BACKGROUND
Accurate prediction of hospital length of stay (LoS) is essential for efficient resource management. However, conventional LoS prediction models with limited covariates and non-standardized data tend to experience limited reproducibility and generalizability when applied to the general population.
OBJECTIVE
In this study, we aimed to develop and validate machine learning (ML)-based LoS prediction models for planned admissions using the Observational Medical Outcomes Partnership Common Data Model (OMOP CDM).
METHODS
Retrospective patient-level prediction models were developed using electronic health record (EHR) data converted to the OMOP CDM (version 5.3) from Seoul National University Bundang Hospital (SNUBH) in South Korea. The study included 137,437 hospital admission episodes that occurred between January 2016 and December 2020. Covariates from the patient, condition occurrence, medication, observation, measurement, procedure, and visit occurrence tables were included in the analysis. Lasso regularization was applied in logistic regression to perform feature selection. The primary outcome was an LoS of seven days or longer, and the secondary outcome was an LoS of 3 days or longer. The prediction models were developed using six ML algorithms, with the training and test set split in a 7:3 ratio. The performance of each model was evaluated based on the area under the receiver operating characteristic curve (AUROC) and the area under Precision-Recall curve (AUPRC). SHapley Additive exPlanations (SHAP) analysis was performed to measure feature importance, while calibration plots were used to assess the reliability of the prediction models. The developed models were externally validated at an independent institution, Seoul National University Hospital (SNUH).
RESULTS
A total of 129,938 patient entry events were included in the planned admissions. The XGB model achieved the best performance in binary classification for predicting an LoS of 7 days or longer, with an AUROC of 0.89 and an AUPRC of 0.82 on the internal test set. The LGB model performed the best in multi-classification for predicting an LoS of 3 days or more, with a mean AUROC of 0.83 and a macro-average AUPRC of 0.56. The most important features contributing to the models were the operation performed, frequency of previous outpatient visits, patient admission department, age, and day of week admission. The RF model showed robust performance in the external validation set, achieving an AUROC of 0.80.
CONCLUSIONS
The use of the OMOP CDM in predicting hospital LoS for planned admissions not only demonstrates promising predictive capabilities for stays of varying durations but also underscores the advantage of standardized data in achieving reproducible results. This approach could serve as a model for enhancing operational efficiency and patient care coordination across healthcare settings.