| Summary: | The Emergency departments (ED) are the major entry point to the healthcare system. With the growing demand due to the increase of life expectancy and the greater number of diseases, it is mandatory for the ED’s to have a more efficient resource management in order to try and provide the best experience possible to its patients. If the resource demand is greater than the resources available, then ED crowding occurs. This phenomenon leads to several problems that affect the patient experience, like longer waiting times, lack of beds, patients in hallways, etc. One of the ways to improve patient satisfaction is through patient waiting time prediction, since it would allow for a better resource management in the ED and providing patients with a waiting time estimation on the triage increases patient satisfaction. The author collaborated with a Portuguese hospital near Lisbon using real ED data and built a prototype to predict the ED waiting time. The researcher complemented the ED original dataset with external data like weather information, DGS Announcements and number of football games, to try to find the most accurate model. To perform the prediction, the Naïve Bayes (NB) and Random Forest (RF) algorithms were applied in three different scenarios: the first one only with data from the original dataset, the second one where the number of football games and DGS announcements attributes were added and finally, a third one with the same dataset as the previous scenario but added weather information (temperature, wind, humidity and precipitation). The RF algorithm was the one with the best performance, especially in the third scenario. For this reason, the author used the RF algorithm with the variable inputs from the third scenario to perform the predictions on the prototype. The author concluded that the external data attributes added in both second and third scenarios were not the most important attributes for the waiting times, being the most important variables, the triage colors, disease category.
|
|---|