Key result
XGBoost and logistic regression predict hip arthroplasty length of stay with 0.81 AUC using patient-reported outcomes.
Why the study?
Hospital length of stay varies widely across hip and knee arthroplasty patients, and prediction methods are needed to improve hospital capacity planning and identify patients at risk of long stays.
Do machine learning models accurately predict hospital length of stay in primary hip and knee arthroplasty patients?
Observational (n=4,688)
Yes
Do machine learning models accurately predict hospital length of stay in primary hip and knee arthroplasty patients?
Machine learning models, particularly XGBoost, demonstrate good to fair performance in predicting hospital length of stay for arthroplasty patients, with patient-reported outcome measures serving as relevant predictors.
Answers are grounded in this paper and cite their sources.
ML models may aid LOS prediction in arthroplasty; hypothesis-generating for clinical integration pending prospective validation.
Hospital length of stay (LoS) varies widely across knee (HA) or hip arthroplasty (HA) patients and depends on multiple factors. Prediction methods are necessary to improve hospital capacity planning and identify patients at risk of long LoS. This study aims (1) to compare the performance of previously applied ML as well as regression methods (XGBoost, NB, LogReg and LinReg) for either LoS classification or regression in a multi-hospital setting for primary HA and KA patients. In addition, the study aims (2a) to assess which variables are the most important predictors for LoS prediction and, specifically, (2 b) whether patient-reported outcome measures (PROMs) collected before surgery act as important predictors. 2611 primary HA and 2077 primary KA patients from eight German hospitals were included to train and test an extreme gradient boosting (XGBoost), naïve bayes (NB) and logistic regression (LogReg) for classification, and XGBoost as well as a linear regression (LinReg) for regression. Area under the receiver operating characteristics curve (AUC) and mean absolute error (MAE) were used as primary performance indicators for classification and regression. For classification, the highest AUC was reached by XGBoost and LogReg (AUC = 0.81) in the HA sample, whereby NB was statistically significantly outperformed by both other. In the KA sample, no statistical difference between any method was found, and AUC was lower for all models. For regression, MAE was lowest for XGBoost (1.21 days for KA and 1.43 days for HA). PROMs and hospital indicators tended to be the most relevant predictors in all cases. The study demonstrated good to fair performance of ML in predicting LoS. PROMs reflect relevant features for prediction. They should be routinely collected and used for practical applications. XGBoost can act as a superior prediction tool compared to regression or other ML models in certain circumstances.
No takes yet. Share an insight, caveat, or question.
Benedikt Langenberger (2023) conducted an observational in Primary hip and knee arthroplasty (n=4,688). Machine learning models (XGBoost, Naïve Bayes, Logistic Regression, Linear Regression) was evaluated on Area under the receiver operating characteristics curve (AUC) and mean absolute error (MAE) for length of stay prediction. XGBoost and Logistic Regression predicted length of stay with an AUC of 0.81 in hip arthroplasty patients, utilizing patient-reported outcome measures as key predictors.
Synapse has enriched 10 related papers. Consider them for comparative context: