Analysts Surf the Tsunami of Healthcare Data
John Wadsworth: It may be winter outside but given the title of this webinar, I almost feel like I should start by saying aloha.
Sarah Stokes: Aloha.
John Wadsworth: These webinars are great, aren’t they? The Catalyst webinar venue is such a great way to build comradery and to share information. Hopefully, if I do my job over the next hour, we’re going to learn together and do that in a fun way as we explore a really important topic for healthcare analytics. Our time will be spent together as follows. We’ll walk through some principles that I have observed and learned from the dynamic world of surfing. Then, we’ll explore the utility of those principles overlaid in the world of healthcare analytics. Last, we’ll end with a review of key lessons learned.
The extraordinary financial investment in healthcare technologies has been done with the predominant belief that technology can help improve care delivery and simultaneously lower cost. While that’s true, what’s probably not understood well enough is that healthcare systems will never fully realize the potential of those technical investments if there are not equally skilled operators that can leverage those procured technologies. For our discussion today, I’m going to refer generically to those technically skilled operators as healthcare analysts. In that term, I would be including data architects, data engineers and data scientist. With all the talk around technology in the healthcare industry, I’m speaking today really because of you in the audience, you attendees.
You actually chose the topic. Feedback from our marketing team consistently has rated this discussion as a high priority. My intent today is to help nudge the industry a bit towards the human aspect of the analyst and we’ll reframe the role of technology. Then, we’ll get tactical. I’ll share with you some important areas of domain knowledge that are essential to adding value to the organizations that you serve. On a warm morning in about a year ago, December, I joined a group of friends for my first ever surf lesson.
After some beach side instruction, our surf coach put us in the water. We padded out to meet the waves. I was thrilled because surfing was something I had always wanted to try. Honestly, Sarah, this picture doesn’t quite do it justice. In my mind, I remember it more like this.
Sarah Stokes: A little more intense there.
John Wadsworth: It was. Two days after the surf lesson, my wife and I traveled to the north shore of Oahu. It just so happen that the pipeline masters were kicking off the very next day. This was to be he culminating event of 2017 for the best surfers in the world. The competition feature perhaps, the most famous surf spot in the world known as the pipeline, aptly named for its quick rising waves and the perfectly formed curls along a dangerously shallow sea floor. With the competition a day out, the beach for my wife and I was relatively unpopulated and afforded us an unobstructed view up close and almost personal with these world class athletes as they practiced getting used to the pipeline.
For me, there were a lot of aha moments that day. As I watch with binoculars in hand, it became clear, I had no clue how to pick a good wave. When I though I saw an awesome wave rising behind the set, I’d comment, “Oh, oh, this is a great wave.” Then, none of the surfers would take it. Much to my dismay, a wave would rise up seemingly out of nowhere. Two or three surfers would paddle like crazy and one or two might make the drop on to a wave. I hadn’t even seen coming. Sarah, it was super frustrating. It turned into a bit of a game for me to try and pick the waves that surfers wanted to ride but I couldn’t do it. I couldn’t make sense of what they could see and that bothered me because I was the one on higher ground. I could see farther than they, especially with them lying down on their surf boards, bobbing among the wave peaks and the valleys.
Time and again, I could not pick the waves that they could see. I pick loser waves that no one wanted to pursue. Since I couldn’t pick out the winning waves, I spent the rest of my time just looking for patterns. Face it. I’m a geek, a geek to the core. I found three that day. They are shift, 20 to 1, and positioning. Months later, I learned that these patterns that I could see actually prove to be principals. They were validated and shared with me as I was fortunate to learn directly from a true surf legend, Gerry Lopez. You see Gerry and his friends actually made the north short pipeline famous.
The first observation that I made that day was that the waves seemed to be affected by external factors. As we watched, a slight breeze developed blowing to shore with the waves. This made the rise on the waves quicker and have proved more challenging for the surfers. After some time, the wind direction change and actually had a breeze coming offshore into the waves. That slowed the curl resulting in a more dramatic rise. It actually made the waves more dangerous with each swell now amassing more water. The wave’s positioning changed. The breaks on the waves were influenced by the wind or the lack thereof. As I was sitting there, some other professional entourage, I asked why do the waves seemed to be getting bigger this day?
They shared with me that there were storm swells that had been moving toward the north shore for the last week and the front had just arrived. That made sense. The waves were indeed actually increasing as the storm swells finally made landfall. Now, surfers that day were well aware of these external factors influencing the waves. They adapted accordingly. To quote Gerry Lopez, he says, “High level surfing is only possible with an acute state of awareness, a deep connection to the immediate environment and the ability to be totally absolutely in the present.” The second principle on that day is what I would call no, no, no, yes action.
A good surfer is patient. She knows it takes tremendous effort to catch and then, ride a wave. There’s only so much energy that one can expend before fatigue starts to set in. You got to be deliberate in choosing which waves you’re going to go after and so it was for those world class surfers on pipeline. I began counting and on average, these athletes where saying no to about 20 waves before saying yes to the right one. It seemed almost wasteful to me. I thought, “Wow! Look at all these great waves that are going un-surfed.” 20 to 1 and that average held out that day on pipeline.
Obviously, vary from beach to beach and depending on the day but it was intriguing to me nonetheless that these elite athletes were allowing good waves even better waves to pass them by and they would only commit to the best waves of the day. There were less skilled surfers content to take the good and better waves but not the pros, and what set apart the pros from the others was their ability to see, catch and ride those best waves. The third observation that I made that day was that that the best surfers never sat still. This is probably obvious to everybody on the webinar today.
