FDA-authorized digital medical devices have grown substantially over two decades, but regulatory databases still can't track them

The number of FDA-authorized medical devices with digital components has grown substantially over the past two decades, with sharp variation across clinical specialties. That is the headline finding of a new peer-reviewed study published in Nature npj Digital Medicine, which applied text analysis to tens of thousands of FDA regulatory documents to map the digital transformation of the U.S. medical device industry in a way that existing regulatory databases simply cannot.

The research arrives as the FDA has been actively building out its digital oversight infrastructure. The agency established its Digital Health Center of Excellence in 2020, has since published more than twenty guidance documents with digital health content, and maintains a regularly updated list of AI and machine learning-enabled medical devices, according to the FDA. Yet the study makes clear that the regulatory data architecture underpinning all of that activity has a structural gap that carries real consequences for operators.

A two-decade surge the databases can't see

The Nature study pioneered a new application of text analysis, combing records from tens of thousands of regulatory submissions for newly authorized devices to reconstruct how digitization has spread across product categories and clinical specialties over twenty years. The authors found not only that digital device authorizations have risen substantially, but also that the growth pattern is uneven. Some specialties have digitized far faster than others, a distinction that existing tracking tools obscure entirely.

The core problem the researchers identified is that FDA regulatory databases do not flag whether a device has software components at all. They also do not distinguish between a standalone Software as a Medical Device, known as SaMD, and a combination product that pairs digital functions with physical hardware, known as Software in a Medical Device or SiMD. Without that distinction baked into the underlying data, post-market surveillance activities cannot reliably apply software-specific criteria to the devices already circulating in clinical settings.

For health system procurement leaders and clinical engineering teams, that gap is not abstract. A device cleared years ago may now carry connectivity or algorithmic functions that trigger cybersecurity obligations or post-market monitoring requirements, but those obligations cannot be systematically identified from the regulatory record alone. The Nature study quantifies the scale of that blind spot for the first time.

What the FDA has built so far

The FDA's Digital Health Center of Excellence, as described on the agency's site, is organized around three objectives: building partnerships to accelerate digital health advancement, sharing knowledge to drive best practices, and innovating regulatory approaches to provide efficient oversight without adding unnecessary burden. Its stated anticipated outcomes include harmonizing international regulatory expectations, achieving consistent application of digital health policy, and ultimately reimagining the medical device regulatory framework for software-native products.

The center has been active on multiple fronts. It has launched cybersecurity initiatives, published guidance on augmented and virtual reality in medical devices, released draft guidance on AI-enabled device software functions and lifecycle management, and introduced the concept of Predetermined Change Control Plans, or PCCPs, for AI-enabled devices. The FDA has also recently announced the first participant selected for its TEMPO pilot program for digital health devices, a pre-submission program designed to reduce friction in the authorization process, according to the agency's Digital Health Center of Excellence pages.

More than twenty guidance documents with digital health content are now publicly available. The regularly updated AI/ML-enabled medical devices list gives procurement teams a starting point for identifying which cleared products carry algorithmic functions, though the Nature study's findings suggest that list captures only a slice of the broader digital device universe.

The heterogeneity problem for health system operators

One of the more operationally relevant findings in the Nature study is the heterogeneity across clinical specialties. The digitization rate has not been uniform. Some areas of medicine have seen their authorized device portfolios shift heavily toward digital and software-reliant products, while others remain largely analog. For integrated delivery networks and health systems managing diverse equipment inventories, that uneven distribution complicates both procurement planning and cybersecurity risk stratification.

A radiology department, for example, may face a very different software exposure profile than an orthopedics or physical therapy unit, even within the same facility. Without a reliable way to query the regulatory record for software components by specialty or product category, clinical engineering and IT security teams are effectively working from incomplete inventories. The Nature researchers' text-analysis method offers a potential model for closing that gap, though it would require adoption at the database infrastructure level to become operationally useful at scale.

What comes next in the regulatory build-out

The FDA has signaled its intent to continue modernizing oversight through the Digital Health Center of Excellence's stated goals, which include developing a reimagined regulatory paradigm specifically tailored for digital health technologies and aligning that framework with international standards. The agency's interactive Digital Health Policy Navigator tool is designed to help industry and operators map applicable policies to specific device types.

The Nature study effectively sets a benchmark: a methodologically grounded count of how many digital devices have been authorized, how fast that number has grown, and where the growth has been most concentrated. For regulatory affairs, procurement, and clinical technology teams, it frames the monitoring challenge precisely. The next concrete marker to watch is whether the FDA incorporates software-component tagging into its regulatory databases, a change the study's authors identify as necessary for post-market surveillance to keep pace with the devices already on the market.