BING BONG NEW PAPER

We pulled together a huge team to look at how we can bring NHS data analysis into the 21st century.

THREAD! https://journals.sagepub.com/doi/10.1177/0141076820930666

Academics love to run regression analyses and publish academic papers as PDFs in obscure academic journals, all driven by the magical belief that this will change practice (lol). But that's only one small niche use-case for NHS and health data... https://journals.sagepub.com/doi/10.1177/0141076820930666

...The NHS is full of people doing practical, coalface data analysis: monitoring variation in care to spot quality improvement opportunities; modelling waiting lists, or optimum service locations; evaluating services; monitoring activity volume and cost; and lots more...

There are an estimated 10,000 analysts in the NHS doing this work. Their work is vital for improving patient care, spotting problems early, and building efficiency. They require similar skills, methods and tools to traditional epidemiology research. But, and it's a big but...

... But this practical analytics workforce in the NHS has been neglected, to a staggering degree. You will find almost no formal training, no textbooks, no CPD, no clear career paths to senior roles. At worst there is a culture of manual labour in Excel...

.. there are pockets of excellence, but they are mostly hidden, because there is no "commons of knowledge", no shared literature, no culture of sharing code, workbooks and methods in the open. In short: there is a HUGE amount of pent up NHS analyst talent waiting to be unlocked.

We assembled a vast group of bigwigs from a huge number of different sectors to sit down and sketch out the problems, and the solutions. In short: how can we help make NHS coalface data analysis great! https://journals.sagepub.com/doi/10.1177/0141076820930666

You can read the full paper for free here. In short: better career paths and training, based on other analytic professions; encourage, facilitate, and reward the use of modern, open, computational data science methods; and train bosses to ask better q's. https://journals.sagepub.com/doi/10.1177/0141076820930666

Does that all sound too much like wordy policy talk? Here are some concrete examples. Data analysts are currently classified in the NHS workforce under "admin/clerical" rather than "scientific/clinical". This speaks volumes!

Here's another concrete example: the costs, and dangers, of analysts working behind closed doors. At present there is almost no culture of NHS data analysts sharing code on GitHub, in Jupyter notebooks, and so on (with occasional fab exceptions!). Why does this matter? Well...

If you share your analytic code, everyone else can see exactly what you've done. They can helpfully notify you about mistakes or improvements. They can learn from your methods, your code, and approach. They can re-use your scripts, to avoid inefficient reduplication of effort..

Now reaching this glorious valhalla needs some prior work. We need to train people to use modern open computational data science methods, Python, R, Jupyter etc, not manual labour in Excel. But we also need to create a culture where sharing is ok, and indeed the norm...

.. barriers to this will vary. One barrier to sharing code is angst, which is very relatable. But your adequate code is perfectly adequate for sharing, as we have argued many times before!! https://twitter.com/bengoldacre/status/1197482152151175168

Another barrier might be your boss (or funder) fantasising that they're somehow going to monetise your data management scripts. There are also incentives, training, practical support, and so on. Anyway. READ THE PAPER. It's really good, and PRACTICAL.

https://journals.sagepub.com/doi/pdf/10.1177/0141076820930666

Just to finish, there are some "we'll know we've won when" statements. My favourite: "We'll know we've won when there are staff at board-level in managerial / strategic roles who once did an inner join in SQL, just as we have clinical leaders who once treated 1000s of patients."

THREAD BUMP NEW PAPER

How we can make NHS coalface data analysis great!

Avoid manual labour in Excel! Embrace modern, open, collaborative data science built around shared code! Unlock the pent up NHS analyst talent with proper career paths!

Onward! https://journals.sagepub.com/doi/10.1177/0141076820930666