A Health Catalyst Overview: Building a Data Warehousing and Analytics Strategy (Webinar)

Introducing Health Catalyst

June 11, 204

[Mike Doyle]

This is Mike Doyle, I’m the Vice President of Business Development for Health Catalyst.  I’ve been with the company for about a year and I have a great opportunity here to share with you a little bit at a high level about Health Catalyst and I’ll turn it over to Eric who will go into a little more detail.  But I just wanted to say thank you.  I’m very grateful that you guys can join us today.

Eric, go ahead to the next slide.

Our Story…

Our company started in 2008.  We’re founded in 2008 but our story really began back in 1980.  At that time, Dr. David Burton was an ED physician at Intermountain Healthcare and he sustained an injury that would prevent him from practicing medicine.  Around that same time, Intermountain was interested in starting an insurance plan.  They wanted it to be a little bit different and they wanted a physician to lead it.  So they asked Dr. Burton if he would be their first CEO of what later became Select Health.  In that role, Dr. Burton started looking at how he could help improve the outcomes and manage the costs for the members of that insurance plan and he realized pretty quickly that one of the roadblocks to that was going to be the variability in terms of the utilization and the costs to treat those patients.

So he hired Dr. Brent James, a physician with vast experience in clinical quality improvement, to help identify some of the opportunities for reducing that variation and both helping to improve outcomes and to lower costs.  As Dr. Burton and Dr. James started thinking about how to do that, they realized they were going to need data.  So they turned to Intermountain’s IT Department and specifically a man named Steve Barlow.  And Steve set about to take approaches for data warehousing, the integration of data from other systems within Intermountain and bring them together for the purposes of looking holistically across the care for their patients.  Steve looked to these approaches, outside of healthcare, these traditional data warehousing approaches, and found that through one or two different iterations of this, he really struggled to achieve the success that he wanted to, based on the complexity of healthcare data.

And around that time, Dale Sanders joined Intermountain as an IS leader and brought with him some knowledge of doing similar thing but for his data in the Air Force and through his work with the Department of Defense.  And he and Steve started looking at whether he takes some of those principles, those Late-Binding ™ principles, and apply them to healthcare.

And to make the long story short, they were very successful at doing that.  And with the addition of Tom Burton, they found successful ways to integrate with Dr. James’ interest in reaching out to those clinical improvement teams to provide them with the data that they needed.

So, a couple years down the road, there had been a lot of success at Intermountain, integrating the data that they have within their organization, providing it to these care improvement teams.  And Dr. James has founded his advanced training program to do that and help train other physicians and clinicians across the country, which many of you probably know all about and maybe have even attended.

One of those organizations like you then was Allina Health in Minneapolis, Minnesota.  And Allina attended the events training program, they heard all about some different strategies and approaches to helping identify wasteful variation.  They came back to Minneapolis to attempt and apply those there.  But they struggled because they really didn’t have the foundation for integrated data around quality, costs, outcomes, patient satisfaction, that Intermountain had benefit from for years through the work of Steve, Dale, and Tom.  So they asked Steve and Tom, “Would you guys mind helping us get off the ground?”  And that’s really how Health Catalyst was started.  Steve and Tom left their job at Intermountain, struck out on their own, and they didn’t really intend to start a company like Health Catalyst but they had a lot of success with Allina.  And Allina told others, and pretty soon we had a network of some of the best healthcare organizations in the country as our customers.  And from those humble beginnings with about two or three employees, we now have a company of about 170 employees and we conservatively estimate that about 30 million patients are impacted by the Health Catalyst platform in some way today.

So that’s a little bit about how we got started.

What does Health Catalyst offer?

Now, I’d like to talk a little bit at a high level about what we do.  At the core of what we offer is that Late-Binding ™ platform.  It’s been informed to Dale’s work through the best practices that Steve and Tom learned and it helps to integrate and organize your data.  We also stand that platform up really quickly, and we’ll talk a little bit about that.

Then we have three different types of analytic applications that help to integrate data from that platform and accelerate insight within your customer base.  Then two types of services, installation services and care improvement services. Installation services help to get you up and running quickly.  Care improvement services help to ignite teams within your organization and it’s informed significantly by the clinical improvement methodology that we’ve developed.  We’ll talk about some of the success we’ve had with that.

Next slide please…

High Level Timeline

So I mentioned that time to value is one of the things that we think differentiates Health Catalyst from other companies in this space.  And I just wanted to share a very high level timeline.  Eric is going to go into more detail about how we actually implement and how our products work in practice.  But this is a pretty representative timeline.  We get your platform installed in about three to four months and that includes training your IT staff and your analysts to bring data into the platform, as well as between three and five data sources, typically your electronic medical record, patient cost accounting system, patient satisfaction system.  And then the application starts going in.  And by the end of about three to six months, most of our customers have achieved what we call the achievement level 1, which is an initial statement of work intended to help set a really solid foundation for clinical quality improvement, it includes many different analytic applications and a really solid platform that has had success across the country within our customer base.

