Timestamping and keeping track of events

Indeed, in a blockchain, records or data are periodically aggregated into ‘blocks’ which represent every transaction that has happened within that time frame. These blocks are linked (‘chained’) to each other using a cryptographic hash of the previous block and carry a timestamp.

Let us abstract from our argument for a moment. If we consider any transaction recorded in the blockchain as an event, then we can timestamp said event and order sequential events in time, so that we can ascertain that an event indeed happened and that of a group of events, each event happened in a precise sequence in time. This is done in a near-incorruptible way, enabling us to consistently trace events.

Asserting and proving events

In the early history of Bitcoin technology, developers tweaked the data structure to store small pieces of information inside the blockchain. Known as a ‘hash’, this short string of characters is the result of putting a document through a hashing function. Any two identical documents will always produce an identical hash; change even a single character, and the outcome of the hashing function is radically affected. Because of these features, this little thing has important functional consequences. A hash can stand for a digital signature of any information: for instance, a document, however long, can be shortened into a hash, which then becomes its one and only ‘signature’. Thus, a person receiving a document can hash it and compare the hash to that of the original document (which in the current context is stored on the blockchain, so it cannot be altered). If the hashes match, this guarantees the integrity of the document relative to the state in which it was hashed.

In addition, each block in a blockchain contains a timestamp. Because the hash of each block includes the timestamp of the previous block, the blocks are chained together sequentially in time. In summary, hashes are digital summaries of data which are calculated on the content of each data block that makes up a blockchain. Importantly, blockchain can store as many of these proofs of data as necessary.

Automating processes

One of the greatest promises of blockchain is related to the development of script language that enables programming on top of the blockchain architecture and so gives all the flexibility of automating processes. These pieces of code are called ‘smart contracts.’ A smart contract is essentially a piece of code which executes on the fulfilment of certain predefined, user-determined criteria. For example, a smart contract might be written to automatically upload trial data to a trials registry, if and only if certain conditions obtain: (1) that all patients have consented and (2) each phase of the trial protocol has been registered as successfully completed.

This is a potentially very powerful tool, though currently in its infancy.

Public and private blockchains

Because blockchain technology relies on several component technologies, there are different variations on blockchains which reflect differences in the component technologies. A major distinction is between blockchains in which anyone can participate, therefore called public, and blockchains which require permissions to enter, therefore called private. The reasons for these are many and varied, but it is important to consider which architecture is best suitable for the context of healthcare broadly and research and consent in particular.

The first operational blockchain network, which underlies Bitcoin, is a public blockchain which anyone can join. It prevents fraud by forcing each computer in the node to solve hard calculations, which are calibrated to ensure a high energy cost resulting from the computational complexity. By making each computation costly in terms of resource costs, the proof of work system guarantees that messing with the previously calculated blocks becomes prohibitively expensive.

However, this dynamic is not scalable in its current form, as it consumes unacceptably large amounts of electrical energy. There are, however, other consensus mechanisms than the proof of work which are being explored but are beyond the scope of this paper.

Data enclaves and homomorphic encryption

Distributed computing architecture may lead to the greater decentralisation of study conduct. The importance of real-world evidence and patients’ reported outcomes are in line with the contemporary sense of a need for greater patient centricity of research studies.

Blockchain architecture could help define and concretise such an infrastructure. Moreover, to open to a distributed data-sharing ecosystem with patient-level fine-grained ownership control, some research teams are designing new ways to process data. The idea is to push the algorithms rather than pull the data: algorithms process the data remotely without breaking privacy. This is especially interesting in an artificial intelligence era, where federated learning techniques would let algorithms jump from one data warehouse to another, increasing each time the spectrum of its machine-learnt knowledge.

Blockchain technologies may play an important role in these future architectures of distributed computing and federated learning. Other systems for data aggregation could be envisioned, such as safe houses or physically secure databases, also relying on public and private key cryptography. However, blockchain technologies offer several advantages, including speed of information transfer, no single target for breaches and various automations. These strong privacy-respecting systems may provoke more consent and participation to studies. For example, some entities such as the US Food and Drug Administration (FDA), and private companies, are working on a data brokerage system where a blockchain-based system would help consent to share the data to some dedicated research entities and to trace back the value created resulting from the data processing, for the purposes of distributing the value created from the analysis of data to those who have made those data available. In general, blockchain technologies could be used to manage data access, ensure transparency, reduce or prevent fraud and tampering, increase efficiency and connect stakeholders in a learning healthcare system.

