Building an AI community in Healthcare at NHSX
For the last 6 months, I’ve been on a loan to NHSX working in the NHS AI Labs programmes for the Skunkworks team.
The NHS AI Lab creates an environment for collaboration and co-creation by bringing together programmes that address significant barriers to the development and deployment of AI systems in health and care.
The Skunkworks programme aims to find new ways to use AI for driving forward the early adoption of technology to support health, in both clinical and business contexts. The team provides free short-term expertise and resources to public sector health and social care organisations to support AI projects and develop capability.
My brief when I joined was to help explore if a community can be built in this space and to bring my community expertise and experiment. I’m really proud of the work I’ve delivered over the last 6 months. This blog will go into the different projects I’ve initiated.
My 6 months were divided by three approaches I decided to take, research, activate and reflect. I was tweetnoting the first few months but when it got to the activity phase, I was too busy doing. You can catchup on that thread here.
Research
AI Virtual Hub
In September 2020 the AI Lab launched the AI Virtual Hub, a space for the NHS and industry to collaborate and learn from one another on the adoption of AI in health and care. Set up on the Future NHS platform, the AI Lab hopes to offer opportunities for consultation on AI Lab programmes and encourages best practice sharing on what good looks like in AI. The virtual hub helps to build trust and confidence and demonstrate the potential of AI-driven technologies for health and care to build understanding amongst the public.
Launched during the pandemic the community has flourished online. As of September 2021 the AI Virtual Hub has 1,321 members and has seen an average growth of 117 members joining per month. The virtual hub has been a great place for members to learn about a range of things from the programmes to developing AI tools and technologies for Health and Care.
Community Deep Dive Discovery
Each programme team in the NHS AI Labs has a different need and use for the community and there are also many reasons why people from NHS organisations or industry join the community from wanting to understand regulation to sharing best practices.
In order to do this and develop the community further, it is important that we dig deep and understand the community. As part of this one of the first things I did when I started my role was do a discovery into the community. The research objectives were:
- To understand the AI healthcare community.
- To understand the current uses of the virtual platform and members currently on it.
- To understand what different stakeholders would want from the community and what capability needs are needed.
As part of the discovery, I spoke to many people from members of the team but also analyzed how the current platform is used.
It helped us to try and figure out exactly where we want to position the AI Virtual Hub. Analysing the professional role data to regional data really helped us to understand the community and how we want to shape it going forward.
The current AI Virtual hub is a great space for the community to share but most of all for the NHS AI Labs to share its news, put up webinars, and a central place for resources. The membership, however, is diverse with 60% of the makeup being people working in NHS Trusts, Academia, and Arms length bodies with the other 40% being made up of suppliers (individuals and organisations developing AI tools and technologies). With such a diverse community it means people’s needs and wants can be quite different. People join the community for different reasons.
A platform though does not make a community and the current platform the community sits on from a community management perspective isn’t the best and can be improved to be more user-friendly for members but also for community managers. How you work with the community is key and they can really help shape and influence our work.
By analyzing and running the discovery I was able to see where the gaps are but also see how I could work with the community better.
Accessing the community database meant that I could explore how to use community membership data in a segmented way to help with better outreach and community building within the Skunkworks programme. This led to running surveys, followed by exploration chats to understand what projects the community was doing generally in AI and see how we could potentially work together.
Activate
Building Community in Skunkworks — Taking proof of concepts and helping others to experiment with AI
Within the Skunkworks team, we’re really keen on developing the community to help trusts be more open. If one trust has a problem, it’s more than likely another trust has a similar problem. A good example of this is a project the team worked on which was about using machine learning to identify patients at risk of long term hospital stays. The team worked with Gloucestershire Hospitals and ACE through our problem sourcing call to help develop a proof of concept over a 12 week period. This project was blogged about and code was open sourced through our github.
Once the project was completed, as well as the relevant documentation and write up it was essential to share what we’ve done with the community. This led to a public webinar, followed by a workshop. The long-stay problem is not just unique to Gloucestershire hospital but one that many hospitals can relate to it. By being open and presenting our approaches and running discussion-based workshops, it has led to trusts coming forward to explore how they can implement the code in their organisation and re-use and adapt to their environment. Currently, our team of Data Scientists is working with Nottingham University Hospital to help them adapt and implement their code which is very exciting to see because not only is this a great way to build AI capability, build confidence with AI technology but it could lead to big potential cost savings in the future. This is leading to developing a catalogue of use cases to exemplify where AI-driven technologies can be most useful in the health and care system. Building an evidence base is crucial to constantly improving, demonstrating benefits, and creating regulatory frameworks.