And then testing and validation is happening throughout the way to ensure that the data is of high integrity and that your clinicians, your administrators, your caregivers all trust the data that’s coming to them through these applications.

Next slide please…

Success Stories

As a result a result of that, our customers have been very successful.  We’re enormously fortunate to work with some organizations that have really embraced the utilization of data and analytics to help providers deliver high quality, lower cost care.  Most of our success stories incorporate both a clinical or an operational improvement component and a cost component.  So you’ll see for example MultiCare reduced sepsis mortality by about 22% in one year, with an associated $1.3 million cost savings.  And North Memorial reduced elective early-term deliveries by 75% and saw a modest $200,000 bonus payment from their payer.  Texas Children’s recently focused on labor cost and they didn’t want to reduce their workforce in order to help reduce their salaries and benefits expenses, and they saw a 2% reduction in doing that using our labor management explorer.  And that 2% sounds like a small number but most people know for organization the size of Texas Children’s, it’s probably a $10 to $15 million annual savings using that one application.

Success Stories (continued)

And if we can go to the next slide, our other areas of success, asthma outcomes, appendectomy outcomes, streamlining operations, and even helping to make your reporting teams more efficient at delivering data.  One success story had a documented average time to create a report and deliver it to an average user from about 97 hours to 30 hours.  We have more of these success stories out on our website.  We encourage you to visit www.HealthCatalyst.com and there’s a ‘Success Stories’ tab there.  And I’ll stop now after this sort of high level introduction and turn it over to Eric Just, a very good friend, a dear colleague, who will help you learn a little bit more about the way we do this and then we’re really looking forward to your questions afterwards.

Roadblocks to Success

[Eric Just]

Alright.  Thanks Mike.  So our goal is to really have one of those success stories that Mike just presented for every single one of our clients.  And one of the first things that we do with a client, even before we install our technology and doing assessment with them, we try to understand where they are in terms of their use of analytics and the ability to use those analytics for care improvement.  And we’ve identified some common roadblocks to success that we see with many clients in the industry today.

One of the first roadblocks to success is we’ve got a great team of analysts who are really skilled at looking at data and interpreting data but they spend so much time collecting data that they really cut into their time adding value by interpreting that data.  So we’re talking about Excel documents and Access databases where the analyst is using those tools to collect data when they could be analyzing data more effectively.

Another common roadblock we see is that we’ve got all kinds of reports and dashboards but we’re not using them to help improve care.  We’re not really looking at relevant data.

And lastly, we seem to see a problem of we can improve care, we got a team together and we have a great improvement but it’s not sustainable.  As soon as we move to another project, the previous project suffers.  So we have trouble sustaining this improvement.

Poll Question

Which common roadblock resonates the most with you?  98 respondents

Before we move on, I want to ask a quick poll question.  So which of these common roadblocks resonates the most with you?  Analysts spend too much time gathering data, reports and dashboards are not showing relevant data, or there’s difficulty achieving sustained improvements.

[Tyler Morgan]

Alright.  We’ve got that poll question up.  We’d also like to remind you while you’re answering this poll question that you can type in your questions to the presenters in the questions pane of your control panel.  We’ll go ahead and close this poll in just a few seconds.

Alright.  We’re closing the poll now and here are the results.

Poll Results

[Eric Just]

Oh thank you, Tyler.  Okay.  So it looks like most people are seeing their analysts spend too much time gathering data.

Three Critical Elements of Success

Alright.  So, in implementing our success stories with our clients, we have identified really three critical elements of success that help us to get around these roadblocks.

Three Critical Elements of Success – Analytics

The first roadblock of analysts spending too much time collecting data is solved by having a solid analytics framework, and an analytics framework really creates the single source of truth for the organization and makes it really easy for analysts who need data to know where to go and know how to find that data.  It also provides broad distribution, automated broad distribution of information so that you don’t always have to go to an analyst for information.  So there’s a self-service component here as well, and through that we save the analyst time from having to answer more routine questions.  And that allows enough time for the analysts to do what they do best and discover patterns in data and decide how we are going to improve based on this data.

Three Critical Elements of Success – Content

To get up the issue of we’re not looking at relevant data, there’s the notion of content.  It’s not just about the analytics.  Content is how we define clinically driven patient populations.  Are we using the latest evidence-based medicine to identify wastes?  And how are we identifying high and rising-risk patients.  All of the algorithms, all of the criteria used to define cohorts and the measures that we’re looking at using evidence-based medicine comprise the content piece.  And quality content, when it’s combined with analytics, results in reports and dashboards that are showing relevant data and they’re making out to a broader audience

Three Critical Elements of Success – Deployment

And finally for the last issue of not being able to sustain improvement, we have the concept of deployment.  And deployment is about how do we organize around these analytics, how do we create workgroups that are focused on defining content that’s accurate for our institution and creating actionable data, and how do we skill these workers so that we can organize for scalable improvements across a variety of clinical and operational areas.  That’s really what the deployment system is concerned with.  And in general, going back to my platform application and services slide, analytics is really part of the platform.  The content is displayed in the applications and our services arm really helps with the deployment aspect of this.  But these are really the three critical elements of success – Analytics, Content and Deployment working together.


So for the rest of the presentation, I just wanted to walk through what I’m going to be showing.  The first will be a discussion of the platform.  Then we’ll do a demo of one of our applications, called the Key Process Analysis.  We’ll do another demo that was used to create one of those success stories in reducing heart failure readmissions.  We’ll look at a host of other applications, very high level.  And finally we’ll end with a conclusion.  And throughout the agenda, we’ll be tying our analytics content and deployment system concepts into what we look at.

Late-Binding ™ Data Warehouse Platform

Fast-tracking Analytics and Content

So, the first discussion about our platform is really about fast-tracking analytics and content.  And what we do with the data warehouse platform is we create that analytics backbone to make it really easy for an analyst who needs data to know where to go and have ways for them to get that data.  And we do that using our Late-Binding ™ Data Warehouse Architecture.  The goal of the architecture, and again with our engagements with clients in general, is to really get a rapid time to value.  And to do that, one of the first things we do is we identify the key data sources to load into our data warehouse platform.  And when we do this, when we load the data from these systems, they are indicated here, these (14:55) at the bottom represent the operational data collection systems, things like the medical record system, financial system, human resources claims, all sorts of data sources that you can think of.  When we map those data sources into our platform, we’re not making a lot of business decisions about how that data is going to be used.  We’re getting that data in to our platform with as little transformation as quickly as possible and that allows us to be very quick and have more flexibility later on in our process when we really do need to make those business decisions.  And because we’re not doing large scale transformation, it makes it very easy for us to create tools to automate this process.

So we have a tool called Source Mart Designer which accelerates the growth of this analytics repository by creating an automated way to reaching these source systems, pull out information about what data elements are in those systems, and help us to map those data elements into our source mart area of the data warehouse and you can think of a source mart is really a corner of the data warehouse that holds data from a particular data source.  A Source Mart Designer captures all that we call that metadata, those mappings, it captures all that metadata so that the feed these systems and pulling that data every night could be created very quickly and with very little maintenance.

We have another tool on the right-hand side, and the tools are indicated in purple because they’re part of our analytics system, where they facilitate the analytics system, is Atlas.  And Atlas is a web-based tool that helps us to look into all of that metadata that we collected with Source Mart Designer, and it’s a key tool in helping analyst to locate data within the data warehouse.  So I can type in a very simple query in Atlas for phone number and I could find all the tables and columns that deal with that particular data element.  We likened it to Google and Wikipedia because we can also edit descriptive information about the fields in the data warehouse.  But that’s a critical element in making this data easy to access.

The next layer in our platform is about linking, standardization, and content.  So by linking, we mean how do we take data from the EMR and link it to financial data.  Well there’s a series of identifiers that are available to us to create those links.  Those are called common linkable identifiers.  We also have standardized data structures for things like patients, labs, encounters diagnoses, and medications.  This (17:24) confused with a larger scale enterprise data model but it’s very focused on these key elements that we see time and time again and we know are going to have a lot of value in creating additional data structures based on these concepts.

And we also have placeholders and provide content at this level.  So, one of the things that we bring with the Health Catalyst platform is definitions for over 800 populations out of the box.  Now, many of our clients want to get their hands into fine-tuning these definitions and that’s allowable by our platform, but starting with the starter set really accelerates the ability to look across a variety of disease conditions, and most organizations don’t have time to define 800 definitions themselves.  So this really accelerates that content.

We also have the concept of hierarchies which are used for stratification in our application and we’ll look at one of those today, as well as models for comorbidities, risk stratification, and patient and provider attribution.

The set of tools that come with the platform to help us manage this is, one, the Subject Area Mart Designer.  This is a tool that’s used to create and manage the content.  So if we want to change the definition of one of our populations or add a definition, we will use our tool called Subject Area Mart Designer.  And that tool, the data that’s created by that tool, is also viewable in Atlas, so that our users of the data warehouse can view content definitions and a lineage in Atlas along with the other information that we talked about from the Source Mart layer.

Building on top of this layer and using the same tool set, we have the concept of Subject Area Data Marts.  And Subject Area Data Marts are really focused on specific areas of care.  So one might be heart failure, it might be a clinical area, it might be an operational area, like financial management, or a dashboard that looks at chronic care like our community care dashboard.  But the point is when we’re creating a Subject Area Mart, here is what we’re building in those more complex business definitions and we’re engaging clinical and operational people to help us to build these data structures and validate them as we build them.  This is where we’re starting to quote “bind” information and that binding is best on at this later stage when we have engagement from the people who will really be using this to improve care.

