Making more of health care data12/01/2018
Four local execs discuss what’s working, where they’re looking to do better still and how Nashville’s practicality could use a little pixie dust
The Post team in late October welcomed about 200 people to City Winery for a breakfast and panel discussion on how care providers, technology companies and patients are making better use of health care data. The gathering — which was sponsored by DVL Seigenthaler, a Finn Partners company, and organized in partnership with the Tennessee chapter of HIMSS — featured the following speakers:
- Joshua Douglas, CTO of Bridge Connector
- Gillian Hooker, VP of clinical development of Concert Genetics
- Kevin O’Hara, CEO of Syus
- Elizabeth Ann Stringer, Chief Science and Clinical Officer of Axial Healthcare
Their conversation was moderated by Michael Heinley, a partner and deputy of the New York Health Group at Finn Partners. Here are lightly edited excerpts from the morning.
Heinley: Let’s jump in. When it comes to using health care data more effectively, let’s talk a little bit about what’s working well these days. Even just a few years ago, some of this technology and the amounts of data that we have and the way that we manage it was not even imaginable. Josh, how is Bridge Connector making health care integration easier for health systems?
Douglas: Historically, data has been siloed; we hear that a lot. EMRs held all the data. And really, the new models of integration and integration platforms as a service have brought us forward and unlocked those silos so we can access data within those systems. Really going through the clinical workflows and using different pieces of software — not just the EMR — for different things has really helped unlock that data out of the silos.
Heinley: It’s sort of the dream of interoperability becoming a reality.
Douglas: It’s getting close. It’s getting closer every day, but we’ll see.
Heinley: Kevin, what about from your perspective? How is that working in the hospitals, the health systems? Are there any that are doing a particularly good job and connecting the dots?
O’Hara: First, taking a quick step back: The economic stimulus package that pushed adoption of EHRs got everybody pretty much caught up. And then over the last few years, you’ve had people migrating systems. And people are used to moving. When you move, you get things organized. So when people go through these migrations, there’s a lot of opportunity to clean data up and to get on new systems.
There’s also been — particularly in our area — a lot of hospitals might have had a kind of a best-of-breed operating room documentation solution and that’s now generally been absorbed, without any surprise, by Epic or Cerner. So that move over the last four or five years — to everybody being on an EHR, a lot of people being on a consolidated single system and then really getting down that Cerner and Epic and Meditech — has greatly simplified what data sets look like. I think that’s been helpful.
In terms of who’s doing well, I would say it’s kind of the usual suspects, the Intermountains, Cleveland Clinics. And some of them are systems that have been on these larger integrated systems for some period of time, and so you can almost see the future of what happens when you get very organized data, clean data, and you can get out of that realm.
People spend an inordinate amount of time just getting data organized and clean and it doesn’t leave them a lot of time to do something with it. I think the folks that are doing well have already passed through that over the past few years and are now applying the resources to doing something with the data as opposed to just trying to figure out how to make it say something.
Heinley: Are you also tapping into the EMR, EHR systems in terms of the data that you’re funneling in and out?
O’Hara: Yeah. We get data about essentially the who, what, where and when of a surgical case. That can help you answer the basic questions that you need to answer about how many operating rooms you need to run and how to allocate time and how to standardize. But what ultimately you need to start to do is tap into other systems — supply and implant tracking systems, systems that deal with HR data. So that’s really kind of the next step and companies like Bridge Connector and others that are helping to bring that data together and get it out of those silos, that’s really what’s important now.
Stringer: [There’s also] a sort of behind-the-scenes grassroots efforts. Think about data science: It’s a new field that’s sort of responding to the aggregation of large data sets. And you have some really scrappy people who are just hacking away and trying to make things work. I think that ecosystem is really, really important for then getting those ideas up and adopted to a more enterprise level. You’ve got so many people sharing information down at this grassroots effort and I think that’s really, really important.
Heinley: That’s a really interesting point. Does anyone have a perspective on maybe coming in as a square peg and trying to fit into a round hole? Because there are legacy systems and these are new technologies and platforms. Is that a challenge that you ever have to face?
