Data-Driven Care: The Key to Accountable Care Delivery from a Physician Group Perspective
Data-Driven Care: The Key to Accountable Care Delivery from a Physician Group Perspective
Today’s agenda, we’re going to have Dr. Spencer share about Crystal Run and why they’re so invested in data analytics with some experiences they learned, their strategy for how they’re using care improve care and also improve their financial viability, how they’re using data to manage patient populations that’s universally enough covered by a large body of different payer entities, the experiences and challenges and successes that they had in terms of engaging clinicians to use data, and then discuss the importance of the Adoptive Data Architecture, where using that platform to turn clinical questions into actionable results is the objective of what Health Catalyst does as a company.
Poll Question #1
So with that we have our first poll question .
What best describes the group you belong to?
Okay. Our first poll question is to kind of help us get a sense of the constituency that’s on the phone. So if you wouldn’t mind taking a second and identify a fit for these groups or in this broad other category and they’ll help us kind of target our comments to make sure that they’re specific to your particular area.
Okay. So, Joanne, we have, it looks like 4 health plans, 8 physician groups, 4 provider organizations, 5 vendors, and 11 in the other category. So that’s a pretty good mix of audience and that certainly does help us kind of target our comments during the presentation.
So with that, I will be glad to turn it over to Dr. Spencer.
Crystal Run Healthcare
[Dr. Gregory Spencer]
Thank you. So I will get back here. And so, I’m Greg Spencer, I am the Chief Medical Officer at Crystal Run Healthcare. We’re a physician-owned multispecialty group in upstate New York. We’re about an hour or so northwest of Manhattan in Orange County, though the East Coast Orange County. We’ve been around for about 18 years. I was the second internist and the seventh doctor who joined the group in 1996 and we’ve grown mostly through kind of organic growth, hiring one at a time many out of residency. Up to this point now, we have about over 300 providers. Average age of our group is about 41. And so, we’re young both in age and in our provider’s age.
We are getting to be pretty full service. We have a Joint Venture Ambulatory Surgery Center, we have multiple, now four urgent cares, a diagnostic imaging, we do in-house sleep center, we have a high complexity lab. We are very early adopters of EHR. We’ve been an EHR as an organization since 1999 and we don’t just use it as a substitute for dictation but we actually use the templates. We’ve been kind of moving in a direction towards population management for a long time. We’ve had our own internal care managers that initially were kind of more oriented towards getting people to do their health maintenance stuff but eventually have morphed into complex kind of case management and we’re one of the first pilot practices in the country to be certified by the Joint Commission and we’ve been medical home certified as well. And medical home, for those of you who aren’t kind of familiar with that, it’s a certification of primary care that kind of again recognizes the importance of population management in addition to the patient in front of you but all the trappings that you need to manage groups of patients.
Crystal Run Healthcare
We are a single entity ACO. We are in the first batch of Medicare Shared Savings ACO in April of 2012. And so, we don’t have a hospital in the mix. Our local hospital was in the business of opening the first new hospital, building the first new hospital in New York State in 25 years and didn’t have tremendous bandwidth at that time, and it’s okay and you know, that is another topic for another discussion. But we are the early adopter credentialed NCQA ACO. We take care of several hundred thousand patients but we have about 25,000 commercial lives and about the 10,000 or 12,000 under Medicare Shared Savings that we are actively managing. New York State is kind of behind as far as the west of the country anyway as far as the population management risk contracting, but we kind of moved everything in that direction. Of the people that we have in our ACO, about 82% get their primary care within the ACO.
Crystal Run Healthcare
So this is our copy of our mission statement that we’ve designed back in 1997. Our CEO went at that time immediately after I joined of course and went to get his MBA at Colombia and came back and we needed a mission statement, mission and vision, etc, and this is what we came up with. So basically improving the quality and availability and satisfaction with the care, medical excellence, so quality, and also service excellence and empowerment.
The Goal: The Triple Aim
And if you look at this back in 1997 and compare this to the triple aim, you know, increase or enhance the experience of care, improve the health of the population, which is quality and then reducing cost, or at least control the rise of the cost on a per capita basis.
Crystal Run Strategy and Objectives
So we were kind of aligned earlier than we knew it and that’s been helpful for us going forward. We’ve kind of embraced it and brought it in to that kind of fabric of our practice. We really feel, again, my worldview, because I’m in a multispecialty private physician-owned physician wide practices, that physicians have an important role to play in driving change in healthcare, and one of the reasons we decided to grow is to become unavoidable and kind of have a see at the table and be one of several entities, including payers and hospital systems but to have a role in changing and driving the change in healthcare.
We’ve really focused on coordinated care. Again, we are a group in every sense of the word. So we are not a product of multiple mergers, although we have done more mergers in the recent past where there’s kind of backend merger in some of those other models where there’s almost independence at the clinical level. We are very much coordinated and integrated and we embrace that and we feel like that’s the best way that we’re going to be able to move forward and kind of improve the health of the whole population, which is very much critical, and a skill set that is not quickly learned. You go to medical school to take care of the person in front of you and you still need to do that but you also need to kind of think about the people that aren’t in front of you that are under your care as well.
We have competition both within the community with other physicians but also from hospitals and health plans that are buying providers, employing providers. New York is not on the leading edge of the health plan but that is of them getting, employing physicians but that is very quickly to be changing, as what I understand.
The coverage area, and we’re growing, we feel like in order for us to do what we need to do in the value-based care world, so we’re not just getting paid for doing things, we’re not getting paid for healthcare, we’re getting paid for health, so in order for us to do that, we have to have systems in place that allow us to manage population. So we’ll talk about some of the things that we need today as far as the IT data warehousing, clinical, organizational infrastructure. But just to kind of make it work, to manage populations, the bigger the population you have, the more it behaves in a predictable way. The lower numbers, you’re then at the risk of, you know, kind of rare random events occurring that really put a wrench in the work. So, in order for us to do this well and effectively, we have to expand in size, scope and geography. And again, I alluded to this earlier, our physicians provide coordinated care, and that, again together with kind of the underpinnings of the medical home, is all about that physician certainly being in accordant to the head captain of the ship, quarterback, whatever metaphor you want to use, but also part of the team and that team, to kind of leverage that time, requires that we kind of evolve and continue to build the skill sets of the members of the team. And we want to be the practice of choice not just for patients but also our employees and our physicians.
Crystal Run Governance Model
This is Crystal Run Governance Model. It should be kind of more data governance model. So Hal Teitelbaum is our CEO and managing partner. He’s a physician. He kind of oversees things. So again, the management structure or the practice involves a management committee, as well as a kind of a C suite of executives that help manage things on a day-to-day basis. But the warehouse, the data warehouse that we’re speaking about, the kind of the place where all the different source systems kind of pull together to be able to get a wider view of things has executive sponsors, myself. Michelle Koury, who’s the chief operating officer, she started the same day I did. She’s an internist and our chief HR officer and an accounting senior VP. And then there are kind of the working governance people of the warehouse. And what the governance kind of involves is to kind of develop the policies of who gets to see what, what information needs to be getting to people. You certainly want people to get information but only the stuff that they kind of need to know. So there’s privacy concerns as well as workflow concerns. There are just the mandating things of what do you call something, is this really the systolic blood pressure, you know, those kind of things, is it right arm or left arm, to kind of make it make sense from a clinical perspective – because if the data is not gathered properly, a lot of times it doesn’t make sense. And then we also coordinate with other partners in the practice, as well as the providers and the rest of the staff, not certainly limited to the physicians or clinical staff.
Crystal Run Care Management Strategies
So our management strategies for kind of in general implementing value-based care – again, value-based care on my mind is providing high quality care, doing good things that are data-driven, preferably evidence-based medicine, and not doing things that haven’t been shown to benefit the patient. So, eliminating wastes and it’s not wastes. It’s a little bit of a pejorative term, so you don’t kind of say that exactly but wastes are things that are currently being done that don’t help the patient. And so, everything is kind of viewed with that, is this going to help and how is it going to help, has it been shown to help. And to try to eliminate the things that don’t kind of contribute to that bottom line of advancing the patient’s health.
So we have care managers, which are RNs, in each of our medical home sites. So we currently have six medical home sites. We have a care manager in those sites to meet with patients as needed. We also have a care manager in the hospital who’s paid for on our dime who kind of coordinates the transition of care, and as well we have a team that would visit the patients in the home if need be if they can’t make it to the office. So we identify high-risk patients either from registries, that is with the patient’s lists, that a patient that has certain conditions, diabetes, coronary artery disease, CHF, and you need systems in place to identify those, to create those lists. They are the people that are appropriate in those lists that need management and are sent to the care managers. And then team referrals, people that are sick in and out of the hospital, the people that we think need complex care management are sent to the care managers.
We have evidence-based protocols that we try to use. We’re going to be going over kind of how we do these projects within the practice, variation reduction projects with the clinicians, but we use things that again have been noted to work in the literature. We use an EHR, as I said, but we also have certain patients, again, with CHF, severe COPD, multiple admissions. We send them home with the Bluetooth-enabled scale pulse oximeter, blood pressure, and they do these things daily. It goes to a web service that sends it to us if there are changes and then we can reach out to them if this patient with CHF has a sudden increase in their weight, for example, so we can reach, you know, intervene.
Crystal Run Quality Structure
As far as the quality structure, we have 27 clinical division heads and each of those are headed by a physician, usually it’s a specialist in that specialty, sometimes it’s a grouping based on kind of a geographic consideration, but the quality committee is headed by physicians who manage the quality efforts and information to call the best practice council who kind of work through, review the registries, review the quality data, and then report up to a practice-wide committee, which is called Quality And Patient Safety, and that’s kind of a joint commission committee.
Why Crystal Run is heavily investing in analytics and data warehousing
So why are we investing in the analytics and data warehouse?
Poll Question #2
Here is the second poll question. I guess I’ll hand it back for that to Joanne.
If you are a health plan, physician group, or provider organization, do you currently exchange clinical and claims data with these other constituents?
Alright. So if you are a health plan, physician group, or provider organization, do you currently exchange clinical and claims data with other constituents? And if you could answer, we’ll get an idea of what everybody does.
Okay. Alright. Survey stopped. So I’m going to take it back here. So it looks like about 85% of people do exchange claims data with others. So that’s interesting. You’re maybe not in New York.
Crystal Run Data Analytics Strategy
So this is one of the reasons that we’re doing this but I’ll continue here. So our data warehouse analytics strategy, we have had for a long time the ability to report. We have been in this intelligence department to do quality reporting and mine data very intense from a personnel perspective when you’re kind of pulling reports and joining things across multiple source systems. So, we knew we needed to have a warehouse to put everything in to be able to simplify, automate, and scale the reporting that we haven’t been able to do currently. And also as things became more sophisticated, they have kind of dedicated analytic applications that would kind of enable people to that expertise and actuarial science and more statistical things of sorting out things that are closer and more grey to help us with that and that’s not possible with a more ad hoc system.
We do daily financial reporting and that it’s becoming more and more complex. Order tracking both inside and outside. And then claims aggregation where the claims that we do get, we would like and need to bring it in and kind of marry that up with the clinical data claims. There’s a big lag from the time that our service is provided until the claim clears and sometimes that is too much time. If you have that together with the instantaneous provider submitted data, at least we know what we did. We may not know what comes from outside the organization, patients getting care outside the organization but if we can really improve our knowledge by kind of marrying the claims with the medical EHR data. And then RVUs to get an idea of how much work is being done so that we can plan both for staffing, as well as for providing care in certain areas.
Crystal Run Analytics Current State
So again, I alluded to this current state. We really depend on data, we always have, but getting more so. We have people that do SQL that run these reports a lot using reporting services and other tools including the simple analytic tools of Excel, Access, Tableau as a front end for kind of displaying data. But that take a lot of time, there’s a lot of manual both entry as well as extraction and not scalable. We just have to keep hiring and hiring. And as a lean organization, you’d like to do what you can to have people practicing really at the top of their license. And as the less automation we have, we always have to, you’ve kind of come to places where you’re like, well, decisions were do we really, how bad do we need this particular measure because the amount of time that it takes or disruption both to your BI team but also clinical teams to enter data, at what point does that become too much disruption versus the value that it provides. And again, the less automation you have, the more you’re asking yourself that question and the more it gets pushed towards the, you know, the work is onerous. And in addition to collection of quality, there is the reporting requirements that we have both within to maintain our quality and advance our mission but also outside to the ACO to other different entities.
So this is an example of a physician dashboard that we had for the last 3 years where we’re able to kind of present user RVUs, OB-GYN, very busy OB-GYN with benchmarks of different things, but very kind of intense to do this.
These are fairly automated but over a long period of time we could see the day-to-day, do people turn in their charges from the hospital when they saw somebody outside the organization. We also have care quality measures that are displayed, no show rates, the E&M variance. So we were pretty sophisticated and this would give you an idea where we were but again this was, there’s a lot of work to do in maintaining this.
Turning data into improve care & ensuring financial viability in the future
Physician Variation Analysis
So, going forward into the future to kind of ensure our viability, as well as to be able to provide best care for our patients, we decided to partner and develop a more modern data warehouse enterprise data warehouse. And this is an example of, you know, you kind of saw prior what, you know, fairly static lines, mostly one-dimensional. These are two to three-dimensional, and in this case, this is just an example of some of the potential reports. So each of these dots is a doctor. The X axis is the cost, in this case, is the vascular procedure, the cost, increasing from left to right. And then the size, this is called a swim lane here. So the size of the circle has to do with how many cases. So you can get an idea of how much variability there is, how much variation in a particular procedure and you can have multiple swim lanes for different procedures and then concentrate on those areas that are high cost or high variability first because no matter how efficient you are, you only have a fixed number of personnel to work on things.
Crystal Run Results
So using different ways, we’ve been able to have pretty positive results. I’ll do the results and then kind of walk through the programs that we used to get there. So we’ve reduced our admission rate by about 4% and that’s absolute 4%. So it’s about 20% overall from where it was before and we kind of, we’re at the national norm before we kind of knew where we were. So we’ve gone down to less than 17% overall. Improved our mammogram rates, still above 75%. A1c. So Hemoglobin A1c is a measure of sugar control and the higher, its hemoglobin has sugars stuck on to it, and the higher your sugars are, the more sugars stick on to the hemoglobin, and 9 is high, that’s 9% of your hemoglobin molecule which have sugars stuck onto it. And so, most diabetic measures, they like it to keep it under about 20% and we’re below 9% to A1c. So that’s very good control.
And then people with high blood pressure greater than 75% have well-controlled blood pressures defined by less than 140/90.
Breast Cancer Screening – Mammography
And here you can kind of see where we started some of these efforts and then when it occurred. So kind of the second quarter of 2012, we implemented a more data-driven approach and you could see the improvement in mammogram.
Outcomes: Avoidable Admissions
The trend again in 2012 down – this is avoidable admissions. This is the data that we got from Shared Savings, so this is all claims, down to now less than 17% at this point. Overall definitely an improvement.
Total cost difference
So this is kind of a little bit different thing. We had an innovation contest within the practice where we kind of put the word out that bring you ideas of how you would kind of eliminate, you know, standardize things, decrease variation. And so, we have this variation reduction project. So what it involves is you get all the doctors together in a certain specialty and you say, okay, our problem that’s kind of worked out where your specialty society says, you know, the data shows that for this thing, all things being equal, you should probably do these things. And so, all the specialists kind of got together and said, we all practice, we’re all covering for each other, let’s all agree to approach this thing in this way. If you want to vary, that’s’ fine, but you should have a reason because the patient was allergic, or whatever, not because that’s how you did it where you trained.
And so, this is an example of a contest where we did that more broadly. So we did the variation reduction at each specialty but this is above and beyond where one or two of the physicians kind of got the idea, well, let’s do a special project. So they looked at patients who were diagnosed with breast cancer and agreed on a pathway of things that you should do and when they should get done and how often, in this case, you should use this PEG-filgrastim, which is a colony-stimulating factor. Kind of when you give people chemotherapy, sometimes you have to give it to make their blood counts come up. But sometimes you don’t. And in many places they kind of give it at a certain point, whether or not their counts get down. So they all agreed we’re going to do it only when certain criteria are met. And you can see a before and after and the numbers are pretty good actually. We have a busy oncology department. So there’s 791 patients before and 817 after and the savings just on, this is just for this one medicine, was over a quarter of a million dollars and this project is being kind of written up and there are other things.
Reducing Pharmaceutical Costs
But you can kind of see here, this before the blue is the doctors before and then after they agreed, and actually all the quality scores went up.
A Culture of Efficiency: Improving Access
So just by standardizing and thinking about what gets done, there’s a lot of wasteful things that can be eliminated. So just agreeing that patients need to be seen, say, for well-controlled diabetes every 6 months rather than every 3 months, you can actually create virtual doctors, you create capacity.
So this is a slide where we have, I believe this is for diabetes, where we have the sum of the visits per patient. So the visits per patient are going down but because of the capacity, we’re able to see more new patients quickly. So the visit count actually went up and the number of distinct patients under our care went up as well. So if you figure that a primary care doctor sees, you know, 3500 to 4000 visits or so, you can kind of divide 41,000 by 3500 and you come up with 12, it’s like we hired, we’re able to kind of create the capacity of 12 physicians without having to pay any additional money.
Variation Reduction Spread
Reduction in Charges
So this is a list of the different variation reduction projects that we’ve undergone in each of the departments. In almost all of them there was a change in the charges per patient. Sometimes there isn’t and this is just full disclosure. There was some issue with Pulmonology and I think this is because a lot of the medicines went – they were killing the ozone layer. So a lot of generics came off the market and so there were more expensive prescriptions.
Variation Reduction Spread
But you could see, overall, a lot of changes of eliminating visits, etc. that weren’t necessary – again, all of our quality measures went up. We’re not giving, no care, and the visits went down but the number of absolute patients went up.
Managing patient populations across payer entities
So that’s how we kind of more or less approach from a clinician level of getting by, and as we meet, we decide, we measure, and then give the feedback.
Crystal Run’s payer mix
As far as how we kind of relate within the care management community, in the past we would say we were blessed because we have 24 payer entities in our market, none of which are hugely dominant, which is great in a fee for service environment where you don’t have kind of from a vendor or from a, you know, working with payers, you don’t have all your eggs in one basket. So that was good in the days when we were fee for service. It is not so good now where because there’s no dominant payer, it’s hard to again leverage the power of numbers, where 3 to 5 to 10,000 patients per payer, that’s really not adequate to do a full population risk arrangement with confidence. You end up getting these things and payer guys know about this here. The whips of the error rate is so high that you have to save 5%, 6%, 8% in order for it to be statistically meaningful and not due to variation because it can swing in either direction just by statistics alone. But the bigger the patient population, the less likely that is to happen. So again, the payer mix and this is the other reason why we are really trying to get the claims data, is by trying to gather that in a warehouse and that makes sense from an analytics perspective. We need to have the power of numbers and have it be in one place where we can see over larger groups of patients.
So we’re having a lot of data-focused discussions with payers. There is not tremendous enthusiasm on payer’s part to give claims data. I think some of it they want to and there is just not the expertise to do it because there hasn’t been any meaningful managed care, you know, risk arrangements, I should say, in New York State since the mid-90’s, and even national firms that do it in other markets, there’s fairly, there’s not a lot of cross-pollination that I’ve seen of skill sets with that. We do need the claims to support risk contracting and like I said earlier in the discussion, the statistical modeling is very difficult with small numbers and we like to collaborate across plans to leverage that.
Living in Two Worlds
So, we live a little bit in two worlds currently where we are really trying to get risk arrangements built into contracts, at least at the shared savings level, but realizing whatever savings we create by practicing value-based care, eliminating tests, eliminating visits, our previous arrangements, that’s how we got paid, and we got paid for doing things and we don’t want to be in the business of being a colonoscopy salesman or a chemotherapy salesman but for providing, for being paid for health. Right now, the urgency is to get the contracts towards a value-based and in a value-based world and so some sort of a risk arrangement. Because right now whatever we say, it’s to the benefit of the payer. And we also have the world of we eliminate visits, we eliminate tests, we eliminate our volume. We’ve been lucky to be able to backfill that volume, which is great, but at some point we need to get paid for doing the services that we have been providing to get to this point.
Hospitals are interesting. They are in a difficult spot right now. At least in New York, there is overcapacity of hospital beds. In the west coast, something like 180 admissions per thousand Medicare patients. In New York, it’s about 400 or so. So about half of the number of admissions for Medicare anyway. So certainly some of those beds have Medicare patients and (35:44) would be willing to guess. And so, the better job that all of us should do to keep patients well and out of the hospitals actually threatens volume that they’re currently getting. So they need to do restructuring as well in addition to everybody else in the equation to try to align the kind of value with their reality of their infrastructure and staffing that they have as well as the reimbursement.
Clinical Engagement: Challenges & Successes
So as far as clinician engagement, we really try to use data wherever possible. The more data you give people that’s coherent, the more they trust you for other things, that things are going to be okay, and the more that they trust the systems, so it helps promote recruitment and retention. We also use it to the same skill sets as we bring on new practices to look at practice patterns of the practices or the doctors that were bringing in to decide if that knowing what we know using the analytical tools are, is it going to be an uphill battle to have them practice in this value-based model.
We do feel the more open we are, the more transparent we are, that the behavior that has been going on just because that’s how it’s been happening, how people practice, it gets better or it changes. So we share data about outliers, we share it with everybody, and people tend to, when they see their name, kind of out on the edge of a curve. The physicians tend to be a little competitive and it does help. Again, not all physicians focus on value. And interestingly, young physicians, when you’re young, and this, I found it to be and it’s not always, I just found it to be an interesting thing as I have discussions about new programs and doing new things. The people that are particularly receptive to new ideas of things are people that have like seen their mortality flash in front of their eyes. So people that have been doing poorly in the community, who have had trouble paying bills and making ends meet before they merge with us, they are very receptive to do things. And independent of their age, young doctors sometimes are trying to establish a way and a confidence and if you haven’t been in practice a long time, you kind of go with what you know, and sometimes there actually more there is to change in people that are chronologically older. So it’s not, I don’t think any of this has to do necessarily if you’re over 60 or it’s going to be a hard go to convince people about a new way of care. It’s very individual.
It’s All About the Data
Again, reiterating this, all about the data, getting the data. The limit is not just getting to it but putting it in a place where you can manipulate it and present it to physicians and others in the organization, the business, the frontline people, the nurses, even the phone team. So, we try to measure care wherever we can to provide people with meaningful feedback. Again, one of the big reasons we’re doing this project is to provide some degree of automation and scalability. We spend about 80% of our BI’s time of just gathering and making reports and less time analyzing data.
Crystal Run EDW Requirements
We looked at different ways and we wanted something that was going to be hitting the ground running, giving us material early, reportable data, early returns, not spending a year of creating a perfect data warehouse after which maybe you’ll start getting data out of it. We wanted a library of measures of analytical applications that we could apply as needed. We wanted a healthcare data model that was not a black box but something that we could see and that could change over time, and we wanted to be taught how to fish and not rely it on every change to have to run back to a vendor to make that change, and we wanted to have a long-term relationship.
Adaptive Data Architecture: Turning Clinician Questions Into Actionable Results
So with that, I’ll hand it back off.
Thank you, Dr. Spencer. I think it’s very interesting, your comments about how increasing your sufficiency, it reduced the number of visits per patients short-term but really increased your capacity and (41:21) physicians. And second, I thought your point about younger doctors resisting change more than older ones was kind of contrary with the experience that I’ve had but I think that’s a very universal observation and given the fact that you’re (41:33) age about 41 for your practice, that I’m seeing that there’s a lot of change going on if your physicians are being very collaborative with that.
Okay. So, I’m going to shift the focus of our discussion. What I’d like to do is to go back for a moment to the introductory comments I made about the shifting paradigms between payers and providers. And payers on the phone know more about your experience as an organization in terms of where your activities are shared with your providers and the providers likewise have your own perspective on where your activities overlap with your payers. But there are certainly some activities which line the domains of providers, some would line the domain of the payer and then there’s just overlap, where collaboration occurs.
Now, what we’ve seen and I am just glad to see the poll that we did that there’s a high degree of sharing of clinical claims data between the members of our audience. Dr. Spencer mentioned that in New York that’s probably with a different situation. But there is I think an increasing awareness about the benefits of payers and providers collaborating and sharing claims and clinical data to provide a more comprehensive longitudinal view from a claims perspective and a much richer data set from a clinical perspective. And we’ve certainly seen that with our partner clients in terms of those that have health plans as part of their organizations where they’re beginning to – they come up with the same organizational (43:03) for of sharing data that are now coming around to that point of view. And obviously the poll shows that this is a case that can be achieved in this discussion. And obviously the kinds of things that Crystal Run has been able to achieve is going to be augmented by that sharing of claims data of payers once we get to that point.
Crystal Run EDW Architecture
Now, my portion of the discussion is to talk a little bit about the data architecture or the adoptive data format and I would kind of do that with a little bit of resistance because I’m not a technical expert and I’m sure my technical colleagues at Health Catalyst are getting choked a lot of seeing me try to present this. But basically what I want to do is to kind of give the audience an overview about this Adoptive Data Architecture and what we believe is to be the most effective way to getting this integration of data, clinical and claims data, into the hands of the people who really affect change.
Direct binding data warehouse is really the hallmark of the Health Catalyst platform. It is a Late-Binding ™ in the sense that we don’t have an enterprise data model that we force all of the data into at the beginning of our relationship with our clients, but that we actually bind data as it’s needed to support specific business users and that leads us to be able to take data sources from EMRs, financial data, patient satisfaction data, etc. to very quickly map those to what we call source marts, in which we draw the data that we need for specific analytical users. We take that then to create specific population of the patients or just sort patients’ comorbidities, and then use that with the visualization layer to put in the hands of the users, which is what Dr. Spencer would mention about putting data in the hands of the users rather than having to pile them next to where you have to wait for somebody to run the reports for you.
We feel that the nature of healthcare data is much more complex than other industries. The volume of data is much more rapidly expanding in healthcare than other industries. The sense of urgencies for physician groups and payers and providers to get access to the data is greater in healthcare than it is in other industries. And that as Dr. Spencer mentioned, it’s higher for healthcare entities to be able to take ownership over managing their data, to being able to add additional data sources, to be able to expand their use of the data and not be dependent on a vendor to make those kinds of changes for them. And we feel that the Late-Binding ™ Data Architecture is unique and is able to meet those needs. It is a very flexible, adaptable approach to mapping data and it provides the type of flexibility that’s required to keep up with the complexity and rapid changes in healthcare data.
Catalyst Apps and Claims Data
So next slide, despite the fact that on our poll the majority of people indicated that they were sharing clinical claims data, we do find that there are situations where the user only have claims data available to him. And in those situations, at Health Catalyst, we do have a number of our applications that can utilize the claims-only data source and to use it to things like identifying where key processes are that require some focus in terms of quality improvement and yield the greatest results from a quality and cost perspective, to support the development of integrated dashboards to use with drill-down capabilities, to look at comorbidities in how they occur in different patient populations, to look at defining cohorts of patients which are important from a population health management standpoint, looking at readmission rates, we’ve been able to partner with hospitals in terms of trying to develop effective strategies for reducing readmission rates, and then again from population health’s perspective, to be able to define populations of like patients that are going to be used for analytical purposes. So we live in a world where there are situations where claims data is the sole data source and personally we have a number of applications that support that.
