It's time for open source design in healthcare

Juhan Sonin at UXDX Community: Open-Source Health and Being Vulnerable. Video: https://youtu.be/z3_rDLXscEE

Readable transcript: edited from the recording's captions for readability (fillers and false starts removed, punctuation and section headings added). Wording is the speaker's own. Timestamps are positions in the video. Names marked [?] could not be verified against the audio.

Disclosures and following the money

[00:00:00] I'm Juhan, and yes, I do work at a design studio in the Republic of Boston, and I do teach at an academic center in Cambridge. Because we're talking about health care, and there isn't enough disclosure about where our finances and where our biases are coming from, I list them here. Yes, my patents and all my genome are open source. You can download it and plant it in the crime scene if you want. To get in touch with me, feel free to email me. I'll try to get back to you within a few business hours, and you can publicly critique me on the Musk mediocre machine.

[00:00:42] Today I'd like to talk a little bit about healthcare. I think it's a noble profession, and it has been for centuries, and yet here we have this idea that it has now become a pretty noble business. Many of us, I think, are in the discipline. It's been a huge force for good over time, but we're using that nobility as an umbrella over some of the weird things that are happening in it.

[00:01:11] Where I live, in between the private equity firm called Harvard and the venture capital firm named MIT, this one square mile in between the two occupies about 25 percent of healthcare GDP. That's an insane number. It's over a trillion dollars in the U.S., meaning it's about 4.5 trillion this year. That's a lot. Is there something more there? Probably. That's one percent of global GDP, all in that small concentration. Nature typically doesn't like that kind of concentration, so maybe part of this exercise is: let's follow the money.

[00:01:52] I think part of what we as designers and engineers have to do is identify, hey, there are profits over people. What's happening? How do we start to bend that? Here's one example of that in the flesh, with a very good cancer institute in New York City called Sloan Kettering, where we have José, who's the chief medical officer, really running rampant on ethics, from breaking conflict of interest rules to really modeling himself after Sackler. You've probably heard of Purdue. There was an enormous amount of funkiness in how he was operating at the hospital.

[00:02:33] The New York Times and ProPublica went a little deeper, and this illustrates a few ideas that we should think about. There was a startup that came out of the hospital called Paige, an interesting one actually, that had an exclusive deal on the actual patient tissue samples. And yet guess what? It was paid for by the public. The collection of those tissues and the patient data was owned by the hospital, and it wasn't just one or two clinicians that put that together. It was hundreds of them that encoded and decoded those slides. And of course the cherry on top is that the board of the hospital had the major stakes in that startup.

[00:03:15] Great. It gets a little better: this is not just work a man asked [?] in a lab in a small amount of time. It was 60 in progress [?]. It was the patient data, our data, that was owned specifically by the hospital. And I think dot orgs are just as culpable nowadays, unfortunately. That signal is no longer clear of where their intentions are, meaning .org, .com, etc.

[00:03:43] Now, okay, yes, Thomas is hiding in plain sight here, but it's always interesting when someone's a research intern and CEO at the same time. More importantly, there was a data use agreement that kept the patient tissue slides in control of the startup and not in the control of the patients.

Who profits from patient data

[00:04:03] What kind of design is this? It is one where we, as patients, don't have ownership rights. The IP is owned by the executives, yet we are funding it with our tax dollars. This is really just a perfect example of corporate welfare at its finest, and it's the rush for money. Ethics is situationally optional.

[00:04:29] Look, this happens all the time in healthcare. It's not just indigenous to that one institution. It's everywhere. I don't care if you're at the NHS, if you're in California, wherever you are, it's everywhere. Let's look at a case in point, where we have the former CMO of Moderna, you may have heard of that company with its vaccines, and Michael Milken, who has his own history. They're on a podcast together right at the beginning of the pandemic. What I thought was interesting here is what does Tal Zaks mean: "I want to move to private industry from academia in order to be one to three logs more effective." What does that mean?

[00:05:11] Well, in actuality, if you look at how he's selling his Moderna stock, he was selling it at about a million bucks a week, right after the U.S. public threw a billion dollars at Moderna. I don't think this is exactly a great signal for health care either. Now, a year later, their CEO gets a billion dollar payout. I think this is a crime, and yet this is duriger [?]. I think it's insane.

[00:05:39] Okay, let's keep going here, but we are all complicit in this. We are, as global citizens, as United Statesians, just pushing our data everywhere without rights, and it feeds Kendall Square, Mountain View and Moscow.