Of course, they’re going to keep moving. They have to. The ocean’s moving and so must they. It was more than that. The movement was thoughtful. It was deliberate even purposeful and the reason for the movement was to position themselves so they could capture the best waves. I’m going to say that again. The movement was to position themselves so that they could capture the best waves. These pros as I watch them moved in packs. At one point, they’re 100 yards offshore to my left, and 10 minutes later, they’re 75 yards offshore directly in front of me. 20 minutes later, they’re about 150 yards out and much further left.
As the conditions changed, so did their positioning. They never get stuck thinking, there were great waves here 10 minutes ago. I’m just going to stay here and wait for those waves to come back to this spot. Instead, they chased the opportunities amid the changing conditions. I just kept asking myself how are they able to pivot and adapt. It was Gerry later that taught me. There were really three reasons. First, they could actually identify the best waves, which was principle two. Second, they could read the changing conditions, which is principle one and recognize where those best opportunities were surfing or surfacing. Third, they were totally in shape physically.
You see they’re extreme fitness coupled with their domain knowledge, allowed them to repeatedly get into position that would allow them to now catch the best waves. In Gerry’s words, “Surfing is 90% paddling, which is just gut busting work and all for a brief ride that’s only seconds long.” We’re a few minutes into this webinar now. We’ve got to make these surfing principles real for analyst. What do they have to do with healthcare analytics? I’m going to tee this up by asking three questions.
First, I think I jumped ahead. Sorry about that. Principle one, do you understand as an analyst the shifting environment of healthcare delivery and reimbursement? Principle two, as an analyst, are you saying no to the wrong opportunities so you can say yes to the right ones? Principle three, are you as an analyst developing the important skills and knowledge-base to position yourself to identify and then capture or help your health system capture those opportunities? Let’s dive deep into each of these principles individually. We’ll start with understanding the shifting healthcare environment.
There are external factors that are reshaping the healthcare delivery landscape in which each one of us work. Mergers and acquisitions at risk contracting between payers and hospitals, a mixture of say reimbursement models and complex payer mix. This is the reality of today Sarah. As an analyst, it’s insufficient now to simply model and forecast increasing volumes and charges for our patient volumes. It is that and it’s exploring care models that deliberately are going to drive profit away from the hospitals and into an ambulatory setting, while being mindful of the impacts whether good or bad on the at risk contracts that your health system has entered into.
It’s also about getting much further upstream in a care continuum to support pop health initiatives. Not just treating them once they’re there but actually preventing some of that admission. There’s clearly a need for the analyst today to understand the pressures on the system so that the analysis can help leadership now develop strategy, which will improve care delivery and keep your clinic doors open and profitable. Okay, principle two. This past year I can think of a handful of moments where this no, no, no, no, yes just really came into its own. In one of these, I sat in almost reverent awe watching an analyst lead a very senior executive team for one of the largest integrated delivery networks in the country through a thoughtful analysis of care delivery for their system.
Now, this is a system I knew that’s generating literally billions of dollars in annual revenues. She made a compelling case for them to stop chasing less impactful work and invest in analytic effort along with process improvement discipline. In areas, the organization wasn’t sure they can actually improve. Fundamentally, it changed the executive strategy as well as the direction of their accountable care organization. Man, it was such a gutsy move on her part, Sarah. Let me illustrate this a little bit further with a graphic. On the screen, a rise on the graph before you in the y axis represents an increase in analytic value or utility. Along the x axis, notice, this is a combination access. It has both technical skill and it’s coupled with a contextual understanding of how the information’s actually going to be used.
The simplest kind of analytics that we see in the industry are what I would call reactive analytics. What are these Sarah? Functionally, they show counts of activities or lists of patients. They include very basic calculations on industry accepted metrics such as a heart failure admission. Reactive analytics do an okay job answering pre-anticipated questions. These are generally confined to a single source of data. Many reactive analytics are included as part of a vended system. This is particularly popular within electronic medical records. Consequently, what we see is the person, the analyst delivering these reports. Actually has a contextual understanding of what’s being measure in a really remarkably low level.
Fundamentally, often does not understand why that matters. Reactive analytics explain some of the what happened but they don’t explain the why. Now, there’s value in these reactive analytics. It’s a great entry point for healthcare analyst getting into the market. Those middle area I would call reactive or above reactive analytics are descriptive analytics. These are moderately complex. They attempt to describe the healthcare world around us. These descriptive analytics leverage highly customized data models. These data models are populated with multiple sources of data. You may pull from your EMR, from your claims, HR, lab, professional billing, et cetera. These are all organized around a given domain. Now, best practices for data provisioning and data integration into an EDW afford a more comprehensive view of the activities within the health system as a whole in this descriptive space.
These get us much closer to understanding the adherence to or the departure from clinical best practices. That measurement allows us now to better understand the financial impacts of consistent clinical outcomes. The achievement of descriptive analytics requires significant interchange between both technical experts and the domain expert. Again, I’m going to reorient the audience to the x axis here. We’ve moved significantly from left to right where we do require technical coupling of skill plus an understanding of what’s actually being measured. As a result, analyst who work in this kind of analytic function developed a strong understanding of how analytics integrate into the actual workflow.
In a macro level, they understand how these analytics address external pressures that are being exerted on their health system as a whole. In the upper right quadrant, the most valuable analytics are what we would call prescriptive. Simply put, Sarah, this is where the why gets explained. We have clinical improvement, waste reduction, financial opportunities. These are discovered through applied prescriptive analytics. Now, in this space, root cause analysis is a core function. I want to call that out because once we understand the processes that are contributing to the waste, we now have information to begin to address meaningful change.
