FDA Clinical Decision Support Software Guidance Refresh : Clearer Rules, Broader Innovation

On January 6, 2026, the FDA issued a Guidance for Industry and Food and Drug Administration Staff regarding Clinical Decision Support (CDS) software. The January 2026 guidance supersedes the previous, and less detailed, guidance document on the topic issued on September 28, 2022.

The non-binding recommended guidance builds on the 2022 guidance with refined interpretations, practical examples and material enforcement discretion. For investors, counsel and product leaders, the message is simple: if the software relies on ordinary clinical information, offers options rather than commands, and makes its reasoning clear enough for a clinician to evaluate, you are likely outside device regulation. If it interprets signals or images, issues time‑critical directives, or obscures its logic, expect medical device oversight.

As a reminder, the 21st Century Cures Act excludes certain CDS from the device definition if four things are true.

• First, the software does not acquire, process, or analyze medical images, IVD signals, or patterns from signal‑acquisition systems.

• Second, it works with “medical information” that clinicians already use—diagnoses, discrete lab results, demographics, discharge summaries, and trusted references like guidelines and peer‑reviewed studies.

• Third, it supports clinical judgment by suggesting or prioritizing options, rather than issuing definitive diagnoses or treatment commands—especially in urgent settings.

• Fourth, it is transparent: intended use and users are clear; inputs and data quality are described; methods, data sources, validation, and limitations are explained in plain language; and outputs include patient‑specific context so clinicians can independently review the basis of the recommendation.

FDA’s January 2026 clarifies, among other things, the following as related to each criterion:

• Criterion 1: Input boundaries. The 2026 guidance expands on what FDA considers a “signal”, “pattern” of information as distinguished between a single, discrete measurement and functions that assess or interpret the clinical implications or clinical relevance and provides helpful examples for each.  

• Criterion 2: Definition of “medical information”. FDA clarified that the type of “medical information about a patient” indicative of data inputs used in non-device CDS is interpreted to include patient demographics, symptoms, certain test results, discharge summaries, and well-understood, accepted sources (e.g., clinical guidelines, peer-reviewed literature, FDA-required labeling). FDA also clarified that a single, discrete test result can be “medical information”, while continuous or repeated measurements are considered a “pattern or signal”, indicative of the type of data inputs used in devices under Criterion 1. 

• Criterion 3: Support vs. directive. FDA indicated enforcement discretion in cases where only one option is clinically appropriate, and the software function otherwise meets the remaining criterion. FDA additionally provided examples of software functions which may fall into the enforcement discretion bucket:

  • Software function that creates a recommended treatment plan, including possible medication(s), for patients diagnosed with cognitive impairment for a health care provider to consider based on the patient’s diagnosis and reviewed and finalized by the provider (but not the same functionality which also analyzes positron emission tomography (PET) scan images).

  • Software function that recommends a specific FDA-approved antibiotic agent for a health care provider to consider based on the patient’s history (but not the same software functionality which also analyzes spectroscopy data to diagnose bacterial infections). 

  • Software function that analyzes a radiologist’s clinical findings of an image to generate a proposed summary of the clinical findings for the patient’s radiology report, including a specific diagnostic recommendation based on clinical guidelines and reviewed and finalized by the provider (but not the same software functionality which also analyzes the image to generate the clinical findings and/or make measurements). 

• Criterion 4: Transparency. FDA detailed concrete expectations for enabling independent review for the basis of recommendations in order to meet criterion 4. FDA recommends that:

  • The software or labeling includes the purpose or intended use of the product, including the intended user and patient population. 

  • The software or labeling identify the required input medical information, with plain language instructions. 

  • The software or labeling provides a plain language description of the underlying algorithm development and validation that forms the basis for the CDS implementation, sufficient for the provider to understand the basis of the recommendation. 

  • The software provides the health care provider with relevant patient-specific information and other knows/unknowns for consideration that enable the health care provider to independently review the basis for the recommendations and apply their judgement. 

FDA concludes the January 2026 guidance with several illustrative examples of non-device CDS software functions applicable to each criterion, as well as device software functions. Developers who wish to remain outside device regulation should design software functions to algin with FDA’s clarifications, and by investing in labeling and user interface transparency that supports independent clinical judgment, above all.