Hooker: This is very much a challenge we’re facing in the genetic testing space. Right now, most genetic tests are ordered in the EHR using a miscellaneous code and maybe a free test field. But more often than not, there’s a piece of paper that exists in parallel to the EHR. So, there’s no tracking of whether a test was ordered, let alone what test was ordered.
And when tests come back, they come back as PDFs. They get uploaded somewhere in the EHR — often not in the same place every time. You could walk into some of the most sophisticated systems in the country and say, “How many of your patients have been tested for X, Y, or Z?” and they can’t answer that question. Even systems that have sequenced everybody in their whole system couldn’t say who was having a clinical test. So the tracking is really poor and there are poor standards. And that’s one of the things we’re trying to do: Build standards for tracking that through the whole process.
O’Hara: Just to jump on that: There is a strange dichotomy between really interesting things that are going on at a high level with AI and machine learning and these really low-level issues like things being entered into a miscellaneous category. We deal with this constantly. We’re introducing benchmarking now and just the names that people use to describe surgical procedures…
Even within the most sophisticated health systems and facilities, this is typically a mess. They’re just free text entries. Your typical community hospital may do 800 to 1,000 different types of procedures a year, with a knee replacement being one and an open heart procedure being another. And there should be 800 to 1,000 names. We have customers that have 18,000 different names and trying to determine what the average length of a total knee replacement is is sort of like getting out of bed and having to go find your clothes somewhere else. You can’t do the thing that you really want to do because you’re first hamstrung by trying to figure out which of those 18,000 codes describe a knee replacement.
Heinley: Josh, what’s your take on the challenges that you have to face in this space?
Douglas: I think there are two ways to solve it. One is with standardization, right? So you have code sets trying to solve those problems but they only go so far. We’re finding new things and constantly having new procedures and new tests — especially genetics that these code sets don’t typically cover — or there’s not enough expertise at the health care facility to apply the correct code.
So one way standardization is coming down is through government regulation. But the other way is through data and modeling and natural language processing and actually using technology to comb over the data and really learn what the intent is. And I think that’s probably the better way of doing it. I don’t know that the governmental regulation is going to solve all of our problems. It’s brought us so far and it’s good in some ways, but I think we’re really looking at technology to solve those problems.
Heinley: What about HIPAA and patient confidentiality? There are a lot of constraints around that, not the least of which is just data security. Elizabeth Ann, how much of an issue is that in your situation?
Stringer: This is a huge issue with the opioid epidemic and we always want to start with the patient, right? We’re all here in health care because we feel strongly about trying to help patients get better. So we also want to protect their information. But one of the problems with sharing information across systems is you do have some of these extra security and privacy rules. One of those, particularly in the opioid epidemic, that many people might not know about is that not all providers are allowed to know about diagnoses of the patients they’re caring for.
At Axial, we gather claims data so we understand about the diagnoses from multiple providers of an individual patient and we would want to share that with all the providers that are touching that patient. But we’re not allowed in some states and in some markets to share information saying that this patient has opioid use disorder. So a doctor might be trying to make a decision about the best course of care — is an opioid appropriate or not? — and they aren’t allowed to know that that patient is in fact diagnosed with opioid abuse disorder.
These are some of the challenges that we’re up against in solving the opioid epidemic. But [it’s similar] in health care more generally, right? What is appropriate for a provider to know? Don’t we want to give the provider that holistic view of the patient so that they can make the best decisions for that patient?
Douglas: The HIPAA law is 300-and-some-odd pages. And what you’re getting at, I think, is also patient consent, right? Having to deal with patient consent across integration of data — whether it’s from HIEs or whether it’s coming out of EHRs — consent’s a hard problem to deal with. And each state has different laws — is it opt-in, is it opt-out — and it’s a really complicated problem to solve and everybody’s worried about it. Nobody wants the Office of Civil Rights knocking on your door.