And then once those data marts are built actually in parallel to those data marts being built, we lay on visualizations.  And the visualizations represent the applications that are used by the deployment teams to actually improve care.  They are the main vehicle for giving this data out to non-technical users.

Where Do We Start?

Key Process Analysis

So as we’re building the analytic repository and deciding where we want to create our first success story, this is where our key process analysis comes into place.  And a key process analysis is an application that sits in our platform and is used to help prioritize areas where we think we can have the biggest effect on improvement.  I’ve got a couple of background slides that I’ll show on the key process analysis and then we’ll jump out to a demonstration of the tool.

Clinical Hierarchy

Organize codes into a clinically meaningful hierarchy

One of the things a want to point out that you’ll see in the tool is this critical piece of content called the Clinical Hierarchy.  And the Clinical Hierarchy was actually developed by our clinical leadership led by Dr. Burton and it serves to organize codes into clinically meaningful hierarchies that really align with the way that care is delivered.  So a lot of hierarchies that we see at our clients are based on more administrative or financial functions.  This is really getting our Clinical Hierarchy that fades around the way that care is delivered and you’ll see how it appears in the tool but the important concepts.  So we map a variety of codes, tens of thousands of codes, into groups that we call care processes.  And there’s about 455 distinct care processes.  We roll those care processes up into care process families and there’s about 92 care process families.  And those 92 care process families roll out into about 12 clinical programs.

KPA:  Measuring Opportunity

Another concept I want to introduce before the demo is how we use the KPA to measure opportunity.  And as Mike discussed in the very first slide about our beginnings at Intermountain Healthcare, one of the roadblocks was understanding variation and the KPA is really designed to understand that variation and this slide is meant to illustrate why we look for variation, how we use it to measure opportunity.

So if you imagine that this blue dot in the middle of the screen represents an individual provider who is performing vascular procedures and this provider does 15 cases per year and does those cases at an average cost of $15,000 per case.  Meanwhile, it appears as institutions are performing the same procedures for about $10,000 on average per case.  What if we could take what Dr. J is doing and standardize his care closer to what his peers are doing?  In this case, he is providing care at about $5000 above the mean multiplied by 15 cases.

If we could take what he’s doing and move him to the mean, that’s about a $75,000 cost savings opportunity.

And then if we layer on all the other physicians in his organization and see there may be another one who’s doing 25 cases a year at $4000, that equates to a $100,000 opportunity, and we keep adding these numbers up, we get a total opportunity for the ability to take all of the providers who are performing care above the mean cost and moving them to the mean.  Now, this is not an ROI calculator but it’s a very good relative measure of how much opportunity there is in different clinical areas for reducing this variation and this forms the basis of the key process analysis tool.

Demo:  Key Process Analysis

So I’m going to jump over to a demo here.

KPA Pareto Analysis

And the first thing we’ll see in the key process analysis is this Pareto Analysis.  And in the Pareto Analysis we’re looking at data that is being stratified by the clinical hierarchy that I introduced in the slides.  So you can see the clinical programs, the care process family and the care processes listed on the left-hand side here.  The grain of this middle chart is the care process family level.  So in this chart we’re looking at things like heart failure, pregnancy, ischemic heart disease.  That’s the grain of the care process family.  And we’re looking at variable direct cost by default and we have variable direct cost stratified by care process family so that the most expensive in terms of variable direct cost is heart failure and at $9.9 million that’s about 7.5% of all of the variable direct costs of all of the care process families added together.  And pregnancy is next, followed by ischemic heart disease, and so on, all the way down our list of care process families.

And these blue dots indicate, each blue dot indicates a care process family and the position on the Y axis is the percentage of the total variable direct cost that that care process family made in the whole.  The red dots indicate the cumulative running total of the blue dots and this is the pattern that we see at every single one of our clients where there’s about 10 care process families that account for over 50% of the variable direct cost of the institution.  So for prioritizing where we think we can have the biggest effect, just based on this information alone, this would lead us to believe that there’s about 10 care process families that should probably be high on our list.

We’re looking at variable direct cost by default.  We can look at other metrics listed here, length of stay, same analysis based on length of stay, based on charges, or case count.

And then these metrics over here on the right, labeled opportunity, equate to the slide that I showed you where the bubbles were moving from above the mean to the mean.  This is how we calculate where that biggest opportunity is.  And according to this analysis, heart failure shows up with the most opportunity based on that provider opportunity analysis that we reviewed in the slide.

KPA Bubble Chart

So the other thing that I mentioned that we look at in this KPA tool is variability.  And in this chart, we’re looking at variability and we’ve moved one level down in the hierarchy.  So we see a lot more data points on this chart because we’re looking at care processes now, not care process families. The position on the X axis is the total variable direct cost for that particular care process.  The bubble size is the case count for that particular care process.  And the position on the Y axis now is the measure of variation.  So we do a calculation of a coefficient of variation to allow us to look at the most variable processes.  So we really not focus our attention on the upper right-hand quadrant of this chart.  These are the highly variable, most expensive processes and these are the areas where we think there’s great opportunity for further investigation and we can mouse over any one of these and get information about which care process is behind it.

