This transcription is provided by artificial intelligence. We believe in technology but understand that even the smartest robots can sometimes get speech recognition wrong.

 Welcome to this Week in Health It where we discuss the news, information and emerging thought with leaders from across the healthcare industry. This is episode number 51. Today we're gonna do something a little different. I'm introducing a case study episode, our first one. In this episode, we're gonna take a health system from chaos to an effective data and analytics program.

Over the next two weeks, we're going to, uh, be airing the best of this week in Health IT episodes I've gathered some of the most, uh, commented on, liked, and discussed short videos from our YouTube channel. I've brought them together with our production staff to, to highlight some of the great content over this past year.

It also gives me a chance to take two weeks off and, and have, uh, and, and the people I usually have on the show to take two weeks off over the holidays and just let you enjoy some of the great, uh, commentary from this past year. So I hope, hope you, uh, hope you enjoy that and hope you share it with your staff as well.

So as our last full episode that I'm gonna be doing this calendar year, I wanna do something a little different, uh, in a digital world. The foundation for, uh, transformation is really data, and this is where our guest lives on a daily basis. The application of data to solving some of the healthcare's biggest challenges.

Dale Senators is the President of Technology for Health Catalyst and a former guest of the show. Uh, good morning Dale, and welcome back to the show. Hey, bill, how are you? Thank you. Good to be back. Yeah. I'm looking forward to, uh, our conversation, but let me, let me share a couple of stories. So these are things that came up on, um, on my feed this morning and I saw in Becker's, uh, just highlight some of the, the headlines that are going on.

So, uh, the first one was in a move to expand its presence in healthcare space. Amazon is selling software that mines patient health records for information that will help phys physicians improve treatments and hospital cut costs. Uh, the other one was Apple and the Department of Veterans, uh, the VA, are reportedly in talks to allow veterans to store their health records on their iPhones.

Quest Diagnostic Diagnostics now supports Apple Health Records feature, making it the second clinical testing laboratory to join the project. Uh, Microsoft released an open source project called Fire Server, very creative fire server for Azure to assist developers in exchanging and managing healthcare data stored on the company's cloud.

Uh, the n I h, uh, is partnering with technology company Nvidia to create artificial intelligence tools to support clinical trials. Uh, so the big hitters in data are, are circling the space. Yet in healthcare we really have fits and starts on the data front. What, what, what are some of the things you, what, what are some of the reasons you think we've had?

Um, you know, we've had some successes, but, you know, we've, we've really, uh, I don't know, we can't get out of our own way in some, um, cases. So why do you think we haven't been as successful with data as, as maybe we could be? Well, I, you know, for one thing lately, last 10 years or so since meaningful use came along, We've been so consumed with meeting the compulsory measures, right?

Who's got time to do anything creative with data really, I mean, right. It's, uh, it's an ocean of compulsory measures that is also undermining clinician satisfaction and contributing to their burnout. So we, it's, it's the, it's a classic case of distraction. Uh, for one thing, we just don't have time. You know, if you go back to the early days, you know, when I started my career at Intermountain, the compulsory measures at that time, for the most part were limited to joint commission, and they were a pretty small number.

And so what it gave us as an analytics team and a culture was the time to do cool things with data, um, which put Intermountain on the map and, uh, and literally there's no bandwidth left. So I'm really actually encouraged in spite of all the craziness of the Trump administration. Sema Verma and, and others.

I'm starting to see some common sense return to those compulsory measures. Um, but we need to, we need to slash and burn compulsory measures and give the industry some time to perform analytic, um, ingenuity and creativity on their own. Yeah, we need a chance to breathe. I, I had a similar, when I became the, uh, c i O for a health system, our internal auditor kept doing, uh, security audits on our team.

And when I went to our team, I'm like, all right, we need to get ahead of this. They just looked at me and said, we can't, yeah, , there's just, there's too much here. So I had to go to our internal auditor and say, Hey, I need to reprieve. You need to gimme a reprieve for six months, like, like no audits for the next six months.

A a year would be better. So our team can actually put some things in place and then you can come back and do as many audits as you want, but we just need some time to, to, to really get our house in order. And that's, uh, essentially what we're getting from the federal government of just, uh, continually changing.

But I agree with you. Uh, I'm seeing some, some positive things there. Yeah. Alright, so we're gonna divert our, our normal show, uh, format and explore a case study. So I shared this case study with you. I'll share with our audience and then, and then we're just gonna dive into it. So, Here it goes. And this, if, if any, by any chance, this reflects your situation and your health system.

This is not based on your health system. I'm just Names have been changed to protect the innocent Right. . Exactly. So, uh, this health system has 18 acute care hospitals on two different EHRs with regional customizations operating in four states. The health system participates in two ACOs that utilize point-to-point integrations between disparate EHRs and separate systems to produce reports and house uh, registries.

Uh, the health system has several medical groups on different, on different, but a single E H R platform and a clinically integrated network that has 50 distinct EHRs represented. The medical group utilizes their own e d W to generate, generate metrics for the clinically integrated network and for their other constituents.

The health system participates in three different regional health information exchanges that transfer data with CCDs, obviously across C C D A format, uh, the system level ed, uh, that's at the system level. They also have an E D W that was created five years back, and the demands on the team are really overwhelming.

The team spends a majority of their time cleaning the data in order to produce a set of reports for their constituents. Every year, the data team makes a request to grow the staff by about 20%. The executive team is asking for a set of reports across the system to measure clinical and operational performance.

Today, the variability of the data and the the definition of the terms makes comparison very challenging for the executive team. There's a feeling amongst the leadership team that their investment, uh, in systems is making, uh, their, their investment. In systems is in analytics and the E H R, that they should be getting a lot more value from the program for the money that they've invested.

Um, their plan is to hire a hotshot data, c i o or C uh, chief data officer to save the day. And Dale, congratulations. You've just been offered and accepted the job, so we appreciate you joining the organization to, to help us sort of move through this. I'm excited. Yeah, this is a great scenario, bill. Nicely done.

