Short answer

The EU MDR does not prescribe how to run a support desk, but every support interaction with live medical software has to fit its post-market obligations: complaints feed post-market surveillance, potential serious incidents (Article 2(65)) must be assessed for vigilance reporting, and fixes go through change control. A bug fix becomes a regulatory event when it is a substantial change to the device or the QMS covered by your certificate (Annex IX, sections 2.4 and 4.10), as set out in your Notified Body’s change procedure. ISO/IEC/IEEE 14764:2022 gives a useful structure for maintenance types, and support response times are defined in your own complaint-handling procedure.

1. What changes after CE mark — the maintenance lifecycle

Most SaMD teams spend many months focused on getting to CE mark. Very few spend equivalent effort designing what happens after. The result is predictable: a support model built for a standard software product collides with post-market obligations built for a medical device, and neither works well.

EU MDR does not describe how to run a support desk. But it creates a framework of obligations that your support model must fit inside:

  • Complaint handling — every user complaint about safety or performance must be captured, investigated, and documented per your ISO 13485 QMS complaint procedure.
  • Vigilance reporting — serious incidents and field safety corrective actions must be reported to the relevant competent authority within the timeframes of MDR Article 87.
  • Change management — every software change must be assessed before deployment to decide whether your Notified Body must be informed or must approve it (MDR Annex IX, sections 2.4 and 4.10, and your NB's change procedure).
  • Post-market surveillance — support data is a primary input into your PMS plan, PSUR, and PMCF plan.

None of these obligations are compatible with an ad-hoc support process. They require a structured system where every support interaction is documented, classified, and routed according to predefined procedures. ISO/IEC/IEEE 14764:2022 provides a useful framework.

2. ISO/IEC/IEEE 14764:2022 — the maintenance standard for SaMD

ISO/IEC/IEEE 14764:2022 (Software engineering — Software life cycle processes — Maintenance) is a general software engineering standard, not a medical device standard and not harmonised under the MDR. It describes the processes, activities, and tasks for software maintenance after delivery and distinguishes categories of maintenance, including the four below — each with different regulatory implications for SaMD:

Corrective maintenance

Fixing defects

Reactive modifications to correct discovered faults — bugs, errors, failures identified in production. The defect exists in the delivered software; the fix restores intended behaviour.

Regulatory implication: Must go through change assessment. A bug fix that brings the device back to its original specification is usually not a significant change, but if the defect affected clinical output or safety functions, it may also trigger vigilance reporting — independently of when the fix is deployed.
Adaptive maintenance

Adapting to environment changes

Modifications to keep software operational in a changed or changing environment — OS updates, new browser versions, API changes in connected systems, hardware upgrades at client sites.

Regulatory implication: Usually non-significant if device functionality is unchanged. Must still be assessed. Verify that V&V evidence covers the new environment — regression testing required.
Perfective maintenance

Improving performance or usability

Modifications that enhance performance, maintainability, or usability without correcting faults or adapting to environment. Feature additions, UX improvements, performance optimisations.

Regulatory implication: Most likely to need NB involvement. A change to the intended purpose, the clinical output or the device classification is likely to be significant regardless of how small it looks in the backlog.
Preventive maintenance

Preventing future failures

Modifications that detect and correct latent faults before they become operational failures — code refactoring, security hardening, dependency updates before known vulnerabilities are exploited.

Regulatory implication: Security patches are a frequent type — assess against MDCG 2019-16. Proactive SOUP updates require documented known-anomalies review for updated versions.
ISO/IEC/IEEE 14764 and IEC 62304

ISO/IEC/IEEE 14764:2022 and IEC 62304 address the same lifecycle from different angles. IEC 62304 defines the development and maintenance processes required for medical device software specifically. ISO/IEC/IEEE 14764:2022 provides the broader framework for classifying and managing maintenance activities. In a mature SaMD QMS, both can be referenced: IEC 62304 clause 6 covers the software maintenance process, and the 14764 categories give your change management procedure a taxonomy to route each change to the appropriate regulatory assessment.

3. Support tier structure for medical software

Standard software support models (L1/L2/L3) apply to SaMD — but each tier carries responsibilities that go beyond what these tiers mean in non-medical software. The key difference is that every tier must have a defined interface with your QMS complaint handling procedure.

