Artificial Intelligence to Get Insights of Multi-Drug Resistance Risk Factors during the First 48 Hours from ICU Admission

Abstract

Multi-drug resistance (MDR) is one of the most current and greatest threats to the global health system nowadays. This situation is especially relevant in Intensive Care Units (ICUs), where the critical health status of these patients makes them more vulnerable. Since MDR confirmation by the microbiology laboratory usually takes 48 h, we propose several artificial intelligence approaches to get insights of MDR risk factors during the first 48 h from the ICU admission. We considered clinical and demographic features, mechanical ventilation and the antibiotics taken by the patients during this time interval. Three feature selection strategies were applied to identify statistically significant differences between MDR and non-MDR patient episodes, ending up in 24 selected features. Among them, SAPS III and Apache II scores, the age and the department of origin were identified. Considering these features, we analyzed the potential of machine learning methods for predicting whether a patient will develop a MDR germ during the first 48 h from the ICU admission. Though the results presented here are just a first incursion into this problem, artificial intelligence approaches have a great impact in this scenario, especially when enriching the set of features from the electronic health records.

Keywords: Intensive Care Unit; antibiotics; artificial intelligence; feature selection; machine learning; multi-drug resistance; risk factors.

Figure 5

(a) Rate of episodes for each department of origin, normalized for non-MDR and MDR patient episodes; (b) rate of episodes for each reason of admission, normalized for non-MDR and MDR patient episodes.

Figure 6

Rate of patient episodes for both MDR and non-MDR when three clinical scores are considered: (a) Apache II, (b) Charlson and (c) SAPS III.

Figure 7

Rate of MDR and non-MDR patient episodes associated with: (a) each group of diseases; (b) illness presence.

Figure 1

Workflow diagram of the proposed methodology to get insights of multi-drug resistance (MDR) risk factors and to predict whether a patient will develop an MDR germ during the first hours from the Intensive Care Unit (ICU) admission.

Figure 2

Three possible scenarios for the confidence interval (CI) of $\Delta s_{AB,x_j}$. The feature $x_j$ will be selected in Scenarios 1 and 3.

Figure 3

Schematic dataset description.

Figure 4

Histogram of age for: (a) non-MDR patients; and (b) MDR patients.

Figure 8

Rate of MDR and non-MDR patient episodes per family of antibiotics administered during the first 48 h.

Figure 9

Average of the p-values for the 95 initial features when considering bootstrap and the proportion/median test for binary and numerical features, respectively, with a significance level of $\alpha$ = 0.05.

Figure 10

Averaged mutual information (MI) values when bootstrapping the patient episodes for each feature. Features with very low MI values are not shown here. In green, the 18 features with higher MI values.

Figure 11

CI for numerical features ($C{I}_{\Delta m}$ ) and for binary features ($C{I}_{\Delta p}$) when bootstrapping MDR and non-MDR patient episodes. The selected features are represented in green.

Figure 12

Description of the selected features with the three Feature Selection (FS) methods: Proportions and Median Test, MI and CI. The final set of selected features is the union of the features identified with each FS strategy.