Photo Credit: David Gyung
The following is a summary of “Development, Deployment, and Implementation of a Machine Learning Surgical Case Length Prediction Model and Prospective Evaluation,” published in the December 2023 issue of Surgery by Zaribafzadeh, et al.
For a study, researchers sought to build a machine learning model that can predict the length of surgery cases for multiple services at different places using only the limited data available when the case was created. The operating room is one of the most expensive parts of a health system. It’s thought to cost between $22 and $133 per minute and bring in about 40% of the hospital’s income. It was important to make accurate predictions about how long a surgery case will take for efficient scheduling and cost-effective use of the operating room and other resources.
They used a similarity cascade to show how complicated cases are and how the operator affects the length of the case. Then, they added that to a gradient-boosting machine-learning model. The model loss function was changed to get a better mix between guessing the case length too much or too little. A production process was made so the model could be deployed and used across their school without any problems.
The future results showed that from August to December 2022, the model output did better than the schedulers’ predictions for case length. When looking at 33,815 surgery cases across outpatient and hospital platforms, the practical execution forecasted 11.2% fewer cases that were too short, 5.9% more cases that were within 20% of the real case length, and only 5.3% more cases that were too long. The model helped schedulers guess the length of 3.4% more cases within 20% of the real length of cases and 4.3% fewer cases that were longer than what was forecast. Conclusions: They made a one-of-a-kind framework that is used every day to more correctly predict the length of medical cases at the time they are posted. The framework could also be used to apply future machine-learning models.
