Microsoft: The Waking Giant in Healthcare Analytics and Big Data (Webinar)
Microsoft: A Walking Giant in Healthcare Analytics and Big Data
Okay. Well thanks everyone. I really enjoy spending time with you again. These webinars are always a lot of fun but they would be really boring if it weren’t for your participation. So thank you.
So I’d like to get the poll results there. I’m a little disappointed that only 7% of the attendees are clinicians and 93% non-clinical. That’s very interesting because the data wave, this revolution we’re going through is going to have a big effect on clinicians. So the more clinicians know about this, the better. So that’s kind of a troubling metric there. We’ve got to figure out how to get a few more clinicians involved in the awareness of this kind of topics that have traditionally been associated with the IT crowd. But we’re all in IT now. That’s the reality. Healthcare has become a technology-enabled services industry, just like any other services industry then. Technology is going to play a larger and larger role. So, we’ll talk a little bit more about that here in just a second though.
One of the reasons that we’ve put this webinar together and one of the reasons that I am talking about it so much lately is, you know, Health Catalyst has a big partnership and relationship with Microsoft but there’s a lot going on in analytics in the big data space. So periodically we evaluate all of our partnerships with all of the vendors that we depend so heavily upon to deliver products to the industry to make sure that we’re still partnered with the right folks. And so, for the past year, I’d really been digging into the details of Microsoft, especially prompted by the new CEO, Satya Nadella, and wondering what kind of effect he was going to have on the company.
And I’ll share with you here in just a little bit. My history with Microsoft has been very up and down as an IT professional. And so, I just wanted to make sure that we were still hitched to the right wagon going forward and hence the scrutiny on Microsoft and the new CEO and what’s going on. So this is kind of a summary of that year of in-depth exploration and diving deeper into the new Microsoft under Mr. Nadella. So that’s kind of the prompt behind this in case any of you are wondering about that.
Why should C-levels care about topics like this one? [02:36]
So, taking on that theme that I just mentioned – why should C-levels care about topics like this? Because in today’s world, business moves at the speed of software. That’s the reality. Great people and great facilities are not enough anymore. Software is the enabling or disabling factor to success in today’s business world.
So for the same reason that a good CIO needs to understand the basics of healthcare delivery, a good CEO needs to understand the basics of healthcare technology and software – because if you don’t do that, you’re not going to be able to participate as a C-level executive in the wave of technology infusion with the healthcare right now. What you’re going to do is you’re going to pump that and dismiss that to the rest of the organization and hope that someone takes care of it, but the reality is your entire company is dependent on software now. So that’s a big delegation.
For Example [03:38]
I’ll give a few examples. If you think about the impact that good and bad software are having on critical business transformation in healthcare alone, regardless of industry, Healthcare.gov, an entire national policy almost in the tank because of bad software. We’re all struggling with better software support, population health management, accountable care and value based reimbursement, EHR interoperability is causing us all sorts of problems, we don’t have any detailed cost accounting software, we’re having a difficult time engaging patients technically, and we don’t have analytics at the point of care that can actually help raise quality of care, cost of care, and speed of translational research. All of these are transformational business processes in our industry that are completely dependent on software. And so, the message to C-levels that I’m dropping as often as I can is that you have to start becoming more aware, you have to start becoming more hands-on in your understanding and your configuration of the software that runs your business or else you’re really starting to put your way out of the success, or potentially the failure of the organization in the future.
On a scale of 1-5, what is your overall perception of Microsoft as a company – its people, products, and culture? [05:00]
Okay. So let’s go on here. Let’s ask a poll question really quickly while we’re talking about Microsoft. I am really enjoying knowing what your overall perception of Microsoft as a company is –people, products, and culture. From 1 to 5. I would be very interested to see this.
Thank you, Dale. I’ve got that poll question up right now for everyone. On a scale of 1 to 5, what is your overall perception of Microsoft as a company? Its people, products, and culture? We’ll leave that open for just a few more moments and then share the results.
And while everyone is filling that in, we would like to remind everyone, you can enter in your questions or comments in the questions pane of your control panel.
Alright. We’ll go ahead and let’s close that and show the results. Alright. Dale, our results are we have 3% responded as very negative, 10% negative, 32% neutral, 45% positive, and 11% very positive.
Well, you know what, those are higher numbers than I would have expected. So basically what is that, 56% of folks have a positive view of Microsoft, which I think is probably higher than it would have been a couple of years ago, for sure. Lots of neutral sitting on the bench probably waiting to see but not as much negative as I expected. That’s pretty interesting. Well, I think that’s good and I think that the more positive impression of Microsoft is justified. And again, that’s based upon my recent experiences working more closely diving into the details of Microsoft. So that’s interesting.
Okay. So let’s go through the agenda here. I’ll talk about my up and down experiences with Microsoft over the past 32 years. We’ll talk about Microsoft’s cultural and technological transformation. And by the way, this story is as much about their cultural transformation as it is their technical transformation. And I think it will probably go down in business history as one of the most significant turnarounds of a giant company in a short timeframe certainly in the last hundred years. We’ll talk more specifically about Microsoft’s analytic options and how those are emerging as really elegant solutions in healthcare analytics but also analytics across the industry. We’ll talk a lot about this unusual product called Azure and the perception of Azure is generally a little different, I think, than the reality of Azure and I have had a lot of fun over the last year digging deeper into that product and understanding not only the technology but its strategy. And I’ll talk just a little bit about the PowerX product line and these visualization tools that lay over the top of Azure as well as the analytic platforms that Microsoft offers.
My Life on Microsoft [08:13]
So, my life on Microsoft. This was kind of a fun little exercise. I went back in time and I plotted out my career on the bottom axis and I plotted out kind of significant events from Microsoft that affected my career on the top of that axis.
A few times, I wanted to poke my eye out [08:35]
And as the cartoon indicates, there were a few times I wanted to poke my eyes out because of Microsoft’s problems. So my original exposure to Microsoft came in the early 80’s when I was a young CIO information systems officer at strategic air command and I think we had the first 50 IBM, 286 DOS machines and basic computers that were available off the assembly line there. They had a whopping 20 Megabyte hard drive. It was not a bad experience. It was a kind of cool and interesting experience. It is the first time that I’ve been exposed to the most significant IBM-based, DOS-based programming and we had our own little network set up there. So it was a good experience in general.
After that, I got out and went to work for a company called TRW and the National Security Agency under a contract there with TRW and under the leadership and the membership of my good friend, Ron Dalt we were involved in software safety, threat analysis, security analysis, all that kind of thing, especially at that time as it related to the nuclear command and control system of the US from the president down to the weapon system. What we watched was this growing infusion of commercial off-the-shelf software into military and very sensitive computing operations. And it was about that same time when those three came out and all of the security problems associated with Microsoft started to emerge.
And so, the security headaches around Microsoft went on for the next 25 years. It was a long time before Microsoft really got a handle on secure computing firm and OS perspective and plugging all their vulnerabilities.
Another interesting thing at that time was TRW was an investor in the development of Sybase, and those of you who will remember, the original version of SQL Server was based around the Sybase code. And so, we got a lot of exposure very early on to not just relational databases but Sybase in particular, the evolution of SQL Server starting with Sybase.
After my stint at TRW and NSA, I went to work for Intel Corporation where I designed one of the first Enterprise Data Warehouses, what we came with kind of an integrated logistics data warehouse for Intel with a big success. I had had kind of an Oracle and UNIX background up to that point. Despite my exposure to Sybase and SQL server, it just wasn’t quite scalable enough when it came to data warehousing. So, I propose at Intel that we developed an Oracle and a UNIX-based data warehouse for them, and of course I was very naïve. Intel was not at all interested in that.