So do you, uh, I mean you see elements of this across the industry as you go out there, right? Oh, yeah. Yeah. I mean, this is the norm now. As complicated as it sounds, it literally is the norm. And that's all right. So the, the stated goal from the executive team and the clinical executives and whatnot is, uh, that they would like to utilize data to rapidly and continuously improve clinical, operational and financial performance for the health system and the care providers.

So for our organization, we're gonna organize it into really six topics. We're gonna start with discovery. You're starting your first day. We're gonna look at discovery, what you need to discover. Second thing is preparing the organization, uh, for change. The third being, uh, the components of a healthy data utilizer.

Fourth being operationalizing the program, fifth being the technology, and sixth being ongoing care and feeding of the program. It's interesting 'cause so many times people wanna start with the technology. If I just bring this in, all these problems will be solved. But, um, but we know that that's not the case.

So let's start with discovery. Where, where are you gonna start? What are you looking for? What data do you need to gather? Who do you need to talk to? You know, I literally putting my hat on is if this were my first day, um, I would start in the usual way, which is you've gotta get out and meet all the, the major influencers and constituents, um, and start building relationships.

I always say that there's nothing more political, especially in healthcare than data and its utilization. Um, people think that deploying an E H R is brutal culturally, but you'll, you'll eventually get past the pain of deploying an E H R. But, uh, the politics of data and the politics of utilization of data and the, the light that it can shine both accurately and inaccurately on people and their careers and processes is what makes it so political.

So, Um, you know, for the most part we've almost commoditized the technology, right? The public cloud has made the infrastructure pretty easy. You know, I'm, I'm, uh, humbled by the progress we've made with Health Catalyst. Other vendors have, you know, technology that sits on top of public cloud infrastructure that makes the tech almost a commodity.

So I would plan on spending about 80% of my time in marketing the value of data to the organization, and also, um, making people feel trustworthy and trusted and, uh, and building those relationships. I, I gave a lecture out in LA yesterday to, uh, a group of folks that are interested in collaborating and sharing data There.

Um, they're brought together under a common mission, but a disparate governance structure. And I shared with them that I spend most of my time now advocating the soft side of data and the human side of data. Um, and, and, and I'm gonna reference an old Harvard Business Review article I read years ago, so long ago, I can't even remember now when it came out.

Uh, but it talked about the ways that, um, the characteristics that are important when you're trying to influence and, and collaborate with people. And it boils down to three characteristics. Uh, the first characteristic is the person on the other side of the table has to feel that you have overlapping values in some areas, that in essence, that you're honest and trustworthy, and that you have their interests in mind.

So they have to feel good about overlapping values. And of course, if you're working across cultures and countries and things like that, it's sometimes hard to find those overlapping values. But for the most part, honesty and trustworthiness and having a genuine empathy for the person across the table is a pretty common way to establish, um, overlapping trust in that first characteristic.

Yes. So that has to be genuine. It can't be manipulative, right? Yep. The second thing is, um, you have to have a direct sense of empathy for the role that the person occupies. You know, and, uh, the classic, you know, example is an IT person talking to a physician about how we're gonna improve their lives with technology.

Um, it's not unusual for us as IT people to alienate those clinicians because we've never been in their shoes. I. So if, if you can't engage in a conversation with a clinician, with a sense of direct first order empathy with them, then at least bring a partner with you, bring a physician partner with you to those conversations, or bring a C f O or someone from the finance team with you.

But the person sitting across the table from you has to feel that you got first order empathy with the situation that they occupy. And, and so there has to be a big overlap in that capacity in addition to the values. But then the last thing, the last of the three traits that's really important is the person on the other side of the table has to feel like you have expertise and skills that they don't have, but they value.

So you have to bring something to them that's common in the first two characteristics, but you have to bring something to them that's unique and valuable to them. And, uh, so. That's how, that's the first thing I would start doing is engaging with people in the organization, in those positions of influence, um, reminding myself constantly of those three characteristics that are so important to affect change.

So, alright, so I'm gonna take you through some groups and, um, so first group you're gonna meet with is the head of the clinically integrated network. And what they're concerned about is, hey, we have an e D W, it's different than the system. E D W, we like what we're doing, our team's doing a good job. It's outside of it.

We have our own little analytics. You're not going to, you're not gonna impact my program, are you? You're not gonna take that away from me, are you? Um, and the likely answer to that would be pr no. And, and I had this conversation with, uh, one of our most important clients the other day. We're kind of moving into the next generation relationship with them and trying to figure out how to expand.

The, the impact they have mul, they still have multiple data warehouses. Um, in the organization we've managed to reduce from five down to three, including ours. Um, but the reality is, you, you can't convince people to give up the data they have unless you can add value to them that they appreciate. So, and, and usually that value is a sense of control over the data, um, agility with the data.

Those are the two most important things. The third is the cost of the data. But, but people will actually hold on to their expenses if you don't show value in those other two areas around control and agility. So you, you'll, you're never gonna win with a mandate to say, we're gonna go in and we're gonna, for the sake of it, consolidation and reduced expenses and licensing, we're gonna take your data warehouse away.

That is absolutely losing strategy. You have to go in and offer them something that's more appealing and attractive than what they have. And until you can do that, Um, you're not gonna win. So now I'm gonna take you into the CFO's office and the CFO's saying, um, you know, look, we've, we've got so many different, uh, systems out there.

I can't even, you know, I'm having trouble closing the books and getting the kind of reports I need out of this thing. What are you gonna, what are you gonna do? And by the way, don't, don't ask me for $5 million, don't ask me for $3 million. Um, what are you gonna do to, to sort of streamline this? So that, first of all, that I'm getting, you know, what I think I should get for the money.

But second of all, um, that, you know, we're not sitting in these board meetings, in these executive meetings anymore, having arguments over the, the definition of the data. I mean, we need to be comparing these 18 acute care hospitals across their performance. And I can't even do that with the disparate systems that you have.

So, what whatcha gonna say to the CFO around those things? Well, it sounds like there's two battle fronts there, right? One is the, the, um, The disparate data and data governance. And the other is just the, the, the plain old expense of having all those disparate systems, right? Yeah. Um, and, you know, you, you can go back out to the organization and the folks that are in the field and appeal to them from a, a perspective of cost and appeal to their larger sense of belonging to the organization.