So if I mouse over this red bubble, I see that that is actually heart failure, which showed up the highest in our opportunity analysis.

So let’s click on the heart failure bubble and learn a little bit more about the variability behind heart failure.  This chart really is very similar to the chart that I showed you in the slide.  Each bubble here now represents an individual provider.  The bubble size is that provider’s case count for heart failure.  The position on this X axis is that provider’s average variable direct cost for heart failure.  We’ve added a Y axis here and that Y axis is the severity scale.  So the sickest patients are up at the top of the chart here and as we move down, and we expect to see this large variation for the sickest patients.  There’s a lot of complicating conditions.  But as we move down the chart, we should really start to see these bubbles start to stack up on top of each other.

And that’s not necessarily what we see here.  We still see significant variation even at these lower levels of severity.  So this is how the tool really helps to focus in on where are the biggest areas of opportunity for us.  Certainly in the demonstration it appears that heart failure is a very interesting opportunity for focusing on care improvement.  Now, we know that there’s a lot of subjective criteria as well that’s just bringing the voice of the data to that conversation but there’s a lot of subjective criteria that have to be applied.  So what we usually do with the KPA is we narrow it down through about 4 or 5 focus areas and then we apply some organizational subjective criteria, such as quality of leadership in a particular area or variability to take on a new project and we combine the subjective with this, with the data here, with the KPA.

Key Process Analysis

So just in summary, the KPA is a tool as part of our analytic solution that uses the data warehouse and relevant content to determine the greatest opportunity for quality improvement and cost reduction.  Key content pieces are the clinical hierarchy that we use to stratify according to classifications that match care delivery, again in contrast to classifications that might match more administrative or financial groupings.  And then the calculations that we use to identify the variability are also a key part of the content.  And this tool is used by deployment teams to prioritize improvement efforts.

Heart Failure Readmissions


So we’re going to jump to another demonstration here around heart failure readmissions.  Heart failure readmission is an example of what we call an advanced application.  And advanced applications are applications that are typically used for the actual improvement of care.  So the KPA was giving us kind of a guide post to helping us create the roadmap.  Now, an advanced application would be deployed to help us understand where we actually can improve those processes and how do we actually achieve that improvement

When we implement an advanced application, part of our deployment recommendations are that implementation workgroup comprised of both clinical and technical people is assembled and that implementation workgroup helps define the improvement Aim statement.  So what are we trying to accomplish with this data mart?  In this case, we’re trying to reduce heart failure readmissions.  How do we define our patient population, and you know, there’s a population definition that comes out of the box with Health Catalyst, but through this process, we will refine that and make it more clinically accurate.  And how do we identify what interventions we’re going to perform as a group to support the Aim statement of reducing readmissions.  And all three of these elements, the Aim statement, the population definition, and the specific interventions to support that Aim, are part of the content piece of this implementation.  And our analytic tool is a data mart, underlined data mart, and a visualization on top that allows the visualization of these particular metrics and our ability to stratify by risk.  So that’s just kind of an introduction to the tool.  Now, we’ll go and look at the tool.

Heart Failure Readmissions

So this is the front page of our heart failure readmissions dashboard and the deployment workgroup that was set up around this decided that they really want, of course they wanted to get these high level overviews of their heart failure readmission numbers, so the 30-day and 90-day readmissions, but they also wanted to balance those readmission metrics with utilization metrics for ER and observation space.  Oftentimes if we push too hard and really try reductions in those readmissions that we see an increase in ER utilization and observation space.  So this is really – these two metrics on the right serve to balance our 30-days and 90-day readmission numbers.

What we see in the middle of the screen here now are these specific interventions that have been proven through evidence-based medicine to reduce heart failure readmissions.  What metrics show up here is the key content decision by that implementation team.  Medication reconciliation, do patients, are they having their medication reconciled both by admit and discharge, are our patients receiving a follow-up phone call after their discharge, when they’re discharged from acute care hospital, do they have an appointment for their primary care provider, and finally, do we have all interventions done.  So have all three of these been done for a particular client.  So we have this high level dashboard view on the front screen.

Heart Failure Readmissions

We also provide the ability to stratify by a number of different filters and those are listed on the left here.  One of the most important ones is the population filter.  So I mentioned that our deployment team developed a clinically validated cohort definition and that was based on iteration and with the analytic team and coming up with criteria, that very solid clinical definitions and the heart failure patient, and we can look at this definition alone and we can also look at a specific population that was defined using CMS core measure criteria.  So it’s important to note here that we provide a system to capture the content, these cohort definitions.  We also provide a way to toggle between two different pieces of content that apply here.

We can also mix and match individual rules here, some of them are based on ICD9 codes and some of them are based on medications to identify specific patient population on the fly.