Hooker: In genetic testing, there are a whole host of emerging problems that aren’t even addressed by HIPAA. Right now, your genetic information isn’t considered personal health information. It isn’t considered protected even though — think about it — it’s inherently identifiable information. You can look at stories like the Golden State Killer who was tracked down using ancestry databases.
There are a lot of concerns that arise when it comes to the fact that your genetic information may not just identify you, but it could identify your family members. So how are we going to handle permissions in that context? And then it expands even more broadly to who owns that data. A lot of genetic information is being used for research and a lot of folks are buying genetic information from the consumer market — the companies like 23andMe — or from research institutions to use it for great purposes, for the development of new technologies, for the development of new drugs. But I think there are a lot of questions to be resolved around permissions and whether folks should be made aware when their data is being sold, whether they should have a say or not.
Heinley: Elizabeth Ann, you mentioned providers earlier and I’m really curious about how they are reacting as a prescriber, a physician, somebody who has taken the Hippocratic Oath, when you’re putting in front of them data that could be to the contrary.
Stringer: This is a question we get asked a lot. How are providers actually taking the information that you’re using? Are those friendly conversations? When we think about what’s best for the patient, it’s really about getting to the provider. How do we change that provider behavior?
At Axial, we have a model where we’re doing outreach to providers, educating them about their patient population. In these outreach efforts — phone calls, onsite visits — it’s really important to think about relationship building and gaining the trust of the provider, and a lot of that starts at the level of the data. When providers start to understand you have some patients that are really high-risk that probably need a different care path than they’re currently on, providers can get really defensive. I mean, I don’t want somebody to come in and tell me how to do my job, right? And so, I think the data can be one of those keys for helping us establish trust early on.
The data is true. We can’t deny that. And beyond the data, one thing that’s really helpful to us in the opioid space is the fact that providers feel like they’re under a lot of scrutiny right now. They are really looking for those trusted partners to help them do a better job of delivering care. It has to be individual patient-based analysis first and then we roll that up to the provider level.
Heinley: It’s the axiom that the data is only as good as how you can really use it and apply it in your situation.
O’Hara: Yeah. Going back to what Elizabeth was saying a minute ago, I think what you’re describing is that, in the health care and medical fields, there’s a lot of faith that providers, doctors and surgeons are scientists. Right? They’re trained scientists and if you put the data in front of them, they’re going to use the data and respond to it.
But one of my favorite areas of what we do is really also looking at them as human beings and understanding that there is a lot in the way you frame the data and present the data [that determines] how it’s going to be received and how it’s going to be used. As Elizabeth Ann was describing developing these relationships, I think there are a couple of interesting things: One is how do you do that in scale and also how do you use different communication methods.
In our area of operating rooms, there is this concept that you have a wall of fame or a wall of shame. Something that hospitals focus on is surgical cases starting on time. What percentage start on time? So there’s this idea that, in the surgeon lounge, you’re going to post either a wall of fame or a wall of shame. And our approach is to do neither of those things. You figure out how to provide them the information. You treat them like grownups. You talk about your partnership with them and how you’re trying to help them do their job better and you try to find some third thing. It’s framing it about the patient having to not eat overnight, not drink, showing up in the morning, their family’s there and not wanting to wait around. And then you translate it into something like that as opposed to kind of being about the hospital or the surgeon.
There’s a lot you can do with raw data. But if you have the best data in the world and you come at it completely wrong in terms of presenting it, you’re going to fall flat. There’s a whole realm around that that I think’s really interesting.
Douglas: One of the things that is important for me is to understand from a customer perspective — and that customer could be a clinician, it could be a registration person, it could be a surgeon — what they want to do. What do you want your new ideal workflow to be? Or how can we make this data actionable in your current workflow? I think that’s what we were getting at. You can collect all the data you want. You could even present the data. It doesn’t make it actionable. Tying into current workflows and delivering communications how doctors are used to getting communications is key in making that data actionable.
