this shows the baseline characteristics and just a a quick thing number one, the size of our uh our um our our hypertensive lung a launch of the hypertensive cohort was uh almost 331,000 patients. Uh, you can see that, um, that, uh, this is, this displays it according to SBI quartile, and you can see that in quartile one the most. Advantaged patients who lived in the most advantaged neighborhoods. African Americans were only 15% of that group, whereas, uh, in the SVI quartile 4, the most disadvantaged neighborhoods, African American Americans, uh, comprised of 54% of people living in the disadvantaged neighborhoods. There was also some um differences in in baseline BMI and baseline uh blood pressure and so clearly we had to adjust for those factors. This slide shows uh characteristics baseline characteristics by race. You can see that um within the longitudinal hypertensive cohort among white people, white people were uh older and more likely to be male uh there were differences in some of the comorbidities so we clearly had to adjust for age. Gender, um, and uh morbid illnesses when we did our analysis. So this slide shows the Kaplan Meyer curves by social vulnerability index quartile and uh the, the, the overall SVI is in the upper left and the theme one, theme two, and theme 4 are in the other, uh, in the other graphs. You can see that there is a step-wise worsening of uh event free survival uh as your social vulnerability index becomes worse. Uh, so there's, there's a, there's a step-wise, uh, association between your social vulnerability index, your neighborhood level social vulnerability index, and your, uh, your probability of developing one of the composite endpoints. So this slide shows unadjusted and sequationally adjusted Cox models for the association between the social vulnerability index and the composite endpoint of myocardial infarction, heart failure, and, and stroke. And so you can look at the SBI quartile 4. It's the third panel. You can see that um there is uh a a highly significant association between social vulnerability index and uh the composite endpoint. And and that association persisted despite adjusting for demographics, BMI and baseline BMI and blood pressure, as well as comorbidities, but roughly 30% higher chance of having those uh those one of the composite end points if you lived in a neighborhood with a social vulnerability index of 4 in the fourth quartile uh as compared to the first quartile. We were missing social vulnerability and basically we missed, uh, we didn't have an address, but, uh, in some patients they had a uh Uh, they, they may not have had an address recorded or they may have had a, uh, post office box rather than an address that consisted of less than 5% of the total population. So we had, uh, we had social vulnerability annex and greater than 95% of the patients that we analyzed. The slide shows the association between social vulnerability index themes as well as the area deprivation index. We did a sort of a sensitivity analysis by looking at a different type of index, and you can see here that theme one, which is socioeconomic status, and theme 4, which is housing type and transportation, showed the greatest, um, greatest association, the, the largest, uh, uh, um, odds ratios and the greatest association. Between between the SVI and the and the composite endpoint and the area deprivation index, a different way of looking at vulnerability also showed a marked and highly statistically significant association between social vulnerability or area deprivation and adverse outcomes, complications in hypertensive patients. We did sequential models excluding the social vulnerability index. We found that African Americans were younger than whites and so at baseline in the univariate model, African Americans appeared to do better actually, but it was because of the composition of the uh of the longitudinal cohort, uh, so we had to adjust for that. And so when you adjusted for age, uh, African Americans were more likely to have uh the composite end point. Um, but then when you added back all of the, uh, various, um co-morbid illnesses, um, uh, race, self-identified race sort of washed out of the model, whereas social vulnerability, which, which is clearly entangled with race, social vulnerability persisted as a predictive model. So this gives us an opportunity. You can't go and change someone's self-reported race, but you can go in and do something about social vulnerability. So this slide shows a heat map of uh in the top panel it's the social vulnerability index with the darker, darker census tracks having worse uh social vulnerability index, and the bottom panel is the outcomes in gray are places where we just had a few too many few too. Too few patients in that particular area to make an analysis you can see that the areas of social vulnerability pretty much track with the areas of uh of worse outcomes. We also looked at control rates, blood pressure control rates by social vulnerability quartile, and we looked at whether you wanted to use the 140/90 or 130/80. The 140/90 is in the dark curves. You can see there's a stepwise increase in uh um in uncontrolled blood pressure rates as you uh look at more socially vulnerable neighborhoods and uh that's highly statistically significant using either cutoff criteria of 140/90 or 130/80. You can see that the, the uncontrolled rate for 130/80 is really high in the 60% range, so we've got work to do. So in a longitudinal cohort of real world hypertensive patients, social vulnerability index was strongly associated with the composite endpoint of incident myocardial infarction, heart failure, or stroke in sequentially adjusted models, SVI remains strongly associated. after adjusting for demographics baseline uh BP and BMI and comorbidities, SVI theme one socioeconomic status and SVI theme for housing type and transportation were most strongly associated with the risk and the hypertensive launched through the cohort, white patients were 6 years older and more likely to be male than black patients, accounting for the negative association between black race and outcomes and the unadjusted analysis. Clearly race and SVI are