Leveraging large, real-world data through machine-learning to increase efficiency in robotic-assisted total knee...

Area of Science:

• Orthopedic surgery
• Data science in medicine
• Surgical outcomes research

Background:

• Operative time in total knee arthroplasty (TKA) is influenced by patient, surgeon, and surgical variables.
• Predictive modeling using machine learning (ML) offers a potential solution for optimizing operating room (OR) efficiency.

Purpose of the Study:

• To identify demographic, surgeon, and surgical factors influencing operative times in TKA.
• To develop and evaluate an ML model for estimating operative time in robotic-assisted primary TKA.

Main Methods:

• Retrospective analysis of over 300,000 primary TKA cases (2007-2020).
• Evaluation of demographic and surgical variables using statistical tests to identify predictors for ML models.
• Development of two ML algorithms to predict operative time for robotic-assisted TKA, with performance assessed using RMSE, R², and prediction error rates.

Main Results:

• Factors associated with increased operative time include male sex, BMI > 40 kg/m², and cemented implants.
• Factors associated with reduced operative time include age > 65 years, cementless fixation, and high surgeon case volume.
• Robotic-assisted TKA impacted operative time differently based on surgeon volume; ML models outperformed historical averages in predicting operative time.

Conclusions:

• Surgeon case volume, cementless fixation, manual TKA, and patient characteristics (female, older, non-obese) are key factors in reducing operative time.
• ML-based operative time prediction demonstrates superior accuracy compared to historical averages.
• Accurate ML predictions hold potential for optimizing operating room utilization and resource management.