Health data is more than you think, and you don't own it

[00:05:57] One: health care data includes more than you think. This is a definition we started for Wikipedia several years ago, and it's finally there. The way we put it together is that it's any information about me that assists in making decisions about my health. Clinically, we may think of our health data as genetics and devices and what's at the hospital. It's much more than that, as many of you know. It's really your ZIP code. It's how much money you make; your salary is a vector here. So is the ability to vote.

[00:06:32] We have a whole mapping of this, by the way. It's an open source licensed technique called determinantsofhealth.org that you can zoom in on. You can print it out, you can download it, do whatever you want. It's on GitHub. But your health data is a lot more than you think it is, and you don't own it, and you damn well should.

[00:06:54] Now, for those who are in civilization, more than the United States, because we're backwards here, there's a thing called GDPR: how do you protect your information? They actually have a pretty good start on this, where they say, hey, you control the access, you control where the data goes. Great, but they don't articulate ownership. If you go deeper into the right of access recital, there's nothing here about data ownership, and there really should be.

[00:07:23] Why is that? Well, one reason is that there are a lot of people who sniff and whiff your data. This is an open source diagram we put together on all the little ants that pilfer through your information. It could be pharmacy benefit managers, it could be the EHR, the electronic health record. Sundari [?] talked about acronyms; I should stop that. The bank, the government, all these health care middlemen have it. It's a large list of humans and businesses who have it. If you want to take a look at this, you can download it here.

[00:08:04] Now, the hope is that this age of surveillance is going to change a bit, because my thinking is that maybe health data ownership will turn it into a net positive. It's not a new idea. Hell, we were talking about it in 2006. Other people talked about it in the early 1990s. Alex talked about it even at the World Economic Forum a few years ago. But it's really the proliferation of this silicon brick that has got us closer to really thinking about it. Back in 2005 we were saying, okay, what do we need? A simple health data standard, and we need to own it. This is back then already.

[00:08:47] Enough of that. We do have some good people on our side. The former head of Brigham Health here in Boston has said we should own it. The former head of the Centers for Medicare and Medicaid Services in the United States, Seema Verma, said we should own our data. Okay, yes, that's good. A little side note here: both of them have had little dings on their reputation recently. Ms. Nabel was on the board of Moderna, and Seema did some other things that were a little funky.

[00:09:23] But guess what? I mention that because I'm dirty too. I've worked for all sorts of different institutions on different kinds of national guidance for healthcare. I've worked in the military and healthcare industrial complexes across the U.S. Okay, I could be better, but there is a spectrum of dirtiness, and I just want to be very upfront about that.

What data ownership requires

[00:09:50] Now, back to owning your data. There's a very good paper, it's like three pages long, so someone like me can get through it, that outlines what has to happen. Of course, JAMA doesn't want to include four authors, so George is left out of it, and he was part of it. The three components that you need for data ownership are: you need common data elements to describe the human and code; you need a patient encounter receipt, of what happens when you go to a clinic or when you have a telehealth visit; and then you need some kind of contract between the parties to say, okay, you did this, I did this, here's what happened, and here's the governance for that data.

[00:10:32] Just to put it in visual terms, because I like seeing it: you have an encounter. I don't care where it is, on your phone, in the middle of a mall somewhere. You have an encounter, and then you get this receipt of what happened in it. You get it, the clinician gets it, and there's this thing called a patient data manager, which is like your Tetris bot that structures the data, puts it together and then injects that back into your longitudinal health record. Then you have different ways of looking at that depending on what your role is and what your responsibilities are.

[00:11:11] Overall, you have a governance, some kind of agreement: yeah, I agree to share it, or I agree to share it for a little bit, or I don't agree to share it at all. That's what says how your data is used. One little bit of this is that this patient data manager goes and collects your information from different encounters, from what you do at home, and then brings it up so that you can actually see what's happening and make better and better decisions over time.

[00:11:40] What does having a patient data manager and patient data ownership get you? One thing is how complete your data is. That's something we have no idea about. How good is it? Can you make a good primary care decision on what to do next? Usually clinicians have just a handful of data points, versus Meta or Google or Amazon, which have hundreds and thousands of data points about you. Who would you actually trust to make the decision based on that data, not who would you trust to hold the data? Well, if it's a better sampling size, if you have a couple hundred thousand... I think this is an important concept to think about: how do you know what data is in there, how much is in there, to make better decisions?

[00:12:25] By the way, another side note: it took a bunch of over-educated MDs and PhDs to say, hey, drawing pictures helps people get stuff. Well, no kidding, but there are papers on this with metrics to say, hey, if you use techniques like this, like I'm showing, that I use all the time, there's backup for you here. We use this across our work, whether it's open source work we do for the US on how COVID is working or not working, or how science works. We use visual techniques in our storytelling all the time. I think most designers know this already, but sometimes you need paper evidence that someone studied this.

