From DOCSF17: We’re following up about EVERYTHING from Dr. Hyde’s keynote presentation at #DOCSF17
A transcription of the entire interview is below:
Breanna Cunningham: Hi there Bri Cunningham here with Dr. Hyde with DRG who just gave a fascinating keynote presentation. I’m honored to be able to sit with you and excited to have you elaborate on some of these concepts that you couldn’t touch in the 15 brief minutes that you had.
One of the places I’d really like to start is HIPAA. You had mentioned that there were some HIPAA-isms and some things that you didn’t get to in your presentation that I’d like to hear more about.
Brigham Hyde: Let me start at a high level. I think protection of privacy is incredibly important for the patients. I think somebody had a quote earlier today about if there is no privacy, then data quality drops. I actually agree with that at a holistic level. I think that the law was written with that intent and unfortunately it may have had unintended consequences. I think the unintended consequences are that it kept data very separate.
If you look at the true intent of HIPAA, of Obamacare, of even GINA, it’s actually to help patients. But we’re not doing that by keeping data separate. My approach to this, and sort of aggregating US healthcare data, has been that I have to be indigent about what my requirements for data detail are, and I have to come up with an innovative way to still protect privacy while enabling long-term patient study in the data, track people over time in all the different data sets.
From that perspective, we set out a couple rules. One, I don’t want data that is aggregated at all. Meaning I don’t want it aggregated by procedure. I don’t want it aggregated to a physician level. I don’t want it aggregated to a payer. I want raw data. Raw 835, 837 and UV04. I don’t want it if it’s not that. You have to have that indignation because if you don’t you sacrifice the value of the analytical approaches. You cannot apply AI, you cannot do the longitudinal studies if you don’t do that.
Second thing was, okay, so I want patient-level data. That’s scary, right? I could hold somebody’s medical history. I could misuse it or sell it to somebody who misused it, I don’t want that.
We thought about a way to get around that and a way to sort of maintain safety and privacy while accomplishing what our standards were. The way we did it is that we work with third parties that now exist. The big one out there is Verisk Health. They’re public. We work with another one called Universal Patient Key. There are others out there.
What happens is that all my data vendors, and I buy data from over a dozen different sources, at the patient identified level like John Smith. They send all that information to the third party. The third party uses statistically valid, published and denoted algorithms to de-identify John Smith. Takes his name out, takes his address out to the zip three and any other information like allergies or things like that that might identify him.
We’re lucky because in the last 10 years there’s been a lot of academic study on what has to happen to do that. They take all that out but they give John Smith a key. John Smith gets the same key in all my data sets, so XYZ123, which is now John Smith.
Cunningham: Ah, okay.
Hyde: I can see across all the data sets which solves the US problem of fragmentation. This is a safe and secure way to do it. It also means that I never have HIPAA-level data on my servers. From a HIPAA data IT requirement perspective, it lightens the load.
Cunningham: That’s so great.
Hyde: This, by the way, is going to be huge at some point for the app companies and the consumer companies because they don’t need or want to deal with that. That’s how I’ve done it.
Let me add a little cynicism. Right now I know, I don’t purchase this data, but I know that I could buy, attached to that same key, the credit history, the online behavior and the social digital link to that patient. It’s available. You can do that. I don’t do that because we’ve not decided that it’s valuable and we’ve not decided that that’s risk we want to approach, but it is there.
Cunningham: Is there anyone that is doing it that you know of?
Hyde: I’m not going to comment on that, but it is available.
Hyde: To be clear, if you’re a patient out there, there are people that can now analyse your risk-taking credit behavior, your online comments, likes, follows, and potentially even what you clicked on on the web as it relates to your actual healthcare activities. The reason to do that, there could be a lot of reasons to do that.
Hyde: To market your products. Ideally it would be so you could find other interventions and correlations that might lead to a better outcome. Again, I’m not sure.
I’ll also make another comment. If you look at the app universe for healthcare, whether we’re talking about digital diaries, or we’re talking about simply Google and Facebook, they are storing behavioral PHI right now. PHI is personal health information. If you said “I’m sad” on Facebook and they’re storing that attached to your name, in the strictest reading of the law, that is HIPAA level data. I would argue that that has not been completely approached by the regulatory bodies. We know that there’s been some discussions, there was a study that Facebook did on its own members without an IRB looking at depression.
Hyde: They got a bunch of flack for that and backed away from it, but it goes back to what I said in the talk about Motorola. These data points dwarf the procedures, drugs, other encounters that are in the healthcare data, and may be just as valuable to looking at-
Cunningham: Sure they’re underutilized.