interrelated. Black patients were almost 54% of the patients in neighborhoods where the social vulnerability index was in the fourth quartile, but only in 15, less than 15% of patients in the uh in the advantage uh uh neighborhoods with a social vulnerability index in quartile one. The effect of race was attenuated in multi-variable models, but the effect of SVI had a persistent effect in all models. Worse SVI quartiles were significantly associated with worse blood pressure control. So, um, SVI is strongly associated with cardiovascular complications in this large diverse cohort of hypertensive patients. Addressing area level social vulnerability may be an important. Uh, for it may be important for identifying high risk hypertensive patients who live in high risk neighborhoods who may benefit from more intensive interventions. Health systems like ours can use this information to target vulnerable areas and then identify high risk individuals within those areas as a strategy to address racial disparities, improve cardiovascular outcomes and hypertensive patients. So I mentioned that um the the the social vulnerability index theme that was one of the most prominent ones was housing and transportation. So maybe if people have adequate inadequate transportation and they have, they're on 3 blood pressure medications, they gotta change buses twice to get to a pharmacy. Maybe that's an impediment. Maybe you can, and, and so just last week VCU Pharmacy, um, uh, published this where they looked at pharmacy deserts. And so this on the right you can see that there's 7 pharmacy deserts according to their analysis in Norfolk and you can see that 3 of those 7 pharmacy deserts match up with our, our, our areas that are particularly troublesome. According to our analysis, so it might be possible to put satellite pharmacies in particular neighborhoods where there's pharmacy deserts. We might be able to address some of these disparities by doing things like that. This is what a a health system like ours is capable of doing. This has been submitted for publication and it's under uh under review currently. So there are other studies we have in progress. We've got a study led by Yuan Liu, uh, hiding in plain sight, leveraging in the electronic health record to assess delay in the diagnosis of hypertension. There's a subgroup of people who have a delay in hypertension and ends up they have a higher complication rate. A 6 month delay in hyper and making the diagnosis has an impact. Um, we've got another study where we're looking at within visit variability, uh, among, uh, and, and looking at that's impact on outcomes. If you've got multiple blood pressures during a visit and, and their blood pressures are all over the place, that's associated with the worst outcome. We're doing also using this OMA platform to do some studies outside of cardiology in conjunction with maternal fetal medicine. Uh, at, uh, at Tutsuya Kaiyukita and George Sade looking at the association of the SVI and stillbirths, um, we're, uh, we're getting ready to start a study with the endocrine folks looking at association with SVI and diabetes outcomes, and we've got another study that, uh, we're getting started with using our OMA platform to see if we can calibrate the prevent. Uh, prediction models based on our local characteristics using AI to uh maybe uh improve the calibration of prediction models. So we've got a number of different studies that are up and running now using our OMO data. It's a, it's an enormous resource that we have now that we've created and enabling us to do big data analytics. So how about the boots on the ground approach? We received the Peoria grant to study hypertension in vulnerable populations. We're in good company on this. We're working with Yale New Haven Hospital, Houston Methodist, and Mass General. And um this is our local uh team with our our community PIs Iris Lundy, Chuck Lavelle, Keith Newby, and Mike Charles and uh and I, I don't, I don't have a picture here of Jen May, but Jen May is here and she is our project manager for this and who's just done. An absolutely wonderful job on this project. Our aims are to uh compare the effectiveness of remote blood pressure management program with and without support of community health workers in comparison with a community standard. The outcomes that we're looking at is systolic blood pressure at 6 months as the primary outcome in 18 months, but also looking at uh how well we engage these people. Remote blood pressure monitoring and community health workers are ways that you can create better engagement. With someone and it's absolutely important for hypertension to have that engagement with the patient. There's a lot of self-care that has to occur with hypertension care, and if you're not engaged with the patient, you're just not you you're given the right prescriptions, but it's gonna go nowhere. So and we're gonna also evaluate the implement implementation strategies using a a very organized approach. So to do this, our community-based organizations in our area are 10 African American churches, um, and you can see the the churches in in Norfolk, Virginia Beach, uh, Newport News, and Hampton, um, it's just been a pleasure and a privilege to go around and meet the people at these churches and see what they're already doing related to not only spiritual health but physical health, uh, in their congregations. Um, Jen May and I find ourselves going to church on Sunday morning where we're invited to the pulpit. We're invited the, the welcome that we receive at these churches is absolutely amazing. We are invited to the pulpit to make announcements, uh, about our project which has given us an unbelievable, uh, uh, access to and, and, and, and connection with the patients, uh, who were trying to recruit but it's also it's just been, um. It's been an amazing experience. I wish I had done this study 32.5 years ago when I came to this community because I think that I would have been a better physician knowing about where people come from, the people that I see, where they come from, what communities are they living in, um, and, uh, particularly communities that may be different from the community