We call these prescriptive because they start to make the needful action clear. Sustained prescriptive analytics require technical experts and domain experts to work side by side almost in a permanent fashion. The analyst mentioned a moment ago, had helped her leadership team see that they were constantly … they had leaders constantly asking for reactive analytics, the lowest level of analytics. That they as a leadership team and managers throughout this organization need to actually raise their own analytic acumen. This would take some time and some training on patients but she helped them see there was a better way. How?
It was all because she said now to 20 request so she could actually say yes to the right one. Okay, last principle here. Great analyst like great surfers are really good at helping their organizations. Not just themselves, the organization. Get into position so they can capitalize on opportunities. Now, to the untrained eye, they almost seem to have an act for where the next opportunity is going to crop up. Somehow, they’re on the right place at the right time. There’s actually a science behind this and that is the result of two things, Sarah. First, they’ve obtained a fundamental knowledge of their healthcare system. Second, they’re technically, I’m using their quotes here, “in shape enough to maneuver to where the opportunities might arise within their system.” Sarah, I’m going to give you some concrete examples of that knowledge and technical fitness later, okay.
All right. I’m going to say something to our webinar audience today that might surprise some of them. Brace yourselves. Your health system did not hire you to run reports. They didn’t hire you to build dashboards, nor did they hire you to leverage some fancy expensive technology in which they’ve invested. Those are duties that maybe assigned to you but that’s not fundamentally the reason for your hire. As I work with health systems around the country, I’m going to tell you my observation is why these health systems hired you. It boils down to this. Your health systems hired you to solve problems. My conversations with executives and directions of analytics this past year has certainly validated that their top analyst use a common approach to adding value.
It’s a pattern of thinking to solve problems and this is not new. It is not rocket science but this is the pattern that they’ve all described. First, their top analyst asked lots and lots of questions. Seeking to understand what is the problem we’re trying to solve and why? Why does it matter? These deliberate questions help them tease out the best opportunities from only the good or the better. Next, the analyst ask, “Well, was an understanding of what’s the problem, I need to know what information would be needed to help solve this problem. Top analyst turn data into information so what they’re really getting at is what data do I need to begin to address the issue? Where do I find it?
Then, after finding the data, these top analyst then asked how does this data need to be organized, analyzed and presented to address the problem and to whom? To whom do I to actually present this information so that they can make a decision based on the information that I’ve shared. Does that make sense? Okay, the last step these top analyst take is to reach for tools. Can you see how this reframes the role of technology here. It flips it on his head. When analyst see their role as problem solvers, they effectively become partners for clinical, financial, even operational teams. The paring of technical fitness with domain expertise becomes a sustainable model for your analyst to become a tremendous asset to be leveraged.
Now, while this model is simple to understand conceptually, it’s rather difficult to implement practically. One of the most challenging aspects of this model is the interchange between technical experts and domain experts. This is something I’ve witnessed for 20 years. It’s fascinating to observe. The manner in which clinicians are trained to think about data or your MBA and leadership folks is very dissimilar to the way analyst are trained to think about data. When you put analyst in a room with management or analyst in the room with clinicians, there’s a real risk that they’re going to end up talking past one another particularly when it revolves around data.
Now, great analyst have learned Sarah that for them to add value, it’s not about what you say but it’s about what the audience hears that really matter. I want to play a little bit of a game. It’s distilled version of this game to illustrate that point. Folks, on the phone here, I want you to just look at the screen. Sometimes as healthcare analyst, do you ever feel like you’re communicating perfectly, the thoughts and the ideas that you have to help your customers. You may feel passionate about this but they look at you as if you’re speaking a foreign language. It’s as if the string of words that you’ve chosen to put together just doesn’t create the same image in their minds that you can clearly see in yours.
I’ve got a few examples of what I mean using a derivative of a popular family game we like to play. Let’s look at number one here. Can you figure out what the audience is actually hearing when you read this out loud? I’m going to read it out loud. We’re going to walk through the examples. If you want to play along, just close your eyes and listen to what I’m reading here. Concentrate on what you hear. If you want to read the words with me, great. You’re likely to get thrown off but that’s okay if you want to read along. Here we go. Number one, hoe pom mick hair. Number two, eye seed deny nor ten codeine. You’re looking at me with this quizzical look. This is great.
Let me say it a little faster. Eye seed deny nor ten codeine. Number three, you shoe wools us pecks. You shoe wools us pecks. Number four, eggs egg you tiffed ease is shun. That’s hard to say. Eggs egg you tiffed ease in shun. Okay, let’s go to the answers here. How many did you get right. That first one is Obamacare. I’ll even scroll back and forth so people can see. Hoe pom mick hair is Obamacare. Eye seed deny nor ten codeine is ICD9 or 10 coding. You shoe wools us pecks, usual suspects. The last executive decision. I’m curious, was that confusing for you following along? Were you able to get some of those right? What do we learn from this exercise? When I do this with live audiences, Sarah, we usually play. We got a couple hundred in the room and we play it by table for about five minutes. It’s super confusing and it’s a ton of fun.
Here’s what I love about the exercise. The most obvious learning is it’s really not about what you are reading. It’s about what the team hears. I know I’m really harping on this today but this is such a distinguishing factor between great analyst and just good ones. To make this applicable in the real world, analyst, it is not what you say it really is what your customer hears that matters. If that’s true and I believe that it is, analyst, are you taking the time to ensure that you understand what the consumers of your analysis are actually hearing? Okay, so we’ve got a poll question. Let me turn it over to you here.
