Predicting Prolonged Length of ICU Stay through Machine Learning

Abstract

This study aimed to construct machine learning (ML) models for predicting prolonged length of stay (pLOS) in intensive care units (ICU) among general ICU patients. A multicenter database called eICU (Collaborative Research Database) was used for model derivation and internal validation, and the Medical Information Mart for Intensive Care (MIMIC) III database was used for external validation. We used four different ML methods (random forest, support vector machine, deep learning, and gradient boosting decision tree (GBDT)) to develop prediction models. The prediction performance of the four models were compared with the customized simplified acute physiology score (SAPS) II. The area under the receiver operation characteristic curve (AUROC), area under the precision-recall curve (AUPRC), estimated calibration index (ECI), and Brier score were used to measure performance. In internal validation, the GBDT model achieved the best overall performance (Brier score, 0.164), discrimination (AUROC, 0.742; AUPRC, 0.537), and calibration (ECI, 8.224). In external validation, the GBDT model also achieved the best overall performance (Brier score, 0.166), discrimination (AUROC, 0.747; AUPRC, 0.536), and calibration (ECI, 8.294). External validation showed that the calibration curve of the GBDT model was an optimal fit, and four ML models outperformed the customized SAPS II model. The GBDT-based pLOS-ICU prediction model had the best prediction performance among the five models on both internal and external datasets. Furthermore, it has the potential to assist ICU physicians to identify patients with pLOS-ICU risk and provide appropriate clinical interventions to improve patient outcomes.

Explainable time-series deep learning models for the prediction of mortality, prolonged length of stay and 30-day readmission in intensive care patients

“…Thus, using a relatively narrow time window, which is 24 h, to predict long-term outcomes theoretically resulted in a weak predictive capability. However, the result is still competitive in all three outcome predictions because of the application of deep learning models with a small quantity of time-series variables ( 8 , 9 , 31 , 37 ).…”

“…However, traditional scoring systems, even some machine learning methods in predicting these outcomes, especially in stratifying the risk of readmission, have shown only modest results (7)(8)(9)(10). Although part of the existing work based on machine learning models seems promising (11)(12)(13), few of them are able to take advantage of the characteristics of features collected in the ICU, which are time-series forms.…”

Machine learning and medicine- a brief introduction

“…Length of stay has also been studied to predict the length of stay in the context of traumatic brain injury (TBI), 39 predictions were also made using patient vital signs 40 as well as a gradient boosted decision trees algorithm trained using data available in the eICU and MIMIC III datasets. 41 A model was able to predict readmission from information in the MIMIC III dataset significantly better than both the Stability and Workload Index for Transfer score (AUC = .65) and the Modified Early Warning Score (AUC = .58), with an AUC of .76. 42 While more work is clearly needed, these studies nonetheless provide a proof of concept for how future versions of these models could benefit intensivists.…”

Development and validation of predictive models for mortality and length of hospitalisation in adults with community-onset bacteraemia under the COVID-19 pandemic: Application of early data dynamics

“…Additionally, the accurate prediction of LOS may allow hospitals to increase capacity and better predict healthcare costs [16]. Previous studies have reported numerous predictive scores for LOS using machine learning (ML) techniques; some have analysed patients who underwent coronary artery bypass grafting [17,18] and others those with critical illnesses [19,20]. However, the majority of reported predictions have been developed with clinical data gathered at the time of initial hospitalisation [19,20] or during surgery [17,18].…”

“…Previous studies have reported numerous predictive scores for LOS using machine learning (ML) techniques; some have analysed patients who underwent coronary artery bypass grafting [17,18] and others those with critical illnesses [19,20]. However, the majority of reported predictions have been developed with clinical data gathered at the time of initial hospitalisation [19,20] or during surgery [17,18]. Little research has focused on predictive outcomes for individuals with sepsis.…”