- Published
- Updated
- Sources checked
Software verification shows that code behaves as specified; clinical evidence must also show that the specified medical output is appropriate.
Medical device software — Define the regulatory scope
MDCG 2020-1 distinguishes valid clinical association, technical performance and clinical performance. First determine whether MDR or IVDR applies to the intended medical purpose.
Medical device software — Build the working record
Create an output-to-evidence map recording algorithm version, input population, reference standard, operating conditions and clinically relevant error consequences. Separate independent evaluation data from development data and describe any generalisability limitations.
Medical device software — Check before release
- Freeze the evaluated algorithm and data lineage.
- Test clinically meaningful errors and subgroups.
- Reassess evidence after relevant software changes.
Medical device software — Record the decision and handover
Approve the evidence conclusion with clinical and software owners. Connect release decisions to change impact and monitoring; a successful unit test or high aggregate accuracy cannot alone justify every medical claim.
Frequently asked questions
- Does every wellness application require this framework?
- First assess qualification and intended purpose; do not assume all software is a medical device.
- Can one accuracy figure support all settings?
- No. Assess population, intended use, reference standard and clinically relevant performance.