How health insurance DAOs can break up RWD silos
KdT has recently been exploring alternative data sharing incentive structures, since as a firm we believe the greatest advancements in science and medicine this century will come from our understanding of pre-existing, siloed data as well as data produced and enabled by the scientific revolution we are currently undergoing. This question led to a trip down the web3 rabbit hole led by our associate Benji L. who put pen to paper to share some of his thoughts below.
The holy grail for human health is accessibility of all health data related to each person into a single data lake, structured into a relational database in the same spec for each patient, for the entire world’s doctors and scientists to mine for insights that contribute to our understanding of disease and improving the standard of care. An App Store of biosensors, if you will, vying for the average patient/consumer to wear and share. Single payer systems (political and personal opinions aside), do benefit from centralization of clinical data, arguably the most valuable of all human health data at this point. Answering questions like, “does taking a booster dose of the COVID vaccine reduce deaths caused by infection with the Omicron variant given prior infection with the Delta variant for individuals aged 20–29?” should be readily answerable with any curious hypochondriac looking to make decisions that impact the health of them and their loved ones. Instead, we are left to consume the political or capitalist motivated talking points fed to us, unable to be audited or questioned, by the entities supposedly created to protect our health. If you think that the average person would never have the know-how to perform the necessary biostatistic considerations to ask this question properly, consider that brands and reputation can form around the generation of these epidemiological analyses to provide analysis legitimacy the same way brands provide legitimacy to any other product or service. What no entity has been able to figure out yet is, how to create the right mechanism design and incentive structure to integrate all human health data into a single platform such that it can be queried to answer questions like the one mentioned above.
To create a health database from disparate data sources requires consensus on a single data specification and interoperability standards. Many health data interoperability standards have emerged, the foremost being FHIR. Initiatives such as these solve the problem of creating standards for data structure, allowing for more interoperability, but most of this inter-hospital data sharing happens for hospitals in the same network to reduce admin costs, not for scientific / medical research. Ultimately, projects such as HealthLinkages have died trying to get hospitals to share data with each other. The root cause of this aversion to share health data between entities is because the RWD companies understand the high intrinsic value of such clinical and omics (genetic / imaging / phenotypic) data. The exception is selling this data to pharmaceutical companies.
Real-world data (RWD) companies who aggregate data from hospitals and with point-of-care diagnostics such as Flatiron Health, Foundation Medicine, Tempus, and ConcertAI, have viable businesses predicated on selling health data to pharma and (sometimes) smaller biotech companies. There are two primary reasons that drug developers make the vast majority of the health data market: 1) they have deep pockets and 2) only they can realize the value of the insights that are found in these health data aggregators. Even if you are the best computational biologist in the world, the only chance you’re going to get access to the biggest health databases in the world at one of these RWD companies is if you work for one of them, or a pharma company that happens to purchase a small subset of their database. This results in an underutilization of the data both from a capital perspective to these RWD companies, but also from a clinical perspective. There is vastly more data sitting in databases than is being studied for insights because RWD companies place multimillion dollar moats between you as a scientist and the keys to the data castle.
The title of this article is Health Insurance DAO, so why are we talking about RWD? The answer lies in answering how the FDA and US health insurance organizations make decisions whether to approve, or cover, diagnostics and therapeutics. To receive favorable decisions from both the FDA and payers, the diagnostic or therapeutic must possess the following: analytic validity, clinical validity, and clinical utility. In the case of a diagnostic, analytic validity ensures the performance of the test in the lab; clinical validity ensures the test itself is associated with the clinical disease of interest; clinical utility ensures that the diagnostic, when used in the clinic, improves patient outcomes. To receive reimbursement from private insurers requires yet another claim to be made — cost-effectiveness — to ensure the diagnostic, if used, lowers overall healthcare costs.
Building an evidence base for these claims requires — a combination of lab data, clinical data, and cost data.
- Analytic validity: Lab data is easy to generate, but requires significant time to prepare and publish for peer-review, where ultimately, the data and code are not necessarily published, and thus, auditable. What would it take to force all lab data out into the open for anyone willing to audit it?
- Clinical validity: Clinical data has been historically generated with trials data, but is now becoming more amenable to using RWD to provide the evidence for clinical validity. What would it take to expose this clinical data for all to verify the authenticity?