L1 — Client support

First response

  • Initial contact, triage, basic troubleshooting
  • Collect full incident description and context
  • Log every interaction in the complaint system
  • Apply severity classification (P1–P4)
  • Provide workaround instructions where available
  • Escalate P1 and P2 immediately to L2
  • Communicate acknowledgement SLA to client
L2 — Technical investigation

Root cause analysis

  • Reproduce the issue in test environment
  • Identify root cause and affected components
  • Assess patient safety impact
  • Determine if vigilance reporting is triggered
  • Initiate change request in QMS if fix required
  • Escalate to L3 for engineering fix
  • Keep client updated on investigation status
L3 — Engineering & regulatory

Fix and compliance

  • Develop and test the fix per IEC 62304
  • Run the regulatory change assessment (NB change procedure)
  • Update risk management file if needed
  • Inform the NB / obtain approval where required
  • Update technical documentation and SOUP list
  • Deploy via release procedure with regression testing
  • Close complaint with documented resolution
The gap most teams have

L1 and L2 tiers operate well in most SaMD companies. The critical gap is the L2→L3 handoff: L2 identifies a bug, raises a Jira ticket, and the engineering team fixes and deploys it — without anyone performing the required change assessment before release. In our experience, this is one of the most common QMS non-conformities raised in post-market NB surveillance audits.

4. Incident classification and SLA by severity

Every support incident must be classified at intake. The classification drives both the internal SLA and the regulatory assessment path. EU MDR does not prescribe support SLA timeframes; your QMS complaint procedure should define them, and they must be consistent with the vigilance reporting deadlines. The SLAs in the table below are an illustrative example, not a regulatory requirement.

Level Definition SLA — response SLA — resolution Regulatory path
P1 Critical. Patient safety at risk, clinical function unavailable, incorrect outputs affecting active clinical decisions. 1 hour Same day workaround or suspension Immediate vigilance assessment. Potential serious incident report to the competent authority — at the latest within 2, 10 or 15 days depending on the type (MDR Article 87). FSCA assessment.
P2 Major. Significant clinical workflow impact, degraded accuracy, data integrity issue without immediate patient harm. 4 hours 5 business days Full complaint investigation. Vigilance assessment. Change assessment before fix release.
P3 Minor. Limited clinical impact, workaround available, no patient safety risk identified. 24 hours Next planned release Standard complaint record. Change request in QMS. Change assessment before release.
P4 Enhancement / no clinical impact. UI issues, cosmetic defects, performance improvements with no clinical output effect. 5 business days Roadmap Logged in product backlog. No mandatory complaint record unless customer formally raises as complaint. Change assessment at release planning.
P1 and vigilance — the reporting clock

If a P1 incident meets the EU MDR definition of a serious incident (Article 2(65) — an incident that directly or indirectly led, might have led or might lead to death, serious deterioration of health, or a serious public health threat), Article 87 requires a report immediately and at the latest within 15 days of becoming aware of it — 10 days in the event of death or an unanticipated serious deterioration in a person's state of health, and 2 days for a serious public health threat. The clock starts from awareness, not from when a fix is ready, and an incomplete initial report can be followed up later. Deploying a hotfix does not suspend or reset this obligation: the serious incident report to the competent authority must still be filed.

5. Change management: when a fix becomes a regulatory event

Every software change deployed to production — regardless of how minor it seems — must pass through a documented change assessment before release. For MDR-certified devices, the MDR requires you to inform your Notified Body of planned substantial changes to the QMS or the device range it covers (Annex IX, section 2.4) and, where the NB issued an EU technical documentation assessment certificate, to obtain its approval for changes that could affect the safety and performance of the device or its conditions of use (Annex IX, section 4.10). What exactly must be notified, and how, is set out in your agreement with the NB. MDCG 2020-3 Rev.1 defines "significant change" only for legacy devices still on the market under MDD certificates (Article 120), but its software chart is a useful reference when structuring your own criteria.

The assessment is not a bureaucratic formality. It is the mechanism that determines whether you can deploy a fix under your existing certificate, or whether you need Notified Body involvement before the change goes live. Our guide to change control for AI/ML SaMD covers this decision in more depth.

1
Change request raised L3 engineering or product raises a formal change request in the QMS. Describes the change, affected components, and rationale. Assigned a unique change ID.
2
Significant change assessment Regulatory or QA function assesses the change against your documented criteria: Could it affect safety or performance? Does it change the intended purpose or the conditions of use? Does it affect GSPR conformity or the device classification? Does it change an algorithm, the architecture or the way data is interpreted by the user? A "yes" means the change is likely to need NB involvement — follow your NB's change notification procedure. Documented assessment required
3A
Non-significant change → internal process Update risk management file and technical documentation as needed. Run regression testing as required by your IEC 62304 maintenance process (clause 6). Document V&V results. Update software version and release notes. Deploy.
3B
Significant change → Notified Body involvement Notify your Notified Body before deployment. Prepare updated technical documentation covering the changed sections. The NB assesses the change and, where required, approves it — for example with a supplement to the certificate. How long this takes depends on the NB and the scope of the change, so plan it into your release schedule. NB notification required before deployment
4
Release and post-release monitoring Deploy via defined release procedure. Monitor for regressions. If the change requires a new UDI-DI (for example a change to performance, safety, intended use or interpretation of data), register it in EUDAMED before release; minor revisions only need a new UDI-PI, which is not registered. Feed release data into the PMS system for the PSUR.
Is your change management procedure NB-ready? A change procedure has to work for every release after certification, not only on audit day.
See EU MDR certification →

6. Client communication protocols

Communication with healthcare provider clients after CE mark has two layers that must not be conflated: commercial support communication and regulatory communication. Mixing them — or handling regulatory obligations through informal channels — creates compliance risk.

