Modeling the risk of fluoroquinolone resistance in non-severe community-onset pyelonephritis.

Abstract

Estimating whether the individual probability of being infected by a fluoroquinolone resistant isolate is higher than 10% may help to choose the empirical treatment of pyelonephritis. We aimed to model the risk of fluoroquinolone resistance in women with community-onset pyelonephritis. Women with non-severe community-onset pyelonephritis were prospectively recruited in 4 French emergency departments of 2 districts. Adjusted odds ratios (aOR) and 95% confidence intervals were estimated through multivariate logistic regression. Among 190 patients, 19 (10%) were infected by a fluoroquinolone resistant isolate. Fluoroquinolone resistance was more frequent in district #2 (17%) than in district #1 (3%). Independent risk factors for fluoroquinolone resistance were district (adjusted OR, 7.0 (2.2-31.9)), and in the 6 previous months, urinary tract infection (UTI, aOR, 3.9 (1.3-11.5)) and vesical catheterization (aOR, 4.7 (0.5-33.3)). A specific model was derived to identify district #2 patients with a low (10% or lower) probability of being infected by a fluoroquinolone resistant isolate. Independent risk factors were residency in long-term care facility (aOR, 3.3 (0.7-13.5)), and in the 6 previous months, UTI (adjusted OR, 3.1 (0.9-10.7)) and home nursing care (aOR, 3.4 (0.6-17.0)). For 63 (67%) patients, the predicted probability of fluoroquinolone resistance was 0.10; among these patients, 6 (10%) actually had a fluoroquinolone resistant isolate. Locally derived predictive models may be used to identify patients with a low probability of fluoroquinolone resistance and guide the empirical antimicrobial therapy of non-severe community-onset pyelonephritis.