As technology evolves, multiple sectors including healthcare are looking to harness the immense potential of artificial intelligence (AI). For the healthcare industry, AI has so far been mostly utilized to manage patient medical records, diagnostic screening, identification and treatment patterns, and track and solve nonadherence. There have been instances of highly complex usage of AI, such as the robotics surgical platform developed by Intuitive Surgical, which recently gained FDA clearance. Such instances of highly specialized applications of AI in healthcare are rare but, considering the rapidly evolving technology, more AI-based products interacting directly with the patients are expected to be introduced in the market over the next few decades.
Why Regulations Could Deter AI-Based Product Developers
As the technology evolves, the biggest obstacle for AI adopters will be regulations. Traditionally, healthcare devices were basically hardware that were evaluated by regulators for their performance characteristics. Software codes were also evaluated and the go ahead was then given by regulators. Now consider a scenario where a device learns from its experience and mistakes, modifies its program and improves itself. In such cases, the process of gaining regulatory approval for such a “dynamic” AI-based product is a grey area for developers. By the way, the above scenario is a real-life example — U.S.-based Freenome is developing a colorectal cancer screening test that will be able to learn from its mistakes and evolve to become more accurate.
Regulatory Landscape in the U.S.
From a regulatory standpoint, the U.S. is currently much better placed compared to other countries when it comes to AI-based products. To meet new expectations from the industry, the US FDA implemented the 21st Century Cures Act in 2016, which distinguishes between normal medical software and “clinical decision support” software. Its Digital Health Innovation Action Plan also promises clarity and strong regulations governing digital heath products. For AI, the most notable approval pathway has been the De Novo Premarket Review Pathway. Under this pathway, the FDA cleared a novel medical device, where an AI-based software was developed to detect, interpret and decide the course of action for retinopathy without the assistance of a healthcare professional. Therefore, the De Novo pathway may be the right move forward for AI-based healthcare innovators for quickly accessing the U.S. market. In contrast, the European markets (including the U.K.) are poorly prepared to support AI innovators. Such a scenario is hindering the growth of novel AI-integrated medical devices in the European market.
Such uncertain scenarios create further complications for innovators as they grapple with other issues such as showcasing clear benefits of their technology and products. Then, there is a concern related to post-marketing complications after the product gets approved by the regulators. Further, recent developments relating to user privacy data — such as concerns about unauthorized data collection by technology companies, and the data privacy regulations such as GDPR — are making the situation even trickier for healthcare AI innovators.
The healthcare industry is on the cusp of the AI revolution. The technology has tremendous potential to drastically improve the overall healthcare ecosystem, but it can be only achieved if the innovators feel confident about getting timely regulatory approvals. The current healthcare product approval regulations were best for evaluating 20th century products. This age requires a fresh approach toward such technologically advanced products and appropriate regulatory guidelines for such technologies to find greater uptake. The biggest bottleneck for applications for such path-breaking innovations will sadly be the regulators themselves.