Support communication

Standard incident updates, workaround instructions, and fix ETAs. Delivered through your support system (ticketing platform, email, dedicated portal). The key requirement under ISO 13485: every substantive communication about a complaint must be documented and traceable to the complaint record. A verbal phone update that is not logged does not exist for QMS purposes.

Regulatory communication — Field Safety Corrective Actions

When a defect poses a risk of serious harm and a correction needs to be deployed across the installed base, this may constitute a Field Safety Corrective Action (FSCA) under EU MDR Article 2(68). The FSCA must be reported to the competent authorities without delay, and affected users must be informed through a Field Safety Notice (FSN) — a formal written communication to all affected customers. Except in cases of urgency, the draft FSN is submitted to the evaluating competent authority so that it can comment before distribution.

An FSN is not a support email. Its content requirements are set out in Article 89(8) MDR — in essence, it must identify the affected devices (including their UDIs) and the manufacturer, explain clearly and without understating the risk why the FSCA is being taken, and state all the actions users must take. If your team has never drafted an FSN, drafting one for the first time during an active safety event is not the moment to learn the format.

Software update notifications

Every software update deployed to clients must be communicated with sufficient information for clinical users to understand what changed and whether their workflows are affected. For significant changes (3B above), we recommend that the update communication summarises what changed and why. For AI/ML updates that change model behaviour, clinical users should be told about any changes to the output characteristics or limitations that affect their use of the device, and the instructions for use must be updated accordingly.

Best practice

Maintain a documented communication template library: incident acknowledgement, investigation update, resolution notice, software release note, FSCA notice, FSN draft. Templates that have been reviewed and QMS-approved in advance shorten your response time in a safety event — and help get the regulatory language right from the first draft.

7. Connecting support data to EU MDR post-market obligations

Your support system is one of the richest sources of post-market data you have. Under EU MDR, this data must flow into your formal post-market surveillance system — not sit isolated in a ticketing platform that no one reviews for regulatory purposes.

Here is how support data maps to each EU MDR post-market obligation:

PMS plan and PSUR

Your post-market surveillance plan must define how complaint and support data is collected, aggregated, and analysed. For Class IIa, IIb and III devices, the PSUR (Periodic Safety Update Report) summarises the results of that analysis — including complaints, serious incidents and FSCAs — and the conclusions of the benefit-risk determination; Class I devices prepare a PMS report instead (Article 85). A PSUR that does not reference support system data is likely to be incomplete.

PMCF plan

Post-Market Clinical Follow-up gathers clinical evidence on device performance in real-world use. Support data — particularly patterns of unexpected use, unexpected clinical outcomes, or performance in patient subgroups not well represented in pre-market studies — can be an important input to PMCF analysis. If your support tickets consistently show a specific patient population experiencing different outcomes, this is a PMCF signal that must be investigated.

EUDAMED and vigilance

The EUDAMED vigilance and post-market surveillance module is not yet mandatory. The four modules that became mandatory on 28 May 2026 (actor registration, UDI/device registration, Notified Bodies and certificates, market surveillance) do not include vigilance, so serious incidents and FSCAs are still reported to the national competent authorities through the national reporting processes (using the Manufacturer Incident Report form). The Commission's EUDAMED timeline (November 2025) indicates mandatory use of the vigilance module in Q2 2027, subject to audit results — see our EUDAMED guide. Either way, your support classification (P1 assessment) must trigger the vigilance assessment that determines whether a report is required. If the serious incident assessment lives only in your support system and never reaches the regulatory function, the report will be late — which is a breach of the Article 87 deadlines.

The integration requirement

ISO 13485:2016 clause 8.2.2 requires documented complaint handling procedures, including deciding whether the information must be reported to regulatory authorities, and clause 8.2.3 covers that reporting. In practice, this means your support system and your QMS must share data — either through a direct integration or through a defined escalation procedure with documented handoffs. A Jira ticket that is never linked to a QMS complaint record, and a QMS complaint record that is never linked to the PSUR, are both likely to be raised as non-conformities. If you have no regulatory function in-house to own these handoffs, part-time QA/RA support after launch can fill that role.

Official sources & references