Open-source ISO 13485 quality management systems template from GSTT-CSC
View repo
|
Report Error
|
Request Feature
This repository is an example of a quality management system (QMS) that can be used for the development of software as a medical device (SaMD) and AI as a medical device (AIaMD).
It's an empty version of the QMS that has been developed by, and is currently used by, the Clinical Scientific Computing (CSC) Team at Guy's and St Thomas' NHS Foundation Trust (GSTT), London UK. So far, it has survived 18 internal audits and 3 external ISO 13485 audits by two different notified bodies in the UK.
Having an ISO 13485 compliant QMS for medical software means that you:
- build safe SaMD / AIaMD that are appropriate for clinical use, and generate the appropriate technical documentation
- conduct clinical risk management throughout software development with appropriate clinicians
- establish, risk assess, and continually improve your processes
- train staff, verify their training, and keep it up to date
- continually evaluate your compliance to your standard operating procedures (SOPs) through internal audit
- have an established Corrective and Preventative Actions (CAPA) process to effectively solve issues at the root cause
- have management review meetings to evaluate effective and assign resources
- store documentation in a controlled way
The CSC will update this repo periodically as new challenges arise during our QMS journey.
- NHS Institutions looking to implement a QMS for in-house developed SaMD and AIaMD e.g. Medical Physics / CSC teams
- Life Science departments at universities who are looking at spinning out companies from research groups developing AIaMD and SaMD.
- Small- and Medium- sized Enterprises (SME's) starting their regulatory journey
The CSC team within Medical Physics at GSTT aim to build SaMD and AIaMD to the same standard as commercially available medical software. We are a team of clinical scientists, AI engineers, doctors, and NHS Scientist Training Programme (STP) trainees.
We have adopted an ethos of 'radical transparency' to make available as much of our output as possible, for the benefit of the NHS and the wider health tech sector.
No, this is a template QMS for you to use as a way of performing and logging SaMD / AIaMD development activities during software development. You would need to edit the policies / SOP's and templates to reflect your team, your institution/company, your resources, and the products you are developing.
| CSC Project Template | A template repo that all CSC projects are started from. Integrates with the QMS for appropriate documentation, and MLOps to integrate with our machine learning pipeline for training ML models on our infrastructure. |
| CSC MLOps | A CSC open source python package for machine learning operations - Training ML models using docker, and logging these models and training artefacts in MLflow. |
The UK Medicines and Healthcare products Regulatory Agency (MHRA) will be involved in the development of this repo going forward, as a practical means of two-way communications between the regulator and healthcare institutions / SMEs / start-ups / academia so updates to the regulations in the UK are sensible.
| Apache-2.0 |
|---|
Please fork (online) / clone (local) this repo and see if you can implement in your institution. Get in contact with CSC on gstt.clinicalscientificcomputing@nhs(dot)net if you want to know more.
New Users
Consult the link for the work instruction to set this up for your institution. The CSC will provide consultancy on setting this up and QMS training for NHS departments.
| QMS SETUP INSTRUCTIONS |
|---|
Established Users
Please engage in conversations in the issues on how to modify this repo to comply with updated regulations / standards. Any suggestions you have recorded from internal and external auditors would be much appreciated!
Regulatory Consultants
We would greatly appreciate input from medical device regulatory consultants to rectify any glaring holes.
NHS Clinical Engineering Departments
We would love to hear from clinical engineering departments in the NHS who assess the technology readiness level (TRL) of SaMD in development. We wish for this repo to automatically provide estimates of minimum documentation and software performance required to conduct clinical evaluations.
In the spirit of quality management we are striving for continual improvements to the repo to assist UK healthcare institutions, academic institutions and SME's in developing USEFUL and SAFE medical software by increasing access to clinical and regulatory expertise collaboratively across public private and regulatory space.
- Share your auditor experiences without naming notified bodies or individual audits.
- Share SaMD and AIaMD incidents without explicitly naming healthcare institutions or patients.
- Reflect on types of SaMD/AIaMD products without explicitly naming them when suggesting improvement.
- Polite debate is encouraged, and discussions will be fed back to the MHRA.
We will provide updates to this repo periodically to reflect new risks and new functionality.
- Implement MHRA guidance for AIaMD in SOP's and templates
- A wiki to help set up QMS and useful info on regulations and deployment platforms
- Example repo with full technical documentation for a MONAI Application Package (MAP): a custom Docker container that works on common AI deployment platforms
- Automation of trend analysis (CAPA resolution times, code review times, stakeholder engagement etc.) through GitHub Actions
- Automation of ISO 13485:2016 / ISO 62034:2006 self-auditing through GitHub Actions
- Generation of audit reports for internal and external auditors
- Provide suggested schedules for QMS activities for NHS CSC departments
-
Open Regulatory - We sourced many of our SOP's from the ISO 13485 templates made available by OpenRegulatory, you will see their license at the bottom of many of our documents.
-
Innolitics RDM - We use Innolitics
rdmpython package for generating the technical documentation for individual project repo's. Another fantastic company who build SaMD/AIaMD for FDA approval in the US. -
All past and present CSC team members who have contributed to the development of this QMS, and your effort to get it off the ground and supporting it as a team.
- CSC Website
- MHRA webpage for Software and artificial intelligence (AI) as a medical device Latest considerations from the MHRA on SaMD and AIaMD
- OpenRegulatory Slack Channel A great community Slack channel hosted by OpenRegulatory - connecting regulatory professionals across Europe.
If you engage clinicians and hospital staff early in the development of SaMD and AIaMD, you will prevent wasted time, effort, and money building unusable products with no clinical value. Clinical and IT engagement is required at the start of development in our QMS for this reason.
Early stage QMS activities can be burdensome to non-regulatory people, so instead we encourage the use of the Sense About Science "Responsible handover of AI" framework: a list of digestible questions to be answered at project handovers to maintain the information that will be valuable when technical / QMS documentation needs to be filled out.
| HANDOVER FRAMEWORK |
|---|
The Handover framework maps well to the GSTT-CSC SaMD / AIaMD development process:
