Time series forecasting in an outpatient cancer clinic using common-day clustering

Abstract

The use of forecasting methods in healthcare settings can lead to operational improvements and improved patient care. However, many outpatient care facilities do not engage in demand forecasting and those that do often use rudimentary methods without exploring the best technique to forecast their patient demand. This research study examines the application of time series forecasting techniques to daily patient volume levels at an outpatient cancer treatment clinic. The work focuses on the optimal methods for accurate day-ahead forecasting in this healthcare setting with particular attention given to the differing forecast performance characteristics between traditional calendar sequencing and common-day clustering of the time series data. Through the construction of various forecasting models across multiple patient treatment duration categories, it is found that modifying a time series to a common-day clustered sequence can provide a statistically significant improvement in the accuracy of a forecast.