On January 2026 FDA updated Clinical Decision Support guidance. Updated guidance: Clinical Decision Support Software. This updated guidance supersedes “Clinical Decision Support Software” issued on September 28, 2022 guidance document.
Definitions
To make the guidance easier to follow, here are a few key terms used throughout this article:
CDS (Clinical Decision Support)
Software intended to support clinical decision-making related to the diagnosis, treatment, prevention, cure, or mitigation of disease.
Non-Device CDS
Clinical decision support software functions that do not meet the definition of a medical device under Section 201(h) of the FD&C Act.
Section 201(h) – Medical Device Definition
Under the FD&C Act, a medical device is an instrument, apparatus, machine, implant, in vitro reagent, or similar article that is intended to diagnose, treat, prevent, or mitigate disease, or affect the structure or function of the human or animal body and that does not achieve its primary intended purpose through chemical action or metabolism.
What is the purpose of this guidance?
This guidance clarifies when clinical decision support (CDS) software is regulated as a medical device and when it may qualify as Non-Device CDS.
When does software meets the Non-CDS criteria?
Under the 21st Century Cures Act (U.S. law), some clinical decision support (CDS) software functions may fall outside FDA medical device regulation but only if all four of the following criteria are met:
- No medical signal or image processing
The software must not acquire, process, or analyze medical images, signals from in vitro diagnostic devices, or continuous physiological signal patterns. - Uses existing medical information
The software may only display, analyze, or print already interpreted medical information, such as patient records, discrete test results, or established medical knowledge like clinical guidelines or peer-reviewed studies. - Supports but does not replace clinical decision-making
The software may provide recommendations to healthcare professionals, but it must not make diagnoses, dictate treatment decisions, or otherwise replace clinical judgment. - Enables independent clinical judgment
The software must allow clinicians to understand and independently review the basis for its recommendations, so they do not rely primarily on the software when making decisions for individual patients.
All four criteria must be met.
Failing any single criterion means the software remains a medical device under FDA oversight.
Criteria 1. No medical signal or image processing
The software does not acquire, process, or analyze medical images, signals from in vitro diagnostic devices (IVD), or patterns from signal acquisition systems
Under the FDA guidance, a software function cannot qualify as Non-Device CDS if it is intended to acquire, process, or analyze medical images, IVD signals, or continuous physiological signal patterns.
In simple terms:
If the software takes raw medical images or signals as input, it remains a medical device and is subject to FDA oversight.
What counts as medical images and signals?
FDA interprets these concepts broadly. Medical images include those generated by imaging systems such as CT, X-ray, ultrasound, and MRI, as well as images used for clinical purposes in areas like pathology or dermatology.
NOTE: Even images not originally captured for medical use become medical images if they are later analyzed for clinical decision-making.
Signals and patterns refer to repeated or continuous measurements over time and may include physiological waveforms such as ECG signals, genetic sequences obtained through next-generation sequencing, continuous glucose monitor data, as well as breathing patterns recorded during sleep.
What clearly fails Criterion 1
What clearly fails Criterion 1 is software that analyzes or interprets medical images, signals, or patternssuch as detecting abnormalities in images, identifying arrhythmias from ECG waveforms, analyzing genetic sequences for variants, or diagnosing sleep apnea from breathing patterns, in which case the software remains classified as a medical device.
In all of these cases, the software remains a medical device.
Allowed (potential Non-Device CDS):
A CDS tool that uses discrete lab results, diagnosis codes, and clinician-interpreted reports from the EHR to provide guideline-based reminders or treatment considerations.
Not allowed (still a device):
A CDS tool that processes a raw ECG waveform to calculate heart rate or detect arrhythmias.
Criteria 2. Intended for the purpose of displaying, analyzing, or printing medical information about a patient or other medical information
The software function is intended to display, analyze, or print medical information about a patient or other medical knowledge, such as peer-reviewed clinical studies or clinical practice guidelines.
If the software uses interpreted medical information not raw signals or images, it may fall outside the medical device definition, provided that all other CDS criteria are also met.
What FDA means by “medical information” refers to a broad range of content, including patient-specific information commonly used in clinical care and decision-making such as: demographics and symptoms, discharge summaries and clinical notes, discrete test results like blood pressure or laboratory values, and clinician annotated reports (e.g., ECG interpretations).
Medical information may also include general medical knowledge, such as peer-reviewed clinical studies, clinical practice guidelines, textbooks, approved drug or device labeling, and recommendations from government agencies, and is typically information that can be understood and discussed in a clinical conversation between healthcare professionals.
Important is that this is the type of information that can typically be understood and discussed in a clinical conversation between healthcare professionals.
Examples of allowed medical information inputs
FDA provides clear examples of medical information that may be used by Non-Device CDS software, including: radiology report summaries (e.g. “BI-RADS category 4 lesion present”), clinician-annotated ECG reports (e.g. “signs of atrial fibrillation”), blood pressure readings (e.g. 120/80 mmHg), and laboratory test results recorded in the EHR (e.g. potassium 4.0 mmol/L). But note, these software functions are not considered medical devices only if all other applicable CDS criteria are also satisfied.
Simple way to remember this
Allowed (potential Non-Device CDS):
Software uses EHR-based lab results and patient history to present guideline-based reminders.
