Consumer Technology Association Issues ANSI-Accredited AI Standards for Healthcare
On February 25, 2020, the Consumer Technology Association (CTA) released the first ever ANSI-accredited standards for the use of Artificial Intelligence (AI) in the healthcare industry. CTA brings together a wide spectrum of consumer technology experts – from regulators, engineers, doctors and scientists to industry leaders – to help develop specifications that define how products work and the ways in which consumers interact with them. In the case of the healthcare industry AI standards, CTA collaborated with 52 organizations to develop a set of definitions and characteristics that will provide better guidance and common understanding of terminology so that consumers, healthcare providers, payers and healthtech companies can more effectively develop and implement the use of artificial intelligence based technologies within the healthcare industry.
Artificial intelligence can be an important driver of innovation in care delivery, diagnosis and treatment, and can create efficiencies across the healthcare spectrum. For example, MIT has recently announced that researchers were able to identify a powerful new antibiotic compound using artificial intelligence in a fraction of the time and expense of previous methods for screening antibiotics.
However, as new and innovative ways of using AI have been introduced by both start-ups and large technology companies, there has been uncertainty and ambiguity in certain AI-related terms, causing confusion and delays in actual broad based adoption and implementation of such technology. For example, in a November 25, 2018 article in HealthcareITNews, Leonard D'Avolio, a Harvard Medical School professor and CEO of Cyft, noted that, "If I describe what I do as cognitive computing, but a competitor describes what they do as AI or machine learning or data mining, it's hard to even understand what problems we are trying to solve."
Creating clarity around such foundational definitions is necessary for adoption of broad-based telehealth and remote patient monitoring solutions, as well as technology used by physicians to diagnose disease, create more individualized treatment protocols and improve outcomes. To address these issues, CTA issued two reports this week. The first, ANSI/CTA-2089, contains AI standards which are meant to apply across the entire consumer technology industry, and defines general terms relating to the definition and characteristics of Artificial Intelligence, including distinguishing “artificial intelligence” and “big data” from “machine learning” and “deep learning”.
The second report, ANSI/CTA-2089.1, seeks to address the unique challenges of using AI within the healthcare industry, including: data availability and data interoperability; health and safety repercussions that may occur as a result of incorrect decision-making when using AI technology; and data distribution shift, where the performance of AI algorithms are impacted when the statistical data encountered in practice differs from the data used to train the algorithm.
While in other non-regulated industries, it may be easy to simply recalibrate models, the various regulatory regimes around using and sharing healthcare data (FDA regulations, and HIPAA), as well as the length of service that may reach decades for many medical devices, can impact the performance and efficacy of AI driven products – putting patients at risk of incorrect diagnoses or treatment. Additionally, to further complicate these matters, devices are not simply hardware, and the FDA has issued a proposed regulatory framework for “Artificial Intelligence and Machine Learning in Software as a Medical Device.” CTA also pointed out the challenges of deploying AI in healthcare because of the various overlapping state and federal regulatory approaches which must be aligned to ensure patient safety. Thus, in ANSI-CTA-2089.1, CTA creates standard definitions such as “de-identified data” and “pseudonymized data”, “software as a medical device”, “remote patient” monitoring and “patient decision support system” which relate to these concerns.
In an increasingly complex world of technology, simply having a common lexicon will help to improve the safety and efficacy of digital health innovations, medical devices and information and hopefully accelerate adoption and implementation. The next steps should include further standardization of technologies and processes.
If you have any questions or would like additional information, Dana Petrillo (firstname.lastname@example.org, 215.864.7017), Richard Borden (email@example.com; 212.631.4439) or Lori Smith (firstname.lastname@example.org, 212.714.3075).