Controlling access and health autonomy

[00:13:11] Back to owning your data. Three: when you own your data, you control access. We talked a little bit about this, so let's just skinny this down a bit. We actually have a reference implementation of this on GitHub, and you can download it to your heart's content. It shows what you can do, how you tell the story in a fifth grade manner, in highly visual ways: what consent is, what happens to your data. We actually built this on top of HealthKit, so we weren't reinventing the wheel of how you should think about what your health record should act like.

[00:13:50] Now we're redoing this with OpenAI and Stable Diffusion and Dance Diffusion, so we have a new version of this evolving as we speak, but this is what we were thinking about before. There's your health receipt on the left, and even if you're a real nerd about it, you can see where in the social determinants of health wheel your data is. That's really pretty nerdy, but we put it in there. It goes beyond words. We also do different techniques. You need to build prototypes, because that's what people see, because they can't necessarily articulate it for you. We also do these graphic visualizations of use cases, so it's not just written in words but you see it happening. There are different fidelities of this depending on where you are in the process and who the audience is.

[00:14:31] Okay, four: owning your health data gives you health autonomy. What does that mean? What are the ownership rights? You can possess it, you can share it, you can sell it, and you can destroy it. Yep, the copy that you have, you can nuke it. Now, what we see here, one of the few tenets of data ownership, is that it's not solely owned by you all the time. You own the whole kit and caboodle, all of it, but the second part is that the generated health data about the patient, by a provider, by a nurse practitioner, etc., is co-owned by both parties.

[00:15:12] And this generated data is fully owned by the patient, so they can do whatever they want with it. They can sell it, they can do whatever they want. And all of these uses should be consented in advance by the patients, other than uses required by law. If you have a communicable disease, the hospital has to report it, for instance. That's one example.

[00:15:31] Just as a flip to that, major hospitals have done the opposite of asking patients. They don't ask patients; they just start to make companies, in this case called Truveta, where they're collecting all the health data from all the people who have been in the Puget Sound region, out by Amazon headquarters, to say, yeah, we're going to create a company that takes all this data and sells it to the highest bidder. Great. We had no say in that as patients.

[00:16:03] I think this is a framing: patient data is really like a commodity. It's like Coke, Pepsi, coal or coffee. Oh, let me feel better. But we may need to think about IP a little differently: that it's time limited, like copyright. Maybe there's a way that public ownership of anonymized data is the way to go, and maybe there's some kind of trust, a third party that works on our behalf as patients, that allocates, with our commands, where the data goes to.

[00:16:38] Now, there is some good progress in this. This is coming out of Boston here; I'd highly recommend checking it out if you're in the space. Where's my medical record? That starts to get to, hey, I can now edit my clinical notes and get a feedback loop going. Fantastic, but it still doesn't have data ownership yet. The FTC is actually handing out fines if companies screw up. That's good. Do I think 43 grand per violation is enough? I don't know, maybe it's a tickle for most big companies. You have dot orgs and older companies actually stating that we need data ownership as part of a national digital health strategy. That's good.

[00:17:22] If you want to learn more about this, here are a few resources. I will make this PDF available to anybody who wants it; I'll post it on GitHub. And then finally, while I'm not a big blockchain fan in terms of technology being the major output, the interesting thing here is that if you start to own it, you actually start to own the AI model, the machine learning model, against your data. That means it's a little more portable, and that to me is interesting.

[00:17:51] Foundationally, any healthcare system that doesn't give you your data is ruining you, because this last frontier of capitalism is to sell your experience, to sell your mind. These surveillance capitalists love to think about, wow, how do we exploit these mother scratchers, because they won't even know what hit them. This is why we demand patient data ownership. If you want a little more information about the data use agreements that we're using, they're on GitHub at datauseagreement.org. Most of us here probably know what GitHub is, but all these things I'm talking about here are open source.

Why healthcare should be open source

[00:18:32] Why are all these goodies open? Well, my take is that you can't solve this healthcare systemic problem, at least here in the Dis-United States, until you ask the people to participate. There's a classic line from the Declaration of Geneva that says, hey, I'm going to share my medical knowledge for the benefit of the patient and the advancement of my colleagues. And then we have PBMs going, oh yeah, I'm going to risk adjust all these cats in my black box.

[00:19:04] I think I'm under the illusion, or have been, that we're getting closer to care as a right, a human right. We're not there yet, because when you use a health care service you don't know how it works, why it works, or if the results are true, because of the closed nature of this. If healthcare is so noble, if we don't really have a choice here, we really need to demand health care be open.