Sarah Stokes: Okay, perfect. We’ve reached our first poll question as John said and the question is as an analyst working with your customer business unit, how often have you found yourself saying one thing but your customer is hearing something different? Your options are number one, never. Number two, infrequently. Number three, somewhat frequently. Four, frequently. Five, almost all the time. I’m going to hope most people are in that category but you never know, we’ll see.
John Wadsworth: If anybody is answering never, I want to clone them. I’d like to meet them in person and clone them.
Sarah Stokes: All right, the votes are pouring in. This is a good time to remind you all, you will get a copy of the slides once the presentation is concluded. We are recording the webinar today. I also want to remind you to ask your questions. Any questions that come up that you have for John throughout the presentation, go ahead and submit those in the questions pane. We will revisit those in the Q&A session once the presentation is wrapped. Okay, looks like the votes are tapering off.
John Wadsworth: All right.
Sarah Stokes: We’re going to go ahead and close that poll and share the results.
John Wadsworth: Fascinating. Yeah.
Sarah Stokes: Only 1% reported never.
John Wadsworth: Okay, I do want that 1% to reach out to me. Please give me your contact information. We need to meet. You’re doing something truly extraordinary. We need to clone that and share that knowledge with the industry.
Sarah Stokes: 19% reported infrequently or majority, 56% said somewhat frequently. 22% said frequently and 3% said almost all the time. It’s pretty even distribution there. Is that what you expected to see?
John Wadsworth: Absolutely. Yeah. Every audience where we’ve taken this poll, it is a bell curve like that. I’m going to share as we now get tactical in our conversation, what are some things we can do to actually reduce that so it’s not so frequent. I’m going to talk about some skills and some knowledge that were really helped with that communication effort. I got control back here now.
Sarah Stokes: You sure do.
John Wadsworth: Okay. This is where we’re going to get tactical as I said. Over about two decades now of working directly with healthcare analyst and being one myself, I’ve learned that there are really two key areas of domain knowledge and five core technical skills that must be present in order to achieve some sustained outcomes improvement. For the domains of knowledge, you have to understand what matters to the business and why. Simply put, this is healthcare operations. The second domain of knowledge is to deeply understand the healthcare data elements that we surrounding the care delivery experience.
Now, the five technical skills are data query, data movement, data modeling, data analysis, and data visualization. Let’s take these principles of shift, 20 to 1, positioning and drill into how these could logically be represented actually in a real world use case. I want to do this today, Sarah by taking you through the development of the rules of inclusion and exclusion for an actual diabetes registry. The design of the registry begins with why. Why do we need to build this diabetes registry. As we went through this exercise, the client partner, the question, the right stakeholders we learn that the organization had embarked on a population health initiative.
The initiative also included a marketing campaign to raise community awareness and healthy living. Now, the timing of this pop health initiative happen to also coincide with some significant growth where they were buying up or creating more affiliate primary care clinics for the physician network. Our focus for the risk initial registry build is in support of population health and our target audience is going to be the primary care physicians who have diabetics on their panel. These are the business drivers behind the example of this pop health initiative. Through further questioning, we learn primary care physicians were excited to have a much more robust set of clinical definitions for those who have diabetes on their panels.
The expressed concern about not being held accountable for diabetics on their panels. If the patient provider attribution really was not clear, if this was something that they the physicians were bought into, they wanted to be able to fingerprint that attribution model. We also learn that a small portion of the PCP compensation was now tied to the effective management of the diabetic population, which made the attribution model accuracy even that much more important.
Okay, remember the best analyst have learned, it’s not what you say but it’s what their consumers hear that matters. Through listening carefully and asking a lot of questions, there was a rule set of inclusion and exclusion developed. It was driven by the primary care providers. I’m going to walk the group through this real use case. For rule number one, we had our primary diagnosis on an inpatient encounter. Rule two, a primary diagnosis for an ED encounter. Rule three, a diagnosis of an outpatient visit to any of the network clinics. Rule number four was a hemoglobin A1c result greater than or equal to eight.
Rule five, diabetes listed on a current and active problem list within the EMR. Rule six, a certain number of insulin orders and fills. Rule seven, was the exclusionary set of criteria. We didn’t want to include any gestational diabetes encounters. As we went through the list of qualifiers, as webinar attendees, do any of these rules seem peculiar? I want you to think about that. Remember the business drive for this registry is a population health initiative where the focus on primary care physicians treating diabetics where? In the ambulatory setting.
As attendees today, what do you think about rules one and two? Understanding our business drivers, don’t they seem out of place? Yes, knowingly. As rules one and two included was actually a brilliant move on the part of this analyst, especially given the ambulatory focus. Here’s why. The physician leadership in this case have called out that an inpatient or an ED visit for diabetic could be viewed as potential failure with the pop health effort. Said one physician, “If we’re serious about improving care, we need to keep these patients out of the hospital as long as possible.”
The analyst listening to his took that to heart by creating the rules through careful questioning. The analyst was able to establish a baseline that was illuminating for the pop health clinical leadership. It showed a number of diabetics were slipping through the cracks, much more than the health system had anticipated. Through a simple and elegant analysis of just the rules, not the entire registry. Just the rules of inclusion. There were a few learnings that came out. I want to share those with the audience today. First, it showed many patients were diagnosed for the first time with diabetes in the inpatient or ED setting. That was shocking.