So we deployed it on SQL Server for quite few and at that point all the codes associated with Sybase had kind of been pushed aside. SQL Server was becoming its own new product and we deployed it on MT and MT turned out to be the limiting factor for scalability there but we did okay with it, it worked out alright. It wasn’t too bad. It was successful and Intel got a lot of value out of that analytics platform.
I went on to work for Intermountain as regional director of medical informatics there and one of my first jobs at Intermountain oddly was to convert the organization to Microsoft Office. And I have to think about this for just a second. At that time, Intermountain, being a very Utah-based company and Utah-based culture, had very deep roots with NOVEL, another Utah company, and WordPerfect, another Utah company. And so, the notion that we would transition to Microsoft Office at that time was heretical. I mean it was very controversial. So, very painful experience but Microsoft worked very closely with me to make that a success and it turned out to be the right decision, of course, in the long run but certainly not very popular.
I went on to be CIO at Northwestern when the two CIOs there on the campus there, along with Tim Zoph, and took a bet on SQL Server as the platform of choice for our data warehouse. And at that time, that was, again, a very unpopular decision because Oracle was the dominant data warehouse platform in all industries at that time. SQL Server was unproven. But I could tell that there was a trajectory. I could see a trajectory for Microsoft and I could see a trajectory for Oracle in the BI analytics space. And you could see that Microsoft was developing for analytics and data warehousing the same sort of product suite mentality that they brought to the desktop. So, very efficient, very simple, high value integration of a lot of products required for business intelligence and data warehousing. Oracle was struggling. The total cost of ownership for Oracle was very high. They didn’t have an integrated product environment. And so, I took a bet on Microsoft and it turned out to be hugely well placed then. It accelerated our development by, I would say, years, certainly months, and the total cost of ownership for SQL server was about 70% lower than it was for Oracle. So, as a CIO, that was also very appealing.
I went on to be the CIO for the National Health System at Cayman Islands, and on my first day on the job in the Cayman Islands, actually the day before my first day, there was a complete viral outbreak of the Conficker virus that affected every single Windows device, both servers and desktops in the organization. So my first day on the job there, there was zero computing, nothing worked. We had no access to anything. And that was again because of this ongoing problem with Microsoft’s security environment. Now, you could argue that as the IT team and the Cayman Islands kicked off with patches that the Conficker virus would have been relatively easy to maintain and avoid. But they didn’t do that. And the reality is tax management in Microsoft was never very good, which was always standing in the way of keeping up with security. So really bad experience around the Microsoft environment from that Conficker virus.
In 2011 then I joined and we got Health Catalyst going and I kind of got reintroduced to SQL server 2012 and 2014 and the things that Microsoft was doing. And then started slow and becoming introduced to Azure. And so that’s what brings us to today’s webinar.
So again, kind of the message here is my career with Microsoft has been very up and down and if I would have taken this poll maybe two or three years ago, I would have been kind of neutral on Microsoft, maybe slightly negative. But I’ve pivoted that opinion and it’s unlikely to pivot that quickly on an opinion but I’m convinced that what I’m seeing with Microsoft is real and is impressive and worth the admiration actually.
Analytic Roadmaps for Healthcare [16:41]
Okay. So, now let’s talk about a couple of different analytic roadmaps for healthcare, and I think some of the folks have seen this before. One, there’s a new framework and we’ll talk about that new one here in just a second.
Healthcare Analytics Adoption Model [16:54]
So this is the healthcare analytics adoption model that a number of us developed over the last few years. Denis Protti and I published the first version of this in electronic healthcare back in 2011. The idea being here that this gives you a roadmap for assessing yourself. It definitely gives you a roadmap for assessing vendors, which was the primary motive – hold this up against vendors and ask them how do you meet each of the levels with your product and your functionality? What’s your offering at each of these levels? And it also offers us an opportunity to do an industry-wide assessment of where we are as an organization. And again, I plagiarized very purposely from the look and feel of the HIMSS Analytics Adoption Model for the EMR adoptions because that was a very effective tool. So I think it goes out to Mike Davis and others at HIMSS Analytics and Dave Garrets who developed that model and it’s been very effective before. It’s the baseline in the industry.
So I’m going to ask you to respond here in just a second and ask yourself where do you reside and where does your organization reside on this model. And by the way, what I’m also implying here is that we ask ourselves, where on this stack does the relationship between Microsoft and Health Catalyst reside? And so, Microsoft, being kind an industry-agnostic vendor, you know, they’ve had inroads into healthcare a few times have been very successful. For the most part, Microsoft’s products in Azure and in Analytics stopped at about level one and a half. It certainly don’t go deeply into level 2. So, everything that a vendor like Health Catalyst provide is essentially from level 2 up, with Microsoft satisfying level 1 in a variety of ways.
Now, you’ll see later on, as we talk about Azure, I think there could be a community of development that occurs in Azure that rounds out sort of Microsoft-centric solutions in this other level. But that’s a little waist off. Right now, Microsoft is a directing level 1 at the technology platform level. And then organizations like yours and Health Catalyst and others take off at level 2.
At what level does your organization consistently and broadly operate in the Analytics Adoption Model? [19:36]
At what level does your organization consistently and broadly operate in the Analytics Adoption Model?
So, at what level does your organization consistently and broadly operate in the Analytics Adoption Model? And don’t feel bad if you’re down at level zero or below. Someone yesterday mentioned that they were double zero. And it’s kind of fun. We would like to look at these numbers and polls over time and see if it’s changing. We’ve been doing these now for quite a while. So, see the numbers coming in there.
Okay. We will leave this poll open for just a few more moments. We’ve got quite a few questions of folks coming and asking if they will have a chance to get these slides or the recording of this webinar and I would like to remind everyone that, yes, we are recording the webinar and after the event, we will send out an email to everybody with the link to the slides, as well as to the recorded webinar.
So we’ll go ahead and let’s close this poll now and let’s share the results with everybody.
Okay. 14% stated zero, 34% levels 1 to 2, 33% levels 2 to 4, 14% levels 5 to 6 and 5% at levels 7 to 8.
Awesome. You know what, I think the numbers are starting to move up. We’ve been doing this now for a year and a half and I think the numbers are starting to move up. It would be interesting to do a historical on it. But those of you that are consistently operating at levels 7 and 8, that 5%, the top 5%, we would love to help showcase you as role models. And it doesn’t matter how or what vendor you’re associated with and this is certainly not any attempt to sell a product to you. If you’re operating at that level, we want you to stay there and keep getting better. But we’re trying to highlight in the industry the role models are operating at those levels. So we’d love to hear from you. If you feel so inclined, please reach out to me and Tyler and let us know and let us help you become role models for the rest of the industry. Thank you so much. Okay. Good data.
“Closed Loop Analytics” [21:49]
Alright. So, the Healthcare Analytics Adoption Model is one framework that we used to help kind of define roadmaps and evaluate vendors and define strategies and things. The other concept that I’m starting to advocate with the help of some dear friends in the Vancouver Island Health Authority and University of Victoria is this concept of closed loop analytics. And so, here’s my definition as of HIMSS 2015, and that is, the concept is presenting data in the workflow of decision making, such that the data optimizes the outcome of the decision. And it’s worth knowing that physicians are 15 times more likely to modify their decisions about patient orders and protocols if they’re presented with data at the point of care, as opposed to offline in clinical quality improvement meetings. And I think that just validates commonsense. There’s an interesting side story to this but the US has invested billions of dollars in smart electric meters for homes, with the idea being that we were going to change consumers’ consumption patterns or consumption purchasing patterns around electricity and avoid some of those brownouts and blackouts that we’ve had over the last few years.
