Epidemiology and prediction of multidrug-resistant bacteria based on hospital level.

Multidrug-resistant bacteria (MDRB) result in nosocomial infections and a substantial disease burden for hospitalised patients worldwide. However, strategies to control drug resistance at the hospital level are lacking. In this study, we aimed to find important indicators for risk assessment and predicting MDRB infections in the hospital. Using real-world data and machine learning models, we conducted a retrospective study from 2010 to 2020 in a teaching hospital to analyse the trends and characteristics of MDRB infections. Combining 39 hospital indicators, we used a random forest model and cross-correlation analysis to explore the important factors affecting MDRB and their predictive power. We built a decision tree model to predict the number of hospitalised patients with MDRB infection. The number of hospitalised rescues and rate of rational perioperative antibacterial drug use in type I and II incision operations were correlated with the number of patients with MDRB infection after 1-2 months. The number of hospitalised operations and rate of antibiotics use in emergency patients had an effect on current MDRB-susceptible patients. The indicators, including hospital operation volume and antibacterial drug use, had a positive or negative quantitative relationship with the number of patients with MDRB infection, and their thresholds could be fit to the MDRB prediction model. Surgical, emergency, and hospitalised rescue patients showed the highest risk of MDRB infection. Standardised indicators such as clinical pathway rate and rational antibiotic use rate could be used to control the development and spread of MDRB infections in the hospital. Copyright © 2022. Published by Elsevier Ltd.

Citation

Ying Chen, Xingchi Chen, Zheng Liang, Shuhao Fan, Xiaoli Gao, Hansi Jia, Bin Li, Liang Shi, Aixia Zhai, Chao Wu. Epidemiology and prediction of multidrug-resistant bacteria based on hospital level. Journal of global antimicrobial resistance. 2022 Jun;29:155-162

Mesh Tags

Substances

PMID: 35283333

search → result