Discharge planning suite

This project simulates expected discharges and pathway requirements for patients over a 10-day horizon. This can be helpful for planning for upcoming discharge activity within local community settings.

Features

• Simulates patient journeys for currently admitted population using models calibrated on historic data.
• Simulates new arrivals, and their journeys.
• Aggregates to a 10-day forecast.
• Displays model outputs on a Shiny dashboard.

Limitations

Currently, this project is designed for a specific Docker instance with a customised environment. It requires:

• R version 4.2.2
• Specific SQL tables pre-populated with relevant patient data
• Specific environment variables configured

Due to these dependencies, running the code directly (without modifications) on other machines is not feasible.

Package dependencies

• tidyverse: For data manipulation and visualization.
• tidymodels: For building and applying statistical models.
• RODBC: For connecting to SQL databases.
• fitdistr & fitdistrplus: For fitting probability distributions.
• tsibble: For time series data manipulation.
• fable & fabletools: For building and evaluating ARIMA model.
• actuar & extraDistr: For additional probability distributions.
• ggiraph & ggh4x: For creating advanced visualisations.
• shiny & shinydashboard: For building interactive web applications

Code overview

The core functionalities are implemented in separate R scripts:

• code_main.R: This script runs the main simulation model for predicting discharges and pathways.

• code_los_model.R & code_pathway_model.R: These scripts calibrate the models used for predicting length-of-stay and discharge pathways, respectively. Calibration should be done before running the main model.