Medical Devices

How Regulators Treat Software in Medical Devices: Intended Use Comes First

Whether software counts as a regulated medical device depends mainly on what the manufacturer says it is for, not on how the product is built.

TETom Eriksen4 min read
Gloved hands drawing liquid from a vial with a syringe in a laboratory
Key takeaways
  • Regulators look at intended use: the purpose a manufacturer claims for the software.
  • Software can be a medical device on its own, or a component inside a hardware device.
  • General wellness functions are treated differently from functions that diagnose or treat disease.
  • Manufacturers and buyers should check labeling and regulator databases, not marketing pages.

A mobile app that displays a step count and an app that analyzes an electrocardiogram can look similar on a phone. Regulators treat them very differently. This article explains the logic, using the United States as the main example, while noting that other regions apply similar ideas in their own ways.

Intended use drives the category

The central question is what the software is intended to do. The intended use is established by the manufacturer’s claims: labeling, instructions, marketing and the way the product is presented. If the software is meant to diagnose, treat, mitigate or prevent disease, or to affect the structure or function of the body, it is likely to meet the definition of a medical device.

This means that two products with the same code can fall into different categories. A tool that charts glucose values for general interest and a tool that recommends insulin doses are different in purpose, and so in regulatory treatment.

What counts as a claim

  • Statements on the product label or in the app store listing.
  • Instructions for use and training material.
  • Sales material and website copy.
  • Features that clearly imply a medical purpose.

Software as a device and software in a device

Regulators distinguish between two arrangements.

Software in a medical device

The software is part of a hardware device, such as the firmware in an infusion pump or the control software of an imaging scanner. It is evaluated as part of the device.

Software as a medical device

The software performs a medical function on its own, without being part of hardware. Examples can include software that analyzes images to detect a condition, or an app that calculates a dose from entered values. Here the software is the device.

Risk and the level of oversight

Devices are generally grouped by risk. Lower-risk devices face lighter requirements, while higher-risk devices need more evidence before they reach the market. In the United States, those groups are known as classes, and the pathway to market depends on the class and on whether a similar legally marketed device already exists.

The more a wrong output could harm a patient, and the less a clinician can check it, the more evidence a regulator generally expects.

Wellness and administrative software

Not all health software is regulated as a device. Regulators have published policies on products that support general wellness, such as tracking activity or encouraging healthy habits, and on certain administrative or record-keeping functions. These policies depend on the specific function. A product can contain regulated and unregulated features at the same time, and each feature is judged on its own claims.

What changes after the product is on the market

Software changes often, and regulators recognize this. Manufacturers need processes for managing changes, for monitoring problems in the field and for reporting certain events. Regulators have also discussed approaches for software that learns or updates, such as plans that describe in advance what kinds of changes a manufacturer expects to make.

Practical steps for different readers

For developers

  1. Write down the intended use before building features.
  2. Review marketing language against that statement.
  3. Ask a regulatory specialist early, since the answer shapes design and evidence.

For hospitals and buyers

  1. Ask vendors for their regulatory status and the exact scope of any clearance.
  2. Check public regulator databases to confirm.
  3. Treat features without a clinical claim as outside the validated scope.

For clinicians

Know which functions of a tool have been evaluated, and which are conveniences. A cleared function does not extend to every other feature in the product.

Different regions, similar logic

The European Union, the United Kingdom and other jurisdictions have their own rules and terms, but they share the focus on intended purpose and risk. Anyone selling across borders should check each market, since classification and evidence requirements can differ.

The practical lesson is that regulation follows claims. A team that is precise about what its software is for can predict the path it faces, and a buyer who reads those claims carefully knows what the vendor has and has not shown.

References
  1. Device Software Functions Including Mobile Medical Applications (U.S. FDA)
  2. Digital Health Center of Excellence (U.S. FDA)
  3. Medical device (Wikipedia)
  4. Medical software (Wikipedia)
TE

Tom Eriksen

Industry Analyst

More in Medical Devices

See all →
A medical scan displayed on a screen in a radiology setting

AI-Enabled Imaging Devices: What Regulatory Clearance Does and Does Not Tell You

A regulatory listing for an AI imaging tool shows it was reviewed for a defined use, which is not the same as proof it works in every hospital.

Tom EriksenOct 5, 2026
A small white digital glucose monitoring device lying beside a white pen on a table

Continuous Glucose Monitors Beyond Diabetes: Reading the Evidence Carefully

Continuous glucose monitors are established tools for diabetes care, while their use by people without diabetes rests on far thinner evidence.

Tom EriksenSep 5, 2026
A person touching the screen of a smartwatch on their wrist

Wearable Sensors in Clinical Practice: What Consumer Devices Can and Cannot Tell a Doctor

Consumer wearables record heart rate, rhythm and activity, but clinicians weigh that data differently from a diagnostic-grade measurement.

Tom EriksenAug 24, 2026