Heart Failure Readmissions

The other part of this application that’s very flexible is how we identify high risk patients.  These risk filters on the left are various ways to identify patients who might be at risk.  The Physician Flag comes from the EMR and some of our clients have a feel to their EMR that allows them to flag a patient that’s high risk.  The Charleson Index is based on a comorbidity analysis.  And the Health Catalyst Heart Failure Risk Index is based on a predictive tool that we’ve developed that presents the percentage likelihood that a patient will be readmitted.  So if we want to find high-risk patients, we see them on the upper end of this Health Catalyst Heart Failure Risk Index.  So there’s various ways for us to slice and dice the data, so we can look at admit, we can filter by location within the institution as well.

Heart Failure Follow-up Phone Call

Now, in terms of providing actionable data, this is really high-level overview data, we want to know who are the patients that we should intervene with, and I’ll just show one example of a drilldown tab here that looks at our metric if the patient received a followup phone call.  Remember, this is a key piece of content that was defined in evidence-based medicine that says patients who receive a follow-up phone call typically do much better in terms of readmissions.

So what this chart at the top shows here, these gray bars, indicate the number of patients who are discharged in a particular timeframe and the colored dots here indicate various metrics.  So the blue dots indicate how many patients or what percentage of those patients were called.  The yellow bar indicates how many of those patients were called within a certain amount of time and the green dots indicate whether the patient was actually reached or not.  So we really want to understand which are the patients who have not been reached and we can use the filter on the left here that will allow us to zoom in on the number of days since the patient is discharged.  So we find patients who are discharged within the last two weeks or 14 days.

Heart Failure Follow-up Phone Call

What this table here shows is a list of the patients who were not reached and it has information about the medical record number, their name, their age, their unit, as well as their phone number, and the list is ordered by that risk indicator that we talked about before.  So the highest risk patients come off at the top here.  So if we’re really interested in using analytics to provide actionable data, we’re looking at the highest risk patients at the top.  So if we start at the top of the list and move down, we’re going to have this effect.  And I should point out of course that all of the data is completely scrubbed and de-identified.  These are make-believe names and phone numbers but what you can see here is how we incorporate all of this content to really get us to the point where we can take action based on the data and this is a tool that helps at least one of our clients achieve reduction in the heart failure readmissions.

Heart Failure Readmissions


So just in summary, the analytic tools provide a baseline readmission metrics, also showing balance metrics, and provide that drilldown to the patient level reports that help us to decide how we’re going to intervene and we have it ranked by the predictive risk score.

The content system allows us to have multiple cohort definitions and multiple risk stratification models and toggle between various versions of those, as well as making sure that we are using the latest in evidence-based medicine to focus on those interventions that we know are going to improve the outcomes.

And finally the deployment is really handled by these care improvement teams, these workgroup teams, that use agile improvement methodologies to incorporate this content and use this application to identify and take action on those patients.

Product Portfolio

So in our demos today, we saw a very small portion of the Health Catalyst Portfolio.  We wanted to give you a taste of what we can do and how it relates to the three critical elements of success – Analytics, Content and Deployment.  So I put a star next to the tools that we looked at.  We looked at the Pareto tool and we also looked at one example of what we call a Population Suite, and I mentioned that that falls into a category that we call advanced applications where we’re looking at metrics that specifically deal with a clinical area or a workflow area or a patient injury prevention and these tools really (39:27) the tools that provide actionable information for us to improve care delivery.  We have lots and lots of these advanced applications.  We just looked at a couple of examples on the screen.

On the left-hand side, we have our foundational and discovery applications and foundational and discovery applications are meant to provide more basic data to broader audiences and we help our clients get set up very quickly with these so that basic questions like “what is my average length of stay for heart failure patients, how many diabetic patients do I have who haven’t been in for primary care in the recent months”.  These are all sorts of the questions that should be answered very easily and that are answered through our foundational and discovery applications.  So with these, we’re really helping to reduce the burden on the analyst time and having the analyst not have to feel all of these more basic questions.  So that’s what we see in the foundational and discovery space.  The Pareto tool or the KPA that we looked at is one of our discovery applications.

Population Explorer

One of our foundational applications is Population Explorer.  I’m just going to do a couple of screenshots here to give you a taste of some other tools that we have.  And there are demos of these tools on our website.  So if you see a tool here that you’re interested in, go to that website and click on demos and you will see a recorded demo of these tools.

So Population Explorer is really the window into those 800 pre-defined populations that we talked about that come with the platform.  It allows us to look at metrics across the continuum of care.  So, on this front screen that you see on the slide here, we’re looking at readmission rates over time and trending, we’re looking at cost data over time, as well as length of stay data.  Other tabs show demographic data, risk profiles, and information about when the last time these patients have been into primary care.  It uses a way to quickly get high level information across any of those pre-defined populations that come with the platform.

