Methods: All hospitalized adult patients with carbapenem order requests while receiving at least 48 hours of broad-spectrum antibiotics between 2015 and 2016 were retrospectively reviewed. Cases were excluded if a carbapenem was requested for: 1) a documented infection during the same hospital stay, 2) penicillin allergy, or 3) deterring drug-drug interactions or adverse effects. Prevalence was estimated using the total number of COS isolates of the total population during the study period. Chi-square tests and Mann-Whitney U tests were performed. External validation of the ESBL predictor tools was performed using the Vanderbilt tool.
Results: A total of 1260 patients were identified. Within a sample population, 82 patients met inclusion criteria. A total of 14.6% of patients had COS organisms (7.3% ESBL-producing, 3.7% ampC-producing, 3.7% both), 29.2% were other organisms, and 56.1% were culture-negative. Of the microbiologically evaluable population, the most common organism was E. coli (44%). More than 80% of patients were considered high-risk in both the Italian and Thailand models and rates of high risk were similar between groups across all tools. There was no difference between groups in regards to prior hospitalizations, prior antibiotic exposure, recent urinary catheter use, or immunosuppression use. Overall sensitivity was 100%, 67%, and 92% while specificity was 9%, 44%, and 6% for the Italian, Duke, and Thailand models, respectively. Positive predictive values were 16%, 17%, and 14% while negative predictive values were 100%, 89%, and 80%, respectively.
Conclusion: Identification of COS infections is a serious challenge to clinicians without microbiological data. This study found numerous limitations to the applicability of these published predictors tools in this population.
A. Hanson, None
P. Gurnani, None
K. Ackerbauer, None
L. Geswein, None
S. Harpe, None
S. Won, None
C. Varughese, None
See more of: Poster Abstract Session