Development, Deployment, and Implementation of a Machine Learning Surgical Case Length Prediction Model and Prospective Evaluation.
Date
2023-06
Journal Title
Journal ISSN
Volume Title
Repository Usage Stats
views
downloads
Citation Stats
Abstract
Objective
Implement a machine learning model using only the restricted data available at case creation time to predict surgical case length for multiple services at different locations.Background
The Operating Room (OR) is one of the most expensive resources in a health system, estimated to cost $22-133 per minute and generate about 40% of the hospital revenue. Accurate prediction of surgical case length is necessary for efficient scheduling and cost-effective utilization of the OR and other resources.Methods
We introduced a similarity cascade to capture the complexity of cases and surgeon influence on the case length and incorporated that into a gradient boosting machine learning model. The model loss function was customized to improve the balance between over- and under-prediction of the case length. A production pipeline was created to seamlessly deploy and implement the model across our institution.Results
The prospective analysis showed that the model output was gradually adopted by the schedulers and outperformed the scheduler-predicted case length in Aug-Dec 2022. In 33,815 surgical cases across outpatient and inpatient platforms, the operational implementation predicted 11.2% fewer under-predicted cases and 5.9% more cases within 20% of the actual case length compared with the schedulers and only over-predicted 5.3% more. The model assisted schedulers to predict 3.4% more cases within 20% of the actual case length and 4.3% fewer under-predicted cases.Conclusions
We created a unique framework that is being leveraged every day to predict surgical case length more accurately at case posting time and could be potentially utilized to deploy future machine learning models.Type
Department
Description
Provenance
Subjects
Citation
Permalink
Published Version (Please cite this version)
Publication Info
Zaribafzadeh, Hamed, Wendy L Webster, Christopher J Vail, Thomas Daigle, Allan D Kirk, Peter J Allen, Ricardo Henao, Daniel M Buckland, et al. (2023). Development, Deployment, and Implementation of a Machine Learning Surgical Case Length Prediction Model and Prospective Evaluation. Annals of surgery, Publish Ahead of Print. 10.1097/sla.0000000000005936 Retrieved from https://hdl.handle.net/10161/28553.
This is constructed from limited available data and may be imprecise. To cite this article, please review & use the official citation provided by the journal.
Collections
Scholars@Duke
Allan Douglas Kirk
I am a surgeon with interest in immune management of transplant recipients. I am particularly interested in therapies that influence T cell costimulation pathways and adjuvant therapies that facilitate costimulation blockade to prevent the rejection of transplanted organs without undue suppression of protective immunity. I am also interested in understanding how injury, such as that occurring during trauma or in elective surgery, influences immune responses and subsequent healing following injury.
Peter Allen
I am a Surgical Oncologist with clinical and research training in pancreatic and hepatobiliary malignancy. In 2018, I joined Duke University as the Chief of Surgical Oncology, and the Chief of Surgery in the Duke Cancer Institute. Previously, I led the clinical and research efforts regarding pancreatic neoplasia within the Department of Surgery at Memorial Sloan Kettering Cancer Center, and served as the Associate Director for Clinical Programs within the David Rubenstein Center for Pancreatic Cancer Research. I also held the Murray F. Brennan endowed Chair in Surgery.
Over the previous ten years, I have been interested in the progression of pancreatic precursor lesions called intraductal papillary mucinous neoplasms (IPMN). These cystic precursor lesions of the pancreas present an opportunity for to both study cancer progression, and potentially prevent the development of this lethal malignancy. My research has focused on biomarker development to identify high-risk IPMN as well as studies evaluating the cause of this disease process. I have successfully completed phase II and phase III clinical trials in patients with pancreatic cancer and IPMN, and am currently the PI of a first-in-human multi-center randomized chemoprevention trial for pancreatic cancer that is targeting patients with high-risk IPMN.
My laboratory includes both pre and postdoctoral trainees, and they play a critical role in the development of our pancreatic cancer prevention program.
Daniel Buckland
Dr. Buckland is an Attending Physician at Duke University Hospital Emergency Department. He is the Director of the Duke Acute Care Technology Lab where he leads research in developing technology for the diagnosis and treatment of acute disease in data science and robotics projects by managing collaborative research projects between clinicians and engineers. His work at involves studying how advancements in autonomy impact safety critical systems, including the healthcare system. As part of his focus on autonomy, he is the Medical Director of the the Laboratory for Transformational Administration (LTA) an Operational Data Science group in the Duke Department of Surgery.
In addition, Dr. Buckland is the Deputy Chair of the Human System Risk Board of the Office of the Chief Health and Medical Officer via an Intergovernmental Personnel Act agreement with NASA, where he determines the human system risk of spaceflight and how standards, countermeasures, and mission design can mitigate risk.
Unless otherwise indicated, scholarly articles published by Duke faculty members are made available here with a CC-BY-NC (Creative Commons Attribution Non-Commercial) license, as enabled by the Duke Open Access Policy. If you wish to use the materials in ways not already permitted under CC-BY-NC, please consult the copyright owner. Other materials are made available here through the author’s grant of a non-exclusive license to make their work openly accessible.