AUTHOR=Lin Chun , Liang Zhen , Liu Jianfeng , Sun Wei 

TITLE=A machine learning-based prediction model pre-operatively for functional recovery after 1-year of hip fracture surgery in older people

JOURNAL=Frontiers in Surgery

VOLUME=Volume 10 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/surgery/articles/10.3389/fsurg.2023.1160085

DOI=10.3389/fsurg.2023.1160085

ISSN=2296-875X

ABSTRACT=Background
The use of machine learning (ML) to build high-performance prediction models has been widely used.This study was designed to develop an ML-based prediction model pre-operatively for identifying functional recovery after hip fracture surgery after 1 year.
Methods
We collected data of 176 elderly hip fracture patients who were admitted to the Department of Orthopaedics and Oncology of Shenzhen Second people's Hospital from May 2019 to December 2019 and who met the inclusion criteria. The functional recovery of the patients was followed up for 1-year after surgery. We selected 26 factors, including 12 preoperative indicators, 8 surgical indicators, and 6 postoperative indicators. 77 patients were finally included according to the exclusion criteria. They were randomly divided into the training set (70%) and test set (30%) for internal validation. Lasso method was used to screen prognostic variables. Comparisons between several common ML classifiers were conducted to select the best prediction model. The area under the receiver operating characteristic curve (ROC), calibration curve and decision curve analysis were used to evaluate prediction performance. The recursive feature elimination (RFE) algorithm based on Shapley Additive explanations (SHAP) values was performed to identify the importance of the predictor variables. 
Results
The AUCs for the testing dataset were logistic regression (Logit) model=0.934, k nearest neighbors(KNN) model=0.930, support vector machine (SVM) model=0.910, Gaussian naive Bayes(GNB) model=0.926, decision tree(DT) model=0.730, random forest (RF) model=0.957, and Extreme Gradient Boosting (XGB) model=0.902. Out of seven ML-based models tested, the RF model had the best prediction performance with 4 features, including postoperative rehabilitation compliance, marital status, age-adjusted Charlson comorbidity score (aCCI), and clinical frailty scale (CFS).
Conclusion
A prediction model for functional recovery after 1-year of hip fracture surgery in the elderly was established based on RF, and the prediction performance(ROC) of this model was better than that of other models. The software application was available. External validation in a larger group of patients or in diferent hospital settings is warranted to evaluate the clinical utility of this tool.