The largest community of pharma leaders

Intellectual Property Protections for the Algorithms Used by the Healthcare Industry to Decode Big Data


As healthcare stakeholders begin to embrace real world evidence, along with many other forms of big data applications, there remains the issue of how to protect the underlying algorithms that derive meaning from this data. An algorithm is a very broad term that typically refers to the equations and/or a series of steps for a computer program that help organize and extract desired information. Protecting an algorithm from being stolen by a competitor or otherwise misappropriated requires enforceable legal remedies. Though algorithms are an absolute necessity when decoding big data, they do not fit well within most of the conventional intellectual property frameworks. For example, an algorithm is classified by courts as a mathematical equation, which, makes it nothing more than an abstract idea. This classification as an abstract idea is problematic because the US Supreme Court has consistently held abstract ideas cannot be patented. (1).

Can an Algorithm be Copyrighted?

An algorithm runs into numerous issues if it attempts to seek protection under copyright law. At first view the requirements seem relatively easy to meet, as the primary threshold is an original idea fixed in some sort of medium (even an internet blog may suffice). However, a copyright only protects the expression of an original idea, not the idea itself. (3). This means the copyright will offer no secrecy rights. Thus, there will still be no way to prevent someone from reading the algorithm and using the information themselves; so long as they don’t claim it to be their own idea. This means trade secrets and process patents are the only viable options for an algorithm.

Is There a Way for an Algorithm to be Protected as a Trade Secret?  

Protecting the economic value of an algorithm as a trade secret has several advantages. Trade secrets are free, potentially last forever, and do not require filing anything with a government agency. In fact, do not file a trade secret with a government agency unless specifically asked to do so. Unless instructed otherwise, a document provided to a government entity will typically be considered a public record accessible to anyone, thus, making it no longer protectable as a trade secret. The legal definition for trade secrets is very accommodating, even for something as abstract as an algorithm. Trade secrets by law will protect any information a company or person uses to derive actual or potential independent economic value from, so long as “reasonable measures” are taken to keep it secret. The term “reasonable measure” is applied leniently; a typical example might be a password to access a locked cell on a spreadsheet or information shared after the signing of a nondisclosure/confidentiality agreement.

Trade secrets are a relatively new area of law that was only codified into US federal statute in 2016 with the Defend Trade Secrets Act (DTSA). Before the DTSA, trade secrets were state and common law remedies, which, meant a company would have to sue an offender in every single state a violation occurred to enforce them or find a way to get the suit into federal court, where there was little predictability. (4). However, with the DTSA, along with an increasing number of cases, trade secrets are becoming better understood as financially critical to businesses. Trade secrets recently made their way to Supreme Court through Sandoz v. Amgen as an underlying matter in a biosimilar lawsuit. (3). The DTSA also applies to conduct of someone outside of the United States if they are a US citizen or a from another country but within the US when they misappropriated the information. However, if the nondisclosure agreement doesn’t include critical language regarding the parties’ rights to disclose the information to their attorney or the authorities if it relates to a criminal matter, they may not be able to seek relief under the DTSA. Instead, the person seeking a legal remedy for their algorithm may have to rely on state courts.

Trade secrets have been successfully enforced even in complex biotechnology cases, such as Salsbury Labs., Inc. v. Merieux Labs., Inc., where the court blocked former employees from disclosing misappropriated trade secrets of the process surrounding its biologic. The court granted lost profits, including a rare decision to reimburse some of their litigation expenses. Under the Economic Espionage Act, trade secrets may have criminal penalties as well. However, criminal actions are at the discretion of the government not the company. But in the recent case United States v. Hsu found an attempted plot to steal trade secrets from Bristol-Meyers Squid did qualify for criminal penalties.

Trade secrets do have some drawbacks though. For example, unlike a patent, trade secret protection cannot prevent reverse engineering (where a competitor acquires the product and disassembles it to discover the information). Nor will it protect against independent development (where a firm discovers the idea through their own research efforts). Also, once publicly disclosed accidental or otherwise, a trade secret’s protection is forever lost. For example, if the algorithm were disclosed at a conference, where not everyone was under a confidentiality agreement. Even so, disclosing the information to a large group is risky as writing “confidential” next to the equation will likely be insufficient, as the underlying policy requires a level of controlled access. Therefore, a common defense used by someone accused of misappropriating a trade secret is to locate the information somewhere available to the public or another place that is not strictly confidential. So long as the information was available prior to the alleged date of misappropriation they are likely to have a valid defense.

Will Patent Law Protect an Algorithm?   

The algorithm itself absolutely cannot be patented as it’s an abstract idea (1). However, the applications of its various uses for it can sometimes be patented as processes. But, these types of patents are difficult to enforce as they can be designed around and prove challenging to detect. Unlike in manufacturing where a finished product can be disassembled, there is usually is not a material good generated directly by an algorithm. Patents can also be expensive to file and enforce, unlike trade secrets.


Ultimately, the best option to protect the value of an algorithm is to preserve the equation as a trade secret, then file process patents on all foreseeable applications for it. Copyrights needlessly risk exposing the idea, while offering no protection whatsoever.

1. Association for Molecular Pathology v. Myriad Genetics, Inc., 133 U.S. 2107 (2013); Diamond v. Chakrabarty, 447 U.S. 2204 (1980).
2. Feist Publications Inc. v. Rural Telephone Service Co., 499 U.S. 340, 346 (1991)
3. Biosimilar Regulatory Policies Issues and Implications: Where are We Headed? Arkells N, Chopra A, Chopra I. 23rd Annual International Meeting of the International Society for Pharmacoeconomics and Outcomes Research, Baltimore, MD, United States, 2018.
4. Arkells N. Why Biosimilar Companies Should Pay More Attention to U.S. State Governments. Life Science Connect: Biosimilar Development & Bioprocess. PA. 2018.

Recent Articles