Most of the time these disparate source systems are funded not centrally by it, but by the business units and the departments. Yeah, that's true. So, um, if you can show them a way to save money on those expenses that they have in those local it, um, disparate systems without giving up their sense of agility and autonomy and

And, um, and, uh, you know, self-governance, then you have a chance of making progress. But, um, the c you, you kinda have to wear the rank of the C F O, but hopefully not in, um, in a way that takes away a person's, uh, sense of control and, and autonomy around the data that they control. So let's talk about the clinical leadership.

So the clinical leadership's gonna look at you and say, um, you know what, we have some reports that we've been requesting. Again, these are retrospective reports we've been requesting for the better part of, you know, three to three to four to five months. And we still don't have those reports. Uh, some of this data that's coming in, I, I don't even wanna see it anymore.

Other of the data. I would rather not have it come in that way. I'd, I'd like to see it in my workflow. I don't need a separate system. Um, what are you gonna do? What are you gonna say to that person? I mean, obviously we're just in the dis discovery phase, we're not in the creating the solution phase, and so I imagine you're gonna show empathy for that person, but what, what are you, whatcha gonna say to the clinical leader?

Well, yeah, I mean, the, the first conversation that I have with clinicians nowadays is just a simple acknowledgement that you're, you're over measured about things that don't matter. And I get that as an IT guy. I totally get it. I totally empathize. We're over measuring all your processes. We're under measuring

And detracting from your ability to achieve outcomes and be personal with the care you provide. So here's what I'm gonna do as an IT guy. I'm gonna try to relieve as much of that compulsory burden from your shoulders as I possibly can. I'm gonna try to optimize the way the data's collected in the E H R.

I'm gonna try to optimize and present the data back to you in a way that's helpful to you and not oppressive to you. Um, and I'm gonna find bandwidth in what we do and how we work together so that I can actually give you data that you really want in the form that you really want. But the first thing we have to do is automate as much of these internal and external KPIs that are burning you out and get as much of that off your shoulders as we possibly can.

And then let's give you some time to work with me and my staff to give you the analytics and the decision support that you need. And, and, and, and going back to your. Know the, the strategy for getting things into the workflow. Um, I always break down my decision support strategy in three tiers. The first tier is the, at working at the population level.

The middle tier is working at the protocol level, and the third tier is working at the patient level. And I, and so you have to have an analytics and decision support strategy that operates at all three of those. Right? So there's different kinds of analytics. When you're dealing with the, the population of a community, different skillset, it looks more like epidemiology applied to chronic disease.

Yep. Then, and you're working in timelines that extend, you know, eight to 10 years. Right. The temporal dimension is very long at the protocol level. Now you're starting to do analytics and decision support in a subset of that population. You're saying for these kinds of patients, these are the analytics that we need to better understand the care that we're delivering, the outcomes and the costs associated with these cohorts of patients.

That's typically, you know, now you're dealing with hundreds of thousands, maybe tens of thousands of patients instead of millions. And in general, what you're saying is the patterns for care for these patients need to look kind of like this. Um, and by the way, lemme comment that what we're, what we're doing at Docs right now is we're trying to force all patients into these evidence-based care guidelines, and we're taking away from the ability to personally treat patients.

So we're putting too much e emphasis on evidence-based care, as if every patient should be treated exactly the same way. And what we should be encouraging is sort of protocol level minimized, reduction of variation or minimized variation and reduction variation. But within that, there should be a lot of variability on a micro level about how we're treating patients.

And right now we're not doing that with e, with evidence-based care and the, and the data strategy we have. And then finally the last, the last loop, which is, you know, ultimately might be the most important, is what data are we going to present to help you as a clinician engage with a specific patient that's sitting right in front of you?

And how are we gonna do that within your workflow? Whether the workflow is on your, your mobile device, when you're, you know, on the treadmill at the gym, or if it's in your car, or if it's at the E H R. Wherever your workflow and your decision making can be supported, let's give you the data about this patient that can optimize the conversation that you have with that patient about their care and their outcomes.

Um, You know, the, I referenced the study that Ken Komoto at University of Utah did a number of years ago, in which he concluded that and found that clinicians are 15 times more likely to adapt their, um, their treatment towards a patient if you give them this substantiating data to do that at the point of care, as opposed to giving that to them in a conference room someplace.

Right. It's what I call the, the folly of conference room analytics. Yep. Um, you can only go so far at the population and the protocol level. You have to push decision, meaningful decision support into the workflow of the clinician. And it's still hard to do, you know, frankly, it's hard to do. I've been in this, you know, I've been passionately pursuing all three of those loops for my whole career, and it's still hard to do.

Um, in part, you know, and I'll put a little pressure on the E H R vendors, the HR software was not designed. To support dynamic, intelligent user interfaces. And so I think, you know, fire is helping get to that last loop and that last mile of decision support. But we've, um, I, I think long term, I think the h r vendors have to bulldozer software and build it in modern software that's more indicative of intelligent user interfaces that we see as consumers now.

And it's enabled by good software in the background. It's, it's, it, I quite often say modifying old software is like trying to modify a house with, uh, with concrete walls. Um, it just, concrete walls weren't just not meant to be adaptable like we're accustomed to as consumers today. Yeah. Um, yeah, we, the, the E H R needs to be much more modular in terms of its architecture so that we can Yeah.

Plug these new things in. Well, uh, wow. That's discovery. We've got five more to go and, but I had a side question and my side question is, uh, it is, it is, I, it's not in this case study, but if, if in this case study you had, um, let's say they're, they were operating as a payer as well. Is that a whole different analytics mindset and approach, or is that pretty similar in terms of how they would view the data and go after the data?

Uh, no, I think it's a great thing to have, um, a payer as part of the environment, right? That's what finally closes the economic model back Yeah. Onto itself. Um, instead of this, This open model that we have right now where everybody's spending everybody else's money. But can I take, can I take hospital, um, acute care, uh, analytics team and just say, Hey, you're also handling the, the payer side of it, or is it a different, different thought process or skillset that I'm gonna need?