AI Deep Dive Workshop Series
Building capability is also a key component of community building and helping the system learn more about AI is key which was another gap identified in the research phase. In the Skunkworks team, I worked with my colleagues Amadeus and Jennifer in developing a series of practical workshops designed to increase confidence, trust, and capability of implementing AI within the NHS. Over a 5 week period through a 75-minute workshop weekly, the team will be working directly with a diverse audience of clinicians, technology teams, operations teams, and other stakeholders from Trusts interested in utilizing AI.
The workshop series is as follows:
- Workshop 1: AI fundamentals
- Workshop 2: Problem Discovery
- Workshop 3: Solution Discovery
- Workshop 4: Practicalities
- Workshop 5: Launching your AI experiment
By the end of the workshop series, learners will be able to
- Be confident in having more conversations about AI in Healthcare.
- Embrace an experimental approach to AI in Healthcare.
- Understand practical steps required for experimenting with AI in Healthcare.
- Create a detailed plan for an AI project.
We built out this workshop using our experience of working with the community and seeing where the gaps were. It was important for us not to reinvent the wheel but use existing frameworks out there to help people think through AI technology to help them upskill.
Although version 1 of this workshop was designed to be delivered per team either online over 5 weeks or a day session in-person, we’re keen to think about scalability and explore how to make this workshop series scalable which will be what the team will work on next.
We’ve run an internal pilot and will start working with a trust from January but if you’re keen to have your team involved then get in touch with the team and sign up: ai-skunkworks@nhsx.nhs.uk
By the community for the community
A marker of a good community is all about how members contribute and get involved.
As part of a member’s journey on the Virtual Hub, we also ask people to express interest in becoming community champions. Over the summer we ran a community champion session to understand the needs and wants of the community and to understand what each of them could offer. This helped us kickstart our first Virtual Hub Community Week. In this week we had 8 different sessions delivered and hosted by community members. We had over 150 attendees from many different backgrounds come together to learn from other members. Some sessions that were hosted included:
- Radiology quality improvement at University Hospitals Plymouth by Dr Mark Thurston a consultant radiologist at Derriford Hospital and honorary clinical lecturer at the University of Plymouth.
- Detecting patient deterioration by Dr Gonem Sherif is a respiratory consultant at Nottingham University Hospitals NHS Trust.
- Machine Learning from desktop to BAU from Simon Mortimore — Assistant Director for Business Information.
It’s important to note that none of these sessions were about us as a programme but instead about providing a platform for the community to really showcase and share interesting AI work they are already doing. Each session was an opportunity to share knowledge and have a fruitful discussion. The feedback we had was super positive from attendees who deemed it useful and all responded said they would like us to run it again. This week was run very quickly with just some logistics and facilitation support from us, as we nurture the community more, with my colleague Pardeep we are working on a Community strategy to help facilitate this sharing on a more regular basis virtually and through events.
The last thing as part of the activate phase was as the virtual hub was all launched during lockdown I was keen to put on a small covid safe in-person meetup which I did in collaboration with one of our partners Foundry4.
This was also followed by our first in-person meetup, where we introduced the AI labs, the Skunkworks team spoke a bit about what they do, followed by Foundry4 and Royal Free Hospital talking about the work they’ve done at NHSX. The rest of the evening was all about networking which led to lots of meaningful connections being made. Running this one event has now led to designing a roadmap of events for next year for the AI Labs which is very exciting!
Reflect and Document
Just like we document our technical releases, it’s just as important to document community activity and how it’s done.
In the reflection phase, all this work has helped build out our community strategy which will be a core pillar of the community work for next year.
I’ve been documenting and developing recipe documents for the team to follow after every project in how to engage with the community with copy etc so that it can be a very quick thing to do and not time-consuming.
I’ve also built out a community database to help track the engagement of different community members with different activities so we can see who is interacting the most with us but also be able to have the community data in one place. Community building is all about relationships and often when a community person leaves, the relationships also go with them. I’m keen for that not to happen with the work I’ve done so building the right operational tools was key to do this.
Lastly, with the AI deep-dive workshops, I’ve been making sure we develop the material and facilitator notes so that they can scale beyond our team in the future.
What’s next?
The last 6 months have been an absolute privilege. I’ve learned a lot about the NHS ecosystem and artificial intelligence and I’m really glad I got to work on building the community. I’m excited about their plans for the community for next year so watch this space. As for myself? I have an exciting new role coming up which will be announced in due course.