- Clinical utility: This has always, and likely will always be demonstrated with clinical trials, though the population is usually not representative of the population it intends to be used for, and often is composed of the patients with the highest likelihood of making the diagnostic appear performant. What would it take to encode clinical trial protocols and randomization to ensure accurate disease representation and prevent cherry picking patients for your trial?
- Cost-effectiveness: This is typically demonstrated with rough approximations based on epidemiology and publicly available cost data. The conclusions of these analyses are dependent on highly sensitive assumptions that are difficult to accurately estimate. What would it take to expose claims data to get exact cost measurements?
As of the time of writing, these are the state of the art methods. Is there truly no way to come together, join our datasets, and improve healthcare outcomes and costs for us all? Don’t our very lives depend on it?
Cryptocurrency markets have seen multiple bull runs over the past half decade, where the 2017 run was fueled by an ICO boom, 2021 was fueled by a decentralized finance (DeFi) boom, and the next bull run shaping up appears to be fueled by a decentralized autonomous organization (DAO) boom. DAOs are networks of individuals uniting under some common goal. That goal can be speculative, like getting a return on investment for DeFi projects (BitDAO), or something more fun like unsuccessfully bidding on the constitution (ConstitutionDAO). ConstitutionDAO raised $49M USD, but still pales in comparison to the size of the BitDAO treasury ($2.5B USD). If you’re wondering whether this sounds like a fanciful GoFundMe, there’s three major discrepancies that make DAOs far more powerful. First, platforms like GoFundMe are mechanisms for crowdfunding, but not governance. You can use GoFundMe to raise money for your cat’s surgery, but not to govern which surgical equipment the vet can use. Second, because DAOs exist entirely within blockchains, all fund transfers are auditable. This means that we as funders of your cat surgery can see exactly how much of the funds were used to pay for the vet, the anesthetic, kitty crutches, and ensure you didn’t just pocket the funds that were supposed to heal Mittens. The third, and most pertinent difference is censorship-resistance. Recently, GoFundMe successfully blocked the Canadian Trucker protest from raising over $9M. Whether or not you agree with the underlying message the truckers are trying to send, GoFundMe attempted to siphon the funds to charities it thought were considered to be of higher moral standard. This incident perfectly illustrated how Web2 platforms can be used to exercise control for political purposes and validates the market need for DAOs, especially for crowdfunding and governance.
Quietly, DAOs are taking root in the biotech arena and forming an international decentralized science (DeSci) movement. This can be seen in various flavors. LabDAO is building a community-operated network of scientists to run experiments. Molecule is building a mechanism for porting intellectual property onto Ethereum as intellectual-property non-fungible tokens (IP-NFTs). VitaDAO is aligned around funding longevity research, a historically underfunded and undervalued area of research. VitaDAO has already accumulated $20M in its treasury and has funded research at the Scheibye-Knudsen Lab at University of Copenhagen to mine the Danish National Health Service Prescription Database, and has discovered multiple FDA-approved compounds that are associated with longer human lifespan.
Despite the multiple companies that have launched to leverage blockchain to promote health data sharing between patients, namely Nebula Genomics, LunaDNA, and Patientory, none have yet gathered a network effect and are struggling to gain adoption. These companies may be ahead of their time, but they might also have the wrong incentive structure and mechanism design to ever garner mass adoption. It may be because they have failed to embrace DAOs as a superior form of governance.
Health Insurance DAO
Imagine a Health Insurance DAO governed by patients, for patients, with two objectives: increase quality associated life years (QALYs), and reduce healthcare costs. Ultimately, that should be what a health insurance company cares about. Unfortunately, in capitalist incentive structures, motivations of insurance companies are often profit-motivated, and it is left for others to care about health outcomes. This misalignment in incentives between patients and health insurers is a textbook example of the principal-agent problem and has left us with a suboptimal and grossly expensive healthcare system. Here in the US, we have the 18th best healthcare system in the world, but it costs us 20% of our GDP! Evidently, there is room for improvement.