Catalyst Data Warehouse Advantages
The advantages of the Health Catalyst Data Warehouse, I think, Dr. Spencer reiterated that very nicely. The fact that it is not a beta model that is created upfront and requires adoption of the business needs to meet the data mile, but it’s the other way around, means that we could be very flexible and create data that supports the actual business and clinical needs of your clients. It has a very rapid development and deployment time which means that the sense of urgency to get access to data can be met.
We work with our clients to build an incremental structured approach to using data. That way, it gives them a chance to build their maturity of using the data in a step-wise fashion, it lowers the expense to the partner client over time, and it provides them as a roadmap that they can use then to look at how they’re going to grow and expand their analytic capabilities over the years it takes to get to the point of sophistication that they want to get to.
We do focus very much on transferring ownership of the data model and the use of the data and the data sources for our client. Actually we teach them how to do that. We feel that’s an important responsibility we have as a vendor, is to make our clients independent of us.
We align our use of the data with various organizational roles. Dr. Spencer mentioned a little bit about that in terms of its government model and I’ll help speak to that in just a second. But I think before we do, it would be helpful for us to kind of talk about the context of data and maturity using what we call the Healthcare Analytics Adoption Model.
Healthcare Analytic Adoption Model
The Healthcare Analytic Adoption Model was based on the HIMSS EMR Adoption Model and it was adopted by Dale Sanders, who is a senior vice president then from a CIO, and he is on the senior (49:13) Health Catalyst and is one of our proponents and experts in this technology. But the point of the analytics model is that it gives organizations a chance to kind of look at where they line up in terms of their maturity.
What I’d like you to do is to kind of take a look at this while I’m talking because we’re going to ask you a poll question to rate where you see your organization in terms of these 8 levels in just a second. But what we use this for is for two things basically – one is to help engage our potential clients in a very instructive dialogue about how they assess their own organizational readiness from an organizational-wide perspective and using that at the departmental level or at an individual level. And we find that it’s not kind of a structured approach where everybody falls into one level but that’s within one organization, there is (50:17) volume at different levels of this.
The second thing that we do is that we build our analytical applications to correspond with these various levels of the Analytics Adoption Model, and again that allows us to kind of customize our integration approach with our clients in consort with where they are in terms of their level of sophistication and maturity in the Adoption Model. So you could see that the Adoption Model goes from level 0, which are very fragmented point solutions, all the way up to very personalized medicine and prescriptive analytics. And we find in our experience that the majority of our clients and the majority of the healthcare industry kind of falls in somewhere between a level 2 and level 3 on this model.
So, if you’re ready for that, Joanne. If you could put the poll question up and ask the attendees if they could rate their organization relative to where they see it on the various levels of the Adoption Model.
Poll Question #3
On a scale of 1-8, how would you rate your organization in terms of level of maturity on the Analytic Adoption Model?
Okay. So, it looks like we have majority of the attendees at level 4, which is fabulous. We have lower level number of people who rate themselves at the level 1 and level 2. And we obviously see some organizations rate themselves at level 7. Very interesting.
I think in general the consensus in the industry is that payers probably have more sophistication with data analytics than providers do. I think the characteristics of payers are that their analytics are done by people who have analytical and financial backgrounds rather than clinical backgrounds. And I think we see probably the opposite case in the provider side of the world where providers tend to have analysts that have clinical backgrounds, not necessarily formal training in terms of analytics. And with the advent of more medical informatics programs and the kind of transfer of roles of clinical people into data analytic roles, I think these levels of maturity at each organization, they’re going to be able to achieve over time (52:51). So thank you for your participation in that poll. Very interesting.
Now, what I will do is to the next slide (53:04) lag times here. I apologize for that.
Organizational Structure Goals
Now, one of the things at Health Catalyst that we find to be extremely important in terms of successful analytic initiative is that it goes beyond having technical tools to having organizational-wide changes and a structure that really encourages and supports an adoption of analytics that’s being a permanent part of the culture in the long-term journey. And we do that in several ways. One is to encourage our clients to develop organizational structures similar to what Dr. Spencer shared with you in terms of Crystal Run, where a senior leadership really provides clear guidance in the use of data to make decisions and that then there is underneath the senior leadership team a variety of different structures that supports the content and the analytics availability of the data, guidance teams that looks for oversight at how different domains are being included in the process, clinical teams and ten individual work groups in those clinical teams that focus on various areas of opportunity. We feel that those groups may be created with what institutional (54:39) that senior leadership has to make a (54:41) and compelling case that this is the way that the organization is going to move forward to be able to manage the challenges that it faces and they have to be involved in actually using the data and helping clinical teams prioritize their areas of focus.
