February 18, 2020
From big data to database rights: the role of technology in the future of healthcare
Technological developments in the collection and usage of clinical data create new opportunities for improving patient care and identifying treatments.
To take full commercial advantage of these developments and keep the edge gained by innovation, intellectual property (IP) and confidential information must be safeguarded.
The strategy for doing so cannot follow the legacy model used in neighbouring fields such as medical devices or pharmaceuticals. A new approach to IP must take into account the technical and commercial reality of these new technologies.
Data collection in healthcare is changing, both in terms of the volume of data that is collected and the level of clinical detail it describes. Clinicians may now record data at the bedside in electronic patient records, and so-called point of care diagnostic testing devices provide a further data stream. Patient records may also include information about drug treatments, health history, and traditional diagnostic information such as radiography, biopsies and so forth. The sheer volume of data available, even about one individual patient, is enormous. Indeed, there is so much data that it can in some circumstances exceed a clinician’s ability to assimilate and use it all.
Big data and machine learning techniques offer exciting possibilities to filter mass data or draw insights from it to support clinical decision making. Data mining also offers a way to uncover new treatments and to
change or improve existing treatments, for example, in how drugs are delivered. This might mean adopting a different dosage regimen for different cohorts of patients.
There is huge commercial potential in these techniques, so there is a need to license them in order to promote their use and to protect the IP.
A challenging IP landscape
Traditional protections for IP in healthcare may not work for big data innovation, and recent changes create issues for the licensing of these technologies.
In the same way that source code is key in software innovation, the training data upon which predictive models are based can also be fundamental to the development of machine learning techniques. Databases of such data, therefore, have significant value in their own right.
One of the most relevant types of IP protection is the sui generis database rights that were introduced by the Database Directive. These provide database owners with the right to prevent the unauthorised copying or extraction of data from their databases in the European Economic Area (EEA). After the end of the Brexit transition period (i.e. as of 1 January 2021), UK citizens, residents and businesses will no longer be eligible to receive or hold sui generis database rights in the EEA. However, database rights that exist in the UK or EEA before the end of the transition period (whether held by UK or EEA persons or businesses) will continue to exist in the UK and EEA for the rest of their duration.
Any IP licence which includes the licensing of database rights must take account of this change. One option, if circumstances permit, is to use neighbouring rights such as copyright and rights in confidential information. Proper drafting of the relevant licence and control of the information exchanged under that licence, may be vital if control of this valuable IP is not to be lost.
Obtaining patents for software can be difficult. Happily, in the field of healthcare innovation, this can be easier than in other technologies. So, patent protection for software in this area should not be ruled out.
Where machine learning techniques are involved, the difficulty of defining how the underlying technique actually solves a particular problem adds a further complication. It might be the case that the innovation lies in the manner in which training data is pre-conditioned, rather than in the design of the algorithm itself. In addition, machine learning innovation is often implemented ‘in the cloud’, and the processing engine itself may never be distributed. In so far as the customer is concerned, the technology is just a ‘black-box’. This creates a difficulty in policing patent infringement, which must be weighed against the need to disclose the details of an invention in any patent. This is a real consideration, and patents must be drafted carefully with this in mind.
This does not mean that patents are irrelevant in this space. At the very least, there is a risk that an infringement believed to be hidden might be discovered, and the financial and reputational damage that would arise
cannot be dismissed. In practice, these decisions are made by individuals – CEOs and General Counsel – who are then accountable to their board/shareholders for that decision. In that context, legal advice which says infringement will not be detectable is not to be given or accepted lightly. In addition, there is the question of what would be done in the event that an invention is kept secret but subsequently patented by a competitor. The original inventor would then be left to decide whether they were prepared to run that risk themselves and rely only on the very narrow defence provided by their own secret prior use. The right decision may well be not to file a patent application, but that decision should be taken positively, with full awareness of the costs and benefits.
Big data and machine learning techniques in general and their application to healthcare in particular, are generating exciting new opportunities. The circumstances of each case are unique, and raise complex new issues as the regulatory, legal and technical landscape evolves. Seizing those opportunities requires an IP strategy which is adapted for those circumstances and is far-sighted enough to see the next challenge coming.
This article was originally published in Open Access Government in February 2020.