The only way to improve healthcare outcomes and lower costs is to first measure them. In order to measure outcomes and costs, we must get all patient data onto a blockchain to ensure privacy, censorship-resistance, transparency, and embedded incentives. Once data is on chain, a health insurance DAO can be created to front the healthcare bills, but demand the health data that necessitated the reimbursement in return, in the form of doctor’s notes, medication administrations, diagnostics results, lab results, etc. Why would an insurance company care about this data? Well, it’s looking to make smarter decisions, arguably has the biggest need for this data, and it has the purchasing power, able to withhold reimbursement in escrow until the data has been transferred. To motivate patients and providers to provide this data could be incentivized with reduced premiums / copays / deductibles to patients, or increased reimbursement to providers. In the long-term, this data will be extremely valuable to the DAO as its members will be able to better make decisions for which therapies, diagnostics, surgeons, pathologists, radiologists, etc. charge the lowest fees and provide the best outcomes for patients. All of this is readily measurable and allows patients to shop for health interventions and medical advice as they would on Yelp for restaurants. This is the radical transparency that patients have been deprived of, but deserve. It’s time we audit the entire healthcare system, and financial incentives are the mechanism to do so. One patient, one datapoint at a time.
One might argue that if a health insurer has access to all of this data, it will rationally discriminate against patients with pre-existing conditions or factors outside of their control to hike their premiums. This is a fair concern and addressing it is going to be a legal requirement; however, on blockchain it requires an intelligently designed smart contract integrating zero-knowledge proofs with private key encryption to withhold this information from the premium-pricing equation. One powerful mechanism design choice is the ability to motivate individuals to engage in healthy behaviors. For example, the DAO could enable the ability to reduce your health insurance premium for proving that individuals completed a workout, ate a healthy meal, or took supplements that are demonstrated (from on-chain data) to reduce healthcare costs to the DAO. This financial incentive to be healthy may not be the end-all-be-all, but adds another incentive layer that benefits both the individual (medically and financially) and the DAO (financially). There is value to be captured for incentivizing healthy behavior that is not yet being capitalized by existing health insurance companies because they cannot yet associate behavior with healthcare costs. A Health Insurance DAO could.
Armed with the ultimate health dataset, decisions can be made completely differently than they are today. Instead of coverage decisions being made behind closed doors, coverage proposals can be made for a new intervention based off real-world outcomes and cost data incurred by DAO members themselves. All of the analytical, clinical, and financial data is directly measurable from on chain transactions, and the level of evidence with which the diagnostic can reduce costs and improve outcomes can be deliberated between white-listed physician, scientist, and biostatistician delegates, and their analyses can be published for all the DAO members to see and vote on if they choose.
Once significant data has been accumulated by the DAO members, it can use this data to outcompete other health insurance companies, because it will inevitably be able to make smarter decisions with better data and more actionable data modalities. However, these datasets will be valuable to others and given the ethos of sharing data, the DAO can sell its data for prices voted on by the DAO, based on the cost of acquisition of the data, and its rarity. For example, the cost of an MRI is greater than the cost of weighing someone on a scale, and the cost of finding a patient with Alice in Wonderland Syndrome will be much harder than finding a patient with COVID, and both parameters should be used to price data accordingly. Once data has been priced, it can be sold with specified exclusivity and time-restriction on a case-by-case basis to entities interested in leveraging it for research, marketing, or otherwise. Pharma may wish to purchase this data to determine the market size of various diseases or market directly to the physicians of eligible patients for their clinical trials. Diagnostics companies may wish to purchase the data to improve their diagnostic algorithms to further improve the outcomes of the members. All of this can further fund the treasury, and go toward research, healthcare expenses, or other causes voted on by the DAO. Projects like Arweave, Nevermined, and Ocean are creating the infrastructure to store data indefinitely, de-identify it, and create marketplaces to securely share that data on terms set by the DAO.
The flywheel of health
The prospect of a group of individuals, motivated by studying real-world health data and capitalizing on the resulting medical discoveries to improve the health of humanity is invigorating. What is most exciting about a platform where health data is collected at a massive scale and being mined by the whole world’s computational biologists and other medical scientists, is that it will exponentially benefit the care of those who share more data. The more that’s shared, the deeper insights people will be able to receive about their own health. Maybe not everyone will care enough about their health and just share data for financial incentives, but for those with privacy concerns or sufficient wealth, optimizing health may be enough of an incentive to share their de-identified data with the world. Once data has been collected at a large scale and mined extensively, to have the best chances at a healthy life will likely require sharing health information. The more data you share, the deeper insights you will receive about your health, enabling you to biohack one day at a time and get direct feedback on how your actions affect your body. You only have one — might as well make the most of it.
Many thanks to Cain McClary and Niklas Rindtorff for their contributions.