EducraftPROFESSIONAL LEARNING
Home / Resources / Educational Articles / Medical Device Software Classification
EDUCRAFT EDUCATIONAL ARTICLES · MEDICAL DEVICE SOFTWARE CLASSIFICATION

How Should Manufacturers Classify Software-Driven Medical Devices in the US and EU?

Software medical device classification begins with the clinical purpose of the software, not the technology used to build it. This guide explains how manufacturers can approach FDA and EU classification decisions, document their rationale, and evaluate changes that may affect regulatory status.

Medical device software interface displaying clinical imagery and monitoring data

Software is increasingly central to medical diagnosis, patient monitoring, clinical decision-making and therapeutic delivery. Yet the regulatory classification of a software-driven medical device cannot be determined simply by asking whether the product uses artificial intelligence, operates autonomously or runs on a mobile platform.

The essential questions concern intended use, medical purpose, patient risk and the regulatory framework in the market where the product will be supplied.

For manufacturers, classification is an early decision with consequences for the evidence required, the applicable regulatory controls and the route to market.

Start with the intended medical purpose

A software function does not automatically become a regulated medical device because it processes clinical information.

In the United States, the Food and Drug Administration (FDA) evaluates whether a software function meets the statutory definition of a medical device, taking account of relevant exclusions and regulatory policies.

The FDA's Digital Health Policy Navigator provides a structured starting point for evaluating software functions.

Consider the difference between three hypothetical applications:

  • An application displaying general educational information about healthy eating.
  • Software calculating a patient-specific medication dose for use in treatment decisions.
  • Software controlling delivery of medication through a connected device.

Although all three use software, they perform substantially different functions. Their intended purposes, possible consequences of failure and regulatory treatment may differ.

The manufacturer should therefore describe precisely what the software does, who uses it, which patients it serves and what decisions or actions depend on its output.

Determine whether the software is regulated

A useful first-stage assessment asks:

  1. Does the software have an intended medical purpose?
  2. Does it meet the applicable legal definition of a medical device?
  3. Does a statutory exclusion or relevant regulatory policy apply?
  4. If it is a medical device, what classification and controls apply?

These questions must be answered for each relevant software function rather than relying exclusively on the marketing description of the complete product.

For example, an application might combine general wellness content, appointment scheduling and a function that analyzes patient data to support diagnosis.

Different functions may require different regulatory assessments.

FDA discusses this distinction in its guidance on device software functions and mobile medical applications.

An important distinction is that FDA enforcement discretion does not necessarily mean a software function is excluded from the definition of a device. Those are separate regulatory concepts.

Understand FDA classification

FDA classifies medical devices according to the regulatory controls necessary to provide reasonable assurance of safety and effectiveness.

The familiar framework includes:

  • Class I: Generally subject to general controls.
  • Class II: Generally subject to general and special controls.
  • Class III: Generally subject to general controls and premarket approval requirements.

These are broad categories, not a reliable shortcut for classifying a particular software product.

The actual classification depends on the applicable regulation, product type, intended use and other relevant characteristics.

FDA's Classify Your Medical Device resource explains how manufacturers can research classifications.

A practical classification exercise should identify the likely product code, classification regulation and appropriate regulatory pathway.

Manufacturers should also examine comparable legally marketed products to understand how similar intended uses have been classified.

Similarity of technology alone is insufficient. Two applications may use comparable algorithms while making very different clinical claims.

Understand classification under the European MDR

In the European Union, medical device software is assessed under Regulation (EU) 2017/745, or the Medical Device Regulation (MDR), when that regulation applies.

For qualifying medical device software, Rule 11 in Annex VIII is especially relevant.

Rule 11 addresses software intended to provide information used for diagnostic or therapeutic decisions, as well as software intended to monitor physiological processes. Depending on intended purpose and potential consequences, different classes may apply.

However, Rule 11 should not be applied before determining whether the software qualifies as medical device software in the first place.

The European Commission's MDCG 2019-11 revision 1 guidance provides detailed interpretation of software qualification and classification under the MDR and IVDR.

The MDR is binding legislation. MDCG guidance helps explain its application but is not itself legislation.

Also, the European Medicines Agency is not the general classification authority for all EU medical device software. Manufacturers should assess the applicable MDR/IVDR framework and relevant competent-authority or notified-body processes.

Why automation and AI increase classification complexity

The use of machine learning does not by itself establish a device's classification.

Nevertheless, AI-enabled functions can make intended use, performance limitations and risk analysis more complex.

Consider a hypothetical software system that analyzes medical images and flags findings for clinical review.

Its regulatory assessment must consider the actual intended use and the role the output plays in clinical decision-making.

Does it merely organize information? Does it identify findings? Does it support a diagnosis? Is the clinician expected to independently review the underlying information?

The answers can materially affect the regulatory assessment.

Manufacturers should describe the workflow rather than using broad statements such as "AI-powered clinical assistant."

A precise description should identify inputs, outputs, intended users, target populations, clinical claims and the consequences of an incorrect result.

When should classification be reassessed?

Classification is not necessarily a one-time activity.

Changes to software can affect its intended purpose, clinical functionality or risk profile.

Potential reassessment triggers include:

  • Adding a new diagnostic or therapeutic function.
  • Expanding the intended patient population.
  • Changing from informational support to automated control.
  • Introducing new clinical claims.
  • Extending the software into a new regulatory jurisdiction.
  • Modifying how a connected medical device is controlled.

Not every software update changes regulatory classification.

For example, a minor usability improvement may leave the intended purpose unchanged, although it can still require assessment under applicable change-control requirements.

By contrast, adding a function that independently influences treatment may require a much broader regulatory evaluation.

Build a defensible classification rationale

A practical classification record should contain:

Product and function description: Identify the software functions being evaluated, including their intended medical purpose.

Intended users and environments: Describe clinical professionals, patients, caregivers or other users, together with the intended setting.

Clinical significance: Explain what decisions the software informs or controls and what could happen if its output is incorrect.

Regulatory assessment: Document the applicable legal definitions, exclusions, classification rules and product-code research.

Evidence and references: Retain the legislation, FDA resources, guidance and comparable device information used.

Approval and maintenance: Assign ownership for reviewing classification when intended use or functionality changes.

This approach helps avoid classification decisions based on informal assumptions or unsupported comparisons.

Practical takeaway

Medical device software classification should be treated as a documented regulatory analysis, not a label assigned by a development team.

A useful starting exercise is to select one software function and write a single paragraph describing its intended medical purpose, user, clinical decision and potential failure consequences.

Then evaluate that paragraph separately against US and EU requirements.

The result may reveal differences that would otherwise remain hidden until a regulatory submission or conformity assessment.

Explore the related EduCraft session

For further discussion of software-driven medical device classification, global frameworks and reclassification considerations, explore EduCraft's Reclassification of Software Automated Medical Devices session. Check the course page for current access and availability.