Of 817 enrolled patients with acute pyelonephritis, 793 (97.1%) submitted a urine culture. Of those 793 patients, 272 (34.3%) were excluded from analysis: 149 (18.8%) had no culture growth; 74 (9.3%) grew >1 contaminant; 17 (2.1%) grew >1 isolate at <10⁴ CFU/mL; 25 (3.2%) grew >1 organism at 10⁴ CFU/mL; and 7 (0.9%) had no fever. The study population consisted of 521 patients who grew 1 uropathogen at >10⁴ CFU/mL. The case finding audit revealed a 66% enrollment of eligible patients. Enrolled and nonenrolled patients were similar for most characteristics, including E. coli susceptibility rates to TMP/SMX, ceftriaxone, and fluoroquinolones (Technical Appendix).

Among the 521 study patients, median age was 37 (range 18–88, interquartile range 26–52) years; 455 (87.3%) were female (Table 1). Most (446 [85.6%]) patients had a community-acquired infection; 74 (14.2%) had a healthcare-associated infection (70 with hospitalization and 9 with nursing home residence in the previous 90 days). A total of 286 (54.9%) patients had uncomplicated pyelonephritis; 235 (45.1%) had complicated pyelonephritis.

Of types of uropathogens in patients with uncomplicated and complicated pyelonephritis (Table 2), E. coli accounted for infections in 453 (86.9%) patients; 272 (60.0%) of the infections were uncomplicated, and 181 (40.0%) were complicated. Among the 286 patients with uncomplicated infections, E. coli accounted for 95.1%; among the 235 patients with complicated infections, E. coli accounted for 77.0% (Table 2).

E. coli antimicrobial drug resistance rates among patients with complicated pyelonephritis tended to be higher than rates for patients with uncomplicated cases, except for TMP/SMX (Table 3). Among all patients, E. coli resistance rates varied by drug: ampicillin, 57.2% (259/453); TMP/SMX, 36.4% (165/453); gentamicin, 9.9% (43/436); cefazolin, 14.2% (52/367); ceftriaxone, 7.7% (35/453); levofloxacin, 10.2% (33/325); and ciprofloxacin, 12.1% (48/397). Among 53 fluoroquinolone-resistant isolates, 8 (15.1%) were susceptible to ampicillin, 34 (64.2%) to gentamicin, and 27 (50.9%) to ceftriaxone. Of 46 isolates tested for susceptibility to cefazolin, 19 (41.3%) were susceptible; all 48 (100%) isolates tested were susceptible to a carbapenem. E. coli antimicrobial drug resistance rates for patients with uncomplicated and complicated pyelonephritis varied by study site (Technical Appendix Tables 2, 3).

Figure 1

Figure 1. Prevalence of fluoroquinolone-resistant Escherichia coli infection among emergency department patients with uncomplicated (U) and complicated (C) pyelonephritis by study site, United States, July 2013–December 2014. Study sites are listed in the...

Among patients with uncomplicated pyelonephritis, 17 (6.3%) of 272 E. coli isolates were resistant to fluoroquinolone. The range of prevalence by site was 0.0%–23.1%; for 2 sites, prevalence was >10% (Figure 1; Technical Appendix Tables 2, 3). Among patients with complicated pyelonephritis, 36 (19.9%) of 181 E. coli isolates showed fluoroquinolone resistance. The range of prevalence by site was 0.0%–50.0%; 8 sites had a prevalence >10%, 4 of which had a prevalence >20%. We found fluoroquinolone resistance associated with complicated E. coli infection, prior use of antimicrobial drugs and fluoroquinolone, hospital admission, and prior UTI resulting from a fluoroquinolone- or ceftriaxone-resistant organism (Table 4).

Figure 2

Figure 2. Prevalence of extended-spectrum β-lactamase–producing Escherichia coli infection among patients with uncomplicated (U) and complicated (C) pyelonephritis, by study site, United States, July 2013–December 2014. Study sites are listed in the <>

Figure 3

Figure 3. Prevalence of fluoroquinolone-resistant and ESBL-producing Escherichia coli infections among patients with uncomplicated and complicated pyelonephritis, by study site, United States, July 2013–December 2014. Each dot indicates a study site; the line...