where I live, so, uh, it's been a remarkable, uh, experience for me and, and all of us. Um, the remote blood pressure monitoring, this is yet another example where Senterra IT came to the rescue. This is actually very difficult. What we do is we, we issue a blood pressure cuff that communicates by Bluetooth to something called a Stell hub. Stell Health makes this hub, which takes this data, transmits it by cell phone. To, uh, central monitoring uh resource and so we, it's important to do it by cell phone because some people don't have internet at home and but everybody generally has cell coverage and so um we were able to include more people doing it that way but to get that data into Epic. Required an ongoing effort by must have been a dozen Centerra uh software engineers and epic managers handling the identity management and and setting up flow sheets and setting up the orders and doing all of these things. I, I had no idea how complicated this was going to be, but fortunately. Centerra IT put together this team, got this job done in a really, really remarkable fashion. So the way the study works, it's a step-wise, uh, recruiting approach where patients are each community-based organization is randomized to one of four sequences, and the uh recruiting of patients occurs within a sequence where initially they were in a phase where they were recruiting into the community standard arm of the study. Then we shifted to the remote blood pressure monitoring arm of the study. And now more recently we're now shifting to the remote blood pressure monitoring plus community health worker arm of the study. The point of this is to try to spread out the recruitment over time so if there are any kind of, um, there are any kind of external environmental factors that could potentially affect people's blood pressure control like COVID or something like that that came along, the recruiting would be set was spread out over a longer longer period of time where that would all sort of wash out. And so this was all extremely well designed by uh statisticians, uh, and, and, um, or and, and the folks at Yale. This shows the enrollment update and I'll just show you uh in red, uh, Virginia is us and Tara and you can see that our, our goal over the course of the enrollment period is to enroll 360 people. We're well on our way. We've enrolled 162 people and you can see that we are ahead of the pack. We're doing extremely well up up top is uh how each church is doing. And we've got a little bit of catching up to do in a couple of churches we're gonna go visit those churches in the next uh week or so, um, but, uh, we're very, very happy with our enrollment, uh, here at, at, uh, at Centera. So that's kind of where we are on this project. More to learn from this project, there's gonna be a whole lot more to learn. I'm betting, I'm betting the remote blood pressure monitoring. And community health workers are gonna have an effect. We, we've been doing the same thing over and over and over again. I mean it's like, it's like Groundhog Day. We, we keep trying, you know, handing out pills, awareness campaigns and what have you. It's not working. There are all kinds of high tech possibilities that might be coming down, you know, um, uh, renal artery, uh, ablations and what have you. I'm not betting on those. I'm betting on the low tech, uh, and I think that dealing with populations, uh, where you can deal with populations using remote blood pressure monitoring, by the way, does it make sense to try to treat people and, and see whether we're doing a good job by having them come in. Like the soonest you're gonna be able to come come back is like 3 months for an isolated blood pressure in the office. We've got patients who in a 10, a 10 day period of time with remote blood pressure monitoring, they've got 60 blood pressure measurements at home. We can look at all of them. We can look at trends. It's a much better way of taking care of of blood pressure and seeing whether you're having an effect because you're looking at what's happening where the patient lives. And also the patient, him or herself is looking at the blood pressure through remote blood pressure monitoring. This is not just remote blood pressure measuring, it's measuring, it's remote blood pressure monitoring. All of these blood pressures are being monitored by our nurse practitioner Leanne Gordon. She all of these, they come, they flow into Senttera and if they're above a certain range, Leanne gets an. Automatic internal email within epic that somebody has a markedly elevated blood pressure. She'll call them, ask them how they're doing, if they're symptomatic, they'll be directed to go to an emergency room. If they're not symptomatic, then she'll make some changes and also work with the primary care doctor to get control of that blood pressure. It makes much more sense. So I'm betting on that, but we'll see. I want to just end by this. You know, I, I mentioned that when I was a medical student 45 years ago, 95% of patients were called essential hypertension. 45 years later it's still called essential hypertension. There was a book written by a guy named Sir George Pickering in the in in the mid 1960s called The Nature of Essential hypertension. Um, and the nature of hypertension was, uh, the etiology was debated by two British clinical researchers. One was Dr. Robert Platt, who thought the blood pressure was bimodally distributed distributed in among patients, and that's what suggests that there is a discrete group of people who have hypertension, maybe uh genetically determined, but they have a biological reason in a discrete group of people. Whereas George Pickering, Sir George Pickering thought that blood pressure was just one end of a normal distribution and in patients and in populations, the whole distribution just shifts so Pickering said the difference between those with the disease and those without it is one of degree, not of kind. The differences are quantitative, not qualitative. This is very different from any other disease. When, when cancer, it occurs more likely as we as we grow older, but it's a discrete group of people that define themselves as having cancer. It's