Don’t try to change what people do day to day — whether it’s a clinician or whether it’s a registration person. Find out what they do, how they work and then present the data as they work through their normal workflow. When you try to completely change how somebody works, especially in the health care space, their initial reaction is to say, “Whoa, I can’t do that.” Or like you said, Elizabeth Ann: “Don’t tell me how to run my business. Don’t tell me how to treat my patients.” Just help them treat their patients at those action points.
O’Hara: I think the challenge to that is that, to get improvement in the areas where our customers want to see improvement, it does require behavior change. Now, some of that may be just educating people, right? But how do you provide the framework for change to be easier? We look at a couple different models, and there’s nothing new in these areas but there are models for change management at organizational level.
There’s a really good model by a guy out of Stanford named BJ Fogg. He talks about individual behavior change and it’s basically, “How difficult is the change and how bad do I want to make it?” Then you have this sort of curve that you want to be on the right side of by making it relatively easy and making the desire to do it pretty high. And if you’re on the right side of the curve, you’re going to get change. And if you’re on the wrong side of the curve, you’re not. But what causes that to happen is ultimately a trigger.
What we looked at is how does the data provide a trigger. How do you put something in front of someone that triggers them to make a change that they probably already want to make and is pretty easy? But the challenge is that we do need to push people out of what they’re doing today or you’re going to get the same result that you got yesterday. But we do need to make it easy.
Hooker: That’s where we’ve looked to take a lot of lessons from the consumer space. There’s been a lot of innovation in [creating] interfaces that are intuitive and that folks can use. We’ve built a genetic testing formulary that looks like a marketplace. I would say it’s like organizing a grocery store where we’ve taken all the genetic tests that do similar things and put them together so that people can find them and they can really compare side by side the benefits and the limitations of this particular test over another. What are the cost implications? What are the insurance implications?
Stringer: One of the things I’d like to add is that to make data and analysis actionable, we’ve got to bring together multidisciplinary teams. Thinking about a marketplace, how do consumers consume information? We need experts in a lot of different domains and bringing them all together to solve a problem can be really, really tricky. So, in addition to the challenge that data provides, bringing stakeholders together in the design process is really key to making it actionable.
Talk versus reality
Heinley: Let’s talk about hype. When you are dealing with different stakeholders, with decision makers, with folks in this room who are representing their organizations, how do you convince them that this or that is a new platform that they need to develop? I understand that health systems switching over technology platforms is a long-term premise. So how do you separate that and how do you approach facilitating a decision-making process with these new platforms as the technology is rapidly advancing?
Douglas: When I hear somebody say “big data,” that is an indicator that it may be hype what they’re talking about. Most data is not that big. When people say “big data,” they usually mean something like a distributed database or they’ll use big data and AI in the same sentence and they’re not the same thing.
It’s hard to separate the hype from reality. I think there are some really good companies and some really good data scientists out there that are using things like unsupervised machine learning — which is kind of the next level of “Let’s let the computers and algorithms figure out their own algorithms and dig through the data to deliver some kind of analytics.” But how does that help a customer? How does that help a doctor make the decision? Well, generally, they can’t tell you, and that’s the key.
The question is, “Can you do human analytics first and deliver something valuable to the doctor to make some kind of decision on a patient?” That’s way more valuable than coming in and saying, “Well, we can deliver a big data and AI platform.” That doesn’t actually mean anything, right?
Hooker: We have a long history of hype in genetics. They even have their own name for it called genohype — instead of genotype. Remember when the human genome was sequenced and the announcement was made in 2003, there were folks saying, “Oh, in a couple years, we’re all going to have our genomes sequenced? We’re all going to be walking around with our genome on a watch, on a phone?” And here we are, 15 years later, and still maybe one or two million genomes have been sequenced worldwide, the majority of them for research or clinical purposes.
It’s taken a long time to get even to where we are in 15 years, and in many ways we aren’t where we thought we would be. In other areas, we’re ahead of where we thought we would be — in cancer and molecular microbiology and things like that. So, it’s a mixed bag. Now, bringing in the AI piece, I think it’s sort of adding onto of a lot of the hype that was already there. Through it all, though, I think that there’s a core group of folks able to work through it and continue to innovate in very pragmatic ways.