Hyde: Yeah, you talk about patient-reported outcomes. They’re a much better way of managing and measuring whether or not these people are actually enjoying what’s happening to them, it’s having a positive or negative effect. It’s just a matter of time until those worlds are connected if they’re not already. I’m in the camp that that is a great thing and I’m a little bit to one side on this, but I believe that data transparency ultimately creates value for consumers who provide it. I think healthcare is being held back because it can’t do that. My solution to that is in the interim, we found a safe and secure way to do that in the short term with the clinical data which is step one from my perspective.
Cunningham: Two questions for you.
Cunningham: One, and they’re two very different questions, one is why aren’t more people talking about this? I feel like I’m pretty in the know with the industry and I haven’t heard about these clearing houses if you will.
Secondly, do you think that the change in administration might open up some of these doors, or do you see any impact that that might have on this kind of taboo subject of HIPAA and PHI?
Hyde: I’ll give two answers to that. I think first of all, the clearing house as you referenced, which is one of the sources that we use, are very interesting companies. Often private held and they’re not really more than holding companies for what is essentially an automated exchange. I’d sort of liken it to sort of the back end of the credit card industry. It’s transactional in that way. There’s very little investment in R&D and trying to develop new things, that may be changing, but in essence, it’s a utility, right?
I just don’t think it’s one that people are aware of because the providers don’t really know about it.
Cunningham: We don’t see them at these kind of conferences.
Hyde: Right. The payers have their own view of it but are very protective of the decisions that they have to make on an actuarial basis. Actuarial health is not very exciting, right? It’s not warm and fuzzy.
Hyde: It’s often very effective, I’ll point out. It’s not something that’s talked about for those reasons. I think another side of it that I really believe in and I’m very passionate about is that the consumers are not being empowered, if they had this information, to make any decisions. If you think of the Mylan EpiPen, the thing that happened last year.
Hyde: It’s a very important story, not because pharma is evil, it’s very important because the payers started forcing patients to pay six times more out of pocket over a three month span for EpiPen than they had before. The Mylan price was pretty much static. It was growing but it wasn’t like they raised the price.
What changed was the payer said, “No, no, no, you’ve got to pay for that extra one you want to keep in your minivan.” That’s why the price went up and that’s what led to consumer outcry.
What was missed in that whole story is that the consumers said, “Oh, I don’t want to pay $600 for an EpiPen,” and they pushed back in a very market way on the supply demand curve. They said, “No, I’m not paying that much. I want more innovation. I want a generic EpiPen and oh by the way, I want you to give me a rebate.” The minute the consumer was engaged in the purchasing process, we had innovation and price correction, the second that happened. We’re at an orthopedics conference, how often does the patient even know what is being paid and to whom for orthopedics?
Cunningham: I would say probably less than 1% of the time.
Hyde: That’s right. The people who are paying all cash because they don’t have coverage and that’s very very few, or doing it somewhat electively as their insurer would consider it. That’s a very small portion. I think, and this goes back to the data side, that the real tipping point is when the consumers start paying enough out of pocket to care, it will immediately kick off this whole data revolution.
Now, when does that happen? Whether we stay with Obamacare or not, put it this way. Trump’s not going to give anybody any money back on paying out of pocket, so Obamacare has led to a rise in out-of-pocket cost, that’s going to continue. At some point, that’s going to lead to the consumer being engaged and the rise and demand for information about purchasing, which is what’ll really kick this off. Then people will know what those companies are because they’ll want to know.
Cunningham: That’s so interesting.
Cunningham: I’m going to change the subject here and talk about data ownership a little bit. Sensitive subject, right?
Cunningham: Is it the hospital that owns the data, the physician that owns the data? How is healthcare data purchased? Again another subject that not very many people are talking about, it’s kind of taboo. In every BAA or contract that hospitals enter into, there’s usually five pages regarding data ownership. Could you speak to that a little?
Hyde: Yeah. Let’s go way back in time and then we’ll jump to current.
Hyde: If you think about the old, old world of paper reports. I mentioned 835s, 837s, UV04s, those are literally the name of the forms that you used to have to fill out in a certain way to submit them to get paid from the payer. If you think about that as our earliest form of data, the hospital owned them, or institution, or it could be provider if it’s a provider practice. They own them. The patient doesn’t. Okay?
Hyde: That’s a fundamental truth. The second thing is that they usually faxed them, scanned them, or otherwise sent them to those clearing houses which then routed them to payers. The payers also own it because it’s stuff they paid for, so they own the record and the clearing house also owns it because it is a way for them to prove that they paid the right people.
All three end up with rights and ownership. This is when we’re talking about paper forms.
