Predicting NHS OPEL-4 Status: A Bayesian Approach

Overview

Welcome to the training module on forecasting NHS operational pressures. The goal of this course is to build a probabilistic model that predicts the likelihood of a hospital trust entering “Severe Acute Patient Harm” (OPEL-4) within the next 10 days.

We will use R and the probabilistic programming language Stan to build a time-series model based on simulated daily operational data. This will allow us to quantify our uncertainty and make more informed, pre-emptive decisions to mitigate risk.

Course Outline

This course is broken down into the following modules. Please follow them in order.

  1. Setting Up Your Environment: Install R, RStudio, and all the necessary packages to get started.
  2. Data Preparation: Learn how to generate a synthetic dataset that mimics real-world NHS operational data.
  3. Forecasting Predictors with ARIMA: Learn how to forecast future values of your predictor variables using time-series models.
  4. Introduction to Bayesian Modelling: A gentle introduction to the concepts behind Bayesian inference and Stan.
  5. Building the Stan Model: Construct the core Bayesian time-series model for our prediction task.
  6. Running the Model and Interpreting Results: Execute the model, analyze the output, and visualize the 10-day forecast.
  7. Next Steps

Additional Resources

Training Developer

  • Dr Josh Tyler (Alan Turing Institute) - Website