Predicting Early Hospital Discharge Following Revision Total Hip Arthroplasty: An Analysis of a Large National...

Area of Science:

• Orthopedic Surgery
• Health Services Research
• Data Science in Medicine

Background:

• Revision total hip arthroplasty (rTHA) is no longer exclusively an inpatient procedure.
• Selecting appropriate candidates for outpatient rTHA is crucial for successful outcomes.
• Predictive models can aid in identifying suitable patients for early hospital discharge (EHD).

Purpose of the Study:

• To evaluate machine learning (ML) and multivariable logistic regression (MLR) models in predicting EHD (< 24 hours) after rTHA.
• To identify key perioperative variables associated with EHD using ML and MLR.
• To assess the utility of national database analysis for predicting EHD in rTHA patients.

Main Methods:

• Utilized a large national database from 2021, including 3,097 patients undergoing unilateral rTHA.
• Employed ML regression and various ML techniques, alongside MLR, to predict EHD.
• Compared model performance using AUC, calibration, Brier score, and decision curve analysis; identified feature importance.

Main Results:

• The random forest ML model demonstrated superior performance in predicting EHD.
• Identified predictive factors for EHD: aseptic indication, operative time < 3 hours, absence of anemia, neuraxial anesthesia, White race, male sex, independent function, BMI > 20, age < 75, and home support.
• All identified factors were also significant in the MLR model.

Conclusions:

• Both ML models and MLR effectively predicted EHD after rTHA, highlighting clinically significant variables.
• ML techniques, like random forest, offer accurate preoperative risk stratification for rTHA patients.
• Optimizing resource allocation and improving patient outcomes are potential benefits of using ML for EHD prediction in rTHA.