Pop health leadership called most of these failures of the care delivery system. The second learning was the analyst showed that some patients had insulin orders and fills without ever having an accompanying diagnosis of diabetes. The third learning was the analyst showed that there was less than 50% compliance from the PCPs putting diabetes or diabetics on their active problem list for those diabetic patients. Even though, this was an agreed upon best practice for the management of all chronic conditions. Learning number four was the analyst also showed there were many, many patients who were rapidly approaching the qualifying criteria of an A1c greater than eight but they hadn’t yet crossed that line.
By tracking the trending of A1c results on every patient, the analyst could show the pop health leadership, which patients were actually trending in the wrong direction. Said one leader and this is a pop health initiative. It was like watching a train. You knew was going to wreck in slow motion but now, we could intervene if we hurried. Now, we spent a lot of time on this slide. Now, I it warrants the dialogue Sarah. The analyst in this example was able to highlight genuine waves of opportunity for improvement. Why? Because the analyst understood the problems that the PCPs and the pop health leadership were trying to solve. That analysis informed the strategy for the pop health leadership. You may also have noticed on the screen the data sources that were approved for use in qualifying a patient for the accompanying rule.
EMR, claims, lab, pharmacy data were used to feed these rules. This part may not have been intuitive but I want to call it out. Any patient could qualify for any rule and every rule. A patient could qualify for the same rule more than once. A patient could qualify from the same rule from different data sources. We’ll talk about why the analyst set this up in a couple of minutes when we talk about the technical fitness. Okay, having data driven rules for the cohort inclusion. This was so important. In order to accurately populate these clinically defined rules, the analyst needed to understand now which data elements can I actually use. What’s available and what are approved now by the key stakeholders for use. Those decisions rest with the pop health leadership and the PCP leaders on the work team, not the analyst.
For these rules, the following codes were used in support of register inclusion. Generically, these were ICD codes, pharmacy orders and fills, lab orders and results and the problem list within the EMR. Now, she didn’t grab every available ICD code. Only those that were specific to diabetes and even more detailed in the definition, only those that were a primary diagnosis for either an inpatient or an ED encounter. She had to tie it to the encounters as well or a diagnosis on an ambulatory visit. Similarly, only A1c lab orders and results came from the lab data system or the EMR. For this analysis, she wasn’t grabbing every available lab order and result. Just those around the A1c.
As healthcare analyst, we have been given an incredible gift, Sarah, through these coding systems. There’s so much embedded meaning associated with these codes. These can and should be exploited to produce powerful analyses. We can only do this if we really understand the healthcare data. We’ve got to wade out into it. We’ve got to swim in it before we can surf it. It’s silly to assume that one could become an pro surfer by simply watching from the shoreline. It’s equally preposterous to assume that one can become an expert analyst without deeply understanding the inherent meaning embedded in coded data. Where’s the capture? Why is it captured in the clinical workflow. Let me give you an example.
Everybody on webinar today is going to feel awkward. Just bear with me. Please close your eyes. I just want you to close your eyes, wherever you are. If you’re at home, at your work, close your eyes and I want you to imagine something. Imagine a green field somewhere in Ireland. Sarah, your eyes are still open. Close your eyes. Visualize the rolling hills, the sky is a grayish blue. Keep them close. You’re standing on this one lane road and you’re looking into a field to your left admiring some grazing sheep. Most of those sheep are just minding their own business. They got their heads down. A few have their heads up.
You count them. There’s 17 sheep and one wolf. The wolf, keep them close, is jumping over the sheep nearest to you and just yards away and appears to becoming towards you. Okay, open your eyes. Is this what you had imagined? It may not be. Our experience has taught us. It’s not uncommon to have sheep and wolves together. We’ve been taught that when they mix, the wolf wreaks havoc. With that belief, based on our past learning, when we look at this imagine, we’re surprised to see some unanticipated behavior from the world. The jump over the sheep doesn’t appear menacing or attacking at all. In fact, it appears to be a game, almost a leapfrog. We’re shocked to observed something we’ve been taught just won’t happen.
The point is without our learning about wolves and sheep, we would’ve missed the humor in this obviously photoshopped image. What does this have to do with healthcare analytics? You remember the diabetes rule we talked about few moments ago, Sarah, with that A1c greater than eight. When the analyst who did this were pulled in all the A1c results and analyze the data, it revealed to that pop health leadership. An alarming number of patients with a 2.0 rise in their A1c over a two-year period rapidly approaching that 8.0 qualifier. Within the pop health initiative for leadership to see that patients could and did jump from say 5.5 to 7.5 in a matter of months. That was startling.
In fact, for the pop health leadership, this observation was every bit as surprising as a wolf and sheep playing leapfrog in a field. The analyst, though not clinical understood the clinical relevance of this 2.0 rise. How? By asking questions. As a result was able to share a shocking insight that many prediabetics were already in the system. Now, on their radar and thankfully because of this analysis, intervention was possible. For an analyst who doesn’t deeply understand healthcare data, Sarah, I would bet you a good steak dinner. They would’ve missed the insights around this A1c greater than eight rule. Okay, next principle here, making this real.
The best surfers in the world are obviously in top physical condition and the best analyst need to be in analogous top technical condition. In the diabetes registry example, I want to walk through how these technical competencies actually shape the build of this diabetes cohort. First, data query and this is affectionately known as SQL. That skill was used to join primary diagnosis using ICD code to encounters as well as patient types for the first three rules. Separate queries were developed for each data source against claim and against the EMR. Data query was also used to tease our assumptions about where to these lab orders live and the results within the EMR. The same using SQL to tease out where the pharmacy orders and fills live?