So we’ve put all these smart leaders in place. We started collecting all this data. It’s a little bit parallel to EHRs. We spent billions on EHRs to collect. But what we didn’t do with those smart leaders in the electric utility world is we didn’t expose the data at the light switch, where people are making their decisions about power consumption. So people’s purchasing habits and power consumption habits haven’t changed one iota since we installed all those smart leaders around the country because we haven’t embraced this concept of closed loop analytics. And so, Google, with their acquisition on this, probably sees that, right? And that’s where we’ll have a big impact eventually on the way we consume power in the US.
‘Closing the Loops’ on Clinical Outcomes to Optimize Quality [23:56]
But the same kind of thing is happening in healthcare. So this is a confidential draft. Again, my friends, Corinne Eggert, Ken Moselle, and Denis Protti up in Canada and I are going to publish a paper on this in the next couple of months. But what this diagram attempts to communicate is the relationship between the electronic health record and the analytics that we can expose through an enterprise data warehouse and how those two have to interact together. And there are basically three loops of closed loop decision making that we believe we have to address in healthcare. Loop C as at the population level – what are we doing to understand and plan and develop practices and policies for population health management. The next level down is protocol management and protocol enhancement. And the final level is patient level analytics that’s specifically addressing patient care. I’ll talk a little bit more about how each of these kind of interact here in the next slide, but you might kind of pause and think about this, right? How do you address the analytics for populations, protocol development and optimization? And what are you doing with analytics to specifically improve patient care at the point of care? And I would argue that we’re just starting to understand the analytics of population health management. We’ve been toying with protocol enhancements for a number of years and a few leading organizations and using data to drive that. But we have largely left the patients out of the benefits of analytics to support their precise care. That’s the new frontier.
And if you’ll think about this diagram, there are companies like Health Catalyst and Explorys and Watson, especially IBM Watson under their new acquisitions, that can address and have been addressing Loop B and Loop C reasonably well, which can always be better but reasonably well. But we’re losing out in Loop A. There is no closed loop analytics to speak of of any significance in the industry. Now, there is a company called Stanson Health and I would give Stanson Health a plug right now. Great group of folks out of Cedars-Sinai. It’s got one garden and others working out there. And they’ve implemented some very cool closed loop analytics around Epic out of Cedars-Sinai, specifically around choosing wisely.
So, we’re starting to see some traction here but we’ve got a long way to go. And as I mentioned, physicians are a lot more likely to change their ordering and practicing behaviors, and patients too, consumers, if you deliver analytics to them at the point of decision making, not in board meetings at Loop C and not in clinical process improvement meetings at Loop B.
“Closed Loop Analytics” [27:11]
So, I want to talk about the sort of Mean Time To Improvement and Span of Population affected too as it relates to this. Loop C, we’re talking about a Mean Time To Improvement that’s measured in years and decades. The Span of Population Affected, you can measure than in millions, several hundred thousand people, lots of folks affected. The analytic consumers tend to be the Board of Directors, the executive leadership team, strategic plans and policy, government bodies, that kind of thing.
At Loop B, we’re talking about a Mean Time To Improvement that’s measured in weeks and months, hopefully, not decades anymore like it used to be. So protocols are being adjusted on a weeks to months basis, forfeiting those clinical process improvement teams with the data they need to reflect on those protocols, and you’re affecting subsets of patients, numbered in the hundreds and thousands to your subsets of patients, chronic conditions, complex procedures, that kind of thing. And the care improvement teams and clinical service lines are your consumers.
At Loop A, what we’re trying to do is drag that Mean Time To Improvement down into minutes and hours, maybe then seconds. The Span of Population Affected is in individual patients. We are now precisely treating that patient according to analytics that informs that decision. And it’s physicians and patients at the point of care that are the consumers.
So what I’m suggesting here is think about your strategy – what are you doing with your data, what are you doing with your vendor partners and hold this up against them and ask Microsoft, ask Explorys, ask Health Catalyst, ask Watson, what are you doing to address each of these closed loops and how are we delivering analytics back to the point of care to support these kinds of Mean Time To Improvements and Span of Population Affected?
Big vs. Small Data
The ROI of data to Population Health [29:14]
So there were a couple of questions and I think it’s really to this context, this big versus small data. That’s a kind of a new term. Everybody now is moving from big data to small data, which I can’t even keep up anymore. But there is some validity to it and that is there’s a return on investment from data as it relates to population health especially.
And so, I just drew this chart together this morning. It probably requires more thought, but I think conceptually it’s there, and that is there’s actually some very low volume data that has extremely high value to healthcare but we’re not collecting it on a regular basis yet. Lab data is really really important to patient care and aggregating that and analyzing that, looking at that. Familial data is a poor man’s genomics, where we don’t collect it. Outcomes data, from patients, critically important, but we don’t collect it and it can be just a few questions. Literally, you know, 20 or 30 questions about outcomes can make a huge difference in our understanding of healthcare, but we don’t collect that data. Cost accounting data, very valuable but we don’t have it on a detailed basis. Simple socio-economic questions, there are probably 10 or 15 questions that are critically important socio-economically to healthcare outcomes but we don’t collect it. And then patient activation measures, the ability for a patient and the willingness for a patient to participate in their own care. These are really low volume, high value data sources in healthcare that we do very little with.
So, there is some sense to this new term “small data” and this is how I look at that term – is this kind of a chart. So if you have any thoughts about how to make this better, it would be very interesting to hear from you.
Volume, Ability, Act [31:03]
I would also like to suggest that the volume, the ability to analyze and the ability to act on data are all interrelated as well. So typically what’s happening, and this is what happens across all environments. I witnessed this in the military, I witnessed this in National Intelligence. The volume of data that we can now collect is exploding at X to the 10 at least. But our ability to analyze that data, I think, is operating more at X-cubed. And our ability to act is linear. And that’s always going to be the case. We’re always going to have a lot more data than we have the ability to analyze that data or the ability to act on that data. But the unfortunate thing is we tend to develop policies and strategies, believing that these are all congruent, that these graphs all overlap, and they don’t. The love affair with predictive analytics is a good example of that. Our ability to analyze and act on predictive models in healthcare is still a way behind the data that we have access to.
So just keep that in mind, as you’re developing strategies for analytics and things like that, remember that we’re always lagging behind, especially on our ability to act on the volumes of data that we have. It’s going to take some time.
Finding optimal data volume [32:32]
There is also this notion of what I call “the needle in the haystack zone”. At some point, you become overwhelmed with data and the value of your data reaches the tipping point and spills over into negative, declining value. And you have to be aware of that and it’s a part. I don’t really know how to suggest, I don’t really know how to get my arms around this problem. Our National Intelligence Committee is struggling with this right now – how do you find the needle in the haystack and all of the data that they’re collecting right now. It’s just beyond our ability to analyze that data and a lot of times it’s beyond our ability, again, to react. So just be aware that you can become overwhelmed with data and at that point, it becomes negative value, not positive value.
Microsoft’s Cultural and Technological Transformation [33:22]
Okay. Finally we’re ready to start digging into Microsoft here a little bit with all that background.