[00:19:30] Look, the internet, or at least the infrastructure, is open source, and a human right, says the UN as well. If you look at a pretty engineering-centric OSI model of how the internet works, most of the stack here is open source, when you look at the protocols, when you look at how it works. You can see that we're using things that use open services, open stacks, to drive this conversation we're having now, or I should say monologue.

[00:20:03] Healthcare should be open source too, and yet, according to Eric Topol, we really have anti-open source in this healthcare world. Same with energy, for that matter. And then we have friends and colleagues on our side who are in high places, and we have a pretty good example of that: my homeland of Estonia has an entire open source backbone, and it uses some of the emerging things that are coming out of the planet and the US. Here are some examples of that.

[00:20:33] There are some good reasons why open source healthcare is good for infrastructure, for leading edge research, but also for this thing we're just in the long tail of right now, called COVID. It's insane. This is a great project, by the way. This came out just two or two and a half months after, I think, March, so it was May in 2020: the basic data definitions, open to the planet, done. Fantastic, in an open source manner. You go to GitHub and you see how many repositories there are about COVID: there are 150,000 of them. That's great. In my news feed I'm seeing that the Linux Foundation worked on it, MITRE worked on a common health work. There are a lot of very good projects. And a note that all the projects we complete here as a design studio are open source, so all of these are available to anyone on the planet to use.

[00:21:34] Another good example is what happens when you have many organizations working at scale. Sundari [?] talked about how she works at Meta, and there are billions of people on that platform. Amazon has millions and billions. That's a different kind of scale, and you need that to fight conditions and disease and outbreaks. What this coalition did around defining the data around COVID is they did it in an open source manner. The license was open across all these institutions, Mayo Clinic, you name it. Fantastic, amazing response.

A designer's agenda

[00:22:12] Now, in the last two minutes here, my call to you is: hey, we screwed up on the vaccine. We spent billions and billions and billions on it and yet did not open it up to the planet. There's no march-in rights. But I think we, as designers, as engineers, have to have some positions. What's your agenda as a designer? The things that I have outlined here are things we're working on now, just FYI. You need an agenda. You also need to say, okay, what the heck are you working on? How are we doing this? What's our focus?

[00:22:51] I publish these, by the way. These are not closed or anything like that, along with my policy positions. You may not agree with them, fine, but at least you know where I'm coming from and we can talk about it. I think that's part of the systems engineering that we need to do as engineers, as designers. I'm still looking out for who's the next Stacey Abrams for Medicare for All. We're not there yet, and we really have to be deliberate in committing to open source.

[00:23:20] We do it as a studio, and we actively go after projects that are open source. Again, these are not huge numbers, but we're a small company, and it occupies a good chunk of our yearly attention. In fact, we publish everything here, so you can go on our about page, and we have a working Google Slide that we update weekly with our agenda. And what's next for owning data? Ta-da: we have an open source graphic novel about it, and it's available here. We have a published article that's actually going to print in a publication on Friday. There's a lot going on.

[00:24:07] To wrap up: healthcare and public health are a utility, a human right, and too important to be closed. To recap: health data is more than you think it is. You don't own it. When you own your data, you control access. Owning your data gives you autonomy. You can't solve the problems that we're having until you ask others and the people to join. And finally, having your data can save your damn life. I have stolen from these fabulous people, and they have injected my brain with smarts. If you have any questions, please feel free to email me here. Thank you very much.

Q&A

[00:24:54] Host: Well, you were insightful when you said you were going on a monologue there. I think it's a very important topic, and I think it's really good to have at least some people fighting for what should be happening instead of the quick and easy money.

[00:25:18] Juhan: Yes, well, money bends our minds, and I am fallible. We all are fallible. So the question is, where do you draw your integrity line? To me, without integrity you don't have much else. It's tricky, though. Once people throw you serious cash and a stable gig for four years or whatever it is, all of a sudden you're like, yeah, I like my sushi dinners, and all of a sudden, I am part of the machine. You do have to fight, and sometimes it doesn't feel like you're making progress, and that's why it just takes lots of time.

[00:26:00] Host: What I'm going to do with my questions that I'm going to throw at you is play a bit of a devil's advocate role, because I'm a European, so that probably just tells you that I already believe in healthcare as a right. It's just free in all the European countries. So I'm actually going to play the opposite. GDPR has given us the cookie consent banners, and the stats on it are that people just click OK and they don't even think about it. They don't care; it's just more of an annoyance. Would asking people to own or make any decisions about their health care record just be a case of another cookie consent panel where people just go okay, okay, okay?

