Neil Pettinger

Headline: "Scientists discover emperor penguin colony in Antarctica using satellite images"

Yesterday it was cows using tools, today its penguins using satellite imagery.

Brexit in its proper context.

"All of these three measurement failings – the wrong doors, the wrong time zone, the wrong connotations – conspire to prevent the rigorous examination of cause-&-effect relationships in the health & care system that’s needed"
@kurtstat.bsky.social for #NHS2048
www.enlighten.scot/nhs2048/thre...

Well, it's a little clearer now why billionaires are so invested in technology that produces better written emails.

Guys this is the ONE client you really need to double check

My AMU 'heaviness' metric came about after I tried showing AMU lengths of stay so far (ALoSSF) as horizontal timelines on top of each other. On the left: a snapshot that captured 60 patients with an ALoSSF of 24.6 hours. On the right: 58 patients with an ALoSSF of 37.8 hours.
#rstats #ggplot2

Stuart Hall's encoding/decoding model (adapted from his 1973 essay) made to look a bit smarter with blue and pink text boxes.

I want to have a go at adapting Stuart Hall's encoding/decoding model so that it works for how healthcare data messages are communicated. I think the key to this is how I flesh out the meaning structures on each side of the diagram.

A scatterplot showing the relationship between Acute Medical Unit (AMU) 'heaviness' (measured along the horizontal axis) and Emergency Department (ED) length of stay (measured up the vertical axis). 366 overlapping grey dots in the chart represent the 366 days of the year. A "green zone of flow" has been (arbitrarily) drawn in the bottom left of the chart. NOTE: AMU 'heaviness' is calculated as each day's average number of occupied beds multiplied by each day's average length of stay so far.

I'm hoping that - eventually - *everyone* involved in acute hospital patient flow will be talking about AMU 'heaviness'! (Meanwhile I can't work out if this chart is a volcano metaphor or a Pandora's Box metaphor...)
#rstats #ggplot2

Good question! I need to try and get hold of data from different types of AMUs to see if there's a difference!
(I've recently embarked on a similar-ish quest following advice from @mancunianmedic.bsky.social about how medical specialties are organized differently from hospital to hospital...)

The important thing would be to try and get the clinicians themselves (with a bit of help from the analysts!) to define the dimensions of the grey zone. "Ought-to-be" zones are probably better calculated 'bottom-up' rather than 'top-down', I think.

A scatterplot showing the relationship between Acute Medical Unit (AMU) 'heaviness' (measured along the horizontal axis) and Emergency Department (ED) length of stay (measured up the vertical axis). Two red dots in the chart represent the 'worst' day of the year (23 Jan 2024, when four-hour compliance was 40%) and the 'best' day of the year (23 Jun 2023, when four-hour compliance was 79%). A "grey zone of flow" has been (arbitrarily) drawn in the bottom right of the chart.

Acute Medical Unit (AMU) 'heaviness' scores are calculated by multiplying each day's AMU mean lengths of stay by the same day's AMU fullness snapshots. The scores are closely correlated with Emergency Department (ED) performance, so I thought a scatterplot would be the way to go.
#rstats #ggplot2

A scatterplot showing the relationship between occupied beds (along the horizontal axis) and no. of "wrong" wards used (up the vertical axis). There are 366 blue dots on the graph, each dot representing a midday snapshot in 2023-24. As occupied beds increase, so do the number of wrong wards.

Here's the graph after the correction. I'm trying out different ways of showing how a specialty's inpatients get more displaced as its workload increases. As @em-dr-jacklin.bsky.social might put it: the fuller the specialty, the greater the "wrongward-ness".
#rstats #ggplot2

ggplot(data = df_resp_midday) + aes(x = total_fullness, y = no_of_wrong_wards) + gg_point()

Error in gg_point() : could not find function "gg_point".

Please tell me I'm not the only person who's made this mistake!
#rstats #ggplot2

A white poodle leans out an suv with neon light streaks flashing across the surface of the car

Tried to sneak a photo of a cute doggo and accidentally had long exposure on and I believe I have created man’s greatest expression of art

Has anyone done 'loathed to say' yet?

Speak for yourself. It's a 'doggy-dog world' for me from now on!

A diagram that is basically a square. The horizontal axis is labelled Reason > judgement. The vertical axis is labelled Emotion > action. Within the square, a path is suggested first up the vertical axis (by using the visceral layer of design), then across the horizontal axis (using the behavioural layer of design), and finally up to the top of the vertical axis (using the reflective layer of design). The diagram is titled "Reason leads to judgement; emotion leads to action."

I'm working on a diagram that shows how we might incorporate emotion into data visualizations. I'm trying to mash together some ideas from Gavin McMahon's new book (www.story-business.com) with Don Norman's three layers of design (www.interaction-design.org/literature/a...).

The result is that hospital beds are managed “reactively” in the permanent-crisis here-and-now: a bed manager sees an empty bed, so they immediately fill it. It would be better if clinicians could see and discuss their flow metrics, to give them at least a chance of managing beds "proactively".
6/6

Preconceptions apart, the real problem is that we hardly ever report average bed occupancy to the clinicians whose patients occupy the beds. Most clinicians – and most managers– are not given the numbers that retrospectively describe demand and capacity in a general hospital.
5/6

There are some “NHS-specific” preconceptions, too. For example, there’s an “aura” surrounding “85%: the cure for all our ills”. In fact, the “right” percentage will be different for each part of the hospital. Some wards/specialties can work optimally at 95%; others need it to be lower than 60%.
4/6

Most of us bring “generic” preconceptions to percentages. We think higher numbers are better. Like the anecdote about the optimist who sees a glass half-full, a pessimist who sees a glass half-empty and a hospital manager who sees a glass twice as big as it needs to be.
3/6

Average percentage bed occupancy is the conventional way we measure bed usage in the NHS. (Here’s a visual example: 8,784 consecutive hour-by-hour snapshots of the number of beds occupied in a 46-bed hospital ward.) But the NHS has problems with bed occupancy as a metric.
2/6

On the edge My week in the NHS

This is a fascinating - and excellent - post by @samfr.bsky.social on his recent experience of being an NHS hospital patient. My heart jumped when I saw bed occupancy mentioned in the article: I have “views” on bed occupancy, so I thought I’d say something about it.
substack.com/inbox/post/1...
1/6

Yes, perhaps we do need to be careful. One of my objectives here is to find a way of engaging clinicians in the specialties by giving them an indicator of dysfunction "of their own".

Thanks for this - I like this WrongWardHours idea a lot. Moving away from snapshots and building in a bit of history (including ED history) to the indicator. This might also allow me to add "number of non-clinical ward moves" into the mix...

As it stands, it's a snapshot metric. I tried doing it for one specialty (Respiratory Medicine) based on hour-by-hour snapshots from two years ago. I posted it on X; can't remember if I posted it here. So here it is - in all its dual-vertical-axis glory...

Man walking into the Post Office earlier: You look like a person who knows about things.
Person behind the til: I do? Well I know about some things I suppose.
Man (showing him his phone): Good. What do I press to answer this thing?

...and called the package {abracadabra}...

I still think of pivot_wider() as magic, basically.