Well, it, it depends, right? If you're utilizing claims data to support a better understanding of clinical operations, cost of care, um, what's happening outside the four walls, the healthcare delivery system, um, and it's rounding out your understanding of the, um, the patient's health, then, then your traditional analytics team in a hospital centric organization can handle that kind of thing.

If you're talking about the analytics associated with risk management, risk projection, more, that looks more like an actuarial function. Then I think you're, you're more than likely gonna stretch, if not exceed the capabilities of most analytics teams in a hospital. Um, I, I will say that there is an opportunity for those folks in the hospitals, those analytics teams in hospitals to bring new data science techniques into risk management projections.

So, you know, the short story is I think we can actually disrupt traditional actuarial techniques with some of our new machine learning capabilities and, and, uh, and algorithms. So there is a chance for those analytics teams in the hospitals and, and traditional delivery systems to actually be better at predicting risk and managing not only risk clinically, but risk financially than what the um, actuarial folks do in a traditional payer, I.

If, if you ever, and I got, I had the opportunity to peek behind the scenes, you know, at Intermountain and um, and in the Cayman Islands where we had this integrated delivery network with insurance and care delivery under the same leadership. And I gotta tell you, man, the actuarial techniques that are being used by payers is way outdated.

And, um, and we pay a premium for that as, uh, patients. Um, so there's a lot of opportunity to disrupt the actuarial techniques that are currently in use today. Awesome. Alright, so last, so we have five, uh, things. The next thing we're gonna do is prepare the organization for change. And that is create a sense of, yeah, create a sense of urgency, build your, uh, coalition of the willing and a vision.

So one of the things you said is you're gonna go out there and educate people on the, the value of data, I guess is where you're gonna start. Um, how are you gonna do that? So what, what is the vision, potentially the vision of like one or two talking points on the power of data? Uh, how are you gonna create urgency and how are you going to build the coalition of the willing?

Well, yeah, it's a great question, Fran. You have to paint a vision that's compelling and motivating and, um, that's, I think if I do anything in the industry right now, I think that's what I spend most of my time on. And, you know, if you describe the use of data to enable better digital conversations between a physician and a patient, right?

I've got one, I've got my aspirational statement hanging on my wall right over here. I should take it off and read it to you. You read those aspirational statements to a clinician and to a patient and to an administrator, and they go, yep, that's exactly what we need to do. Well, what, what, what, what is your aspirational statement here?

Okay, hold on. I, I wish I were smart enough to have it memorized, but I I still don't, so I'm gonna pull it off the wall. Uh, hang on here. Um, well, we'll, we'll just admire the beautiful picture of your family back there by the lake. That looks nice. . That's up in Canada a couple of years ago. Okay, you ready?

Here it is. So this is the aspirational statement that, that we have here at, at Health Catalyst, but I'd have this hanging in my office if I were a C I O. We provide the software, data, and professional services that enable physicians to extend this commitment to their patients. I can make a health optimization recommendation for you informed not only by the latest clinical trials, but also by local and regional data about patients like you.

The real world health outcomes over time of every patient like you and the level of your interest and ability to engage in your own care. In turn, I can tell you within a specified range of confidence, which treatment or health management plan is best suited for a patient specifically like you and how much that will cost.

So, you know, we literally parse this statement and we look at it and we go, okay, what's the technology and what's the data that we have to have to achieve and enable that conversation? Um, and when you read that to physicians and administrators, they go, well, yeah, that's exactly what it is. And when you read that to patients, they nod their head.

Boy, I'd love to have that kind of conversation with my physician. Um, and, and I think we all, we all get this right, so yeah. What we're saying is we're gonna, we're gonna enable you to be a better C f O with the data. We're gonna enable you to be a better c e o with the data. We're gonna enable you to be a better clinician with the data.

We're even gonna enable you to be a better patient or, or consumer of health. Yeah. With the data, it is the enabler for all these things. And that's, that's the vision. We're sort of painting, but how do you build the sense of urgency around, Hey, we need to, we need to start doing things now because things, I mean, generally what people do is they go in there and say, Hey, things are dramatically changing in the space.

You feel it, you feel it with, you know how you, they'll paint the picture of a Blockbuster or Amazon or whatever and they'll say, look, all these things are changing. Healthcare's changing, but um, You know, my experience, I came in from outside of healthcare and my experience was, hey, things are changing rapidly.

And then when I came into healthcare, I said, I think healthcare's gonna change in three years. Yeah. And the things I was, I was saying six years ago that were gonna change in three years haven't changed yet. Yeah. Well, yeah. And I said the same thing. I wrote a paper for himss, uh, 12 years ago that said, analytics and decision supported we're standing on the brink of a revolution.

That was 12 years ago. And we still are a long way from what I described in that paper. I thought we were within a year or two of it. Um, you know, it's funny, right? Build the, everyone has a sense of urgency. I mean, all except the most passive and apathetic people in healthcare, and there aren't very many of those, right?

Everybody has a sense of urgency, but I think we're trapped on a treadmill that makes it incredibly difficult. To implement the kinds of changes that we'd like to the system makes it very difficult, which makes me sort of pessimistic about whether the existing system can change fast enough, or whether it's going to be an Amazon or a Google or an Apple or a b c consortium that pulls it off.

Um, you know, there are always the old guard in every culture who feel like they're either surrendering to the fact that we can't change or they're not convinced that we're not good enough. Um, and you know, quite often as part of change management, you just have to get the right people on the bus. So if there are key people that are holding back the organization's progression towards being data-driven in a smart way and being more digital, I think you have to get rid of those people, quite frankly.

Um, and, you know, I had lots and lots of turnover and it, it's not something you go into with a lighthearted at all. It's brutal, it's horrible. But if you don't get the right people on the bus, um, then certainly your sense of urgency is going to suffer. Now, whether good people on the bus can still operate within the constraints we have, um, is still yet to be seen actually.

