Related Experiment Video
Updated: Jun 6, 2025
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
Machine Learning Algorithms Exceed Comorbidity Indices in Prediction of Short-Term Complications After Hip Fracture
Anirudh K Gowd1, Edward C Beck, Avinesh Agarwalla
1From the Department of Orthopedic Surgery, Wake Forest University Baptist Medical Center, Winston-salem, NC (Gowd, Beck, Godwin, and Waterman), the Cedars Sinai Medical Center, Los Angeles, CA (Gowd), the Department of Orthopedic Surgery, Westchester Medical Center, Winston-salem, NC (Dr. Agarwalla), the Department of Health Policy and Management, University of North Carolina at Chapel Hill, Chapel Hill, NC (Patel), Department of Orthopedic Surgery, the Cedars Sinai Medical Center, Los Angeles, CA (Dr. Little), the USC Epstein Family Center for Sports Medicine, Keck Medicine of USC, Los Angeles, CA (Dr. Liu).
Machine learning (ML) algorithms significantly improve surgical risk assessment for hip fractures compared to traditional methods. These ML models offer a more reliable way to predict patient outcomes, including complications and discharge, aiding clinical decisions.
Area of Science:
- Orthopaedic Surgery
- Data Science
- Machine Learning in Medicine
Background:
- Hip fractures are a major cause of morbidity, often exacerbated by patient frailty.
- Accurate assessment of surgical risk is crucial for managing patients with hip fractures.
Purpose of the Study:
- To evaluate the reliability of machine learning (ML) algorithms for assessing surgical risk in hip fracture patients.
- To compare the performance of ML models against established comorbidity indices.
Main Methods:
- Utilized the American College of Surgeons National Surgical Quality Improvement Program data (2011-2018).
- Trained 3 ML algorithms to predict outcomes: extended length of stay (LOS), death, readmissions, home discharge, transfusion, and medical complications.
- Compared ML model performance (AUC, PPV, NPV) against legacy indices (ASA, Charlson, Frailty, Nottingham).
Main Results:
- ML models demonstrated superior predictive performance across all assessed complications compared to legacy comorbidity indices (P < 0.01).
- ML models achieved higher Area Under the Curve (AUC) values for predicting medical complications (0.65), death (0.80), extended LOS (0.69), transfusion (0.79), readmissions (0.63), and home discharge (0.74).
- The best-performing legacy indices showed significantly lower AUC values for each respective complication.
Conclusions:
- Machine learning algorithms provide a more comprehensive and accurate method for calculating preoperative risk in hip fracture patients.
- These algorithms can predict morbidity, mortality, and discharge destination more effectively than traditional scoring systems.
- Validated ML-based risk calculators hold potential to enhance clinical decision-making for healthcare providers and payers.
More Related Videos
07:25Predicting Amputation using Local Circulating Mononuclear Progenitor Cells in Angioplasty-treated Patients with Critical Limb Ischemia
Published on: September 22, 2020
09:51The Transition to an Anterior-Based Muscle Sparing Approach Improves Early Postoperative Function but is Associated with a Learning Curve
Published on: September 7, 2022