nhs-r-community/statements-on-tools

The NHS-R Community statements on the use of tools including (but not exclusively) R and R Studio.

This is an evolving document which describes how and why to use R and other data science tools and to share and reuse code safely in health and social care settings.

The scope and content are expanding all the time as the community collaboratively produces a definitive statement of the NHS-R way.

Please file issues, make pull requests, and get involved, we're very happy from hear from friends from inside and outside of NHS-R.

statements-on-tools's Issues

NHS England GitHub Way of Working

https://github.com/nhsengland/github-WoW

Cannot locally build book because of child documents

Currently it is not possible to locally build the book due to the use of child documents - while you can get bookdown to render the document, it will only ever pull in the files from the main branch making it impossible to ever verify your own changes. The only way to ever check that the changes you make work would be to commit to main, then check.

I suggest that we remove the use of child documents and move to a single repo (#19, #23 seem to be handling this), then move to use bookdown in a more traditional sense with individual files. I'm happy to sort this.

Options then would be to either name each file as "00-filename.rmd", setting the numbers to order the chapters as we wish, set the order in _bookdown.yml (e.g., like rd4s did in the first edition), or move to quarto (e.g. as r4ds has moved to. My shout would be on the first option, I think this typically makes it clear how the book will be built just looking in the files pane.

Once we have done this we can sort of resolve #18 - though this will not be run once per day but run on merge's into main (usethis::use_github_action("bookdown") will probably be sufficient, and the action will need no tweaking)

Chapter 3 content missing from PDF copy of handbook

Sections 3.1 to 3.7 are currently missing from the PDF copy of the handbook (https://github.com/nhs-r-community/statements-on-tools/blob/main/docs/_main.pdf - starting from page 13), compared to the bookdown version hosted here: https://nhs-r-community.github.io/statements-on-tools/

Move to bookdown

This was started in a README file but will quickly expand and possibly would be better in bookdown. I used the bookdown template to build my own https://github.com/Lextuga007/bookdown-template but the actions need updating and this would be better with the NHSR colour branding.

Create and maintain a curated national open library of NHS analyst code

The Goldacare report calls for a "NHSA 19. Create and maintain a curated national open library of NHS analyst code"

The NHS-R community has already made rather a humble start with such an undertaking and the vision could explicitly call for this to be expanded.

(there is an issue relating to improving the formatting of this repo nhs-r-community/demos-and-how-tos#2)

Glossary of terms

Although there should be descriptions of each technical term within each chapter it would be great to collate them into a quick reference guide.

Don't know if we'd need to have internal links from this to relevant chapters?

Barriers to RAP adoption including open source coding

This article from OSR about overcoming barriers to adoption could be useful to incorporate/reference.

Jonny Pearson's talk at the NHS-R Conference 2022 - need link

NHS Digital How to publish your code in the open

Reproducible Analytical Pipelines - Master’s of Data Science - Bruno Rodrigues

https://apolitical.co/solution-articles/en/open-source-and-accessibility-how-we-can-make-government-digital-services-available-to-everyone

GitHub Action failing - missing image

Error in knitr::include_graphics("images/logo.jpg", dpi = NA) : 

Possibly better to link to https://github.com/nhs-r-community/assets/tree/main/logo?

Add CNAME for url https://tools.nhsrcommunity.com/

GitHub actions remove this currently.

IT security

Highlight that IT security needs to be contacted about site access like GitHub, published pages in GitHub, Slack, Netlify. People often get permissions that change as security is built upon and IT helpdesk may not necessarily understand the implication where IT security staff would.

Define code review

It may be useful to have a formal definition of code review. Here are some links I stole from the ever helpful Jess Morley.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9367192/
https://academic.oup.com/aje/article/190/10/2172/6218064?login=false
https://google.github.io/eng-practices/review/reviewer/

@tomjemmett may find this issue interesting

Addressing other current challenges to modern analytics adoption

On top of issues around the basics of getting open source tools installed onto Trust machines, there are further topics that must be addressed in order to move towards the analytics vision set out in the Goldacre Report, the NHSX Open Source policy and the NHSD RAP Community.

This include but are not limited to:

• Open code: publication of code in public repositories,
• Synthetic data: the creation and authorisation of synthetic/dummy datasets in order to improve reproducibility (amongst other things),
• Balancing analytics processes and software engineering.

Create GitHub Action to `bookdown::render_book()` to keep up with any changes in other repos

Not sure about frequency, but this will avoid having to remember to re-render the book every time one of the md files we read into it is updated in its original repo. Suggest once a day, unless there's a clever way of watching all the connected repos - that might be over-engineering it?

NHSR Vision

The vision was originally written from a workshop in 2021 but was very focussed on research, statistics and refers to clinicians rather than analysts and data scientists. It might be that the vision requires rewriting completely to include other languages like Python, tools like Git/GitHub and relationships to other communities like Government Data Science Campus and NHS.pycom.

Minor formatting questions

• Do we want to replace "document" with "chapter" within each chapter?
• Any strong feelings about whether to number sections/chapters?
• Anything else you'd like to see on the cover page (link to NHS-R twitter account?)

Maybe @bclarke-nes or @Lextuga007 would like to offer an opinion?

404 on pages site

Visiting https://tools.nhsrcommunity.com/ gives a 404 error

Pulling together some thoughts for Purpose chapter draft

I wanted to start bringing some thoughts together about the overarching purpose chapter. Roughly:

• I'll move some of the material from section 3.1 (about tools) into the main purposes chapter. It'll do more work there as part of a general introduction to the handbook, and allow us to cut to the chase on the specific tools/packages issues
• I'll copy some of the points over from 2.1 too
• And make some general comments about R / use in the NHS, based on @wbryant 's discussion in #4

Question: what else should I look at for this? Any suggestions very welcome.

