The Ancient Farm
The last couple of weeks have been good, and we’ve made decent progress. They have also been short weeks, with a two-day hackathon taking up some of last week and four-day weeks for most of us this week.
We’re also down two designers and onboarding a new temporary user researcher. Despite that I think progress has been good.
Planning how to test our user-centred design hypotheses
During a planning session talking about which experiment to prioritise next, we spoke about two constraints: we don’t have any baseline data for the current experience so can’t do comparative studies to show we’ve made improvements; and proving we’ve made the experience better is almost impossible without access to real patients, their data, and a longer period of time to measure outcomes.
We can, however, run experiments that indicate a direction to move in, and may create the opportunity for those more objective tests to be run.
After a couple more conversations planning our research methodology and the first baselining study, we’re now preparing interview guides, recruiting participants and outlining what prototypes we need to run the study.
Planning how to test our technical hypotheses
An event-storming workshop I ran last week helped us discover, in more detail, the questions we still had around how things work now and how things might work. For example, we weren’t totally certain how a service processed a result from a lab. And we weren’t entirely sure about the conditional logic in the app for displaying test results.
So we set up chats with teams, explored their data models, looked at code, and spoke to adjacent teams during the hackathon. It built up a much clearer picture of everything, meaning our tech lead could propose the first technical spike – and by far the riskiest assumption.
Again, great progress. Admittedly we’re limited in what we can learn as we don’t have access to real patient data, but it may prove out an approach.
I’m also on the horn to OpenSAFELY, hoping they can help us run a statistical analysis of pseudonymised patient records held by GPs across the country. This is what would help inform the reliability of the approach at national scale. This analysis will be on the consistency (or quality) of patient record data across suppliers. Everyone knows it’s hugely variable through experience of working with it, but as far as I know there are no definitive statistics on patient record data at national scale.
A testable belief
My hypothesis is that data quality across suppliers is poor, and not high enough to support the national roll-out of any of our ideas. I’m open to being proven wrong, but realistically there’s a reason that teams are only able to make limited, tactical changes to services providing patients their data. And the concept of a Single Patient Record is being discussed for the same reason, presumably.
This is a personal belief based on my experience and not a view of any contracting parties between me and NHS England, but the virtual duopoly of GP IT systems is a problem, and the existing standards and frameworks don’t apply enough leverage to solve it.
It’s a market-shaping problem. Switching costs are massive for GP surgeries, and barriers-to-entry are high for new market entrants. Despite what the Competition & Markets Authority say, existing mechanisms do not mean the market is competitive.
But I’ll get off the soapbox: it’s not my game to play or the immediate focus.
Other things
- Went to Butser Ancient Farm to see some prehistoric, Romano-British and Dark Age buildings in action. Experimental archaeology is dope.
- Enjoyed a long birthday weekend in Lille with friends. It’s only an hour-and-a-half away by train, is a walkable city and has great food and wine options. Hot recommend.
- Wrote a case study on rapid prototyping with AI and pulled together testimonials about Boring Magic.
- I enjoyed The Odyssey, especially the Cyclops scene. More films like that please. Also think it’d be fun to adapt the Mabinogion into a film (no idea if it has been done before).
- Started reading The Living Mountain by Nan Shepherd and it is a treat.
- Another long weekend near Pisa coming up, plus my first experience driving abroad…on Italian roads. I am bracing myself.
Bookmarks
- Stay put, or keep moving?, 4 mins. Great reflections from Dan on working in digital healthcare.
- Finding a jump-off, 4 mins. Another fabulous weeknote from Mike. The NHS App often grows by adding small parts rather than bold, new ideas. Without a clear design vision, teams only think about where to put new links instead of how to combine services well. This causes the app to feel like a list of separate pieces instead of one smooth health tool.
- Traits before tools: rethinking when to insert AI into your ways of working, 7 mins. Many organisations fail to get value from AI because they start with tools instead of understanding how people work best. Barry O’Reilly’s 3T Model says to focus first on traits, then tasks, and only then tools to improve judgement and decisions. Close to Tom’s decision maps.
- No Siri AI in EU, 12 mins
- Siri AI Beta, 4 mins
- Apple has three new smart home products ‘nearly ready to launch’, per report, 2 mins
- Deckchair shuffling is for the unserious, 2 mins
- Plan A: 2040, 12 mins
- Scrapping Digital ID is the wrong right decision, 5 mins
- New research shows honesty about AI use at work is backfiring, 5 mins
- The single patient record: a laudable aim, at risk of mistakes that could derail it, 5 mins
- From the RAT to the Magenta Book, 4 mins
- Why we write, 4 mins
- Learning about clinical assurance, 6 mins
- Lobster pots, complexity and starting small in the right way, 4 mins