Jump forward, where we are today is that that existing process has been replicated in the modern data contracts. When a physician hires a switch to route their claims, they sign an agreement that they still own the data and that the switch does. The payer similarly does it on the other side. The rights of ownership are with the provider, the payer and the switch. All three own it. They have slightly different use rights, I’ll point out, and I won’t get into the details of that, but there’s some nuance there, other than to say the trend is they’re all expanding their use rights. They’re increasingly getting more use and aggregation rights.
The EHR is on the other side, where again, those used to be paper or same thing, where we are with that is because the EHRs are so competitive with each other initially, at times, they sacrificed use rights of that data to resell and aggregate it to the providers in order to win the account. The providers have begun to realize that they wanted to own it as an asset and there’s some of that. Again, not every EHR has the right to completely resell their data, but on a scale of forever, they will eventually get those rights. They are actively, they have legal teams actively renegotiating those rights. You’re going to end up with the EHRs and the claim switches and they are going to basically own the chunks of data and they will probably be between 10, 25 major companies who will own most of it and they will consolidate.
Cunningham: What about data registries? Let’s talk about, because this is an orthopedic conference, the HARR, do you work with any of those data registries or have you been finding value from what’s being the output?
Hyde: No. I’ll tell you why. Because they’re trying to do what I’m trying to do. They’re trying to gather data.
The only problem is, I could bring more to bare. I have money. I have a profit line going on the other side of this. I can share an up side. These other groups exist primarily in order to benefit the consortia that they support.
In fact, I’m working with the Angiogenesis Foundation right now because, as an example, they exist to show that anti-angiogenic drugs are not being used properly. That’s their whole thesis, that that therapy’s important and they need data to prove that. I’m like, “Great, let’s prove that.” We found out that the variation in care and treatment for solid tumors is something like 30% variant, just to translate what that means, it means that the chance that you get the same treatment is one in three depending on where you go to see a doctor. We’re talking about cancer here.
Hyde: There are guidelines, it should not be unclear, but right now physicians perform all over the place, and that’s just oncology.
You could talk about orthopedics too. That’s an example of where we actually become the data partner for them.
Cunningham: With oncology, especially some of these aggressive forms of cancer, there’s a big question on the outside of how much is this benefiting the overall quality of that patient’s life, is it extending it? It’s interesting, having that data, I can see in the future becoming increasingly important for regulatory decisions as well.
Hyde: I think the FDA is signaling that they will be looking at virtual trials and if you look at IMS, which is probably, they are the Microsoft of healthcare data, they have it all, and in Europe. I’m just a fly at their ankle at the moment.
In short, they just purchased Quintiles, which is one of the largest clinical trial companies in the world. Together they are going to be running trials on existing patients just looking at their data. Both to stratify them, identify them, but also look at how certain therapies are working out because right now a phase four trial in the FDA, which is, I am approved, I have to report back on what happens to a certain cohort of patients, are sort of comical because they’re on very small population again and they’re not looking at all the data that’s available on the patients that are actually taking it. It’s kind of crazy, it’s like what are you doing?
We should use the data that’s out there and I think you will see these virtual clinical trials begin to emerge in the next five years, maybe even sooner and the FDA maybe even approving drugs for new indications as an example, for things that are being used off label based on that data. If I see, a good example, any coagulant, Xarelto is approved for AFib, well most patients with AFib also have coronary artery disease, heart failure and may have one or two other things.
What if I just took the data and said, “Hey, how’s Xarelto look in these patients?” They’re already taking it, is it helping? Then, you could go to the FDA say, “Look, here’s our efficacy and boom, you can get a new label. That’s real. That’s happening.
Cunningham: Awesome, I think that’s actually going to benefit healthcare in general. I want to end with kind of a personal question.
Cunningham: Okay. Your first startup, you eluded to having a first startup as you exited, can you tell us about that?
Hyde: I did, yeah, we started in 2008, a company called Relay Technology Management, which is not Relay Health the data company, which, I should have started that one.
But in short, we were really interested in trying to predict. I have a PhD and I started in basic science and I spent a couple years on Wall Street, so I completely sold out, went that way, decided to come back and do something interesting.
We sort of had the thesis that the drug targets that drug companies go after are predictable and the patient opportunities can also be predicted if we use big data. In those days, I think it was more like big spreadsheet, so we were cobbling together whatever data we could get, applying AI and machine learning techniques. We used a lot of Bayesian Models to basically predict what was going to be developed for drug and the idea was that in that case we would suggest and recommend more exciting targets or therapeutics or potential partnerships. As a business, we decided to build software and sell that B-to-B and that worked fine, but after a couple years, we said, “Let’s sell this.” I sold it to DRG and I’ve been there ever since.
Cunningham: This has been absolutely fascinating. Thank you so much for really shaking it up and adding something fresh in the mix and with that, we’re signing off. We’ll be right back with our next interviewee.