Yeah. But, but building the coalition of the willing isn't as hard as it once was because, yeah. Uh, because we've had successes, right? So people are out there going, oh, I, I've, I, I've seen what, you know, this house, I've seen what Asante's done. I've seen what Intermountain's done. I've seen what Geisers done.

And they're able to say, Hey, why can't we do those things? So, and you find enough of those people that are saying, we want to do those things. That's your co coalition of the willing, right there. Totally. Yeah. Find the early and when you go out, you know, and you're building these relationships, find the, the, the folks that are passionate about change, that are passionate about doing it from a data perspective and that have influence in the organization.

Look for those three qualities and then attach yourself as an IT person to those folks. You know, I always referenced back to Don Lappe, head of cardiology at, in the, uh, cardiovascular clinical program at Intermountain. He was my first, um, clinically influential anchor at Intermountain. When I came in, I didn't know anything about healthcare, but I knew a lot about analytics and data warehousing, and I just found him and his staff, they were data-driven and they were research-oriented.

And I said, you know, I think I can make your life easier with the technology that I understand. And Don became the seed for influencing the rest of the company and clinicians when they saw all the cool things that he was doing around data. Then a lot of folks followed, Yeah. And, and that is such a huge principle for CIOs because the CIO shouldn't, shouldn't be influencing as much as sort of coming alongside the influencers and yeah.

So let's talk about the next thing. So components of a healthy data utilizer. So you have some clients, uh, you can name 'em or not name 'em, doesn't really matter to me on the show, but, uh, you know, components of healthy, uh, utilizers, they, they have certain roles in place, certain principles which they operate on and certain types of governance.

I mean, one of the reasons we're having this conversation is I read a paper that you write wrote three years ago on, on data governance. And I thought, man, I wish I had read this six years ago. Uh, or maybe, maybe it's that old and I just didn't, but uh, but yeah, there, there are some principles that good, healthy data utilizers sort of operate with and some roles they put in place.

Why, why don't you talk about some of those things? Yeah. Okay. Well, I will mention a couple of role model clients and I've, I'll, of course, I'll offend. By not mentioning them, but, but Mission Healthcare is Mission Healthcare example. Uh, John Brown, the c i o there, um, uh, North Carolina, right? Yep. Um, Allina, uh, the, um, the culture there at Allina, Jonathan Shoemaker's, the c i o up there.

Yep. Um, those are two good examples of, we have more than that, but those are two good long-term examples of clients who take advantage of data. They're good utilizers and they, they have a great combination of, um, defining their data-driven goals from the top down, sort of the Intermountain model, frankly, which is the board is going to approve some significant financial and clinical goals for the next year.

And those goals have to be backed by, by, um, the evidence of data that we've achieved them. So there's always Intermountain Geisinger's is the same way. Mission. They have really good top-down clinical and financial goals that then they line up analytics resources underneath those, and they, and as well as change agents underneath those.

But they also leave room in the analytics capabilities and capacity of the organization to allow things to bubble up from the ground up. Recognizing that innovation, the true innovation right around data happens at the edges of the organization. Um, so to do that, you have to have, um, good centralized governance, but not too heavy, can't be too top down.

And you also have to have a culture that is, um, I, I would say more risk tolerant than the norm when it comes to data utilization. So you have to knock down barriers to data access. You have to have approvals. Um, when there are needs for data, you have to knock down those barriers for approval so that people feel, um, empowered to get to the data as quickly as they can when they need it.

And then you have to trust people with data. You have to push it out to the edge of the organization. You have to give each person in the organization that wants data, the data they need, as opposed to some organizations who still have a very authoritarian governance structure, a very authoritarian security environment, and a very authoritarian utilization environment.

And you know, I still hear to this day, we can't trust this person with that data because they'll misuse it. We can't trust them with their own ignorance. Well, what you, that's not the right attitude for heaven's sakes. Give them the data, then teach them how to use the data in a proper way. Yeah. You know, so those are some of the components.

The, the, the, the data governance structure at some of these high utilizers are, um, is a very good combination of, um, a balance is what it boils down to. Right. I always paint the extremes of data governance as authoritarian or anarchy. Right there is either very heavy handed governance, which holds back progress or there's no data governance, which has all sorts of impacts on data quality, utilization, and consistency.

And the best organizations, um, hit the sweet spot in the middle and, uh, and they appreciate data as an asset. So it's interesting. So you're saying, uh, from a risk posture standpoint, error on the side of Yeah. More transparency, allowing people to utilize the data. I mean, you can pull it back later. Um, I mean, you're not, obviously you're, you, you're saying it's somewhere in the middle.

It's not, Hey, we're just gonna let everybody in here. But, um, one of the things, one of the things I found interesting was we had a, uh, an organization we were working with to share data with the patients and through our, uh, digital tools and. Um, our internally, they were concerned about sharing it externally.

'cause they were like, Hey, we're not sure how clean that data is. And the, the the partner said, well, give 'em the data. They'll tell you how clean it is. Exactly. Exactly. There's no better way to clean up data than to expose its ugliness. Yeah. And, and if you've gotta have a thick enough skin as a c I O and a, you know, chief analytics officer to recognize that data's never gonna be perfect.

And as long as you release the data in a smart risk management way, explaining the possibility of, and the limitations of data quality, the best way to improve data quality is to expose it. There's no doubt about it. But that takes a little bit of courage that frankly, I think a lot of organizations don't have people worry too much about their careers.

So good governance programs. I, you know, I'm sure people are gonna want more detail and we're gonna be able to give them in, in 45 minutes. But, um, when you think about the best governance, you know, what, what people, you know, how do they meet, what do they discuss when they get together, those kinds of things.

Yep. Uh, the best data governance function has the weight and authority of the Supreme Court. But, um, not every case goes to the Supreme Court. So you, you want a top-down endorsement about data as an asset. Data governance is important. Data quality is important literally from the c e o. So I help, I help craft, you know, emails for CEOs to express their, their commitment to data governance and data as an asset.

Establish that tone, um, announce the data governance function as just another one of our governance bodies in the organization. Announce the membership, which should include. A good chunk of the C-level on the CEO's staff. So the C M O needs to be there. The c f O needs to be there, the c o O, the c i O.

