Explore opportunities to reduce hospital care

A web app to explore opportunities to reduce hospital care via Types of Potentially-Mitigatable Activity (TPMAs).

The app is deployed openly to Posit Connect with no access requirements. Developers can also access a development version of the app.

For developers

This section is aimed at maintainers of the tool who work for The Strategy Unit Data Science team.

Framework

The app is made with Shiny and is an R package following the nolem approach.

Prerequisites

To develop the app, you must first:

• create an .Renviron file from the .Renviron.example template (restart R after making changes to this file)
• run pak::pak() to install required and developmental dependencies from the DESCRIPTION
• install Air, which is used for formatting the code
• install Jarl, which is used for linting the code

Develop and deploy

To launch the app locally for development purposes, run pkgload::load_all() and then run_app().

Before submitting a pull request, run locally:

• an R-CMD check, e.g. from the R console with devtools::check()
• the test suite, e.g. from the R console with testthat::test(), noting that you must reach 100% test coverage
• a format check, e.g. in your terminal with air format .
• a lint check, e.g. in your terminal with jarl check .

When ready to deploy to Posit Connect, step through the dev/deploy.R script. Note that you should deploy to the 'dev' deployment following pull-requests, or to 'prod' following a release.

Data

Location

Underlying data is generated via the NHP inputs-data pipeline in the nhp_data repository and is read into the app from the parquet files in the relevant Azure container (named in the AZ_CONTAINER_INPUTS environment variable).

Version

The version of the data fetched by the app is controlled by the DATA_VERSION environment variable. The value should be dev for local development and for the dev deployment. The production deployment should set this value to the latest version of the data in the form 'v5.2'.

Re-fetch data

Note that the inputs-data parquet files are downloaded to the app_data/ folder when you run_app().

Locally, you can force-redownload the data by (a) deleting app_data/ and re-sourcing app.R, or (b) by running get_all_data() with the argument redownload = TRUE.

On the server, authorised devs can invalidate the current data cache by appending ?reset_cache=true to the apps' canonical URLs (i.e. https://connect.strategyunitwm.nhs.uk/tpma-explorer and /tpma-explorer-dev). The data will be re-fetched the next time the app starts up.

File structure

In:

• app_data/ you can find data downloaded from Azure (if run_app() has been run at least once)
• data-raw/ you can can find code used to generate lookups in inst/app/reference/
• dev/ you can find the deploy.R script to deploy to Posit Connect
• inst/ you can find:
  • golem-config.yaml, which contains configuration (copied originally from nhp_inputs)
  • lookup data files in app/reference/
  • Markdown files under app/text/, which contain body and tooltip text
• R/ you can find:
  • Shiny modules (server and UI components) that are stored in mod_*.R scripts
  • functions to help prepare data in utils_*.R scripts
  • logic for user facing outputs (plots, tables) in fct_*.R scripts