[00:26:45] Juhan: I think that's a brutally pointed question, because it preys well, and on purpose, on how humans operate, which is we really don't want to think about it. I agree with that concept wholeheartedly. Nobody in their right mind wants to think about their health data, unless you're a wonk, and you're speaking to one. No human wants to do that, so I think there is good reason to believe that people will slide through much of this.

[00:27:25] I still want the protection. In this country and in the EU, I want that protection regardless of whether I give a crap about it or not. I want it in law. I want there to be governance. I want there to be rules, and if they're bent, I want people to pay the consequences or understand the consequences. Right now there are no rules on that, especially in the United States and in many places on the planet. I'd rather have the rules started, for us to figure out, okay, maybe this little bot will take over some of the thinking, so to speak, for me, and I don't have to do all of it. I'm just advocating that the little thing is a lawyer on my behalf and I don't have to think about it. So I buy your argument in terms of human behavior. I don't buy it the other way: I need there to be rules over my data that people need to adhere to. Make sense?

[00:28:16] Host: Yeah. One of the challenges, I guess, in something as complex as healthcare, and I've worked on projects where we were trying to define the rules, or, as you would say, the common data elements, is that there's the argument that you only really discover them as you build the system, and then you start seeing more use cases and figure out how you need to change it. And there's an argument that regulation and locking them down will stifle innovation in the space. If you over-regulate, you'll stifle innovation. What's your feeling there?

[00:28:53] Juhan: Well, a couple of things. That same argument was made when GDPR was in its birthing process, and to your point, people skip through it, and it hasn't stopped anything in terms of innovation. You look at what's happening in the NHS in London and the vast amount of data they're collecting because patients want to give back to science. I think it's 55 percent of patients now; three years ago it was much higher. But over half of the population would give their data to science, for instance, and that's pretty remarkable. If you're getting half the population, because I'm US-centric and it's all about the US, there are no other countries out there, that's 150 million people or more saying yes, and that is a huge sample. Is it as much as what OpenAI uses right now, in the billions or trillions? No, but it is a huge sample.

[00:29:55] The second part of that is you can now synthesize personal data pretty damn well. There are services like Synthea, which is an open source patient health data generator, which allows huge swaths, not just towns but states of people, to be generated in order for you to try your health policy or your health ideas or software on. That to me is also very promising.

[00:30:29] Host: Excellent. I'm going to do the last devil's advocate question. These organizations that have all of this data don't have the skill set to do anything with it. Their skill set lies in providing healthcare, not in analyzing data. So is it such a bad thing for them to partner with companies? Those companies need to be able to make some money to invest in it, to be able to do something and get some value out of that data.

[00:31:00] Juhan: Well, you're making an assumption. There are a couple of parts to that. One is, okay, maybe data is now becoming a skill for hospital systems, because they have to report on it anyway. They have state and federal rules, again, here in the United States, so there's some basic data understanding. Have they made full use, or a third use, of what's there? I agree with you, probably not, because we don't see any of that excellence or data understanding when we go back for a second visit. They're like, oh, you're coming for a second time. Oh yeah, we know that because you paid by credit card, but where's the rest of your data? They know your money. I'm being a little facetious, but it's true.

[00:31:49] I think there are reasons for them to hire people who are very good at getting data and doing things with data. I don't have a problem with that, because they also will co-own part of your record, because they're delivering a service to you. That's part of, not bartering, but part of the reason to go to a hospital. They also have a legal reason to keep it, because they have to prove that they did things with you, to you, etc. That's the whole "nothing about me without me." So I think it's perfectly reasonable to have experts come in, people who are very good at data wrangling, and do that with the known permission of the parties involved. I think that's completely reasonable.

[00:32:37] Host: Excellent. Well, that's me playing devil's advocate finished, so I think that's all the questions. Alicia has just said thank you for bringing the information; she's enjoyed it.

[00:32:53] Juhan: So I put people to sleep. That's good.

[00:32:56] Host: I think it... I'm optimistic about what AI can do, just after having seen the protein folding improvements from DeepMind and things like that. I think once you start getting enough of a data sample out there, we're going to start seeing patterns and trends that we didn't even know existed. So I'm optimistic in this space.

[00:33:19] Juhan: The cost of intelligence is dropping like a rock, and that is particularly amazing and scary at the same time.

[00:33:30] Host: Yeah, but I'm an ever optimist, so I like to look on the bright side of it. Anyway, that brings us to time. Once again, thank you, Juhan, for sharing how we're going to get towards open health.

[00:33:45] Juhan: Anytime.