cdu-data-science-team / pharmacyforecasting

Putting the R in pharmacy- a Shiny application to make forecasts of medication use in pharmacies in Nottinghamshire Healthcare NHS Trust

Putting the R in phaRmacy

Inventory management of medicines within hospital pharmacies could be improved by harnessing advanced analytics to inform stock ordering requirements. This should release valuable resources which could be used for higher value clinical work. This project aims to develop, test and implement an open source analytical tool in R that will support and improve existing inventory management practices within our Trust pharmacy.

This Shiny application was too CPU intensive to run in production and this repository has been archived. An RMarkdown based approach was used instead.

This project is part of the Health Foundation's Advancing Applied Analytics programme. The Health Foundation is an independent charity committed to bringing about better health and health care for people in the UK.

pharmacyforecasting's Issues

Incorrect handling of days by forecast function

The forecast function counts backwards from the last date in the data frame and produces forecasts up the the current date in the dataframe. This behaviour is useful for training models but not for live use.

Change this behaviour, possibly leaving the other functionality in too for the purposes of model testing

URL is out of date on README

Due to the change in server the README reference to the hosted app needs updating to https://involve.nottshc.nhs.uk/rsconnect/pharmacy_forecasting/

Add simulation based code

It won't be part of the Shiny, but simulations are taking place to test the performance of different models in realistic ordering situations

make plots beautiful

• labels
• font size
• colours

see some of these resources

Read data live from data warehouse

Connect Shiny to the SQL reporting system

Add stock control functions

Add functionality to consider remaining stock and do reordering

No outstanding orders input to main function

The main stock control function does not have an input for outstanding orders, and they are just loaded from .rda

This is obviously no good for live use. Add in functionality to change outstanding orders in the function

Get rid of red error messages

I don't know why there are red error messages on here- I've put req() in various places but it doesn't seem to work

Add the ability to remove weekends

Forecast may be better for some drugs with the weekends removed

Improve handling of prediction intervals

Prediction intervals can more easily be added with the {fable} package, with the option to use bootstrapped intervals

Add save button to table

The reordering table needs an "Export to Excel" button

Cache the forecasting section

The forecasts are slow, and based on repetitive inputs, and would benefit from being cached https://shiny.rstudio.com/articles/caching.html

Produce simulation based code for the Shiny backend

Modify the existing simulation based code to produce a function that is capable of modelling the next four weeks of inventory to feed through the top line of the Shiny interface