Why my faith and trust, in any company? [33:26]
So why do I have faith in any company? It doesn’t matter if it’s Microsoft or not. By the way, that’s a little cartoon of Satya Nadella. But the reality is my faith and trust in any vendor is based upon their track record, the values, the vision and execution of that company, especially that relates to the CEO, the financial viability of that company, the tactical behavior of the employees, and then ultimately the affordability and the value of the products that they offer. Those are all of the things that make me feel good or bad about the company. And when I look at those criteria, in Microsoft’s case, I feel pretty good about it. Satya Nadella is a very different guy. I was never a big fan of Bill Gates or Steve Ballmer. Bill Gates is a much better person now than he was when he was a CEO. And I think that a lot of Microsoft’s problems were reflected in the leadership style of the CEOs at the time, and Satya Nadella is a completely different leader.
But note out, the financial viability of Microsoft is giant. There is the latest market cap figures, those I mean huge huge money. No problem there. They’re going to be around for a long time. And that money gives them the ability to re-tool both culturally and technically. That’s the other nice thing. The behavior of the employees is now a better reflection of Satya Nadella than Steve Ballmer, which is great. Really good people now and they are behaving differently.
Microsoft has always had pretty decent product value but it’s even more pronounced now with some of these new analytics products in comparison to their peers. The track record of Microsoft has been a little bit spotty but it’s getting better. So that’s how I look at companies.
The Innovator’s Dilemma
Microsoft has shown the repeated ability to overcome this… [35:21]
One of the things that Microsoft has done is they’ve really shown me the ability, repeatedly actually, to overcome the innovator’s dilemma, and that is as Craig Christensen wrote about – Can you sacrifice your sacred products and business model in order to thrive in the new world? Do you have the courage to do that? Not whether how but can you do it? Can you give up that current revenue stream to prepare for the new world? And can you literally bulldoze products and retool yourself? And Microsoft has shown an amazing ability to do that but more recently even more so.
Sacrificing the Sacred Cows [35:57]
So here’s a good example of that, speaking of sacred cows. Microsoft is making Windows free on devices with screens under 9 inches – not licensing that anymore. Giving up that revenue stream. Using that sacred cow. Microsoft’s giving away Office on iPad, iPhones and Android. The two cash cows, fun-intended, of Microsoft, Windows and Office, over the years are now becoming commodities in order for them to compete in the new world. This is really hard to do. When you’re a company, like Microsoft, all those folks that used to make money out for selling Windows licenses and Office licenses, that’s disrupting them very personally, not to mention the microscopic business model of Microsoft. So they are leaving the innovator’s dilemma and overcoming it right now.
Microsoft Openness [36:58]
The other thing that’s very interesting about Microsoft is their openness. And there’s a little cartoon that shows what happens when you have an open mind and that’s what they’re doing. So Steve Ballmer in 2001 called Linux “a cancer”. Satya Nadella last year said “Microsoft loves Linux.” Microsoft has a team that was operating as a separate business entity, now is back in to Microsoft organization formally. It was a bit of an experiment to see whether it would work and it has worked. And so, the whole purpose of that team is to open Microsoft’s products, as well as take advantage of open software in general.
And here are some interesting metrics indicating their commitment to that. They are the single largest contributor of code to Open Source and Apache. 20 percent of the Azure infrastructure now runs on Linux and you’ll see how big that Azure infrastructure is, it’s gigantic. .Net is now in the Open Source community. There’s a tight relationship with Hadoop through Hortonworks. They initially support Dockers containers in the Linux world. Microsoft owns 310 Android patents. It’s a big supporter of Facebook’s Open Compute data center project. And it’s the third largest contributor to the Linux kernel now. This is a very very different company than the one that I grew up on in my career.
Review: Microsoft Azure beats Amazon and Google for mobile development [38:29]
Another interesting thing is that their development platform for mobile products is now surpassing those of Amazon and Google. Now, they still have a ways to go with Apple just because the momentum is behind Apple but Microsoft Azure is an amazingly capable development environment for mobile platforms now and it’s getting better all the time. And who would have ever though that they could close that gap in such a short period of time. Pretty amazing.
Microsoft’s Analytics Options in the New World [39:00]
So let’s talk about their analytics options in the new world and how this plays into healthcare.
The Transition Period [39:07]
We’re in this interesting transition period between relational databases and Hadoop/NoSQL “data lakes”. The reality is relational databases were never designed for analytics. They were designed for transaction processing. Data modeling came out of an attempt to model business transactions and business processes and reflect that in the collection of data and then the development of applications to support those workflows. We stretched relational databases so that they would work in analytic environments and they’ve gone okay but it’s really not a natural thing for them. The big IBM mainframe flat files, MBS and all that of the old school days, they were actually better at big data than relational databases. But we’ve long gone the relational databases, we made them work, and they’ve been okay. They’ve been pretty good. But the reality is they’re on their way out, and Hadoop and NoSQL is on its way up and that’s a good thing – because data modelling stands in the way of good data analysis. That’s the real challenge that we have with relational databases. The more you model, the harder it is for you to analyze in a relational environment, that data. Hadoop takes the restrictions of that data model away and now you have kind of the open freedom to relate data in any way you need to.
So, the challenge here is we’re in this transition world and we’re going to have to figure out a way to straddle both worlds both with our skill set, as well as with the technology platforms that we’re offering. And Microsoft has done a really good job with that.
Microsoft’s Analytics Product Lines [40:53]
So there are about 5 different general product lines in Microsoft stack. There’s good old reliable SQL Server. It does very well. There’s nothing wrong with it. Had lots of success with that in analytics. There is old reliable on steroids, which is the Parallel Data Warehouse Platform for SQL Server. There’s the very cool hybrid architecture that matches SQL Server with Hadoop under the Analytics Platform Services environment. And then there’s the emerging future of computing under Azure. And PowerBI is the visualization layer, the manipulation layer and tools that can lay over any of these products. And I put Excel in here on purpose. We’ll talk a little bit more about that, but it’s still the most popular BI tool in the world, friends.
And so, if you have any hopes of ever getting Excel out of your environment because it creates problems with data quality and it dilutes the source of truth and all of those kind of philosophical discussions, you’re tilting at windmills. You’re never going to get rid of Excel as an analytics tool in any industry during my lifetime. It’s just not going to happen. The cool thing is Excel is improving all the time with very cool real plugins and add-ons. So it’s becoming a better and better tool, as well as the old standby like it always has been.
Price-Performance Numbers* [42:23]
While we’re talking about all these, affordability is really important. And I just want to mention that Microsoft is still the price-performance leader. And these are some of the big data, future data vendors that are in the space right now. And I want to thank Value Prism Consulting for putting this together. The numbers are a couple of years, well about a year and a half old, but they’re still accurate.
Price per TB User-Available Storage (uncompressed)
We won’t go into great detail here. The charts kind of speak for themselves. On just about any metric of affordability, Microsoft is considerably more affordable than Oracle, in the same ballpark as IBM and Teradata and EMC. But Oracle has just worked itself out of being a viable alternative really on a price-performance basis.
This is another way to describe that on a kind of ability to move data in and out of these platforms, how much does it cost on an I/O bandwidth perspective. And you can see Microsoft’s numbers are one-tenth what Oracle is and there was no data on IBM or Teradata. But the IBM and Teradata typically have beenvpretty expensive. One reason I have not ever embraced those platforms as a CIO.
Hybrid Architecture: Analytics Platform System (APS) [43:44]
So let’s look at the hybrid architecture here for a second.