Among patients with uncomplicated pyelonephritis, ESBL production was found in 7 (2.6%) of 272 E. coli isolates; range by study site was 0.0%–8.3% (Figure 2; Technical Appendix Tables 2, 3). Among patients with complicated pyelonephritis, ESBL production was found in 22 (12.2%) of 181 E. coli isolates; range by site was 0.0%–17.2% (Figure 2; Technical Appendix Tables 2, 3). Frequencies of ESBL-producing E. coli isolates were higher among patients with antimicrobial drug resistance risk fac-tors than among those without these factors (Table 5). Nineteen (65.5%) of 29 patients with ESBL-producing E. coli infection had a recognized risk factor for antimicrobial drug resistance (Figure 2; Technical Appendix Tables 2, 3). Sixteen (55.2%) had antimicrobial drug exposure within the previous 60 days. During the previous 90 days, 6 (20.7%) had healthcare-setting exposure and 4 (13.8%) had travel outside North America. We found ESBL-producing E. coli infection associated with complicated infection, prior antimicrobial drug use, travel outside North America, and prior UTI resulting from a fluoroquinolone- or ceftriaxone-resistant organism. Among 37 isolates that grew other Enterobacteriaceae, including Klebsiella pneumoniae, 1 (2.7%) was ESBL producing. Among 29 ESBL-producing E. coli isolates, susceptibility rates to other antimicrobial drugs were 41.4% to TMP/SMX, 18.5% to ciprofloxacin, 21.7% to levofloxacin, 41.4% to gentamicin, and 100% to carbapenem. The prevalence of E. coli fluoroquinolone resistance correlated with the prevalence of ESBL-producing E. coli by site (Figure 3).

We further characterized 26 ESBL-producing E. coli isolates and 1 K. pneumoniae isolate. Among E. coli isolates, PCR identified multiple ESBL types: 22 (84.6%) produced CTX-M-15 (16 [61.5%] produced only CTX-M-15; 6 [23.1%] produced CTX-M-15 and TEM-1); 2 (7.7%) produced CTX-M-27; 1 (3.8%) produced CTX-M-14; and 1 (3.8%) produced CTX-M-14 and TEM-1. The K. pneumoniae isolate produced SHV-1 and CTX-M-15 ESBL types. Sixteen (61.5%) of the E. coli isolates were clonal type O25b-ST131.

Among patients with an E. coli infection, 223 (49.2%) were discharged from the ED and 229 (50.6%) were admitted to the hospital; of these, 13 (5.8%) and 18 (7.9%), respectively, were treated with an antimicrobial drug that lacked in vitro activity against their infection. Of 53 patients with a fluoroquinolone-resistant and 29 with an ESBL-producing infection, 24 (45.3%) and 22 (75.0%), respectively, were initially treated with in vitro–inactive antimicrobial drugs. Among 29 patients with an ESBL infection, 9 (31.0%) were discharged from the ED; an in vitro–inactive antimicrobial drug was initially prescribed to 7 (77.8%). Among 20 (69.0%) hospitalized patients with an ESBL infection, 15 (75.0%) were initially given an in vitro–inactive antimicrobial drug; 1 of those 15 patients was given gentamicin, 13 were given cephalosporin, and 1 was given both.

For 2013–2014, we found that prevalence of E. coli fluoroquinolone resistance was >10% for patients with uncomplicated pyelonephritis at 2 of 10 sites and >20% for patients with complicated infections at 4 of 10 sentinel sites surveyed in the United States. E. coli fluoroquinolone resistance was particularly prevalent in groups with antimicrobial drug resistance risk. These rates exceed thresholds for the 2010 IDSA treatment guidelines, which recommend consideration of an additional antimicrobial drug of a different class and other agents (9). Data from a similar study we conducted during 2000–2004 (15) indicate that, among all healthcare-seeking ED patients with acute pyelonephritis, the prevalence of fluoroquinolone-resistant E. coli increased from 3.9% during 2000–2004 to 11.7% during 2013–2014. During 2000–2004, we found no infections caused by ESBL-producing bacteria. As in other parts of the world, ESBL-producing Enterobacteriaceae are emerging among patients with community-acquired UTI in the United States. For patients with uncomplicated and complicated pyelonephritis caused by E. coli, we found that 2.6% and 12.2%, respectively, had infection caused by an ESBL-producing organism; rates were even higher for patients with risk factors. The globally disseminated, multidrug-resistant clone ST131, which produces CTX-M-15 β-lactamase, accounted for 85.2% of these infections. Of ESBL-infected patients, about one third lacked traditional antimicrobial resistance risk factors (i.e., antimicrobial drug or healthcare-setting exposure or international travel), suggesting that these isolates are now endemic in some US communities. Among ESBL-infected patients, about three quarters were initially treated with an antimicrobial drug lacking in vitro activity, including the sickest patients who required hospitalization. We did not collect outcome data, but lack of in vitro activity of the antimicrobial drug used for treatment has been associated with relatively poor response rates among patients with pyelonephritis (1,10,11).

