Comparing Patient Risk Factors, Sequence Type, and Resistance Loci Identification Approaches for Predicting Antibiotic Resistance in Escherichia coli Bloodstream Infections.

Rapid diagnostic tests of antibiotic resistance are increasingly being developed that identify the presence or absence of antibiotic resistance genes/loci. However, these approaches usually neglect other sources of predictive information, which could be identified in shorter time periods, including patient epidemiologic risk factors for antibiotic resistance and markers of lineage. Using a dataset of 414 Escherichia coli isolated from separate episodes of bacteremia at a single academic institution in Toronto, Canada between 2010-2015, we compared the potential predictive ability of three approaches (epidemiologic, pathogen sequence type, resistance gene identification) for classifying phenotypic antibiotic resistance to three antibiotics representing classes of broad spectrum antimicrobial therapy (ceftriaxone - 3rd generation cephalosporins, ciprofloxacin - fluoroquinolones, and gentamicin - aminoglycosides). We used logistic regression models to generate model receiver operating characteristic (ROC) curves. Predictive discrimination was measured using apparent and corrected (bootstrapped) area under the curves (AUCs). Epidemiologic models based on two simple risk factors (prior antibiotic exposure and recent prior Gram-negative susceptibility) provided modest predictive discrimination with AUCs ranging from 0.65 to 0.74. Sequence type demonstrated strong discrimination (AUCs 0.83-0.94) across all three antibiotic classes. The addition of epidemiologic risk factors to sequence type significantly improved prediction of resistance for all antibiotics (p<0.05). Resistance gene identification approaches provided the highest degree of discrimination (AUCs 0.88-0.99), with no statistically significant benefit of adding the patient epidemiologic predictors. In summary, sequence type, or other lineage-based approaches, could produce excellent discrimination of antibiotic resistance and may be improved by incorporating readily available patient epidemiologic predictors, but are less discriminating than the presence of known resistance loci.