Microsoft APS (Analytics Platform System)
A brilliant hybrid architecture [43:49]
The analytics platform system is a great combination of SQL Server and Hadoop. So it’s bridging that gap, that hybrid architecture between the old and the new. Very smart of whoever came up with this at Microsoft. I don’t know who led this effort but it was brilliant on their part. So what you have is relational SQL Server in a parallel massively scalable environment made right up against Hadoop. So now you’ve got the past and the future, all this non-relational data that’s going to start streaming in as we expand the ecosystem of patient data beyond the four walls of the organization and all these relational sources. For now, we’re going to be able to match those together into this deal of data that looks pretty seamless. And the cool thing is all of the typical development tools and system management monitoring tools that Microsoft has had that we’re familiar with now lay over this same infrastructure. Very very smart architecture. And then PowerBI is the layer that covers both. That’s a really cool architecture for the future.
Polybase Bridges The Skills Gap [44:53]
Underlying this is the ability to bridge the skills gap between the relational and non-relational worlds. So you can issue a select statement in SQL out to this translator called PolyBase that sits between the relational and non-relational worlds and you get the result set back as if Hadoop were a SQL relational data warehouse. So really key that the Microsoft has bridged that skills gap. It’s very cool. So again, brilliant product strategy on their part.
HDInsight = Hortonworks in Microsoft APS
HDINSIGHT / HADOOP Eco-System [45:27]
I won’t go into this detail. You can get this in the slide. But this is a further breakdown of all the different Hadoop/NoSQL components that come in that APS platform. Hortonworks has been one of the best Hadoop platforms from the very beginning. So great choice on Microsoft to partner with them.
Interactive Analytics Delivered to Any Device [45:48]
A very cool thing too is that you can, now with Microsoft on mobile devices, you can now deploy this out to just about any device. So Native apps for iPad, iPhone, Android, Windows, all of these analytics are now available to you on your handhelds and mobiles.
Azure: The Future of Computing [46:07]
So Azure, let’s talk about that. And I really do believe it’s the future of computing.
What is Azure? [46:14]
What is Azure? This is the Wikipedia definition. And one of the key missing concepts in this definition is that Azure is a hybrid cloud – meaning you can bridge data and applications between an on-premise data center and the Azure cloud. So you can have some data in your data center, you can have some data in the cloud, you can have some applications local, some applications in the Cloud, and that boundary between the two is becoming less and less distinct. There are over 3,000 applications in the Azure marketplace now and these are overwhelmingly business level apps, not consumer apps that we’re accustomed to in Apple and Google.
So Azure in short is infrastructure – CPU, storage, servers, data management tools that lay on top of that infrastructure. And then this very open development platform for applications and services, and it really is the future of computing.
Azure Analytic Services [47:20]
So this is a diagram showing how that might work, your data and your on-premise data center accessing and interacting with these Azure Analytic Services, returning that to your data scientists. So the future will be the ability to access these very complex analytic services so as you look, your data scientists don’t have to be programmers capable of developing machine learning algorithms or NLP algorithms anymore. You can take advantage of this emerging Azure Cloud and the services that Microsoft is enabling in that Cloud. But it looks like it’s all happening in your ecosystem.
Azure is Big and Mature [47:57]
I won’t go in all these numbers but trust me, Azure is big and mature. It’s way bigger than I appreciated when I first started researching Azure. They already have over a billion customers. Microsoft has invested $15 billion in this and they continue to build it out. I mean this is a big high throughput, highly capable infrastructure at every one of those three levels.
Huge Azure Infrastructure [48:22]
So to give you some idea of how big that Azure infrastructure is, these are the data centers that are located around the world. And if you get a chance, I encourage you to go out and watch one of the videos that shows the data centers. One of these data centers is the size of 100 football fields laid together. So it’s giant. Really it’s big big and amazingly capable data centers. And I might add that Microsoft has been twice as many and six times as many different regions across the world than Google and AWS. Microsoft has really expanded. And one of the cool things about that is if you have clients that have unique compliance and security and auditing rules, Microsoft just had to deal with that across all these different regions. So for instance, you know, eventually you might want to, if you’re a vendor, work with clients in South America or Canada. Well, Microsoft is going to provide those local data premises so that you can do that.
Azure Marketplace [49:29]
So the marketplace looks very interesting. And these are just screenshots from the website. You can get access to virtual machines, and I might emphasize, as you can see, that this is not just Microsoft products. You can spin up an Oracle Database in the Azure cloud. You can spin up Cloudera. You can spin up Barracuda, SAP, DB2. These are all business level apps that are not necessarily Microsoft, largely not Microsoft, but you can run these now in the Cloud instead of a data center. And I can tell you that CIOs who constantly have to manage and design these data centers or causing the expansion, I have no interest in doing that anymore. None. Zero. And this is the sort of thing that the CIO of the future will be doing – is interacting with vendors and applications at this level, not on your own data center anymore.
Application Services [50:25]
There are also application services. For development of apps, there’s APIs. There’s Azure Active Directory apps that allow you to do different things. There’s access to Dropbox. And you can see how open this is. This is not just Microsoft. This is like a cafeteria, like a smorgasbord of business enterprise services and applications that you can access in the Azure Cloud.
Web Applications [50:51]
You can develop and maintain web applications. Microsoft Dynamics will soon be coming on, so that ERP. And then there’s these generic data services. I want to work with Microsoft to add person merging and identity management in here. They don’t have that yet. Big problem for us in healthcare.
Machine Learning [51:14]
One of the cool things is, again, the commoditization of all this technology is what’s really amazing. And one of the things that they are commoditizing is the Machine Learning capabilities that we could bring to healthcare. And so they have a very nice kind of wizard-driven how to choose an algorithm in Azure Machine Learning. So you can literally go through this wizard. It will help you choose which model and which algorithms to use and you can upload your data and you can return the results of that.
And these are the models and the different algorithms that they have available in that environment right now. It’s very very robust. This is the sort of thing that you would have in the past had to have had Ph.D. on staff to provide this capability for your organization. That’s not the case anymore. It’s really cool.
Azure Security, Privacy, Compliance [52:04]
For those that are concerned about security, privacy and compliance, and again you’re going to have to take this opinion in context. I’m a former Air Force CIO, worked for the NSA, CIO in healthcare. I’ve had to manage top secret code word information. I’ve had to manage the most sensitive information in the United States. There isn’t a snowball’s chance that I could replicate the skills and the capabilities that Microsoft has implemented for security, privacy and compliance in the Azure Cloud. No way could I replicate that in my own data center.
So if you think that your data is more safe, more protected in your data center, I really think you’re kind of an old school thinker. There is no way that you could protect your data as well as Microsoft is protecting it in the Azure Cloud. Amazing.
The Visualization and Analysis Layer [53:08]
We’ll start wrapping things up with a little bit more discussion here about what the visualization and analysis layer looks like, and that is, as I mentioned, this distinction between what’s on your desktop and what’s in a Cloud is going to start becoming less and less capable. I use Dropbox and Dropbox looks like a drive on my local computer. I really have no clue that it’s out in the Cloud. You’re going to start seeing more and more of that. There’s a commoditization to these analytics tools in the Cloud that make really powerful capabilities available to all of us. So you can do things in PowerBI, then interact with your desktop Excel environment, and then you can push that out to mobile and SharePoint environments as well.
Natural Language Queries [53:50]
And one of the cool things here is Microsoft’s working in and they’re borrowing from Bing to do this, natural language queries, so that you can actually enter into this tool they call Power Q&A natural language queries and they’ve taken the machine learning algorithms and the natural language query from Bing, they’ve matched it up with Power BI and you can return the result sets.
Now, you know, the truth is it’s kind of Gee-whiz right now and it would only be appropriate for maybe a CEO that had no data analysis skills or something and wanted to ask a very simple question of your data set. But it’s a really cool start and it’s one of the best starts of this kind that I’ve ever seen.