But then, um, what I've always managed to do and what I advocate is, um, delegate a lot of the day-to-day decision making to the C I o. So he, that's, that he or she becomes the, um, what amounts to the, the day-to-day court system. And when the c i o needs to elevate and resolve a, uh, a situation around data governance and data strategy, then you have the body of the Supreme Court to appeal to and, and support.

Yeah. But are there, are there, are there some roles? So let's, let's go back to your data and your analytics team. Are there some roles that are morphing or changing? So we, we had, we had a bunch of e t L people. We had a bunch of, uh, uh, analysts. We had a bunch of people doing data cleansing. We had, I mean, are there, are there roles that are changing with, uh, the technology as it's, as, as it's evolving, are in, in our health systems?

You know, it's interesting. I, so I'll just air my mistake. So I hire, I advertise and hire three data scientists. They get snatched up in the organization. I go back six months later and they're writing reports and I'm like, ah. You know, and so that's, you know, what, what are the, what are the roles look like and are they changing, I guess?

Yeah, they are changing and we're, we're starting to move up the stack of commoditization, which is great. So in the old days, right? A simple thing like a storage engineer for a data warehouse, data warehousing, storage engineering was really complicated. You don't need that anymore, right? The public cloud has made the infrastructure, the storage, the server configuration, especially around analytics and data warehousing, essentially a commodity.

So the old, the old style data warehouse, d b a, systems engineer, systems admin, storage engineer, those roles are starting to become commoditized and replaced by, uh, by the public cloud. Um, one of the things that I do see, um, that isn't moving fast enough is too many, um, resources in the analytics team spending their time on E T L, um, That's a combination of a couple of things, right?

Over time, the number of jobs, the number of data objects that you have to manage in a data warehouse, in a data platform just increases. So that's just a necessity. Uh, or, or a, um, not a necessity, but just sort of a reality. The other is, you know, you can follow the two extremes of data modeling that continue to drag things down.

Enterprise data models have their own E T L problems. Light binding has its own data problems for E T L. In between, there ought to be some structures that should be more commoditized, um, where you lock down information models and data models around things that are persistent and comprehensively agreed upon.

And by the way, that's the, that's the, the mantra I use is persistent and comprehensive agreement about data bindings and data logic ought to be locked down so that you're not continuously maintaining that you can lock that down. Then you can move your analytics staff and your E T L staff onto other things.

Those, so a big part of what I see with analytics teams right now, they're still spending too much time on E T L because of this extreme of either late binding data modeling or enterprise data models. We need to start facilitating these intermediate data structures that offload some of that work. Um, the, yeah.

You want me to comment on the data analyst versus data scientist thing there, bill? Oh yeah, absolutely. That'd be great. The, um, so what quite often happens, right, is because we're inundated with these compulsory measures, you look around the organization, for anybody that can write a report to fulfill the need, um, you just have to resist that.

There's a great book called Essentialism. Everybody ought to read essentialism. And, um, and quite often it's, it's too easy to say yes to every report that comes along. And if you say yes as an analytics team, every report request that comes in. There's no first order economic pain from the requester of that report, you're gonna spin yourself right into the ground.

So you have to have a data governance structure and a leader of the analytics function that can, that knows when to say yes and when to say no. That's really important. And as mundane as it sounds, I see it all the time where analytics teams get consumed by individual requests funneling in and their request queue never ends.

And so they never get a chance to do cool and innovative things. So you gotta be able to say, no, you have to have the air cover from the governance community to support no, occasionally, and then you have to carve out a team and it, I think right now, I would absolutely carve out a data science team. That was focusing on what amounts to next generation analytics.

You've gotta have that core, that addresses the internal and external KPIs that are required by compulsory measures right now that's inescapable. Try to make it as efficient as you can with these intermediate data structures, but then carve out a dedicated team and don't let anyone touch that team without the significant oversight of the leader of the analytics function.

And put those data scientists on on innovative opportunities and ideas that we've never had a chance to address before. Are you hiring those people? Are you potentially outsourcing or just out staffing that? Well, I'll tell you a great lesson that I've learned actually. Um, I think there's, um, I think the path from data analyst to data science is easier than it's ever been because the models, the algorithms and such are being commoditized by the open source community of data science and machine learning.

So I went into this role about three years ago thinking that I was gonna have to hire a, you know, million dollar a year data scientist to take advantage of this asset that we have. And I'd probably have had that same mindset as a C I O if I were still practicing. But what I found is that over the last three years, the rate of improvement for machine learning and data science is, is far outpacing Moore's law.

You know, if Moore's law is, you know, doubling in capacity every 18 months, we're doubling in machine learning capacity in like every three to four months. And that's, I mean, that quite literally. So you don't need the PhD data scientist like you used to. What you need are solid feature engineering skills, which data analysts are perfectly suited for.

And then what you really need is knowledge around what amounts to experimental design in data science so that you make sure that what you're producing from data science makes sense operationally. Um, and clinically, so it's, it, the feature engineering and the data analysis skills of traditional data analysts.

They've got a great career path going forward. The models, the algorithm things are easily accessible to the citizen data scientists today. I can be a data scientist today in ways that I never could in the past. Interesting. And then if you add that experimental design, um, to your skillset, knowing how to take advantage of, and, and thoughtfully apply machine learning and ai, then you're, you've got a really bright future.

So I don't think you, I don't think organizations have to hire outside anymore for the data science skills. I think you, they ought to build the data analyst skills or the data analysts from data analysts into, um, data scientists. So, operationalizing the program, I had two things under that, handling the backlog and quick wins.

Um, I do want you to, to go into quick wins that make the most sense for a a, a new program. Uh, but before we get there, handling the backlog. So let's assume I have, you know, requests. That are, you know, nine months, nine month kind of backlog. Do you essentially put a stake in the ground and say, all right, here's what we're gonna do.

We're gonna take all of these requests back through governance. Yeah. Or you, you, you, you would, okay. Yeah, absolutely. Yep. And you, and you've gotta, you know, essentially set aside the economics to support the request and say, look, we're gonna dedicate, and I, this is a number I quite often advocate, 60% of the centrally funded data analysts, uh, resources will be, um, assigned to things that we decide as a governance structure.

