A team of Duke researchers within the Laboratory for Transformative Administration (LTA) has published a study in Annals of Surgery Open that describes how they developed machine learning (ML) models to predict, at the time the surgery is requested by the surgeon, the post-surgical length of stay and discharge disposition for adult elective inpatient cases.
While the Duke Department of Surgery already uses ML models to predict how long a surgery will last, this new publication highlights their development of new models for predicting LOS and DD. The researchers believe these models could improve hospital operations in a number of ways, including for case scheduling, resource allocation, optimal bed utilization, earlier discharge planning, and preventing case cancelation due to bed unavailability.
Related: Addressing Barriers to Care Through Strategic, Data-Based Solutions
The models were developed using data from more than 63,000 patients who had elective inpatient surgery in Duke Health. The models were able to predict the length of stay with an accuracy of 81% and the discharge disposition with an accuracy of 88%.
“This study is important because it shows that machine learning can be used to improve hospital efficiency and patient care,” says Daniel Buckland, MD, PhD, Co-Investigator and Medical Director of the LTA and co-author of the study publication. “Since this paper was submitted for publication, the model is now implemented within Duke and we are hopeful that these models can be used by hospitals around the world to improve the delivery of care.”