Previous surveys have suggested that E. coli fluoroquinolone resistance rates are increasing in the United States. Among outpatients seeking care at 30 US centers during 2003–2004, 59 (6.8%) of 862 E. coli isolates were resistant to ciprofloxacin (2). Another analysis of >12 million urine specimens from US outpatient centers found that the E. coli fluoroquinolone resistance rate increased from 3.0% in 2000 to 17.1% in 2010 (3). Such laboratory-based resistance surveillance data may exaggerate the prevalence of resistance because patients for whom cultures are performed would be expected to have received prior therapy and to have had healthcare exposure more frequently than patients without cultures. Isolate-driven studies require retrospective review of records, which have missing and inaccurate data; also, uncertainty may exist regarding whether a specimen is from a patient with an actual clinical infection, rather than being a colonized or contaminated specimen. In routine practice, providers typically estimate local resistance rates on the basis of an antibiogram published by the local hospital laboratory. Resistance rates determined from antibiograms are prone to bias and indicate only whether the specimen was obtained from an outpatient or inpatient location.

In contrast, we conducted syndromic surveillance of patients who sought care at a geographically diverse network of US EDs. We studied acute pyelonephritis because it is a distinct clinical syndrome for which cultures are routinely obtained and because isolates grown are less likely to be contaminants or colonizers, compared with those from patients with suspected cystitis. Historical data were obtained by real-time patient interviews, enabling accurate classification of complicated and uncomplicated pyelonephritis and ascertainment of antimicrobial drug resistance risk factors. To identify biases, we compared characteristics of enrolled and nonenrolled qualifying patients and found their characteristics to be similar, suggesting the validity of our sampling. Consequently, our data and that of other studies indicate that, in some parts of the United States, the rate of E. coli fluoroquinolone resistance among uropathogens is >10% among patients with uncomplicated pyelonephritis and >20% among those with complicated infections. However, variability exists; among patients with uncomplicated pyelonephritis, 3 of 10 sites had fluoroquinolone resistance rates <5%. Consistent with findings of previous investigations, we found E. coli fluoroquinolone resistance associated with complicated infection, prior use of antimicrobial drugs and fluoroquinolone, hospitalization, and prior UTI caused by a fluoroquinolone- or ceftriaxone-resistant organism (16,17).

By using active, prospective surveillance, we found that ESBL-producing E. coli infections have now emerged to a considerable degree among patients with clinically confirmed community-acquired infections in parts of the United States, including among persons lacking commonly recognized antimicrobial drug resistance risk factors. This new observation is not unexpected, given the reported epidemiology of ESBL-producing Enterobacteriaceae infections in communities outside North America. CTX-M enzymes are currently the most prevalent ESBL types worldwide. The ST131 clone is largely responsible for the international epidemic caused by CTX-M-15–producing E. coli, including infections seen in the United States (7,18,19).

A few US laboratory-based surveillance studies have reported community-acquired ESBL infections. Peirano et al. (19) described ESBL-producing E. coli isolates from 30 community-dwelling patients at 5 Chicago-area hospitals during 2008. These ESBL-producing strains represented 2%–8% of E. coli isolates at each hospital. Khawcharoenporn et al. (20) reported that ≈5% of Enterobacteriaceae isolates from ED patients with presumed UTI during 2008–2009 were ESBL-producing, although specific risk data were not provided. Doi et al. (7) reviewed records of patients with cultures that grew ESBL-producing E. coli isolates at 1 hospital in each of 5 US cities during 2009–2010. Among 13,270 E. coli isolates, 523 (3.9%) were ESBL producing. Of the 291 patients infected or colonized with ESBL-producing E. coli as outpatients, infections of 107 (36.8%) were thought to be community associated. Community ESBL-producing E. coli isolates were resistant to multiple agents: 87.5% to ciprofloxacin or levofloxacin and 39.4% to gentamicin. All isolates from that study were susceptible to a carbapenem; we also found that all ESBL-producing isolates in our investigation were susceptible to a carbapenem.