Top down, these are important initiatives, departments, projects that we're going to support from the top down. Then I'm gonna carve out 40% of my team to handle, um, opportunities that pop up that we didn't foresee from the top down as a governance body. So I'm a strong advocate of that 60 40 split and, and essentially keeping those two teams almost separate because if you try to split a person 60 40 towards what amounts to compulsory measures and new opportunities, they always get overtaken by the compulsory measures.

And we all do it. We all sit there and go, we all, we all do it. 60% of your time over here, 40%. And you know what happens is whichever one they like the most, that gets a hundred percent of their time. Yep, yep. So I'm an advocate of creating two teams within the analytics department. One is focused on what amounts to the automation of and the, and the hands free completion of top-down compulsory measures and things that the board's interested in.

And then 40% of the team is dedicated to opportunities that are gonna pop up from, uh, the grassroots. So to reinforce the value of this program, we need some quick wins. What are some places you've seen quick wins? Uh, quick being defined as within the first, uh, three to six months. Uh, oh gosh. The, um, there's the usual suspects of clinical and financial value that we see all the time in orthopedics.

Um, uh, cost control unnecessary and sort of low value care, sepsis, cotti, clabsi, um, and it's usually pretty easy to rally people around those kinds of things. Um, uh, so those are some off the top of my head Bill. What I also try to do is, is, um, take care of a kind of a bright and shiny object in the quick wind.

So you, there, there's the commodity kind of analytics around cost and quality that, you know, there's roughly 10 to 15 things that make up 80% of the opportunity in healthcare. So take care of those things. But the, the other side is you've gotta give people something kind of cool and interesting that keeps them attracted to the data digital strategy.

So I would look around for a researcher or maybe, um, um, an opportunity even in finance, right? We, we have, we've got a, a predictive model in that helps identify propensity to pay and get patients lined up with better financial arrangements ahead of time before they get into trouble, for example. And those, those bright and shiny objects are important.

They have to have real value, but they also have to show sort of the value of what you could do more broadly. So taking care of the commodity stuff is important, but you also have to have a little bit of a marketing and a bright and shiny project or two, um, that keeps people motivated. Yeah. Well, and that bright, shiny object for us was, we created a, uh, sort of a, a.

A front door to our, our data warehouse that the financial people could play with to look at value-based care. And they were able to move different levers and see what the impact would be. And my gosh, I mean, they could play with that thing for, for for days, but it was, it was solving a very real problem, which was when, how, when and how, and what impact is it gonna have to go from fee for service to value-based care and how do we make this transition?

So, very cool. Yep. It's very interesting. Um, you've already talked a lot about the technology. Is there anything else you would say on the technology? Well, the, you know, the big trend in the technology realm right now is these hybrid architectures between SQL and NoSQL and all of the cloud vendors, Microsoft, Google, Amazon, they all offer some version of that.

Um, Yeah. I, I as a c I O, so you're not saying this, I'm saying this, I, I can't imagine building my own e d w on-prem anymore that does the, the cons, like, I don't understand. I, I would not build up my own data center anymore if, if somebody came to me and said, Hey, let's build a $5 million data center for, for our, our health system.

I would say that's insane. Right? But I almost feel the same way about an eed w at this point. It's, it's like, no, let's look, I mean, there's this many layers and if we go to a cloud vendor, we're gonna cut off, you know, six of the bottom, half of the layer that we're not gonna even have to do. And why would we do it?

It's already out there and it's already efficient. Is that, is that what you're seeing or is that what you feel at this point? Well, we're, you know, we're in a weird spot right now, right? Where, um, the E H R vendors are offering their analytics and data warehousing strategy. They're nascent, they're just emerging, frankly, they haven't been around for very long.

You've got companies like Health Catalyst that are deeper. We've been around for a longer time. We're not perfect either. You've got the, the pure technology players like Cloudera, um, Mongo, and then what the public cloud, Amazon, Azure, and, and Google offer. And so there's this weird churn going on right now in the market where if you compared Health Catalyst, for example, against a pure technology player, you'd go, wow, those tech guys are way more sophisticated than Health Catalyst, so I'm gonna go that direction.

Well, then you get over there to the public cloud vendors and you go, they don't know anything about healthcare data and they don't have any tools that are specific to healthcare data. So I'm gonna have to build all those data models, all those APIs, all those applications myself. That's not very appealing.

Right? So then you look at the E H R vendors and E H R vendors are new to this. They offer some tech that's, that's always centric around the data that they have. They don't have the experience working in a disparate, heterogeneous environment. They don't have the tools for that. So it's a, if I were a C I O today, it would be pretty confusing.

I mean, of course it sounds self-serving. I'm, I'm, you know, with Health Catalyst, I'm trying to be that clear answer. But the reality is there's the convenience and the attractiveness of doing business with your core vendor. There's the bright and shiny appeal of doing business with Amazon, Google, and Azure.

And then there's the sort of pragmatic, let's do business with Health Catalyst, and those are the three. Um, those are the three options that I see in the industry right now, and I think it's still working itself out. You know, I, I think it's still, um, yet to be determined which of us is gonna survive as the, as the best, uh, option.

So since, since I've taken up more of your time, and I'm already three minutes over the end of our show here, um, I, I am gonna give you a, a, a chance to answer. So, but doesn't Health Catalyst gimme the best of all worlds? Aren't you built on top of the cloud platforms, the cloud vendors that are already out there and you give me the ability to bring in all that disparate data so I can, I can actually plug into, I, I would assume I can plug into some of those cloud AI and machine learning things that are out there, plus you're, you're, you are sucking in a lot of that E H R data as well, aren't you?

Like sort of that, that middle ground that sort of gives me the best of the cloud vendor and a little, I mean, I understand that the, the E H R vendors are gonna be able to do that, that regulatory reporting and the clinical operations probably better than anybody because they can just bake it right in there.

