A study of Hospitals’ resources in reducing non-urgent elective surgery waiting list through Time Series Methods and Simulations: Analysis of data from NSW Local Health Districts

Abstract

This thesis strives to improve the elective surgery wait list by predicting the future demand and optimizing resources using discrete event simulation, while analysing the 2016 – 2021 Urology, Orthopaedics, and Gynaecology elective surgery wait list data from Nepean Blue Mountain Local Health District (NBMLHD) in New South Wales, Australia. In this thesis, we illustrate hybrid ARIMA – Machine Learning models to forecast the elective surgery wait list demand by predicting the residuals from ARIMA model using machine learning models. The results are then compared with the traditional ARIMA and Machine Learning models’ performances. We found that ARIMA – ANN one week feed-forward hybrid model provides the best predictions in both the Urology and Gynaecology data, while hybrid ARIMA – LSTM one week feed-forward model gives the best predictions for the Orthopaedics data from NBMLHD.

For illustration, we also emulate the NBMLHD elective surgery process by creating a discrete event simulation (DES) using ARENA simulation software. Our results based on ARENA discrete event simulation model showed a significant decrease in wait times when distributing resources to smaller institutions. A reason for this outcome is that the patients from smaller institutions now have the option to elect to schedule their surgery at local medical centers rather than travel to larger hospitals. We also used the ARENA model to calculate the cost of resources used in a simple way for each elective surgery and allow us to monitor the wait times for individual patients throughout the process through the entity mechanism in the ARENA software.