Practical consequences for the consent process

Let us sum up all these functions. Using blockchain, we can trace if and when consent was given, we can bind a consent to a document, for instance, a study protocol, on any of its versions, the proof of which are stored in the blockchain through the so-called hashes. Automaticity through smart contracts enables endless possibilities, some of which may be: automation of aspects of consent collection processes (eg, identification of potential subjects and contact via email), reconsenting being triggered when some conditions are met, for example, major changes to the protocol, and conditioning consent to feedback of results.

Three uses of this system

At least three novel approaches to data sharing become possible using such a system. The first grants data subjects or originators the sole power to determine who may access their data. The second would include some default permissions on an opt-out model. The third would remove the opt-out option, thus mandating data access.

Data owner access control

The first possibility involves giving data owners complete control over access to their data. Data owners can grant, modify or revoke permissions to access data by means of blockchain transactions. Importantly, data owners could treat different categories of data differently and assign varying levels of access protections to them. For example, access to sensitive medical data might be kept private or granted only to select entities, whereas less sensitive data might be put up for donation, or, from the point of view of some start-ups, possibly even for sale. The blockchain thus provides a practical means of implementing meta-consent.12

Consent, minimal risk and default permissions

Alternatively, default settings could be set to allow certain entities access to some data. In an opt-out model, default settings would be controllable and modifiable by the data subject. It would also be possible to have certain types of data shared by default. Table 3 lists some of the potential benefits of implementation.

The motivation for opt-out or mandatory models stems from the effects of consent requirements on research. Consent requirements can be excessively complex, especially where the data involved are not sensitive and might be put to general, open use. They can also lead to selection bias—the systematic distortion of research results due to statistically irremediable deviations from a normal sample—which can seriously reduce the reliability of research.15–17

Significantly, both ethics and the law allow for consent waivers to avoid these problems if the research in question can be shown to involve only minimal risks.18 Using blockchain-based data access management system and multiparty secure computing could reduce the risks of much non-interventional research to minimal, since data would remain at its origin and not be subject to additional breach risks.

Prosent: enabling bidirectional research requests

Consent protects the autonomy of patients and research subjects by allowing them to refuse unwanted treatment or participation in research. However, consent does not enable individuals to go beyond what is offered in terms of participation or interventions. Patients and research subjects might wish to exercise their autonomy by sharing other data or sharing data with other trusted research or healthcare entities. This would be possible through the prosent feature enabled by blockchain.

As manifested by the powerful trend of crowd and citizen science, many people outside the traditional research ecosystem have both the means and the willingness to gather and contribute important data.19 20 Indeed, citizens generally hold positive views about data sharing for public benefit research.21 22 Prosent could help further this trend by allowing data owners to identify each other and request data and/or participation of other citizens or scientists.

By analogy with the word consent, we propose a novel term that captures this ability: prosent. The prefix ‘con-’, originally derived from Latin cum (‘with’), refers to a joining or togetherness in the present tense. By contrast, the prefix ‘pro-’ expresses both a positive affirmation (eg, prochoice) and a forward-looking aspect (eg, prospect). Just as consent implies a current acceptance (con) of some feeling or thinking (sentio), so prosent implies a forward affirmation (pro) of an emotion or cognition (sentio).

Prosent leverages several of the affordances of blockchain to enable much greater communication between stakeholders. In a hypothetical health research ecosystem, there might be several groups of distinct stakeholders. These might include data subjects and data owners (whether individual, institutional or commercial); various data generators (including individual patients, patient advocacy groups, grassroots databases and individuals or institutions with skills in data aggregation and/or scraping); and various data acquirers (individual healthcare professionals, third-party information services, charities, both public and private institutions, hospitals and research centres; the pharmaceutical and actuarial industry; various government actors; and interested private individuals, not to mention interested loved ones).

These many stakeholders have varying degrees of motivation, insight and expertise. In the current system, few of these are leveraged; typically, the physician or institution owns the data and does not share it, except with close colleagues. However, using blockchain, it becomes possible to open up this data exchange to others who may have relevant expertise. Each entity would be represented by a pseudonymous identifier. This identifier could include information which verifies their status (eg, of healthcare professional, institution or individual) in a privacy-preserving manner, such that, though it cannot be said whom a particular identifier represents, it can be determined what their status is. Thus, a patient suffering from a rare disease might include information on her profile to that effect. This would enable others, whether others suffering from the same disease, or researchers interesting in advancing research, to locate a potential patient without compromising the identity of any involved.

This architecture would enable all kinds of interesting interactions between stakeholder groups. A patient with a rare disease but some resources might take it on themselves to request the data from all other patients with that rare disease who can be located pseudonymously through the blockchain. They might then issue a prosent request for the data, and, if obtained in sufficient quantities, they could then release a prosent request to an institution or individual or groups of researchers to carry out analyses on these data.