Closing Thoughts [54:36]
So, we’ll start wrapping up here. Again, I just want to emphasize that business moves at the speed of software and I would claim that older C-levels don’t generally grasp this yet, and I’m old, so I can say that. But I do see this as a generational issue, where a lot of older generation C-levels had decided that technology is not important to them. They’re delegating that. Well the reality is your entire company runs on it right now and your ability to transform is dependent on it. So you better get involved and you better learn some of this technology, especially around the big data in analytics.
I’m not easily impressed when it comes to IT vendors, especially in Microsoft, but I am impressed and I think that Microsoft is going to be one of the biggest cultural and technological re-toolings of all time and it’s going to disrupt organizational IT strategies and the role of the CIO especially, in a good way.
And that concludes the slides here, friends. We’re just about out of time, but as usual I will always stay after the hour and answer questions. After you do that, Tyler, anything that you want to say, friend, or shall I jump right into the questions?
Before we jump in, I’ve actually went back and I’ve got the historical data on that Analytics Adoption Model question that we asked. If you’re interested in hearing, you’re absolutely right, where we do [crosstalk]…
You know what, friend, what I would like to do. Can we send that out in the email to the attendees afterwards rather than talk about it today? Can we send it out in the email?
We’ll get that out in the email to everybody, so they can see that positive movement.
Okay. Awesome. That would be great. Okay. I’ll answer a few questions here.
|I’m interested in machine learning in medicine. Can it help with diagnostics and such, classified and annotative medical images?||The first real foray into machine learning in healthcare came in image classification actually. And so, yeah, I think there are quite a few products now on the market that help radiologists, in particular, with image classification. That was one of my first entries into healthcare as well.One of the challenges of a machine learning in healthcare is that we have no outcomes data, so quite often we don’t know how to perform supervised learning on those algorithms because we don’t know what outcome we’re trying to train to. And we don’t collect data on healthy patients. So what we’re trying to create are healthy patients. So we collect zero data on healthy patients right now. We only collect data on sick patients when you go into the doctor. Every once in a while, in the physical, once a year, once every two years, there’s this very thin slice of data, if you’re healthy. But generally speaking, we don’t know what we’re training these machine algorithms to yet.
Now, there are some phenotyping and pharmacogenetic machine learning success stories and I think that you’ll see that take off considerably.
|Comparison between competitors like Tableau, QlikView, Tibco.||If I had to grade Microsoft right now on a scale of 1 to 10 in comparison to Tableau and QlikView, which I consider as the market leaders for general desktop data analysis, I would give Microsoft a 7, I’d give Tableau a 10, and I’d give QlikView a 9.5. So Microsoft has some gap to close there but they are also very affordable. So from a price performance standpoint, and integration with other tools in Excel and that kind of thing, you could argue that Microsoft is pretty competitive in some use cases.While I’m on that topic, I want to mention, you know, a lot of organizations still believe that you can force all analytic use cases through a single platform, like QlikView or Tableau or Cognos or Business Objects, and that’s just not realistic. There are going to be multiple types of analytics and consumers of analytics in your organization and you have to support at least three different tools and those tool categories are the general category that I just mentioned – Tableau, QlikView, Business Objects, etc.
But then you also have to support statistical tools, Lite Fast and SPSS. Status. And then you’re always going to have to support Excel, I mentioned that. So get used to it and be open to that. There’s a lot of analytic values supporting Excel. And more and more, we’re starting to see people starting using R and Revolution analytics, one of Microsoft’s recent acquisitions is really good at exposing our libraries to non-programs.
|IBM and the new acquisition with Watson and if not combining Phytel, Watson and Explorys.||Well I think it was a very smart move on IBM’s part. They’re IBM data model, healthcare data model, has been struggling. It hasn’t done well in the market. It doesn’t adapt well because of that old relational mentality. You spend all your time mapping into that model and never were actually getting anything out of it. So I think Explorys has had good success. They’re Hadoop-based. They offer some very solid products. Phytel is a great tool for sort of care management, patient engagement. So I think it’s a great move on IBM’s part. I worry that it’s going to be consumed in the IBM machine and it would be interesting to see if that happens or not. IBM has a reputation for gobbling up companies and then sort of beating them into submission. So it will be interesting. I’m not sure.|
|What would Steve Jobs do?||I don’t know. What would Steve Jobs do? He would sell an electronics health record on iTunes and then it would be captured forever in the closed ecosystem of Apple, I think. Tongue in cheek.|
|Microsoft’s willingness to sign BAAs.||Yeah, Microsoft is going to do that. I talked to Microsoft today. They are having some challenges around their BAAs. They believe they have to sign a BAA for every service in the Azure Cloud, which I think is wrong. I think the BAA is not between organizations like a healthcare organization and a service in the Cloud. It’s between business entities.So I’m going to spend some time with Microsoft getting them to change their approach to BAA. So that’s between Microsoft and the healthcare organization, not between products like PowerBI and Azure and the organization. That’s a crazy strategy, as far as I’m concerned. Maybe they’ll change my mind but I don’t think so. But yeah, they’re willing to sign BAAs. They’re just a little convoluted about how to do it.
|Comments about health records making.||It’s a little different than this topic today but he’s a physician, general surgeon who struggled as a physician with having complete health information. And I think the only way that’s ever going to change is for the consumer to own their health record, and then we become the broker of that information and we grant access to care providers and extended care family and that kind of thing. I think as long as EHRs are healthcare provider-centric, there just aren’t any market incentives right now for those EHR vendors or those organizations to share data. That becomes a proprietary relationship between the patient and their data. You’re reluctant to change providers because that means you have to move your data and that’s you could argue that it’s kind of monopolistic when it comes to data. So until we get to a patient-centric EHR where I decide you have access to the data and I grant broker drives for that, then I think we’re going to have troubles with this.|
|How are you handling PHI or HIPAA compliance with processing healthcare data, processing speed with unstructured data and any issues?||Well those are kind of big questions. You have to have algorithms in your analytics environment for de-identifying data and keeping a brick wall between identifiable and de-identified data in the analytics environment. There are tools and techniques for doing that. It’s been around for a long time.Processing speed with unstructured data. We found that providers don’t want to wait for analytics. Well, it’s usually an issue with memory and CPU throughput and that kind of thing, indexing. But if they’re waiting for their analytics, I agree, they’re not going to wait long. So you’ve got to solve that processing power and that’s kind of a sheer hardware horsepower issue that’s worth of, you know, that’s a separate kind of topic, I think.
|Any issues getting multiple proprietary data sets plus pharmacy data into one data sort?||Well that’s what we do and that’s what enterprise data warehouses do – is we take all of that data from multiple provider data sets, we link it across the patient, and that’s the whole idea – is to get that 360 view of patients and that’s fundamentally what Health Catalyst does. That’s fundamentally what a data warehouse must do. So plenty of precedents for that.|
|Why are you preaching Microsoft?||Well hopefully that’s clear enough. Sorry if I didn’t make that clear more upfront. No, they are not sponsoring this. We get no remuneration for this at all. This is as much a story about how kind of amazed I am at the transformation of Microsoft as I am anything else. But it’s also just generally advocating for what I think is the best platform in the industry right now. And I wouldn’t do this if I didn’t believe that. You could say, well, of course I’m advocating this because it’s important to Health Catalyst. I wouldn’t be advertising it if I didn’t think it was best for the industry and I really honestly, sincerely and genuinely believe that Microsoft is offering the best path forward of that enterprise data warehouse level than any other vendor that’s available.|
|HIT is behind the curve on master data management and data governance.||Yeah, we agree with that and I’m working kind of independently but also working under the hospices of Health Catalyst to improve data governance and hit that sweet spot between too much and too little data governance.Danielle Byron mentions that one of her tweets last week was people want little data, not big data. And Danielle the earlier slides I touched on. There’s notion of volume of data versus value of data and why I think we do get kind of backwards on that. And I don’t know that Microsoft is doing anything really to right-size big data. I don’t know really what they could do. I think that’s up to us.
