TY - JOUR
KW - Antimicrobial Resistance
KW - Intensive Care Unit
KW - Prediction
KW - Pseudomonas Aeruginosa
KW - Temporal Data-Driven Modeling
AU - Àlvar Hernàndez-Carnerero
AU - Miquel Sànchez-Marrè
AU - Inmaculada Mora-Jiménez
AU - Cristina Soguero-Ruiz
AU - Sergio Martínez-Agüero
AU - Joaquín Álvarez-Rodríguez
AB - One threatening medical problem for human beings is the increasing antimicrobial resistance of some microorganisms. This problem is especially difficult in Intensive Care Units (ICUs) of hospitals due to the vulnerable state of patients. Knowing in advance whether a concrete bacterium is resistant or susceptible to an antibiotic is a crux step for clinicians to determine an effective antibiotic treatment. This usual clinical procedure takes approximately 48 hours and it is named antibiogram. It tests the bacterium resistance to one or more antimicrobial families (six of them considered in this work). This article focuses on cultures of the Pseudomonas Aeruginosa bacterium because is one of the most dangerous in the ICU. Several temporal data-driven models are proposed and analyzed to predict the resistance or susceptibility to a determined antibiotic family previously to know the antibiogram result and only using the available past information from a data set. This data set is formed by anonymized electronic health records data from more than 3300 ICU patients during 15 years. Several data-driven classifier methods are used in combination with several temporal modeling approaches. The results show that our predictions are reasonably accurate for some antimicrobial families, and could be used by clinicians to determine the best antibiotic therapy in advance. This early prediction can save valuable time to start the adequate treatment for an ICU patient. This study corroborates the results of a previous work pointing that the antimicrobial resistance of bacteria in the ICU is related to other recent resistance tests of ICU patients. This information is very valuable for making accurate antimicrobial resistance predictions.
IS - Special Issue on Artificial Intelligence, Paving the Way to the Future
M1 - 5
N2 - One threatening medical problem for human beings is the increasing antimicrobial resistance of some microorganisms. This problem is especially difficult in Intensive Care Units (ICUs) of hospitals due to the vulnerable state of patients. Knowing in advance whether a concrete bacterium is resistant or susceptible to an antibiotic is a crux step for clinicians to determine an effective antibiotic treatment. This usual clinical procedure takes approximately 48 hours and it is named antibiogram. It tests the bacterium resistance to one or more antimicrobial families (six of them considered in this work). This article focuses on cultures of the Pseudomonas Aeruginosa bacterium because is one of the most dangerous in the ICU. Several temporal data-driven models are proposed and analyzed to predict the resistance or susceptibility to a determined antibiotic family previously to know the antibiogram result and only using the available past information from a data set. This data set is formed by anonymized electronic health records data from more than 3300 ICU patients during 15 years. Several data-driven classifier methods are used in combination with several temporal modeling approaches. The results show that our predictions are reasonably accurate for some antimicrobial families, and could be used by clinicians to determine the best antibiotic therapy in advance. This early prediction can save valuable time to start the adequate treatment for an ICU patient. This study corroborates the results of a previous work pointing that the antimicrobial resistance of bacteria in the ICU is related to other recent resistance tests of ICU patients. This information is very valuable for making accurate antimicrobial resistance predictions.
PY - 2021
SP - 119
EP - 133
T2 - International Journal of Interactive Multimedia and Artificial Intelligence
TI - Antimicrobial Resistance Prediction in Intensive Care Unit for Pseudomonas Aeruginosa using Temporal Data-Driven Models
UR - https://www.ijimai.org/journal/sites/default/files/2021-02/ijimai_6_5_12_0.pdf
VL - 6
SN - 1989-1660
ER -