At the level below the overall organizational structure, we help our clients engage in developing teams that focus on particular areas of opportunities for improvement, and this particular example is for Women’s and Children’s Clinical Programs. And on these teams, we encourage the use of subject matter experts, which tend to be people with clinical experience, like nurses or PAs, people familiar with capturing the data, people who are familiar with how to basically visualize the data and to help users get the value out of the data that’s there, and then the analysts that actually can help support these teams in terms of doing the actual data analysis. And we put them into different groupings, knowledge managers who have kind of more of a data analytic experience with some subject matter expertise, data architects that have certainly a lot of experience in terms of data but also are familiar with data analysis, application administrators that are very familiar with the source data, and then these clinical people who are subject matter experts that kind of provide guidance to these overall teams. These have to be permanent teams, they have to have integrated clinical team members and again they have to be part (56:19) within the actual coach of the organization.
We also spend a lot of time with our clients in terms of helping them define these particular roles on these organizational teams and to help them identify how different personnel in the organization perhaps can be repositioned and to be taught different skill sets to be able to build on the current skill sets they have, whether it’s clinical or technical. So we work with our clients to help them identify these roles up down to again develop job descriptions that will include various skill sets they’re required and to make transitions in terms of helping these people move from current roles into their new roles and give them the skill sets that they need to be effective in those roles. And again, the very ongoing relationship that we have with our clients, each of our clients fingerprint this a little bit different in terms of how they create each roles and the structures and with some encouragement to do that. And as Dr. Spencer mentioned, the government structure at Crystal Run has been modified to support that organizational change in overall level.
Crystal Run EDW Teams
And this is an example of the enterprise-wide data warehouse teams that they created, where they defined them into two groups, the data acquisition teams with the programmers and the clinical subject matter experts and then the data architecture teams that have more technical experts and clinical subject matter experts on that side as well.
Catalyst Resource Deployment
Now, one of the things that we certainly carry the focus as a company and was mentioned by Dr. Spencer is the fact that we look at ourselves at being enablers from a technical and from a clinical perspective with our clients and we really foster independence over time with our clients and we purposely do that in terms of our level of engagement with our clients. And as you can see on this graph, that as the clients incorporate and utilize more of our applications, our level of engagement from a technical perspective (58:31) declines over time. We also see that from a client perspective that they take over more of the responsibility for basically working with our engagement executives in improving projects and (58:45) and give them the resources and tools that they need that they become more independent in our engagement levels from an engagement executive level standpoint decreased.
There is always another line level of support that we provide to our clients and certainly anytime the clients need more support from us, we’re more happy to step in from a technical standpoint or from a clinical perspective to do what we can to help them achieve the objectives that they want to accomplish.
The Analytic Organizational’s Journey
Now, I’ll start my closing with borrowing a slide from Dale Sanders again. I appreciate his working and his effort. But it kind of summarizes one of the expectations that we have to set there early on with our perspective partner client, which is the fact that the analytics journey is a long one, full of valleys and full of peaks, that there’s an initial euphoria that we see where people finally get access to data and they’re able to merge clinical data and claims data and financial data and patient satisfaction data. They’re able to run their own reports, they’re not dependent upon bottlenecks from the IT perspective. They’re able to get actionable data. They’re able to put it into practice and see the impacts significantly in how they care for patients and there’s a very high level of initial euphoria about – man, finally I’m able to use data but I notice that they’re all long but I haven’t been able to until this point to really (61:10).
And then over time we see some level of stagnation and we see that this initial layer of euphoria, (60:19) level of – wait a minute – now, we’re seeing that this part of the organization is really making advances but this other part of the organization needs physicians. They are not improving to the level that we want to. What’s the problem here? Is it bad data? Is it lack of clients? What’s the issues here? How to get ourselves out of this point where we get back to that level where we think we were before? And that with perseverance and commitment and again this permanent organizational commitment to doing process improvement of these data, we see over time this level of commitment paying off, where as they mature and they go up the Adoption Model, the results that they gain return back and (61:04) to the initial level they had using the data. And this is where we start to see success stories with our clients. We start to see a rejuvenation of interest and commitment to using data. We see that this really pays off in the long-term, ROI, for the organization, as well as improvements in patient care and how they manage populations of patients.
And so, that’s our message with our client – is that there are no quick fixes, that using data is a long-term commitment, that it can be a rough and rocky road at times, but that with perseverance and commitment, you can get to a point where you’ll start to see the long-term benefits and that it’s a journey that certainly will be worth the effort that the organization will see in terms of how they care for patients.
[END OF TRANSCRIPT]