Tom Smith

Try this in Qliksense!

Weeknotes s07e25 TL;DR: To the Moon and Back — Taking our Nervecentre EPR live

Weeknotes s07e25

TL;DR: To the Moon and Back - Taking our Nervecentre EPR live

weeknot.es/weeknotes-s0...

Weeknotes s07e23 TL;DR: In the spotlight for various things. Reaching an EPR milestone and heading into the final 2 weeks before go-live.

Weeknotes s07e23

TL;DR: In the spotlight for various things. Reaching an EPR milestone and heading into the final 2 weeks before go-live.

weeknot.es/weeknotes-s0...

Thanks @kurtstat.bsky.social for writing about this and @ludictechnologies.bsky.social for tagging me

But few factories successfully transform their productivity by focusing on their cardboard box stock - they look deeply at the interactions of their process bottlenecks, and spend improvement effort there. Those are the things I think it would be great to talk more about.

In the factory analogy hospital beds are the cardboard boxes of parts sitting between the people doing the work. Important to have enough, and by looking closely it's possible to deduce which machines or processes are bottlenecks (the ones with queues of boxes waiting).

While beds are easier to count and measure, I think focussing on the true capacity bottlenecks (surgical time, diagnostic time, clinical decision processes, information sharing, phlebotomy, and pharmacy, among others) would give more insight than measuring beds.

Beds don't treat patients, and doubling their number would in many cases only succeed in doubling patients length of stay. Patients in beds are usually waiting (for the real bottleneck), or healing. With twice the number of beds and no change to the bottleneck capacity, most will wait longer.

That is an excellent article, and it's really valuable to see it laid out from a patient's point of view. I do think the NHS focuses too much on beds, and bed occupancy. As the article hints, the real problems in the writer's case were information flow, and the emergency surgery timetable.

Version 0.2.2 of the {NHSRplotthedots} package has today been accepted by CRAN. I have taken over the baton as maintainer, and I'm looking forward to helping more NHS data analysts produce XmR plots of their data, quickly and without fuss. nhsrplotthedots.nhsrcommunity.com

If now that #TDF2025 has ended you don't feel your summer has yet included enough 'people doing impossible things on bikes' then may I recommended the 4000km unsupppored nonstop #TCRno11 from the Atlantic to the Black Sea, which started last night...

Personal Goals Review 2025 #2 Here we go again. I set out some personal goals for 2025. It’s now time to review progress at the end of Q2

I set out some personal goals for 2025 and have just completed a review of progress at Q2.

TL;DR: Nailing the cycling. Need to get back on the books and get the research complete.

andy-callow.medium.com/personal-goa...

Thanks this is really helpful. I had followed links but not recognised the significance of the gcc-ASAN one. I think the C++ code puts this out of my reach, but it is a learning opportunity! I agree it does look like the maintainer has moved on to other things. Thank you!

Hi, thanks - I just checked rio out but unfortunately it doesn't support xlsb. The recommended workflow of manually re-saving as xlsx in excel won't work for me because I'm batch-processing about 100 separate files. I'm thankful the readxlsb code is still available on github!

Do any #rstats folks know how I can find why {readxlsb} was removed from CRAN? I've forked the github repo, but we use the code at work to read finance spreadsheets over which I have no file format control. Interested in what would be needed to get it back onto CRAN

This applies to the NHS too. Improvement can only be delivered to patients by the front line. It's the reason I choose to work and make a difference in a provider trust, not anywhere more remote from patients.

The recent history of AI in 32 otters Three years of progress as shown by marine mammals

I wrote a history of recent AI development in 32 images of otters using wifi on airplanes, from images to video to code.

It shows two big trends: rapid improvements in AI models of all types and the growth of open weights AI models. www.oneusefulthing.org/p/the-recent...

The ragnar R package helps you build RAG systems for document information retrieval. You'll need API access to an LLM and an API to make embeddings. Still unfortunately difficult in the NHS, but it will become possible as orgs find and share the news of benefit in specific use-cases.
#databs #nhs

Decoding Data Orchestration Tools: Comparing Prefect, Dagster, Airflow, and Mage IntroductionData is exploding, and so are the tools to manage it. From generating and collecting, to cleaning and analyzing, these tools help create valuable products for customers and give stake

I wish every team published a doc like Free Agent did for why they chose the tool they did. Of course I am glad they chose Dagster+, but I think this kind of clear-eyed proof of concept criteria can help teams choose the right tool for their team:

#dataBS

engineering.freeagent.com/2025/05/29/d...

YouTube video by FORMULA 1 The Evolution Of F1 Pit-Stops! | DHL

I can't read the article, but hopefully they mentioned the improvement context too. I wonder what those in the 1950s would think if we told them that pit stops in 2025 are routinely done in 2 to 3 seconds, not minutes. youtu.be/n_esmAYxE40?...

An exceptional new @nejm.org review on cancer of unknown origin
www.nejm.org/doi/full/10....

Chart from The Economist showing a drop over time In poems that rhyme

Can you spot
What this chart has got
That modern poems have not?
www.economist.com/culture/2025...

A thread about using AI to summarise documents; but it could really be about almost any use of AI.

Right now - in healthcare for example - it's a great timesaver for people who don't *need* it, and potentially troublesome for those who may try to use it outside their existing expertise.

I'm busy reviewing a large number of applications for a position in my team. I'm reflecting on the huge amount of human experience and expertise that is embedded in the applications. Pretty humbling.

Thanks to everybody who chimed in!

I arrived at the conclusion that (1) there's a lot of interesting stuff about interactions and (2) the figure I was looking for does not exist.

So, I made it myself! Here's a simple illustration of how to control for confounding in interactions:>

Git was 20 years old this week. Written by the creator of Linux as a tool to streamline his workflow - it shows the importance of making tools. If you have a better way of doing something, write and share it. It's possible that few others understand the problem, let alone have the solution you do!

Image from Bertin, 1967: A semi-graphic table showing how different kinds of variables can be encoded by various visual aspects--size, value, texture, color, orientation and shape

Another representation of the relationship between the nature of variables and visual variables

#TodayinHistory #dataviz #Onthisday #OTD 📊
💀May 3, 2010 Jacques Bertin died in Paris, France 🇫🇷

In 1967 his Semiology of Graphics became the first
comprehensive theory of graphical symbols and modes of graphics representation --> Grammar of Graphics

Good advice. "Customer first" cuts across almost every domain. I think Adam Wathan called this "programming by wishful thinking" in one of his TDD talks, and that language has always stuck with me.