Forecasting Hourly Patient Visits in the Emergency Department to Counteract Crowding

Morten Hertzum*
Royal School of Library and Information Science, University of Copenhagen, Copenhagen, Denmark

Article Metrics

CrossRef Citations:
Total Statistics:

Full-Text HTML Views: 7035
Abstract HTML Views: 2137
PDF Downloads: 609
ePub Downloads: 302
Total Views/Downloads: 10083
Unique Statistics:

Full-Text HTML Views: 2645
Abstract HTML Views: 1090
PDF Downloads: 387
ePub Downloads: 230
Total Views/Downloads: 4352

© 2017 Morten Hertzum

open-access license: This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 International Public License (CC-BY 4.0), a copy of which is available at: This license permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

* Address correspondence to this author at the Royal School of Library and Information Science, University of Copenhagen, Njalsgade 76, Bldg 4, Copenhagen, Denmark; Tel: 45 3234 1344; E-mail:



Emergency Department (ED) crowding is a frequent problem that causes prolonged waiting and increased risk of adverse events. While the number of daily and monthly patient arrivals can be forecasted with good accuracy, ED clinicians need hourly forecasts in their ongoing scheduling and rescheduling of their work.


We aim to assess whether the hour-by-hour evolution in patient arrivals and ED occupancy can be accurately forecasted using calendar variables.


We obtained data about the patient visits at four Danish EDs from January 2012 to January 2015, a total of 393717 ED visits. The data for 2012-2014 were used to create linear regression models, autoregressive integrated moving average (ARIMA) models, and – for purposes of comparison – naïve models of hourly patient arrivals and ED occupancy. Using the models, patient arrivals and ED occupancy were forecasted for every hour of January 2015.


Hourly patient arrivals were forecasted with a mean percentage error of 47-58% (regression), 49-58% (ARIMA), and 60-76% (naïve). Increasing the forecasting interval decreased the mean percentage error. ED occupancy was forecasted with better accuracy by ARIMA than regression models. With ARIMA the mean percentage error of the forecasts of the hourly ED occupancy was 69-73% for three of the EDs and 101% for the last ED. Factors beyond calendar variables might possibly have improved the models of ED occupancy, provided that information about these factors had been consistently available.


Hourly patient arrivals can be forecasted with decent accuracy. Forecasts of hourly ED occupancy are less accurate and their accuracy varies more across EDs.

Keywords: Crowding, Emergency department, Forecasting, Occupancy, Patient arrivals, Healthcare.