|What hybrid backup programs are in place for predictive analytics in the current healthcare industry setting?||I’m not sure I’m understanding your question there. You asked what hybrid backup programs are in place for predictive analytics in the current healthcare industry setting. I’m sorry, friend, I don’t understand your question. So I won’t even venture to answer that or I’ll do something wrong.|
|Providers here are still learning how to enter data in Epic. That is probably why they are still not focusing on how to get data back out for the most part.||Yeah, I agree. We’re still kind of early. We’re barely getting our handle on EHR use and data collection. We haven’t had time to think about analytics much. But I would also argue that now is the time before you get too embedded in data collection processes and principles. You start thinking about the analytics you want to get out and you work backwards from that and start tuning Epic accordingly, so that your data collection processes are reflected of what you want to get out of it.|
|Business Intelligence Analytics and Informatics are pigeon-holed by various groups and these groups mostly see that these terms are very interrelated. How do you review these terms and the audiences that should pay attention to these terms? These clinicians are just interested in the informatics term and may be the reason they aren’t attending this webinar.||Yeah, that could be. I always say that analytics is the verb, data is the noun. And so, analytics is what you’re doing to that data to understand and take action upon it.Informatics, I mean no industry ever uses the term informatics with healthcare and that’s just – you know, I don’t know why we decided that we have to have it but it basically means sort of this blending of an understanding of healthcare with an understanding of a data of healthcare. But it doesn’t necessarily mean analysis and typically it’s been associated with someone who works with an EHR, like a physician champion and a physician informaticist. But I don’t see informatics and Business Intelligence and analytics being interchanged very much.
But pretty much I use Business Intelligence somewhat interchangeably with analytics and I think that’s pretty common.
|All this data geeks using Microsoft on the presentation today.||Yes, that’s true.|
|What would be the source of your negative expectation?||I assume you’re talking about Microsoft there. I’m not sure. And I don’t know what the source would be other than probably, you know, I’ve had some bad experiences with Microsoft over the years. So I think there’s lingering bad feelings from those bad experiences but they’re starting to fade.|
|We had a similar path in a different way. I started as a dental officer in Osan Air Base and currently working with Intermountain folks on DHMSM project. 1982 at Osan.||Oh that’s interesting. Very cool. Glad to know that.Yes, I’ve been at Osan. I received 135 recon planes – flew out of Osan.|
|Would a product like Amalga would be more successful if started today rather than 7 or 8 years ago?||That’s a good question. I mean Amalgav is definitely a better product care time actually. It’s now a better product than it’s ever been but it may be too late to the game. And I’ll cut myself on the back here. If the amount of folks would have listened to my suggestions about how to evolve that product early on and how to position it in the market, I think they would have had an opportunity to grab and be a very useful part of the market. But I think it may be too late for them now. But to their credit, it’s a better product than it ever has been. So good job on their part.|
|Getting into a doctor’s workflow is very hard. EHRs don’t like to work with HIEs to query for available data which makes access to Loop A difficult.||Yeah, we agree, very much. The APIs that surround EHRs are not easy to work with. So it’s hard to pry those open and actually insert the analytics into that context. Now, xG from Geisinger, Cleveland Clinic, Duke, a few organizations are starting to use Fire to open up those APIs and write applications that are starting to enable closed loop analytics. So I think we’ll see more of that in the future. I think we’re at a good point now.|
|How do you define the point of care for Loop A in that diagram?||Well that’s a great question, friend. For the most part, I’ve been defining it as the traditional point of care in the clinic or the hospital. But in the future, we have to change that definition. It has to be the point of healthcare decision-making and that may be the device that the patient is holding in their hand and it’s encouraging them to adhere to their medication protocol. So I think that definition at the point of care is going to evolve. Right now, all I’m really interested in is improving the state of analytics and Loop A within the EHR because we have so much opportunity there. We’ll evolve to these other external sort of non-traditional points of care later on, I think. But thanks. Good question.|
|In terms of getting the closed loop analytics at the patient bedside, will companies like Epic allow external algorithms to write back data to real time data in Chronicles or will Epic only allow BPAs to be built internally with Epic?||I don’t really know what Epic’s stance is on this. So far, the only thing that I’m really seeing is Epic allowing BPAs. But I have a feeling that Epic, especially as pressure mounts from the industry, they’re going to have to open the API so that you can write back to chronicles directly. The folks that might be better at this asking this would be Stanson Health and companies like Health Bench that have been working around the edges of Epic’s API. But there are programming points that allow you to get access to Chronicles. But whether Epic will allow third-party vendors to do that or not or how easy that will be, I’m not exactly sure yet. But that’s something I’m going to start spending more of my time on.|
|Clinical notes often contain a lot of high value, low volume data.||Well they offer high value in an analogue perspective but not in a digital perspective. And what I mean by that, it’s computable. So we still have a very hard time with NLP making discrete computable sense of a clinical note. So they are relatively low value and they do have very low volume and they certainly have high value to clinicians. They just don’t have high value to population health management right now because we can’t compete with them very well.|
|We have a decent sepsis EPA in Epic but it’s only as good as the data contained within Epic. How might we open up what Epic can consume externally?||But I think that kind of gets back to what I mentioned earlier. There are some things happening in the industry around Fire that I think will allow for that.|
|Are healthcare organizations getting more comfortable to consider using offsite products like Azure with PHI data?||I think they are but I think there is still this misunderstanding among a lot of healthcare CIOs that the Cloud, especially the Azure Cloud, is less secure than their data center. And if they had a chance to see what I’ve seen and if they have the benefit of my background in secure computing, I think they would step back and say there is no way that I can replicate the security and protection that Microsoft has in their Cloud down the hall and my data center. There’s just no way. So I think we’re going to start seeing that more and more, and I think forward-thinking CIOs and CMIOs will start taking advantage of this. And who needs another data center, right? But it’s a commodity. We don’t need to be spending healthcare dollars on data infrastructure that we can rent and buy in the Cloud.|
|I agree with you about the aspects supporting faith and trust. I also have more faith and trust in companies that have been found guilty of monopoly practices, which so far only means IBM and Microsoft. I’m certain that IBM and Microsoft companies know they have to be careful on misusing monopoly or oligopoly power or they have a shorter path than others to penalties with US and European Union and companies that have not been convicted, i.e., Apple, Google, Facebook, etc.||Yeah. Good point. Yup. The monopolistic practices of – you know, they were frankly encouraged by Ballmer and Gates. Microsoft had to pay a price for that. And Google, you know, is going through some anti-trust issues in the European Union right now too. But yeah, that definitely sharpens your game when you go through those.|
|Is there any reason you don’t have Google in there?||Well Google so far really still offers just kind of basic infrastructure, utility computing for the most part. They’re starting to offer some services but nothing on the level that Azure is offering. So I don’t really consider Google in the same space as Azure. Azure is definitely a superset of services.|
|Where do you see Microsoft Dynamic CRM fitting into this model?||I think you must be asking about the Healthcare Analytics Adoption Model. It would fit in a small piece of level 3, probably a level 3 reporting, that internal reporting, for your automating basic internal reports for your CRM function but it’s definitely not an enterprise data warehouse. It’s just a small little niche program there.|
|Can you use PowerBI over Azure cable storage?||If I understand your question correctly, yes, you can. Azure cable storage, I’m not exactly sure what you mean that way but yeah, you can use PowerBI over Azure’s data content.|
|Do you have any opinions on using Open Source Weka versus Azure from machine learning?||I would say that Weka requires a different skillset than what Azure machine learning would. And with Azure, you’re going to get sort of basic services, pretty advanced services actually, but they are going to be the algorithms and the models that are available. Whereas with Weka, you can kind of develop your own. So I think it depends on your skill and what you want to do. For someone like me who doesn’t have really advanced machine learning skills but I want to take advantage of it, machine learning in Azure would be more appealing than using Weka. But for someone like David Crockett, our Ph.D. machine learning guru at Health Catalyst, I would suspect he would prefer to use Weka and on occasion use Azure when it’s convenient.|
|What are the business models and pricing for these analytics should we, as potential customers, think about assessing value?||They’re generally quite a bit less expensive. The only thing is they do operate for the most part, a lot of what I talked about today, on a subscription basis. So sometimes I think we’re all going to be – I mean I can’t even tell you how many subscription services I belong to now and I feel like I’m going to be paying someone for the rest of my life but that’s the reality of the new world. The old days where you bought a license, you own the license, and you pay sort of the maintenance fee for upgrades after that and then if you wanted to cancel upgrades, you could opt out but still keep the license. I think those days are pretty much gone in computing.So these tools and the licensing fees, especially around PowerBI, are generally quite a bit cheaper than the other products I mentioned but you’re going to pay forever. So if it is the total cost of ownership, the Microsoft products would still come out probably cheaper but not dramatically.
