Statewide validation of a patient admissions prediction tool

We validate a proprietary system to predict hospital emergency department presentations. A key advantage in planning health service delivery requirements and catering for the large numbers of people presenting to hospitals is the ability to predict their numbers. Year-ahead forecasts of daily hospital presentations were generated for 27 public hospitals in Queensland, Australia from five years of historic data. Forecast accuracy was assessed by calculating the Mean Absolute Percentage Error and Root Mean Squared Error between predictions and observed admissions. Emergency Department presentations were found to be not random and can be predicted with an accuracy of around 90%. Highest accuracy was over weekends and summer months, and Public Holidays had the greatest variance in forecast accuracy. Forecasts for urban facilities were generally more accurate than regional (accuracy is related to sample size).