Studies from outside North America have identified several characteristics associated with ESBL infection, such as recurrent and complicated UTI; advanced age; recent hospitalization; use of a β-lactam or fluoroquinolone; travel to Asia, Middle East, or Africa; and fresh water swimming (5,21–23). Banerjee et al. (23) conducted a case–control study among adults with E. coli clinical isolates cultured in the Chicago area and found that ESBL infection was associated with travel to India, ciprofloxacin use, and age. We found ESBL-producing E. coli infection associated with complicated infection, prior antimicrobial drug use, travel outside North America, and prior UTI resulting from fluoroquinolone- or ceftriaxone-resistant infection. The importance of investigating past susceptibility data when considering empirical treatment is highlighted by our observation that both fluoroquinolone resistance and ESBL-production were associated with previous resistant infections.

Our study has limitations. We were unsuccessful in enrolling all consecutive patients, which may have introduced bias in our selection of patients. However, our audit of eligible case-patients showed similarity of enrolled and nonenrolled patients, including their E. coli antimicrobial drug susceptibility rates; furthermore, most (>97%) enrolled patients had urine cultures collected, reducing potential bias. The prevalence of ESBL-producing strains may have been underestimated because we did not use ESBL-selective media and used only ceftriaxone instead of several advanced-generation cephalosporins to screen for presence of ESBL-producing strains. However, this method enabled us to have greater site participation; furthermore, screening isolates with a ceftriaxone MIC >1 µg/mL has been reported to have a sensitivity >98% on the basis of phenotypic testing (24). In addition, some patients may not have had pyelonephritis if a contaminated specimen was misinterpreted as noncontaminated, although >97% of patients had urine collected by a technique that minimizes contamination (i.e., clean catch, urethral catheterization, or suprapubic aspiration), and the diagnosis of pyelonephritis was further supported by clinical assessment. Our definition of confirmed infection as growth of 1 uropathogen at >10⁴ CFU/mL may have missed some cases of pyelonephritis, although only ≈5% grew a single uropathogen at <10⁴ CFU/mL or grew >1 uropathogen at >10⁴ CFU/mL. Our hospitals were large US urban centers and may not represent patients in other settings, emphasizing the importance of local surveillance.

IDSA treatment guidelines for acute uncomplicated pyelonephritis recommend that, if the fluoroquinolone-resistance rate is >10%, then in addition to a fluoroquinolone, an agent of another class (i.e., ceftriaxone or gentamicin) should be administered (9). Our findings indicate that fluoroquinolone resistance rates for E. coli are approaching or exceed this threshold for patients with uncomplicated pyelonephritis in many parts of the United States. For uncomplicated cystitis, the guidelines recommend alternative agents if the resistance rate is >20%, which is the current situation for fluoroquinolones in many settings for patients with complicated pyelonephritis. Unfortunately, we found that only one half to two thirds of fluoroquinolone-resistant E. coli isolates were susceptible to ceftriaxone or gentamicin. Rates of fluoroquinolone-resistant and ESBL-producing E. coli infections correlate to geographic location. Prior exposure to antimicrobial drugs or a healthcare setting, travel outside the United States, and a history of an antimicrobial drug–resistant infection substantially increases the chance that a person will have a current fluoroquinolone-resistant or ESBL-producing E. coli infection. Therefore, in settings with high fluoroquinolone resistance rates, in settings where ESBL-producing Enterobacteriaceae infections have emerged, or among persons with antimicrobial drug resistance risk factors (especially patients with or at risk for severe sepsis), healthcare providers should consider empirical treatment with a carbapenem or another agent found to be consistently active on the basis of the local antibiogram. In this study, ≈50% of patients with pyelonephritis were managed as outpatients. Currently, no oral antimicrobial drugs with consistent in vitro activity are available for empirical treatment of pyelonephritis caused by ESBL-producing E. coli uropathogens. Our findings, including the variability in the prevalence of resistance by site, show that increased local efforts to enhance surveillance for antimicrobial drug resistance are necessary to best inform treatment decisions. Furthermore, availability of new antimicrobial drugs must be expedited.