Abstract
We describe a method for predicting waiting times in radiation oncology using machine learning. The patient waiting experience remains one of the most vexing challenges facing healthcare. At our comprehensive cancer centre, waiting periods that arise throughout a patient’s course of treatment are generally difficult for staff to predict and only rough estimates are typically provided based on personal experience. To the patient, waiting times feel long and are seemingly unpredictable. Delays for treatment at our centre depend on the durations of preceding patients scheduled in the queue. To that end, we have incorporated the treatment records of all previously-treated patients into a machine learning framework in order to predict treatment durations to infer an overall waiting time. We found that the Random Forest Regression model provides the best predictions for daily fractionated radiotherapy treatment durations. Using this model, we achieved a median residual (actual minus predicted duration) of 0.25 minutes and a standard deviation residual of 6.1 minutes to retrospective treatment data. Waiting times are derived by summing the predicted durations. The main features that generated the best fit model (from most to least significant) are: Allocated appointment time, radiotherapy fraction number, median past duration of treatments, the number of treatment fields, and previous treatment duration.
Tarek Hijal
Associate Professor, Co-founder of the Opal Health Informatics Group
