Prediction of mortality in Intensive Care Units: a multivariate feature selection

Monteiro, Flávio; Meloni, Fernando; Baranauskas, José Augusto; Macedo, Alessandra Alaniz

doi:10.1016/j.jbi.2020.103456

Cited by 14 publications

(11 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The reasons that we employed the above four ML methods are as follows. Firstly, SVM was a frequently used single ML method to deal with complex ICU data, and it showed robust performance in handling noisy and nonlinearly classified data [ 11 , 12 ]. Secondly, the ensemble learning method could combine multiple ML models to achieve better performance and generalizability than a single one [ 13 ], while RF and GBDT are typical ensemble learning models with different ensemble mechanisms.…”

Section: Introductionmentioning

confidence: 99%

Predicting Prolonged Length of ICU Stay through Machine Learning

Lin

et al. 2021

Diagnostics

View full text Add to dashboard Cite

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.

show abstract

Section: Introductionmentioning

confidence: 99%

Predicting Prolonged Length of ICU Stay through Machine Learning

Lin

et al. 2021

Diagnostics

View full text Add to dashboard Cite

show abstract

“…The authors obtained the best results by selecting the wrapper of SVM and Pearson correlation filter, preceding the training of SVM and logistic regression models, respectively. The paper presented in [Monteiro et al 2020] also addresses the selection of attributes as an 4:5 essential technique for improving predictive performance, however, in the context of mortality in intensive care units (ICUs). Among other multivariate statistical analysis methods, the authors analyze principal components to reduce dimensionality.…”

Section: Related Workmentioning

confidence: 99%

“…However, the hospitalization need of patients is addressed here rather than the need for respiratory support. In addition, although we use automatic attribute selection, as [Cueto-López et al 2019] and [Monteiro et al 2020], to improve the predictive model, we propose a hybrid selection approach, which uses a GA in the search for an optimal subset of attributes, addressing the selection process as a single-goal optimization problem, as in [Pawlovsky and Matsuhashi 2017] and [Maleki et al 2021]. Finally, despite we based our approach in [Maleki et al 2021] work, in addition to a different context, we considered large databases, with information from more than 200 thousand patients, which motivated different technical decisions.…”

Section: Related Workmentioning

confidence: 99%

Predicting COVID-19 hospitalizations with attribute selection based on genetic and classification algorithms

Colpo

Alves

Pereira

et al. 2022

iSys

View full text Add to dashboard Cite

The COVID-19 pandemic has been pressuring the whole society and overloading hospital systems. Machine learning models designed to predict hospitalizations, for example, can contribute to better targeting hospital resources. However, as the excess of information, often irrelevant or redundant, can impair predictive models’ performance, we propose a hybrid approach to attribute selection in this work. This method aims to find an optimal attribute subset through a genetic algorithm, which considers the results of a classification model in its evaluation function to improve the hospitalization need prediction of COVID-19 patients. We evaluated this approach in two official databases from the State Health Secretariat of Rio Grande do Sul, covering COVID-19 cases registered up to October 2020 and June 2021, respectively. As a result, we provided an increase of 18% in the classification precision for patients with hospitalization necessities in the first database, while in the second one, considering a temporal evaluation with sliding window, this gain was on average 6%. In a real-time application, this would also mean greater precision in targeting resources and, consequently and mainly, improved service to the infected population.

show abstract

“…The critically ill patients in intensive care units (ICUs) demand intensive care services and highly qualified multidisciplinary assistance [ 6 ]. Although ICU plays an integral role in maintaining patients’ life, this also implies the workforce shortage, limited medical resources, and heavy economic burden [ 7 ].…”

Section: Introductionmentioning

confidence: 99%

Predicting Mortality in Intensive Care Unit Patients With Heart Failure Using an Interpretable Machine Learning Model: Retrospective Cohort Study

Li¹,

Liu²,

Hu³

et al. 2022

J Med Internet Res

View full text Add to dashboard Cite

Background Heart failure (HF) is a common disease and a major public health problem. HF mortality prediction is critical for developing individualized prevention and treatment plans. However, due to their lack of interpretability, most HF mortality prediction models have not yet reached clinical practice. Objective We aimed to develop an interpretable model to predict the mortality risk for patients with HF in intensive care units (ICUs) and used the SHapley Additive exPlanation (SHAP) method to explain the extreme gradient boosting (XGBoost) model and explore prognostic factors for HF. Methods In this retrospective cohort study, we achieved model development and performance comparison on the eICU Collaborative Research Database (eICU-CRD). We extracted data during the first 24 hours of each ICU admission, and the data set was randomly divided, with 70% used for model training and 30% used for model validation. The prediction performance of the XGBoost model was compared with three other machine learning models by the area under the curve. We used the SHAP method to explain the XGBoost model. Results A total of 2798 eligible patients with HF were included in the final cohort for this study. The observed in-hospital mortality of patients with HF was 9.97%. Comparatively, the XGBoost model had the highest predictive performance among four models with an area under the curve (AUC) of 0.824 (95% CI 0.7766-0.8708), whereas support vector machine had the poorest generalization ability (AUC=0.701, 95% CI 0.6433-0.7582). The decision curve showed that the net benefit of the XGBoost model surpassed those of other machine learning models at 10%~28% threshold probabilities. The SHAP method reveals the top 20 predictors of HF according to the importance ranking, and the average of the blood urea nitrogen was recognized as the most important predictor variable. Conclusions The interpretable predictive model helps physicians more accurately predict the mortality risk in ICU patients with HF, and therefore, provides better treatment plans and optimal resource allocation for their patients. In addition, the interpretable framework can increase the transparency of the model and facilitate understanding the reliability of the predictive model for the physicians.

show abstract

Prediction of mortality in Intensive Care Units: a multivariate feature selection

Cited by 14 publications

References 24 publications

Predicting Prolonged Length of ICU Stay through Machine Learning

Predicting Prolonged Length of ICU Stay through Machine Learning

Predicting COVID-19 hospitalizations with attribute selection based on genetic and classification algorithms

Predicting Mortality in Intensive Care Unit Patients With Heart Failure Using an Interpretable Machine Learning Model: Retrospective Cohort Study

Contact Info

Product

Resources

About