But, but don't you give that sort of middle ground of the best of both worlds. I appreciate you saying that, bill. I mean, that's what we're trying to do and I, um, you know, I'm cautious about using your podcast as a sales channel. The, but that essentially that's, I'm trying to provide to CIOs and to the healthcare system the, what I always wanted as a C I O.

And what I think, you know, we need in the industry is just a commercial version of what we've done, you know, at Intermountain and Northwestern and other places. So, um, you know, if I were still the C I o I would definitely be looking at Health Catalyst. I would not be looking at building my own, I wouldn't be looking at the Pure Tech players in Silicon Valley building my own.

There's no way I would do that. Yeah. Um, I'd be looking carefully at the HR vendors and I, I, what I would probably do is say, I'm gonna use the HR vendors for part of my analytics and I'm gonna use a, um, health Catalyst for the other part. And I think that's the best option in the industry right now, is that combination.

Yeah. So how are you gonna bring all that data together and how am I gonna plug into. How am I gonna plug into the machine learning and AI that is gonna be developed at the, at the level it's gonna be developed the best is at that Microsoft Cloud level and at that Amazon cloud level and Google Cloud level.

And if I can't figure out a way to plug into theirs and rely on the e h R vendor to build that out for me, that, that level of, of, I, I, I'm not sure that makes sense. So I that's, that's my 2 cents. Yeah. Yeah. I would add, I'll add one more comment on this. Sure. It's not as cheap going to the public cloud as I think all of us former CIOs thought it would be.

Yeah. You know, it's still pretty darn expensive and it's, it's actually when you look at the costs of, uh, of our private cloud that we offered, And granted, you know, we couldn't offer the flexibility and the infrastructure that, um, Azure and Google, Amazon offer, but it's still pretty expensive to go to the public cloud.

So I hope that, uh, I hope that we see those prices come down from the public cloud vendors. I'm sure. We'll, yeah. Yeah. And, um, yeah, we're not, I, I, I think about 10 episodes ago I met with Robert Rice and we talked about the five different, uh, cloud architectures that you need to look at, and that's one of the ways that we were able to manage costs, where we were at.

Um, yeah, so ongoing care and feeding we're, we're, gosh, we're seven minutes, I'm probably gonna split this into two episodes now. , but ongoing care and feeding. Is there anything you do, uh, from a culture continuous improvement? So we've, we've addressed, you know, we've sort of stood this up. We've put in governance, we've hired the right people, we've put the right technology in place.

Um, and it, it sounds like, you know, from a leadership standpoint, set the right metrics and. Um, so maybe, maybe where I'll take this is, uh, you recommend a book, I'll recommend a book. So the book I just finished reading is Measure What Matters. Mm-hmm. . Yeah. Which, uh, which a friend gave me. And it's a phenomenal book.

And I, I love that the phrase I get out of this is as measured by. So when somebody says, Hey, this is what we want to do, and the natural follow on question is as measured by what, what's the measurement? And it's really, uh, and I think you talked about this earlier. You know, if, if in healthcare we would stop measuring things that don't impact, I think we'd be better off.

What, uh, what reading would you, would you recommend? Uh, there's two books. One is, it's actually written by, thank you for the book that you gave me, by the way. Um, the, but it's written by the same author, Stanley McChrystal and it's team of teams. Mm-hmm. highly recommend that book. If you, if you peel back the, the, the lines, uh, underneath the lines of that book, It's really about information sharing.

So it's, it's a summary of what, um, McChrystal did as the Joint Operations commander in Afghanistan and Iraq, uh, during the first phase of conflict there when we were, you know, the US was being outmaneuvered by Al-Qaeda in Iraq, and despite, you know, unlimited financial resources and military capability we're being outmaneuvered by this network of terrorists.

And he talks about what he did to transform the way the Pentagon and the way his teams operated in that theater. But, and it's all, it fundamentally boils down to two things, knocking down organizational barriers in pursuit and cultural barriers in pursuit of a common goal and knocking down information, sharing barriers.

Um, it's really, it boils down to those two things, is what he did. Information sharing became critical to the success of. Of the first phase of that conflict. So that's one book I highly recommend. The other is a book by, um, an author. Her name is Tally Sharot. Last name is Ss, h a r O T, tally Sharot. She's a PhD, I think she's a cognitive scientist.

Um, she's at the, uh, university of London. And, um, more and more, you know, as we commoditize the technology of data, I'm advocating you have to be, uh, more in tune with the neuropsychology of data and the behavioral side of data. Going back to those three attributes I described, you know, about common overlapping values and common, uh, empathy and then, uh, unique skills that you bring to the table when you're engaging somebody for change.

The book that, that she wrote recently that I really enjoyed was called the Influential Mind. The subtitle is What the Brain Reveals about our Power to Change Others. And it's, uh, if you combine the right technology with the cognitive science described in that book, that's going to be the key to success with a data-driven strategy in healthcare and engaging people in a different human way than they've been engaged traditionally in the past around data.

You know, I've never, in my 35 career years in this career of data going all the way back to my military and n s a days, I've never seen, um, uh, anything like what I see today, which is the more you try to push facts on people to change their opinion or behavior, the more entrenched they become in that opinion and behavior.

It's not affecting change. So it's tribalism, it's fake news, it's, you know, the climate change deniers, all of that. The more you try to push change through data onto people, the less likely they are to change. So you have to engage your data strategy from a different neuropsychology of data than ever before.

I've never been, I've never seen anything like this in my 35 year career. That's interesting. Um, yeah. Well, awesome, Dale, thank you for coming on the show. Um, best way for people to follow you, uh, uh, I assume is the, um, health Catalyst, uh, blog. I see your articles out there. I also see on LinkedIn, is there anywhere else people can follow you?

Uh, I tweet once in a while, bill and, um, so those are probably the best ways. I guess those are, those are the best three. Yeah. Awesome. Um, you can follow me on Twitter at the patient cio the show at this week. Init our website this week in health it.com. And, uh, shortcut to the YouTube channel is this week in health it.com/video.

Uh, please tune in the next couple of weeks for our, uh, best of, uh, episodes. Everybody have a great holiday and, uh, please come back every Friday for more news information and commentary from industry influencers. That's all for now.