Integrate the content from this book with NHS-R Way

There are a few cross overs on the content with NHS-R Way and also there is now a difference in book type between the two as NHS-R Way is Quarto (it was released many months after this was created).

In order to keep this open to R and Python users should this book be moved to quarto or merged directly into the NHS-R Way?

Front page needs to refer to all data science not just R

The opening page refers to R only and whilst this has been started within the NHS-R Community it needs to be wider in scope to reflect NHS + and R + 😃

Add link for AF Open Sourcing Analytical Code

Suggested link from the RAP community (through Government Data Science Community Slack) https://analysisfunction.civilservice.gov.uk/policy-store/open-sourcing-analytical-code/

Produce GitHub action to render document once a day

it would also be good to use GitHub Actions to render the book once a day so that it stays up to date with any changes in the other repos without needing to manually re-render it

Generalise to Python (and Docker?)

The arguments around R and R packages will be similar for Python and Docker, though their repos (e.g. PyPi and Dockerhub) are run somewhat differently to CRAN as I understand it.

Can this therefore be generalised, but with a few R/Python specific chunks mixed in there.

Not to overcomplicate things but I just discovered RMarkdown parameters; I could just imagine having a document here where you can select Python or R at the beginning and the entire document adjusts accordingly ...

Curriculum for training and learning

NHS-R needs to have a considered view on a "curriculum" or competency framework for an NHS-R analyst, and this document could be a place for that.

@Lextuga007 will have good ideas, as will others no doubt

Set out the book following NHS-R Way Style Guide

https://nhs-r-community.github.io/NHSR-way/community-handbook.html#write-each-sentence-in-a-new-line-line-breaks

Locally hosted Git and GitHub

There are various locally hosted Git platforms with varying costs and implications. These might also require a particular workflow from private to public.

Posit Cloud

Particularly in terms of sensitive data, i.e. never use it for analysis with patient identifiable or sensitive data.

This is often used for training where packages and data are pre-installed. Reasons why this can be helpful. How workspaces are available through organisations with paid subscriptions (like NHS-R Community) but also on individual accounts.

Project file layout

• Get file listing output https://stackoverflow.com/questions/57354989/fsdir-tree-sort-directory-tree
• https://www.r-bloggers.com/2018/08/structuring-r-projects/
• https://github.com/craig-shenton/rapR-package-template

Section on licences

What happens if code/GitHub repositories don't have a licence and suggestions/quick reference to what can be used and any implications.

Setting up git/github safely with .ignore files

A key risk associated with using git/github for healthcare settings, involving working with identifiable or potentially identifiable patient level data, would be in not thinking carefully enough about project/repo structures and which locations to .gitignore within a project/repo. For example, if someone filters on a half dozen records from a secure database it's important no commit contains these records, even though they may want to include the code which performs the filter.
This suggests it's important to have both a clear understanding about how to .gitignore locations, and a priori agreement about which folders inside a project should contain what kinds of data. Some discussion about data security roles within an active repo might be important to include too, so there's not any kind of 'incident' involving this kind of accidental release of data, which could set back progress on collaborative coding and version control quite quickly and quite fast.

Add how-to to README for adding a chapter to the book

It would be great to have a quick bit on how to add a proposed chapter to the book (i.e. where you "register" a .md file to be included).

Properly incorporate the NHS-R vision repo

The vision files are all separate .md files. Do they exist anywhere as one document? If not, we can look into compiling them into one

Rename the book to include the repository name Statement on Tools?

We now have a https://github.com/nhs-r-community/NHSR-way which is closer to the "Handbook" type concept and follows the Turing Way.

Healthcare data

• compiled by AnalystX includes synthetic data https://data.analystx.uk/

Expand side bar in book

The chapters are a bit squashed and go onto two lines. https://stackoverflow.com/questions/73000003/quarto-book-make-body-content-wider

Guide to deployment

Lots of people have questions about deploying Shiny and RMarkdown in the NHS. Someone should write this stuff down

Useful source material

NHSX have got loads of useful material for this repo, such as:

• https://nhsx.github.io/AnalyticsUnit/experiences.html
• https://nhsx.github.io/AnalyticsUnit/toolstackrequests.html

Data in packages

I was writing out the difference between R, R Studio and R packages:

R is a programming language.
R Studio is an integrated development environment which supports R and other languages.
[R packages](https://r-pkgs.org/intro.html) bundles together R code, data, documentation and tests.

And realised that the definition from R packages is that it includes data - that needs to be addressed around concerns for IG/Security. The only data in other people's packages will be (or should be) publicly available but then the issue of our own data should only be around our creating packages.

This probably needs to be spelled out for clarity.

Add contribution instructions

Taken from the quarto-report-nhs-theme repository as it can then be referred to from various repositories.

Is renv necessary?

Sorry if this is a stupid question but is renv necessary in this repo?

GitHub action failed

Check to see too if this relates to #42 where _main files epub , pdf and .tex are not being updated (at least from a local Build)

Practical guidance for adopting open source tooling and open coding

Part of what-is-the-need-in-the-system that could be of great benefit would be a knowledgebase/playbook for adopting and supporting open source tools and mindsets in the NHS. For instance someone shared some great resources addressing (I think) getting authorisation for R installation and running open source tools.

It's quite a broad topic so could be honed for NHS-R but I think it's a great community to be involved in that sort of work.