Also, fascinatingly, individuals could indicate that they wish to acquire, generate or sell data only under certain circumstances. For example, a patient suffering from an orphan disease might add an identifier to their profile, such that they can be located and contacted (still pseudonymously). Another patient might add a different kind of identifier to their profile, perhaps indicating a willingness to share their data, but only with certain entities and not others. We might imagine that many would be happy to share their data for important epidemiological work, but less enthused about doing the same (without remuneration) for commercial research. Alternatively, an individual could indicate that they only wish to share with researchers from their own social, ethnic or cultural background. Finally, monetary or healthcare incentives could be offered for making data available, although this would lead to interesting questions concerning the correct levels of regulation for a healthcare data market, and on what money can and should not buy.

Thus, we use the term ‘prosent’ to refer to the bidirectional research requests enabled by blockchain technologies. Below, we briefly sketch some of the features enabled by prosent mechanisms. This is not an exclusive list.

For one, many stakeholders who can benefit from data access but who have previously left out of the research ecosystem can exercise a greater degree of control over data; both their own data, but also by pooling or acquiring the data of others, for example, to establish a database of rare diseases. Second, these stakeholders can interact with each other pseudonymously, such that they can locate each other as entity types but not as uniquely identifiable entities and can communicate without privacy concerns. Third, various stakeholders respond to various incentives, and using prosent it is possible to offer this variety; data release could be conditioned on the aim of the study involved, the researchers or patients involved, whether or not money or healthcare has been offered and who the likely main recipients of the benefits are.

Although this is a rough description and there have been several other calls for data marketplaces, we believe the concept of prosent has not been adequately captured in the extant scholarship. The possibilities of opening up science to interested, powerful and numerous stakeholders under controllable conditions are vast. Thus, prosent, especially when coupled with smart contracts, has implications both for justice and for beneficence.

Justice

Certain populations are under-represented in research,23 due to worries ranging from additional susceptibility to health risks to the ability to give voluntary consent.24 Others may be reluctant to trust in biomedical researchers due to historical factors.25 26 Still others have very rare medical conditions.27 28 Finally, many individuals belong to groups that are less able than others to pay for the advancement of their interests, making them less attractive targets for pharmaceutical and other medical companies.29 30

Prosent mechanisms offer a novel and potentially powerful means of re-engaging individuals from these communities. Using the pseudonymity of a blockchain-based, prosent-enabled data exchange, groups of individuals with similar conditions could find each other via pseudonymous profiles which could contain tags indicating the preferences and interests of that user.

Smart contracts and beneficence

As mentioned above, smart contracts are pieces of code layered on top of a blockchain which execute automatically when certain conditions are met.

Publication bias, resulting from the preferential publication of positive results, is a known problem in biomedicine.31 Using smart contracts, it would be possible for stakeholders to agree at the beginning to release the data to a public trials registry, which would then happen automatically on study completion. Similarly, publicly minded patients might condition their consent on such data release, either to the public or to themselves; if such a condition were not released, the smart contract would invalidate that person’s consent.

Smart contracts could also automatically trigger a request for reconsent in cases of major protocol changes. For example, it is crucial that primary and secondary outcomes are specified in the protocol before the conduct of the study and that they are not subsequently changed or manipulated.32 In addition, the recording of consent on a blockchain could be fully transparent, visible and auditable for relevant stakeholders through dedicated public websites. Finally, blockchain could be used to put consent management in the hands of the patients. Patients who wish to revoke or modify their consent could do so directly via a transaction on the blockchain, without relying on a third party, such as the study administrator, to document the modification.

This is not a small opportunity, since failure to obtain or document consent is a known problem in clinical research. One review of FDA records found a failure to protect subjects and/or obtain informed consent in 53% of cases studied.31 Blockchain could be used to make the documentation of consent both transparent and traceable.8 Smart contracts could be used to freeze patient data or the progression to the next protocol phase, thus predicating the release of data on unequivocal documentation of consent.7

Similarly, blockchain technologies can be used to document other key components of the protocol. Any revisions to the protocol would be timestamped and transparent, reducing the incentive for fraudulent changes. These might include the data-sharing plan, the version of the analytical code used at the outset and subsequent modifications to it, and documentation of obtained consent.7

Looking towards the future, smart contracts could be used to automate several important bottlenecks in medical research. In theory, smart contracts in combination with secure multiparty computing and the OPAL principles could lead to a situation in which trial results are uploaded immediately on trial completion by one smart contract, and then processed and integrated to existing systematic reviews, all fully automated and much quicker than the current laborious process. Again, these suggestions are not exhaustive; smart contracts could be used in the recall and monitoring of defective drugs and medical devices; in the provenance of surgical tools; in the automated transfer of patient information to relevant healthcare providers in cases of emergency or relocation; and many more.