Next, data movement or ETL, leverage those SQL queries that have been written to only poll the needed data for the rules. The result in data sets then were staged within a dedicated analytic environment. By following ETL best practices, the questions that inevitably arose from the clinical team around data integrity and data lineage were easily addressed. Data modeling rode on the heels of both data query and data movement. By creating and characterize it as a landing zone if you will for patient rule qualifications. That’s included patients to qualify for more than one rule or the same rule multiple times. In other words, every instance, a rule qualification was captured and by so doing, was then available for data analysis.
In this instance, data modeling played a pivotal role in capturing what the business actually cared about. Then model that in the database. Now, it was primed and ready for analysis. when you just call out here Sarah, the data analysis came after data query. They came after data movement. They came after the data modeling. With the deep understanding of the pop health business drivers and the problems that they were trying to solve, the analyst could do what she does best. Analyze the data now within the data model in light of the business drivers. You may have noticed that I didn’t put a star here next to data visualization. That’s not all to say that data visualization is not an important skill. It’s crucial.
However, in this diabetes registry bill up to this point, we were only interested in what skills were needed to develop the cohort for inclusion and exclusion. The analyst used Excel bar charts to highlight the counts of rule qualification that was totally sufficient to get buy in for the rules of inclusion. Now much later, a tool like Qlik or Tableau would consume the subsequent data models and metrics around that broader care continuum and highlight the baked in analysis but this wasn’t necessary for the inclusion criteria. There you have it Sarah, the five technical skills absolutely necessary if your analyst are going to catch the best waves of opportunity within your health systems.
Okay, a small number of tools were used by the analyst to build out this diabetes registry rule set. One, a dedicated analytic platform or environment that allowed the SQL to be written and the ETL to be performed, and the data models to be housed. Second, there was a tool for building those custom data models, which I’ve represented as data relationships here. These were the tools required to developed the clinically driven diabetes cohort. You’ll obviously see dashboards and KPIs or start. Those tools are leverage completely when the clinical work team was ready to track adherence to best practice protocols around that care continuum within the outline clinics for the pop health initiative.
I want to ask the audience something here. Have you ever Googled learn SQL. If you do, over 181 million results come back. A more selective search like learn SQL for healthcare or learn SQL for healthcare analysis and reporting still return to somewhere around 16 million results. Sarah, I’ve had this debate so many times particularly at tech conferences where I hear vendors and senior leader say, SQL is dead. This a skill. It’s a language that no longer holds value or utility. I find Googling SQL and getting these results very interesting for a couple of reasons.
First, the sheer volume of results validates that SQL is indeed a valuable skill. It’s a hot commodity even with big data more mainstream than ever. Second, the overwhelming number of results make it hard to know where to start. Incidentally, every big data vendor has developed APIs that allow SQL to be written. When they came out, they were saying, “No, no, no. SQL’s dead. We do it for you.” The fact that all of them have now opened APIs for structure query language to still happen further validates this is not yet a dead skill. In fact, if you repeat these Google searches and you replace the word SQL with ETL or learn data modeling and learn data analysis. The same deluge of this result in pit plays out.
This is a problem for the health systems for whom I’m able to consult. They believe many of then, that they do need to learn SQL and ETL, and data modeling, data analysis, data visualization. The teams are usually uncertain about where to start and what of this education isn’t really a value add. How much SQL do you need to be effective as a common question? How deep do you need to go in data modeling to make that skill effective as an analyst? I want to have another poll question here. I’m going to turn it over to you. I think it will be informative for us to day to understand where the audience feels their needs to be more emphasis around training and education for some of these skills. Turn it over to you Sarah.
Sarah Stokes: All right. We’re going to go ahead and launch our second poll question here. This question, we’d like to know, which of these skills or knowledge deserves more training or education in your opinion. The first option is data query for outcomes improvement. The second option is data modeling, also for outcomes improvement. The third option is data analysis. Fourth option is healthcare data for outcomes improvement. The fifth option is healthcare operations for outcomes improvement. There’s a lot of leveraging going on in those answer options there.
John Wadsworth: Yeah, those are deliberate the way they’re worded.
Sarah Stokes: The poll votes are coming in.
John Wadsworth: Interesting.
Sarah Stokes: Seeing a pretty event spread so far. This will be interesting to see where our majority comes out. This is a good time to also give you a reminder. We are getting towards the end of today’s presentation. If you haven’t asked a question and you have something that you’ve been pondering about throughout, we encourage you to go ahead and submit that question because we will be reaching the Q&A session soon.
Okay, looks like things are tapering out. Give them just one more second. Get in your votes. We’re going to go ahead and close that and share it. 8% voted for data query, 18% voted for data modeling, 27% reported data analysis, 31% reported healthcare data and 16% reported healthcare operations. It’s not like really clear majority there. I mean we have data analysis seems to be one in the higher as well as healthcare data for outcomes improvement. Is there one that you would’ve thought should’ve bring to a little bit higher?
John Wadsworth: I’m pleasantly surprised to see actually that healthcare data for outcomes improvement has surpassed the data analysis. In fact, I think this is the first time that I have polled and had that one leap ahead of data analysis. I find that encouraging from an industry perspective. There’s been so much hype around the data analysis. I think this demonstrates to me and validates as an industry and as a community of healthcare analyst, we are coming to appreciate more the value of the contextual analysis, which would be reflected in learning the healthcare data.