Community Care

Screening and chronic disease management

Community care is a dashboard that’s based on screening and chronic disease management and that falls into our advanced application category.  And what you see on the screen here is that we’re looking at several different populations, we’re looking at A1 – for diabetics in the middle here you see we’re looking at A1c screening and LDL screening.  We have a variety of preventative metrics that we’re looking at around immunizations and screening.  And on this screen, we’re looking at how the system is doing with respect to these various metrics as a whole.  The tool also allows to drill into the organizational structure, so we can look at metrics work and compare across the different clinics within our system and get laid down to the provider level.  We also have a patient centric view in the community care dashboard that allows us to look at how a particular patient is doing with respect to these measures, and this is the tool that’s highly customizable.  So the metrics you see here are really examples of what’s possible, what we’ve done with one of our clients, and it really provides a framework for that organizational drill through on these metrics that particularly pertains to community care.

Financial Management Explorer

Analyze costing, billing, and payment information alongside financial and volume metrics

And additionally we have tools that are not clinical in nature.  So we have a tool called Financial Management Explorer and it’s based largely on financial data, looking at costing, billing and payment information, looking at financial and volume metrics.  And one of the things you see on the left-hand side here is that clinical hierarchy.  So this tool allows us to slice these financial metrics by our clinical hierarchy, that key piece of content that we saw in the key process analysis.  So there’s a common threat here I think you can see.


In terms of implementation, our implementation is really based on the realization that we can provide value if we focus on specific areas and get you up and running without tackling the entire organization first.  So it’s an agile approach.  Our first achievement level is really designed to get our clients up and running with the platform, with our metadata tool, with three very important source marts, and five foundational apps and a handful of discovery applications as well.  This achievement level 1 is designed for us to get up and running very quickly and start to push this data out to your organization.

Achievement level 2 is really based on rounding out the source marts that flow into the data warehouse and getting some additional foundational and discovery applications but now in achievement level 2 is where the deployment team starts to take shape and we start to tackle some advanced modules around populations and workflow.  And typically achievement level 2 is where we identify those first success stories, like Mike presented on the first slide.

Then achievement levels 3 and 4 are designed to expand that relationship and bring in more applications and help to scale those improvements across the organization and across clinical areas.

…Back to the Health Catalyst Story

So going back to the Health Catalyst story that Mike opened with, our approach to success and these three critical elements, it starts with our creation story and that creation story predates our company.  Our company was founded on understanding what it takes to be successful in analytics and building them to the DNA of our company.  When we first started out, we had the concept of this three-legged stool.  This was in some of our very early slide presentations where we’re presenting these three critical elements to success.  We did decide that the stool was decidedly low-tech for a technology company and moved to a slightly more abstract version of the then diagram here.  But the message is the same and I wanted to show this just to show you that this is part of who we are as a company  We understand what it takes to be successful.  It’s not just technology.  Technology is a huge part of it and that’s really a very large portion of what we bring, but we also bring the concept of content and having valid content and the deployment systems to really ensure that our customers are successful.

Our goal is to make sure that all of our customers have ‘success stories’ and this is the model that we used to help ensure that success.

Poll Question

Which do you think your organization needs the most help with?

So before we conclude here, we wanted to ask a quick poll question.  Which do you think your organization needs the most help with?  The analytics, content, or deployment?

[Tyler Morgan]

Alright.  We have that poll question up.  And while you guys are answering that poll question, it looks like we’ve got a lot of great questions that are coming in.  I would like to remind you, you still have the opportunity to ask questions by typing your questions into the questions pane.  I would like to address that we have had several questions about if the slide deck will be available afterwards.  This webinar is being recorded and we will provide everyone with links to the recorded webinar, as well as the presentation slides and the like.

So we’re going to go ahead and close the poll now and let’s share the results.

Poll Results

[Eric Just]

Okay.  So most of our participants feel that their organization needs the most help with the analytic system.  Great.  And of course there’s a good balance between the content and the deployment system as well.  Thank you, Tyler.

Thank you

Upcoming Educational Opportunities

So at this point, I’d like to thank the audience.  We really appreciate everyone’s attendance and attention here.  I just want to highlight some upcoming educational opportunities through the Health Catalyst webinar series.

Healthcare Analytics Summit

Transforming Healthcare Through Analytics

[Tyler Morgan]

Thank you, Erick.  One thing we would like to highlight as well is our summit that’s coming up.  And as a matter of fact, like I mentioned at the beginning in the introduction of the webinar, we have two passes to give away for the Healthcare Analytic Summit we’ll be holding on September 24th and 25th.  The first is a pass for single registration.  The second is a pass for a team of three.  This drawing is very very simple.  But before we do this, I would like to mention that the registrations for the summit so far has exceed our most optimistic expectations, coming in at over three times our best case scenario, so much so that we’re now working to find extra space because at this rate we would run out of passes by the end of July or early August.  That’s two months ahead of time.  We still have these free passes to offer you but we simply ask that if you enter the contest, that you are confident that you could travel to Salt Lake City on those dates to maximize the chance that the winner can come.

Poll Results

And it looks like that about 64% of you are interested in being able to attend with your teams.


So let’s go ahead and jump in to the questions and answers here.  Now, we’ve got a lot of great questions here.

What is the frequency of moving the data from the sources into the source mart?  Is it real-time, daily, etc.? That’s a great question.  We typically start with a nightly feed.  So once every evening, that’s really the default configuration for our ETL tools.  In certain circumstances of frequency if higher, then every night is required and our tools handle that.  We can set a smaller interval.  And in certain cases we ask for a longer interval.  So in certain cases that data is just not and the source system isn’t refreshed every night and we can go longer as well.  So typically at the day we can go more frequent or less frequent depending on the business requirements.
Please define the term Late-Binding ™. Late-Binding ™ applies to, when we’re talking about the platform, we have the layers of the platform, many data warehouse methodologies attempt to bind business rules and definitions to data as it’s brought in to the data warehouse.  So before we even do anything, we’re already doing a heavy transformation in creating a business rule as part of the scripts that we create to load that data into the data warehouse.  That’s an example of early binding.  That leads to really long time to value in our experience.And Late-Binding ™ is really when I show those data marts at the top of the stack, that’s where we’re building in those definitions and creating those bindings, as we call them, to the business rules.  And the later we do that binding, the more flexibility we have at our source mart layer as we’re not building those definitions through that source mart layer and the more flexible we are if one of those definitions changes.  And that’s certainly our experience in implementing our advanced analytics, is that there’s often changing definitions or tweaks that we want to make and by doing it later in the process, it leaves us that ability.
What point do you integrate master data management? That’s a great question and our master data management strategy is really very institution-specific.  So some organizations are already doing work in managing their master data and they have tools and data sources that we can bring in to the data warehouse as master data sources that we can map to those sources.  Other institutions, we have to do a little bit more work in our platform to get at the master data and create the linkages that we need to create in order to use that master data.  But it’s variable and it really does depend on the institution.  Our strategy is – we finally have got as many implementations to that as we have clients just because different clients are doing master data at different levels.
Do users have access to the source marts or is it restricted for IT developers?  I’m trying to understand if source marts are similar to staging data as one does with mainstream data warehousing methodologies. That’s a great question.  So yes, users have access to the source mart and that’s why we make all that information available in Atlas.  We make it usable by having it searchable and available on Atlas and really there’s a lot of data in those source marts.  We’re not pulling in a focused data set.  We’re really going broad and we want people to access data directly from those source marts when there’s a data element that may not be built in the data mart yet.  So we absolutely encourage access beyond just developers to that layer.[Mike Doyle]Would you mind if I just add to that real quickly.  So part of my background, guys, is I was the data warehouse manager out at Allina Health for about four years and we had a lot of experience with folks getting in and querying the source marts directly.  It also tied into data governance.  And probably one of the things users after seeing, you guys after seeing this presentation, may want to also look at some of the content on our website around data governance, but we use the concept of a data steward or a person to help grant access to those source marts appropriately.  So just in case anybody thinks that it’s a broad open access to all the data in the organization, we have a really great approach to work with you to help you to find the right people to provide access to and the right processes to integrate with your own access granting flow and workflow within your organization to make that happen.  I just want to add that part to that as well.
How is this product licensed? So I could probably take that.  This is Mike.  We have two different high level approaches to licensing.  There’s a standard perpetual license model where you have a license to the applications, the platform, and all of the technology that goes into it, as well as support and maintenance that helps to ensure that you’re able to keep current with our software and professional services that we talked about at the beginning of the care improvement services and the installation services that help you to get that installed and in the later achievement levels to support those clinical improvement teams.But that can also be offered as a subscription model which is basically just a way to take the cost of the perpetual license model and break it up into monthly chunks.
I see the value of the KPA, etc., for cost containment.  How are patient outcomes brought into the system for analysis? That’s a great question.  The KPA is really a tool that – as you mentioned, a lot of it is based on cost data and that’s a good measure, but the variability is oftentimes tied to quality outcomes.  So the more variable a process is, the more we notice that outcomes are not optimal and that quality is lower.  So by reducing variation, there’s almost an automatic result of improving outcomes because we’re standardizing on the best evidence-based care.  So reducing that variation is really how the KPA is measuring outcomes.  Bad outcomes are tied to higher levels of variation.
How do you obtain the clinical data for the interventions? So in terms of the clinical data that we’re looking at, most of that information typically comes from the EMR and we’re using documentation in the medical record to determine whether particular interventions are required or have been done.  So the EMR is the main source of that data.  There are some other instances where we’re pulling from other systems but the EMR is definitely the main source.

How interested are you in a demonstration of Health Catalyst’s solutions?

[Tyler Morgan]

And we just like to say, on behalf of Eric Just and Mike Doyle, as well as all the folks at Health Catalyst, thank you so much for joining us today.  This webinar is now concluded.


Three Systems of Care Delivery

Population Explorer Summary

Community Care Summary