Establishment, evaluation and application of GM(1,1) models for predict-ing prevalence of common carbapenem-resistant bacteria in China

  Objective  To establish, evaluate and preliminarily apply grey model (1,1) GM(1,1) models for predicting the prevalence of common carbapenem-resistant bacteria in China.

  Methods  We extracted the national data of 2015 – 2021 on detection rate of carbapenem-resistant Escherichia coli (CR-E.coli), carbapenem-resistant Klebsiella pneumoniae (CRKP), carbapenem-resistant Pseudomonas aeruginosa (CRPA), and carbapenem-resistant Acinetobacter baumannii (CRAB) from the China Bacterial Drug Resistance Surveillance Network. Four GM(1,1) models were constructed using the data of 2015 – 2019. Posterior difference ratio C, small error probability P, and relative error and residual error were used to assess the accuracy of the established models in predicting prevalence of common carbapenem-resistant bacteria with the data of 2020 – 2021. The prevalence of CR-E.coli, CRKP, CRPA and CRAB for years from 2022 to 2025 in China were predicted using the four GM(1,1) models established.

  Results  For the four established GM(1,1) models, the values of posterior difference ratio C were less than 0.50, the small probability error P were close to 1.0, and the relative errors and residual errors were less than 10%, indicating good efficacies in predicting the prevalence of CR-E.coli, CRKP, CRPA and CRAB. Based on the four established GM(1,1) models, the predicted yearly prevalence of CR-E.coli, CRKP, CRPA, and CRAB in China are 1.668%, 12.208%, 16.663%, and 52.507% for 2022; 1.698%, 12.886%, 15.930%, and 1.549% for 2023; 1.729%, 13.601%, 15.229%, and 50.608% for 2024; and 1.761%, 14.355%, 14.559%, and 49.685% for 2025, respectively, with a upward trend for the prevalence of CR-E.coli and CRKP but a downward trend for the prevalence of CRPA and CRAB.

  Conclusion  In this study, four GM(1,1) models were established for predicting the prevalence of common carbapenem-resistant bacteria and the predictions based on the models could be used in developing strategies on the prevention and control the prevalence of CR-E.coli, CRKP, CRPA and CRAB in China.