just not one end of a bell shaped curve, but hypertension is different. These are the curves. These are the distribution curves, one on the left is from Platt, the one on the right, um, one on the left is from Pickering. One on the right is from Platt. Um, you can see the numbers, you know, they're they're like 40 blood pressure measurements, 30 and 50 and on the, on the right. So we now have the capability of looking at blood pressure using big data analytics that Platt and Pickering would could only imagine. This is a blood pressure measurement on over 198 million blood pressures in our in our database. This is the distribution of over 198 million blood pressures, and you can see that, you know, there are spikes, people tend around to the nearest 1010 millimeters, and so you can see those spikes, but you can see that. Pickering was right. It's a, it's a, it's a bell shaped curve. There's maybe a little skew to it, but there's a bell shaped curve. You the systolic is up top and the diastolic is at the bottom. So you can see here, you can see the distribution of about 1.5 million people. This is distribution of blood pressures averaged by individual of 529,320 patients and it's displayed as by race and ethnicity, but it's also um it's distributed by by age and so top is. I in in white patients, the second one is in black patients. The third is in Hispanic patients, and the 4th is in Asian patients and you can see that as we age, as we age, what happens is a normal distribution just shifts to the right. You can see here on the top you can see that the the mean, the median systolic blood pressure, this is the median of individual means it went from 118 to 126 to almost 130 as we age, and it's a perfect bell shaped curve. Pickering was right. There is no specific abnormality. Something happens to us as populations could be stress, could be environmental factors, could be diet. Uh, it could be any number of things, but in terms of an inborn abnormality, those are rare, those are extremely rare. So this, it suggests to me that that operating on. Environmental exposures or social vulnerability may be a way that we can get better uh uh treatment and also looking at this from a population level we're looking at an individual level trying to take care of each person, but we're trying to reduce the population risk that's our intent and so maybe rethinking how we think about hypertension is in order hypertension is a peculiar disorder. Standard way that we make a diagnosis is through a process of of hypothesis generation and hypothesis testing, not with hypertension. It's diagnosed by making a blood pressure and then it's confirmed by a circular logic of repeating the same blood pressure measurement. There's no independent confirmatory test to um to prove that someone has hypertension. It's compared to, you know, if you think somebody's got cancer of the lung, you do a a CAT scan and. Or if you think somebody's got heart disease, you do a cardiac catheterization, you've got confirmatory tests. So compared to those other diagnoses, a diagnosis of basic 22 blood pressure cuff inflations seems kind of flimsy, right? And so patients are reluctant to accept the diagnosis and there's a lot of diagnostic inertia that occurs in in hypertension. The basic science has revealed this whole tangle of mediators that control blood pressure. We know all kinds of things that control blood pressure, but there's no smoking gun. To explain why patients with some patients have hypertension. 95% of patients are still categorized as having essential hypertension after all these years and all the basic science that's been done. It's inherited, but it's not well explained by genetics. But guess what, social determinants can also be inherited, particularly in places where people live in segregated neighborhoods. People are sort of constrained of those neighborhoods generation after generation, and so those so those neighborhood level social vulnerabilities may translate, um, from generation to generation. The diagnostic cutoff for hypertension is defined by epidemiology and by population level treatment effects. These are abstract concepts, uh, for patients to think about, you know, thinking about long term risk, they, they're not thinking, I mean just helping people think about the concept of risk is sometimes difficult. And so, uh, those abstract concepts can lead to patient denial and diagnostic inertia. Maybe it's time to redefine hypertension. As a population level so socially driven disease, there may be patterns in big data that we're like we've been looking at that may emerge that can add to our knowledge about hypertension, about its ideology and about the reasons why certain populations are more vulnerable than others. A population approach may also result in better blood pressure controls. We can meet people where they congregate in churches and community centers to try to uh create a community-wide process of trying to take care of people with high blood pressure, having them come in every 3 months. for an isolated visit may not be the approach we can see them much on a much more broad way with remote blood pressure monitoring and that may be a much better way of taking care of a broad population of people in a way that is is more efficient. So, um, I think that our our um. Our research at Centera and hypertension has practical implications. I think we have created the capability, the capacity to do big data analytics using hypertension as our initial use case. Um, we've gotten this Peor grant now, a federally funded grant, and we've created the capacity to handle federally funded, uh, grant funding now, and it came just in time we're recruiting a. An academic chief of cardiology is coming with with grants and so we've got we've got the mechanism set up to handle the administration of grants and so research has made us better and uh and hypertension research has been part of that so I really appreciate your attention and I'll stop there and we'll take some questions. Thank you. Published February 26, 2025 Created by Related Presenters John Brush, M.D., F.A.C.C. Sentara Cardiology Specialists -Interventional Cardiology View full profile