Stringer: As a scientist, I say prove it to me. And that’s something that I tell people all the time. Here’s a story that I thought was really interesting and sort of proved my point: I’d get asked all the time by investors and other leaders, “What are you doing to bring in new data sources and shouldn’t you be putting together legal information about somebody who has a DUI or something like that?” And it’s like, “Well, our models are pretty good right now. I don’t know how much more accurate you want them to get.”
But in the latest Stanford Medical Center magazine, there was a profile of a scientist working with a physician on a predictive model for patients that need to be engaged in palliative care conversations and end-of-life conversations. It’s a really great piece. And you think about Stanford and their access to the EHR and in pulling in data from there. What they ended up pulling in — the data they pulled in to train their model — was just this sort of claims-based data.
That’s the data I deal with so I found that very interesting, that I’m sort of challenged all the time about needing to integrate with EHRs and pull in the data from EHRs. And here’s, you know, Stanford — where so much innovation is happening — and they’re just using the claims part of this EHR. It’s something for us all to really consider. What are those validation outcomes? How much better we can get with these hype words?
Heinley: The proof is in the pudding.
Heinley: So let’s bring this back home, to Nashville. I’d like to get your take on the future of the city’s fast-growing, data-driven medical sector.
Douglas: I think it’s really exciting to see Nashville. I was born here and moved around a lot and ended up back here. I love this city. It’s exciting to see it shift from — and I mean no disrespect here — the old guard of health care such as HCA, CHS and so on and really get into the smaller, entrepreneurial, tech-focused companies delivering solutions for niche industries within health care. That’s what I get really excited about and when I come to events like these and meet people and hear about what they either are doing or want to do.
Hooker: We talk a lot about knowing Nashville is both physically and sometimes is metaphorically situated between Silicon Valley and Washington D.C. and how that gives rise, in my perception and experience, to a very pragmatic innovation. It says, “Yeah, maybe we won’t all be wearing our genomes on a watch six months from now. In Silicon Valley, they might. But how do we really innovate at the real-world challenges of health care? How do we use real-world data to drive decisions and how do we deliver systems that can really reach a lot of different people who are in a lot of different places when it comes to health?”
O’Hara: Using the word pragmatic is how I think about it. We can hold ourselves out as the real world. You know, we’ve got the HCAs and the LifePoints and large companies that manage a lot of hospitals. We know what the reality is on the ground. That can be limiting in a way, too, right? But we’re going to be more about iterative innovations, pragmatic innovations. And that’s sort of what we think about ourselves. But that makes us, I think, inherently less of kind of a moonshot town. We’re not hype-y.
We compete against some companies that are in Silicon Valley, that have raised tens of millions of dollars and they’re former Google engineers and former Uber engineers and their take is that health care is really broken. “The people who are doing it today are messed and they don’t understand how to fix it and we’ve got magic pixie dust and we’re good enough to come sprinkle it on you guys in Nashville and fix the messed up stuff that you’ve been doing for a long time.” And there’s some value in that, right? I mean, there’s this notion that, if you ask people what they want, they wanted a faster horse.
So there’s that inherent tension that I think we in Nashville have a lot to offer in the regard of understanding how this stuff lives on the ground. But thinking about how to get out of that trap a little bit to take bigger moonshots is also important.
Heinley: I saw a lot of heads nodding around the room as you were talking. Elizabeth Ann, anything to add to that?
Stringer: I think another important part about moving the technology forward is the money that it takes to do that. When I think about Silicon Valley and the way they’re able to advance and innovate, so much of that is predicated on the money that’s there and the investor community. We have some great investors here in Nashville but we’re nowhere near the size of the capital that we would need here to do the things that they do in Silicon Valley. That does limit us a little bit more.
And so there is more of that pragmatic approach. Data inherently is kind of sterile so how do you bring humanity to that? And there’s a connection here in Nashville with having such a large provider community, a hospital community that helps us really understand the health care ecosystem better and allows us to move forward. But it’s not going to look the same as Silicon Valley.
Via Nashville Post.