P1512: resistance diagnostics tool to combat antibiotic resistance in hematology wards: a model-based evaluation from a large multicenter case–control and cohort study

Topic: 30. Infections in hematology (incl. supportive care/therapy) Background: Pseudomonas aeruginosa is a critical priority pathogen on the antibiotic-resistant pathogen list and is part of the ESKAPE pathogen group. It is intrinsically resistant to many antibiotics and easily develops resistance. The overuse of antibiotics is a significant factor associated with the rapid emergence of antibiotic resistance. Aims: To investigate this, we conducted a retrospective study to develop a more quantitatively rigorous approach, fitting a linear or quasibinomial regression to cross-sectional data about antibiotic use and resistance in three large tertiary teaching hospitals in China. Methods: For the study, we used an 8-year retrospective cohort as a training cohort from Jan 2014 to Dec 2020. The Test1 cohort consisted of the population from Jan 2021 to Dec 2022 in our center. The Test2 cohort included the populations during Jan 2011 to Dec 2022 in Tongji Medical College Hospital and The Third Xiangya Hospital of Central South University in China. We used the C-index to assess the discrimination of the model with censored CRE infection status, and calibration plots of observed versus expected absolute risks with the Brier score to assess the accuracy of predictions. Results: Our results showed that we developed a simple prediction rule based on a few clinical factors, such as prior treatment including carbapenems, piperacillin-tazobactam, and quinolones (time-to-infection within 3 months) and BSI occurring during antibiotic treatment (time-to-infection within 0 days). Each predictor was assigned one point, except for BSI occurring during antibiotic treatment (4 points). The model had an AUROC for MDR infection of 0.755 (95% CI, 0.639-0.871) in Test1 and 0.779 (95% CI, 0.685-0.874) in Test2. The sensitivity, specificity, and accuracy values were 0.692, 0.845, and 0.824 for Test1, and 0.654, 0.832, and 0.808 for Test2. The calibration curves showed good agreement for observed rates of MDR organisms with calibration slope 0.67 and 0.77, respectively. In total, 169 cases were receiving antibiotic treatment at the time of infection in the three centers. Different types of antibiotics had varying effects on the resistance of subsequent infection strains. Meropenem or antibiotic regimens containing meropenem caused the highest proportion of subsequent carbapenem-resistant infection strains, ranging from 43% to 67%, and MDR proportion reached 36.7%. Quinolones or cephalosporin-based antibiotic treatment caused a very high proportion of carbapenem-resistant and MDR strains when used alone, while a low proportion when used in combination (66.7% vs 14%, p=0.01). Summary/Conclusion: In conclusion, we constructed a prediction model on antibiotic resistance integrating antibiotic exposure factors including non-time-interval and with-time-interval, which performed well in the data from multiple hospitals. Among antibiotics used within time-to-infection within 0 days, it appears that combination treatment more attenuated the infection proportion of drug-resistant strains, which should be considered by clinicians.Keywords: Infection