Machine Learning Approaches to Define Candidates for Ambulatory Single Level Laminectomy Surgery
Qiyi Li1, Haoyan Zhong2, Federico P Girardi3
1Department of Orthopaedics, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences & Peking Union Medical College, Peking, China.
Global Spine Journal
|January 7, 2021
Summary
Machine learning algorithms can identify patients suitable for same-day discharge after laminectomy surgery. Preoperative lab values are key predictors for same-day ambulatory surgery candidates.
Area of Science:
- Neurosurgery
- Data Science
- Health Informatics
Background:
- Ambulatory surgery is expanding for spinal procedures like laminectomy.
- Identifying suitable candidates for same-day discharge is crucial for efficient patient pathways.
- Predictive modeling can optimize surgical candidate selection.
Purpose of the Study:
- To compare two machine learning algorithms for predicting same-day discharge after laminectomy.
- To identify key patient characteristics associated with ambulatory same-day laminectomy surgery.
Main Methods:
- Retrospective cohort study using the National Surgical Quality Improvement Program database (2017-2018).
- Trained Artificial Neural Network (ANN) and Random Forest (RF) models to predict same-day discharge.
- Evaluated model performance using AUC, accuracy, sensitivity, specificity, PPV, and NPV.
Main Results:
- 37.1% of 35,644 laminectomy patients were discharged same-day.
- Both ANN and RF models showed satisfactory performance (AUC ~0.77).
- Key predictors included age, operation duration, BMI, and preoperative lab values (hematocrit, platelets, WBC, alkaline phosphatase).
Conclusions:
- Machine learning models effectively identify candidates for ambulatory laminectomy.
- Preoperative laboratory values play a significant, previously unrecognized role in same-day discharge prediction.
- These findings can inform patient pathway design for ambulatory spinal surgery.
