Predictive Modeling of ICU Healthcare-Associated Infections from Imbalanced Data. Using Ensembles and a Clustering-Based Undersampling Approach

Abstract

Early detection of patients vulnerable to infections acquired in the hospital environment is a challenge in current health systems given the impact that such infections have on patient mortality and healthcare costs. This work is focused on both the identification of risk factors and the prediction of healthcare-associated infections in intensive-care units by means of machine-learning methods. The aim is to support decision making addressed at reducing the incidence rate of infections. In this field, it is necessary to deal with the problem of building reliable classifiers from imbalanced datasets. We propose a clustering-based undersampling strategy to be used in combination with ensemble classifiers. A comparative study with data from 4616 patients was conducted in order to validate our proposal. We applied several single and ensemble classifiers both to the original dataset and to data preprocessed by means of different resampling methods. The results were analyzed by means of classic and recent metrics specifically designed for imbalanced data classification. They revealed that the proposal is more efficient in comparison with other approaches.

An explanatory machine learning framework for studying pandemics: The case of COVID-19 emergency department readmissions

“…For severely imbalanced classification problems, Sánchez-Hernández et al [ 47 ] suggest G-mean, the geometric mean of true positive rate (TPR) and true negative rate (TNR) (Eq. (1) ), as an objective measure of predictive power.…”

Applying Machine Learning Techniques to the Audit of Antimicrobial Prophylaxis

“…Both Bagging and boosting adopt this voting approach, but they derive the models in different ways. In Bagging, the models have equal weight, whereas in boosting, the base weak learners receive more weights, and the simple weak learners are combined into a more complex strong ensemble [25]. In this study, Bagging and AdaBoost approaches were implemented on the dataset to improve the classification performance of decision tree.…”

Selecting the Suitable Resampling Strategy for Imbalanced Data Classification Regarding Dataset Properties. An Approach Based on Association Models

“…There is no complete agreement between different authors about when a dataset is considered imbalanced. In this paper, we consider that a dataset is imbalanced when the imbalance ratio (IR) is higher than 1.5, since previous experiments [4,79] have shown that above this threshold, the classification of examples of the minority class(es) is usually significantly lower than the examples of the majority class(es). In addition, many studies in the literature on imbalanced classification start from this IR value to select the datasets [14].…”

“…However, these conclusions cannot be generalized since the datasets used in the study have a low imbalance ratio and most of them have a low number of attributes and examples. In [4], a clustering-based undersampling strategy combined with ensemble classifiers is proposed and tested with only one dataset in a medical application domain. An ensemble approach is also presented in [43] for assisted reproductive technology outcome prediction.…”

“…Healthcare and medical diagnosis are two of them, in which this problem is particularly relevant. The way to deal with it has been addressed in studies about infections acquired in intensive care units [3,4], indication of noninvasive mechanical ventilation [5] and predicting neurological outcome of patients with severe trauma [6,7] among others. In the field of information retrieval this drawback is also very present, especially in areas such as text classification or sentiment analysis, where relevant documents and sentiment polarity/intensity respectively show a clear class imbalance [8,9].…”