I’m also curious about the 8% here for data query. Two observations I’d share with you on that Sarah and the audience. One is I’d like to know and I don’t know if we can tease this out. If people feel they are very good at that already, or do they not believe that query really is an important skill. Maybe they already have it or maybe a lot of data has already been provisioned for them. I just want to call out my experience for why I feel so strongly about this is if analyst can validate their own assumptions about the analysis, then they are able to really move the needle within the healthcare organization. Usually, it boils down to them having a dependency on someone else to actually validate those assumptions.
Data query is a skill where you validate those assumptions on your own. Instead of having to hand that analysis back and say, “Can you answer this question? Can you answer that?” If they can answer it on their own, SQL’s usually a way for them to get further upstream in the data streams fast maybe. Okay, let’s wrap up here. Just a couple more slides. Okay, this is my impassioned plea to our community today. Every year, so for those that don’t know, I live here in Utah. I’m fortunate to ski some of the greatest snow on earth right here. When people asked whether I’d recommend they come to Utah or Colorado, I say go to Colorado because I don’t want you clogging the ski lines in Utah.
Some days here, when it’s really good, I get a choke on powder turn after turn. Wonder struck, I’ve sat on the rim of the Grand Canyon and I’ve looked on that cavernous divide and serpentine water below. I’ve been blessed to walk in the Canadian Rockies and dip my feet in those cobalt lakes beneath those gorgeous peaks. Sarah, I’ve canoed in some of the still waters of the Snake River and watch the sunset on the Tetons and Jackson Hole. I’m sharing this with you because these are moments for me that were breathtaking, and I’ve loved them Sarah. They’ve in a way, even inspired me.
What I find more inspiring and far more lasting are the views afforded me within healthcare systems. I hope that webinar attendees feel the same. As analyst, we have the most amazing jobs in the world. I’m serious about that. We get to work with passionate people. Some of the educated elite of the world who’ve chosen to put their time and their talents to use for the betterment of society truly the cause of the clinician is noble and our role as analyst is equally noble. Then, a real way I believe it’s far more reaching and lasting because your hands, your work, your ingenuity can touch the lives of tens of thousands of patients. As you come to better understand the system of care delivery, you literary and figuratively hold those care givers in your hands with your SQL, your data models, your analysis.
It’s your scrutiny of their data and protocols that will influence their ability to deliver care. After one such analysis early in my career, the medical director for primary care sat quietly studying an analysis that I had done of his comorbid diabetes and depression patients. He looked up at me and with tears in his eyes says, “John, this changes the way I practice medicine.” Sarah for me, that was the moment that I found my career. Analyst, I’m serious when I say I believe you are a God-send in this healthcare crisis today. Your analysis truly can help systems lower costs and make care more affordable. You are the ones who are finding ways to make care more accessible to the masses.
Your technical fitness will help administrators find new waves of opportunity to better serve the patients within their communities. To me, you’re not just an analyst. I do see you as a catalyst for change. I just want to thank everybody that’s on this for choosing this career path. Thank you for your courage and for your compassion. I’ll summarize here. To be an effective analyst is going to require technical fitness. This is not for the faint of heart. The data skills that we’ve outlined coupled with a knowledge of healthcare data and operations will help you get into position to identify and capture opportunities. The more you understand the problems your health systems are trying to solve, so much more valuable will be your insights.
What can you do today? Shift your thinking to be a problem solver. Not a report writer. With regards to the role of technology and helping your health systems become great, I share with you three quotes today from one of my personal heroes, Jim Collins, out of his book, Good to Great. The first is technology cannot turn a good enterprise into a great one, nor by itself prevent disaster. Second, when used right, technology becomes an accelerator of momentum. Not the creator of it. The good to great companies never began their transitions with pioneering technology for the simple reason that you cannot make good use of technology until you know which technologies are relevant.
That holds true for analyst. No technology is going to make you a great analyst. That my friends is a purely human endeavor. My plea here is to leaders at the end. The investment alluded to has to be accomplished best. If with regularity, you will let your analyst stand their ground and say no to 20 less important opportunities so they have time to say yes to the right ones. Sarah, thanks for pulling me in on this today. Thank you all attendees for your interesting in the topic as manifest by joining the webinar today. Thank you.
Sarah Stokes: Okay. We have one final wrap up poll question for the audience before we go ahead and dive into that Q&A. While today’s topic was an educational webinar focused on the important role that analyst play in the healthcare environment. Some attendees would like to know more about Health Catalyst products and services. If you would like to learn more, please answer this poll question. Then, we’re going to go ahead and just dive into the Q&A. We do have a hard stop at the top of the hour, so we’ve got four minutes. We’re going to start with a question? Question, a question.
John Wadsworth: A question. You’re going to…
Sarah Stokes: That came in from Linda pretty early on in the presentation. It’s a pretty meaty one. She says, “What’s the best way to deal with data accuracy or integrity? How do you assess it and what do you do about it? Is there an analytic procedure that can mitigate the impact of inaccurate data?
John Wadsworth: Can you pull that over so I can just read that a little better. Yeah, that’s helpful. Great. Best way to deal with data accuracy and integrity. Okay, great. This is a big problem that I’ve seen addressed in two ways. Linda, thank you for asking. First is there’s got to be repeatable standards that can be adhered to in the movement of that data so you can trace the data lineage. Often, when we report out through an analysis, we are measuring someone with that analysis. If it’s a pop health initiative or even just a list report for a panel for providers. They want to be able to trust the integrity of data. What they’re really asking I think is how do I trust the data lineage and how do I trust your process that what you’re putting into the report is something that’s repeatable. I can now act on the information and not challenge the integrity.
