Abstract:Objective To construct and validate the prediction model for the third-generation cephalosporin-resis-tant (3GCR) Escherichia coli (E. coli) bloodstream infection. Methods 494 patients with E. coli bloodstream infection in a hospital from January 2014 to Deciber 2021 were selected as studied subjects, of which 434 cases were divided as the modeling group and 60 cases as the validation group. According to 3GCR resistance, subjects were classified into the 3GCR group and the third-generation cephalosporin sensitive (3GCS) group. Clinical data and laboratory results of the subjects were recorded, and the independent risk factors were screened by logistic regression analysis. Nomogram was constructed to validate the accuracy of the model. Results Logistic regression analysis showed that invasive operation (OR=19.482, 95%CI: 11.434-33.194), use of cephalosporin (OR=1.843, 95%CI: 1.070-3.173), high level procalcitonin (OR=1.272, 95%CI: 1.159-1.396) and high level C-reactive protein (CRP, OR=1.006, 95%CI: 1.002-1.011) were independent risk factors for 3GCR E. coli bloodstream infection. Nomograph model was constructed. The Hosmer-Lieshow method showed good fitting results of the model with the modeling group P=0.562 and the validation group P=0.742. The area under the receiver operating characteristic (ROC) curve of the modeling group and validation group were 0.883 (95%CI: 0.851-0.914) and 0.857 (95%CI: 0.807-0.907) respectively, indicating that the model has good discrimination. The decision curve analysis showed that the prediction model was of high value. Conclusion The predictive model based on invasive manipulation, use of cephalosporins, PCT and CRP levels has certain practical value in distinguishing 3GCR strain.