In line with the Trust’s aims and objectives, investigation into theatre optimisation is of interest to increase hospital efficiency and decrease patient waiting lists. An area related to theatre optimisation includes theatre scheduling and understanding the reasons why theatres overrun or get cancelled. One small area in this wider picture relates to pre-medication requirements prior to a patient’s surgery. Pre-medication is offered to reduce preoperative patient anxiety and to increase cooperation with anaesthesia. If a patient is not expected to require pre-medication but are then found to experience benefits from undergoing pre-medication, this can lead to surgical delays which can cause knock on effects on the theatre list for that day.

This project aimed to predict the likelihood of a patient requiring pre-medication based on patient and hospital features. An end-to-end extract, transform, load (ETL) pipeline was developed to extract the relevant data from the electronic patient record (EPR) system using the Digital Research Environment (DRE)’s standard data extraction processes. Using R, various logistic regression and tree-based models were developed to predict the binary classification of whether a patient would require pre-medication, (Yes/No, 1/0), prior to them coming in for their surgery. Using R, the models can be optimised and additional features can explored easily and compared to developed univariable base models by analysing model metrics (AIC, AUC). The outcome of this project included recommendations for tools which would improve theatre planning, by highlighting relevant patient features to consider when booking staff are constructing the theatre list.