Machine learning model closely predicts patient waiting times for CT, MRI
Machine learning might be the next step in predicting patient wait times and appointment delays—factors crucial to healthcare’s quadruple aim and its emphasis on quality of care—in radiology practices, researchers have reported in the Journal of the American College of Radiology.
“Being able to accurately predict waiting times and scheduled appointment delays can increase patient satisfaction and enable staff members to more accurately assess and respond to patient flow,” first author Catherine Curtis, MS, and colleagues at Massachusetts General Hospital in Boston wrote. Prior to their current work, Curtis and her team developed a handful of predictive models based on patient waiting line sizes.
The researchers’ previous technique, which was embraced by Massachusetts General when the hospital began broadcasting predicted waiting times on monitors in reception areas, saw success, they wrote—just not enough.
“We noticed that most patients who were dissatisfied with the displayed waiting times were delayed for longer than predicted, so the need for more accurate models became imminent,” Curtis et al. said. “We also wanted to predict not only waiting times for walk-in facilities, but also delays for the scheduled facilities.”
Stepping outside of existing research, Curtis and her co-authors honed in on machine learning. The models can resist noise, adapt to changing environments and run without human supervision, the researchers wrote, which fit the needs of a waiting room to a T.
The team considered CT, MRI, ultrasound and radiography—only the last of which offered walk-in appointments—for the study. They evaluated 10 machine learning algorithms, including neural network, random forest, support vector machine, elastic net, multivariate adaptive regression splines, k-th nearest neighbor and linear regression, to find the algorithm that most closely predicted waiting times.
The group also took into consideration predictive variables like date and time, scheduling conflicts, patient flow and number of open exam rooms, they wrote.
Across the four imaging modalities, elastic net, an algorithm that works particularly well with large datasets, performed best. Comparing the root-mean-square errors of the 10 algorithms, the authors reported random forest, linear regression and multivariate adaptive regression splines fell just behind elastic net, while neural network and k-th nearest neighbor dropped to the bottom of the list.
"The concept of machine learning is certainly not new," Curtis et al. wrote. “This comparison demonstrated that machine learning with elastic net outperformed other ML algorithms in prediction accuracy and model simplicity."
The newfound method also outperformed their previous predictive model, the authors said.
“This definitively demonstrates machine learning’s potential in predicting workflow events,” they wrote.
After graduating from Indiana University-Bloomington with a bachelor’s in journalism, Anicka joined TriMed’s Chicago team in 2017 covering cardiology. Close to her heart is long-form journalism, Pilot G-2 pens, dark chocolate and her dog Harper Lee.