Predictive Analysis in Healthcare: Emergency Wait Time Prediction

Abstract

Emergency departments are an important area of a hospital, being the major entry point to the healthcare system. One of the most important issues regarding patient experience are the emergency department waiting times. In order to help hospitals improving their patient experience, the authors will perform a study where the Random Forest algorithm will be applied to predict emergency department waiting times. Using data from a Portuguese hospital from 2013 to 2017, the authors discretized the emergency waiting time in 5 different categories: "Really Low", "Low", "Average", "High", "Really High". Plus, the authors considered as waiting time, the time from triage to observation. The authors expect to correctly evaluate the proposed classification algorithm efficiency and accuracy in order to be able to conclude if it is valuable when trying to predict ED waiting times.

Predicting Multiple Domain Queue Waiting Time via Machine Learning

“…Waiting time estimation is a challenging task, since it can be affected by a wide range of phenomena (e.g., sudden increase of customers, employee attendance slowness) that often are not directly measured by ticket management systems. Recently, several research studies have addressed this task, assuming traditional approaches, such as: Average Predictions (AP) [18]; Queuing Theory [17]; and DM/ML approaches [7]. In this work, we detail the ML based approaches, since they are more related with our CRISP-DM approach.…”

“…In 2018, a study was carried out in Portugal regarding the use ML algorithms to predict waiting times in queues, assuming a categorical format, thus a multiclass classification task (e.g., "very high", "low") [7]. To validate the results, the authors used a dataset from an emergency department of a Portuguese hospital, containing 4 years of data and around 673.000 records.…”

A predictive model for the post-pandemic delay in elective treatment

“… [20] and Goncalves et al. [21] both adopt the Random Forest Regression methodology to predict waiting time and pinpoint the variables with the highest predictive power.…”

“…While these studies have generated reasonably sound results (e.g. accuracy of 50.09% [21] ), other authors have approached the problem using more holistic methodologies to mitigate the limitations of ‘black box’ machine learning techniques. In fact, Liu et al.…”

Merging Data Diversity of Clinical Medical Records to Improve Effectiveness

“…This integration allowed the prediction of waiting times based on the day of the week, on the month, weather conditions and correlations with special seasons, and events, among others. For details, see Reference [ 43 , 44 ].…”