1511. Utility of Clinical Scoring Models in Predicting Community Acquired Urinary Tract Infections with Extended-Spectrum β-Lactamase-Producing Escherichia coli in a General Hospital in Mexico City

BACKGROUND: Urinary tract infections (UTIs) are among the most common causes for antibiotic prescription. The use of clinical scoring models in predicting infection with extended-spectrum β-lactamase (ESBL)-producing Escherichia coli (E. coli) may help select an adequate empiric treatment. METHODS: This retrospective case–control study included all urine cultures with E. coli from symptomatic patients 18 years of age or more admitted to Medica Sur Hospital from December 2014 to 2016. Cases were ESBL producing cultures and controls non-ESBL. Demographic and clinical information was drawn from electronic file. Sensitivities and specificities were performed at various cutoffs and area under the receiver curve (ROC AUC) was determined for each of the two models studied. RESULTS: A total of 171 cases and 294 controls were included. Table 1 displays the statistically significant variables associated with ESBL in a multivariate regression model. ROC AUC in Figure 1 was 0.691 for Tumbarello and 0.670 for Duke. With a 2-point cutoff, sensitivity for Tumbarello was 71% and specificity 61%, for Duke 58% and 75%, increasing cutoff to 4 points increases specificity to 87 and 93%, decreasing sensibility to 35 and 20%, respectively. Table 2 classifies by type of UTI, shows the percentage of adequate initial antibiotic for ESBL, and the number of cases predicted by each model. Tumbarello’s model predicts all cases, while Duke’s model predicts most cases of cystitis and pyelonephritis and all cases of complicated UTI and urosepsis. CONCLUSION: Clinical scoring models have a high specificity identifying best non-ESBL infections, this aids in the choice of a more adequate empirical antibiotic for community-acquired UTI. DISCLOSURES: All authors: No reported disclosures.