|How is small data different from operational data on a day-to-day basis?||I don’t really know. I’m still trying to get my arms around what small data means. So I don’t really know how to answer your question there.|
|How does Health Catalyst plan on integrating its data model with Azure?||We’re still working on that actually. Most of our clients now host on premise. We do have an in-the-Cloud service for hosting that. It’s our own Cloud. It’s in our own data center. But our CTO, Doug Adamson, is constantly evaluating all this and I would guess in the not too distant future, we will probably start offering an Azure-based Health Catalyst platform for all the reasons that I just mentioned. But I don’t want to speak for Doug. He is the CTO in charge for those kind of decisions. So I will defer to him. But it’s something we need to start developing more formally. Right now, we’re just kind of in the conceptual thinking mode but we need to publish this and be more firm about likely kind of roadmap and definitely have a better message to the market, so the market understands.|
|What do you think is the greatest inhibitor to leveraging analytics at the point of care?||Well I think conceptually it’s still a new thing, so there’s a conceptual learning curve that we have to go through. People often become used to it. And generally when I put that diagram up, people go, “Oh, you know what, this makes a lot of sense,” especially when you talk about physicians being 10 times more likely to adjust.But I can also speak very tactically and tangibly having tried to do this when I was a CIO at Northwestern and a little bit in the Cayman Islands. It’s complicated at the API level because the Epic, Cerner, AllScripts, other APIs, NextGen APIs, they’re pretty close. They’re hard to get to and they were never designed for this sort of thing. So it takes a pretty good skill set to do that.
Now, like I said, there are new tools, like Fire, and there are people and companies that are starting to figure this out a little bit at a time. So I’m optimistic now that a lot of those hurdles around the API are going to start to fade away. But I think the greatest inhibitor is in sort of this closed API around the EHR vendor products, sometimes contextually closed, sometimes technically closed, but closed being the action verb there.
|Am I to understand that Azure will never be your replacement for Excel or will we always have Excel?||No, I think you can have Excel in the Cloud right now. So to some degree, it already has been replaced by Azure, but not quite as functional. But I think eventually, yeah, I think Cloud-based tools on the desktop is where we’re headed. And so eventually I think we’ll have really really think clients running very robust applications in the Cloud.|
|You mentioned data storage and I/O way back computing power. Does Microsoft have solutions to such as StarCluster that creates a virtual cluster on the Cloud? If so, how cheap is the computing power compared to Amazon AWS and what options are there for machines to pick up or create the virtual cluster?||I could probably give you a type two answer there, type one being very clear; type two being a complete guess. I think it would be better for me to give you an answer from Microsoft on that. So when we did a price comparison recently at Amazon AWS, Azure services are cheaper on a utility computing basis than Amazon.Now, I can’t speak to what that means in virtual clusters and what you mean by virtual clusters – I mean clustering in virtual machines are highly available. I mean that’s kind of the whole basis behind Azure at the computing level. So if you have any further questions about that, I should probably put you in touch with Microsoft.
|How do you see wearables fitting into the data capture for a healthy individual?||Well it’s going to happen for sure. Like I said though, our ability to consume data goes up by X to the tenth, our ability to analyze data is progressing at best X to the cube and our ability to actually do anything with that data is still linear. So I think that’s what you’re going to see. We’re going to start consuming data from wearables. Lots of organizations are already starting to do that. That’s a novelty. It’s kind of Gee-whiz. We don’t really know what we’re going to do with it yet, how to analyze it and we don’t know how we’re going to intervene or do anything to help a patient with it yet, and it’s going to take years for that to happen. But, it’s good to see that the data is starting to be collected.What’s really going to be helpful is when we have laboratory results through transdermal lab results that we can upload. I think that’s going to be more important than knowing how many steps I took.
But in some settings, knowing if a person is taking any steps, an elderly patient, congestive heart failure patient, knowing that is very good, but where do we surface that in healthcare right now? There’s no command center sitting around watching that wearable data really. So, I’m not sure exactly what value there will be yet. There is probably more value in the social engagement that those wearables create than in the healthcare engagement right now.
|With the high remodel, how can the developers, users, etc. discover what their data means, such as data dictionary, and where to store it so all can access?||Great question. Yeah. Right. What does the metadata repository look like in the Cloud? So, you know, I have to say I haven’t given that much thought and I’m going to do that. I think that’s a great question. I think what I’ve always assumed is that you would lay a metadata tool over the top of whatever your Cloud looks like but you know, it’s certainly going to get more complicated. That metadata repository certainly is going to become more complicated, especially as we cross these relational and non-relational world. So that’s a great topic. I’m going to write that down and say what does the metadata would look like in the Cloud and the hybrid architecture.|
Okay. My gosh. There’s a lot of good questions here. See, I get a great question like that and then I get addicted and I don’t want to stop but I have to stop. I’m going to breeze through these just really quickly. Sorry that I’m not – Yeah. Gosh, we’re running out of time. A lot of great questions here.
Tyler, let’s see whether we can capture these questions, friend, before we close the – can we do a copy and cut and paste?
Certainly, we do collect all of the questions. So I can make sure, Dale, that you’ll have a chance to go through all of them after the webinar is over.
Okay. Great. Well I think I better stop now. Thanks everyone. Have a great day. And Tyler, back to you, friend.
Alright. Thank you so much, Dale, for all that great information. Now, before we close the webinar, we do have one last poll question. While we rarely focus on Health Catalyst, we have had many requests for more information about what Health Catalyst does and what our products are. So if you are interested in having someone from Health Catalyst contact you to schedule a demonstration, please take the time to respond to this last poll question. Now, shortly after this webinar you will receive an email with links to the recording of this webinar, the presentation slides, and the poll question summary results. Also, please look forward to the transcription notification that we will send you once it is ready.
On behalf of Dale Sanders, as well as the rest of us here at Health Catalyst, thank you so much for joining us today. This webinar is now concluded.
Thanks everyone. Bye bye.
[END OF TRANSCRIPT]