That begins with good clear, standards around how you’re going to move data, model data, store data, and report it. It also has a technical dependency on the analyst to have that fitness. Now, not all analyst have all of these skills. In fact, some organizations are big enough that they deliberately parse out and specialize by skill. You may have health systems in my experience over about three campuses, three inpatient facilities. Ten to specialist those that are more community hospital-based tend to be more generalist. They do need to have all the five skills. The important thing would be if you’re in a specialize model where you’re doing the analysis but someone else is doing the ETL. Perhaps the data modeling is to ensure there are processes that bring that data architect, data modeler and data analyst together so that the seams appears seamless to the consumer of those analysis. If you’ve got a follow on question with that Linda, feel free to email me directly. What other questions have you got Sarah?
Sarah Stokes: All right, we’ve got two minutes and there’s a lot that have been pouring in in the last few minutes. Here’s one I had found. This one looks like a good one. Do you think increasing data literacy for clinicians is a worthwhile endeavor for a healthcare organization to embrace data as an organizational initiative?
John Wadsworth: Absolutely, absolutely. The lift on this in my mind is much, much bigger but the payoff is also much bigger. It’s a different set of skills but could be in large part supported by leveraging the technical fitness of those very technical people. To help broadly create tools that I would say for an organization, you should probably have two or three tools if possible that could be generically applied to help clinicians raise their own data literacy. If the questions that they’re asking is beyond those tools, then I would set up as a potential best practice dedicated lab time, open lab time to let the clinicians come and sit with an analyst. That way, you start getting some shared learning. You’re also building relationships within the organization.
Sarah Stokes: Okay, well, it looks like we’re actually at the top of the hour. We’re going to have to cut off right now. Unfortunately, there have been so many great questions come in. I will …
John Wadsworth: Yeah, I could probably take one or two more.
Sarah Stokes: One more, okay.
John Wadsworth: Are we okay to do that?
Sarah Stokes: Yeah, yeah. As long as you have time, yeah, you’ve got quite a list still coming in. There’s a couple of questions about people who are either switching into an analytic role from where they’ve worked previously, even different field. There have been several questions about what are some either low cost resources that you would recommend for someone to improve that analytics fitness. You had talked about or different sorts of resources that these people can access if they’re starting out in the field.
John Wadsworth: Great. Do you remember that graphic that I showed that had the three levels of analytics. We had the reactive, descriptive and prescriptive. I would say every health system without question has invested heavily in the reactive space. That’s manifest with very large report cues out of their EMRs. If you’re just getting into it, you already have an environment and an incentive to improve those reactive analytics. I would jump right into get involved with the reactive space. If you’re new to it, it’s a great way to understand prebake reports, canned reports that are already answering pre-anticipated questions. It will greatly accelerate your learning in this space. Then, I would be asking your manager for help how you can grow those skills meaningfully into the descriptive space. I would definitely target that reactive space for your health system. That’s a good question.
Sarah Stokes: Okay. You want to look through any of these others.
John Wadsworth: Let’s see. How to enter into the health data analyst industry without formal working experience as a data analyst but many years of experience of clinician, the history of asking the right question? Okay, so if you’re coming from the other side, as a clinician and you want to be perhaps moving more and more into the analytic space, the example that I gave of the diabetes rule set and that pop health initiative, my goodness. If we can have clinicians start moving in to the analytic space, it’s those kinds of initiatives where they can bring their understanding of clinical workflow, clinical best practices and spend the time with those analytic resources.
Even without writing SQL, no health system, if you as a clinician have the time is going to say, “No, I don’t want you learning SQL or data modeling.” You’re going to learn that best if you can get paired with the top analysts. I would just start spending time with them. Then, asking for help in the extracts to do some of your own analysis. It’s probably easiest for you to start learning some of the data visualization and then, the data analysis. Then, probably work yourself depending on how far you want to go technically into writing some of the SQL and doing some of the data modeling.
I would start with the data visualization and analysis on pop health type of initiatives or even regulatory type reporting. If you want to just get involved with those projects and then, work hand in hand with the analyst.
Sarah Stokes: Okay, do you want to do-
John Wadsworth: Let’s do one more.
Sarah Stokes: … one more?
John Wadsworth: Yeah.
Sarah Stokes: I’m not sure. This question comes from Yinkun, what if data cleaning is taking up too much time? There’s a similar question about the realizing that return on investment from all that time spent wrangling data. If you have any comments on that, any tips?
John Wadsworth: You bet. I’m going to make some assumptions here about what is happening. I’m happy to take an email from you directly and explore this further offline. When I hear about data cleansing taking way too much time, this to me speaks to perhaps lack of clarity on the process for the data movement in that ETL. I would work very hard to either document existing standards that just aren’t clear or pushing your organization to help create those standards. Then, I want to take that offline with her. I will say broadly to those who are still on here in the webinar. My observation of this particular skill is of the five skills. ETL is probably the most commoditized skill of the five.
Rapidly approaching more and more commoditization with offshoring because the technology to move data, store data and track the transaction of the data movement, and do it in a secure way has advanced to such a point now that they’re trying to drive down that cost. Health systems are still wanting to pay for access to that skill of data movement. They just don’t want to pay American dollars for that. If you’re role right now is purely in data movement as an ETL developer, I would strongly recommend you start picking up data analysis, data visualization and data modeling skills. Those are definitely lagging in the commoditization of skills to offshoring. Okay…
Sarah Stokes: Okay, great.
John Wadsworth: I’m running out of time here Sarah.
Sarah Stokes: Well, thanks for staying on for a couple of extra minutes.
John Wadsworth: You bet. This has been great.