Not allowed (still a device):
Software that takes continuous physiological measurements (e.g. ECG or CGM data streams), and analyzes them to produce clinical conclusion.
Criteria 3. Supports but does not replace clinical decision-making
The software provides recommendations or decision support to healthcare professionals related to prevention, diagnosis, or treatment of a disease or condition.
Criterion 3 applies to software that supports or provides recommendations to healthcare professionals about the prevention, diagnosis, or treatment of a disease or condition.
FDA interprets this to mean that the software provides condition-, disease-, or patient-specific recommendations intended to inform or support clinical decision-making, while not replacing or directing the clinician’s judgment.
In short, the software may influence clinical decisions, but it must not make them.
A software function fails Criterion 3 if it provides a definitive diagnosis, issues a specific treatment directive, or delivers time-critical outputs that remove meaningful clinical discretion. If the output does not allow for independent clinical review, the software is considered a medical device.
When the other CDS criteria are met, software that may satisfy Criterion 3 includes risk prediction tools based on established clinical variables for long-term planning, treatment planning tools that present possible care or medication options for clinician review, and differential diagnosis tools that offer multiple or a single highly probable diagnostic considerations that are reviewed and finalized by a healthcare professional.
In all cases, the clinician reviews, revises, and ultimately finalizes the decision.
Software still remains a medical device if it: establishes diagnoses from images, waveforms, or signals, analyzes medical images or signals to generate findings, predicts near-term events that guide immediate intervention, relies on unvalidated or poorly understood data sources, or provides outputs intended for patients or caregivers rather than healthcare professionals.
Simple rule :
✅ If the software tells the clinician what to consider, it may qualify as Non-Device CDS.
❌ If it tells the clinician what to do, it is a medical device.
Criteria 4. Enables independent clinical judgment
The software allows healthcare professionals to independently review the basis for its recommendations, ensuring that clinicians do not rely primarily on the software when making diagnosis or treatment decisions for individual patients.
To qualify as Non-Device CDS, the software must allow healthcare professionals to independently understand and review the basis for its recommendations. The intent is to ensure that clinicians do not rely primarily on the software but instead apply their own professional judgment when making decisions for individual patients.
In short: The software may recommend, but it must clearly show why.
FDA does not require a specific technical implementation but expects that the software or its labeling provides clear and understandable information that enables independent clinical review.
In practice, this means the software should clearly define its intended use, including the intended healthcare professional user and patient population, be transparent about its inputs and any data quality requirements, and explain in plain language how recommendations are generated such as whether they are based on clinical guidelines, statistical models, expert consensus, or other approaches along with known limitations. The software should also provide patient-specific context by indicating which factors influenced the recommendation and highlighting any missing or uncertain information.
NOTE! Software intended for critical or time-sensitive clinical decisions generally does not meet Criterion 4.
Automation bias
FDA explicitly considers automation bias when interpreting Criterion 4. Highly automated outputs or recommendations intended for urgent clinical action increase the risk that clinicians will follow the software without sufficient independent review, making such software less likely to qualify as Non-Device CDS.
In some cases, usability testing may be needed to demonstrate that clinicians can understand and evaluate recommendations without being overwhelmed or pushed to blindly follow them.
One-line takeaway
Non-Device CDS must show its work and leave the final decision clearly in the hands of the clinician.
Simple rule:
✅ If a clinician can easily understand why a recommendation was made, Criterion 4 may be met.
❌ If the recommendation must be followed because there is no time or context to question it, Criterion 4 is not met.
Example
Example from the FDA guidance: Sleep apnea software
Note that the FDA guidance includes a wide range of examples. One illustrative case is Example 15, which highlights where the boundary between Non-Device CDS and a medical device is clearly crossed.
Example 15:
A software function analyzes breathing patterns from a sleep apnea monitor to diagnose sleep apnea or other conditions in patients.
FDA considers this software a medical device, because it fails multiple CDS criteria:
- Criterion 1: The software analyzes a pattern derived from physiological signals.
- Criterion 2: The software does not rely on existing, interpreted medical information.
- Criterion 3: The software provides a specific diagnostic output for a disease or condition.
As a result, software that analyzes breathing patterns and outputs a sleep apnea diagnosis fails Criteria 1, 2, and 3 and is clearly subject to FDA medical device regulation.
How this could look as CDS instead
By contrast, a software function that:
- uses existing medical information (e.g. breathing patterns documented in an HCP-reviewed sleep study report), and
- provides a prioritized list of treatment options to support clinical assessment
may meet Criterion 3, provided the other CDS criteria are also satisfied.
The key difference is that the software supports clinical judgment rather than making a diagnosis.
Summary
FDA’s Clinical Decision Support (CDS) guidance draws a clear line between medical devices and Non-Device CDS.
A CDS software function may fall outside FDA medical device regulation only if all four criteria are met:
- It does not analyze medical images or signals.
- It uses existing, interpreted medical information.
- It supports but does not replace clinical decision-making.
- It allows clinicians to independently understand and review the basis for its recommendations.
Failing any single criterion means the software remains a medical device under Section 201(h) of the FD&C Act and is subject to FDA oversight.
Final takeaway
Non-Device CDS supports clinical thinking, not clinical decisions and it must show its work.
Not sure whether your software qualifies as Non-Device CDS or falls under FDA medical device regulation? Let’s talk.