The prevalence of AMR in outpatient Klebsiella spp. urine isolates was high and increased for most phenotypes throughout the study period; NTF NS prevalence did not increase but were as high as 54.0% for Klebsiella spp. and 57.3% for K. pneumoniae in 2019. Overall MDR-2 prevalence exceeded 10% for Klebsiella spp. and K. pneumoniae. Equivalent prevalence in 2019 was markedly lower for K. oxytoca, which demonstrated a NTF NS prevalence of 15.1% and MDR-2 prevalence of 6.5%. The prevalence of the ESBL+/NS phenotype in 2019, however, was higher for K. oxytoca isolates (6.8%) compared with K. pneumoniae isolates (5.3%).

AMR prevalence for K. pneumoniae differed significantly between US census divisions. Some divisions had relatively high AMR prevalence for multiple phenotypes, including the South Atlantic, West South Central, and East South Central divisions. This finding supports previous observations of high AMR prevalence in US southern-border regions, which may be influenced by access to antimicrobials, and their overuse, without a prescription in Mexican pharmacies along the US-Mexico border [21]. Studies have also demonstrated higher outpatient antibiotic prescribing, including inappropriate prescribing, in Southern US states versus other US regions [1, 22], which might reflect regional differences in antibiotic stewardship [1, 22].

AMR prevalence for K. pneumoniae also differed between age groups. The prevalence of AMR was higher among females aged ≥ 55 years for most of the AMR phenotypes; however, prevalence of the NTF NS phenotype decreased as age increased. It is noteworthy that the 2015 American Geriatrics Society recommendations were not supportive of NTF use in elderly patients with renal insufficiency and low creatinine clearance [23].

The rapid spread of pathogens with the ESBL+ phenotype is of global concern, as effective empiric oral therapeutic options are limited and the burden of AMR in acute care settings is increasing. In our study, the model-estimated overall prevalence of ESBL+/NS K. pneumoniae isolates over the 2011–2019 study period was 4.6% (across all census divisions), with a relative annual increase in prevalence of 5.4% over this same period. These data are comparable to previous studies of ESBL+ prevalence in the US [9, 24,25,26]. The estimated relative annual increase in MDR-3 prevalence (+ 5.6%) that we observed was marginally above that for ESBL+/NS (+ 5.4%) and indeed the two may be correlated. NTF NS was the only AMR phenotype for which the estimated prevalence did not increase between 2011 and 2019; this may be partly due to the aforementioned American Geriatrics Society recommendations, which were published in 2015 [23]. However, despite the decrease in NTF NS prevalence during the study period, the overall prevalence of NTF NS was high (56.6%; NTF NS prevalence was > 50% for each year, and > 50% across all census divisions in 2019), highlighting a reduced likelihood of clinical efficacy for NTF when used empirically to treat community-acquired UTIs presenting to the emergency department. This is important to consider when updating clinical practice guidelines for the treatment of uUTI, such as those published by the Infectious Diseases Society of America (IDSA) in 2011, which recommend NTF as a first-line treatment for uUTI [27].

The prevalence of AMR phenotypes among community K. pneumoniae urine isolates in the US has not previously been well characterized, although such studies have been conducted in other countries, including Spain and China [14, 28, 29]. US surveillance of samples from inpatients (clinical setting not specified; taken between 1998–2010) showed that K. pneumoniae AMR prevalence has increased over time, reaching 19.3% for SXT and 16.8% for ciprofloxacin in 2010 [30]. Our study expands on these findings for the years 2011 and beyond, with a focus on outpatients in the US. Over the 2011–2019 period evaluated herein, the estimated overall prevalence of the SXT NS and FQ NS phenotypes among K. pneumoniae isolates was 9.6% and 4.4%, respectively, with a relative annual increase in prevalence of 2.1% for both phenotypes. Due to potentially serious side effects and the availability of other effective antibiotics as preferred first-line agents (e.g., NTF, SXT), guidelines recommend that FQs are reserved for patients who have no other options for the treatment of uUTI [27].

This study describes a large sample of K. pneumoniae isolates from US outpatients over nine years and provides valuable insights into prevalence of AMR among urine isolates in this setting. Electronic outpatient microbiology results were available for all eligible patients, which provided a comprehensive data source for the evaluation of AMR among Klebsiella spp. uropathogens.

Limitations of the study include potential variability in susceptibility testing due to reliance on local laboratory practices. The lack of information on specific laboratory practices within census divisions also precludes more localized resistance data. The study period included various changes in the minimum inhibitory concentration breakpoints and interpretive criteria from the CLSI for cefazolin, cefepime, levofloxacin, and ciprofloxacin; when these changes were implemented by individual laboratories involved in the study is unknown, which may have led to inconsistencies between centers over time [20]. The patient isolates used in the study could not definitely be linked to a uUTI diagnosis (via International Classification of Diseases-9/10 diagnosis codes), clinical symptoms, or pharmacy claims for antimicrobial prescribing. The potential inclusion in our study of outpatients with recurrent uUTI (due to the inclusion of more than one sample from the same patient with an interval exceeding 30 days) could have potentially overestimated the prevalence of AMR. Similarly, selection bias may have led to the inclusion of more resistant isolates due to the inclusion of isolates for patients with uUTI from whom a specimen was collected and analyzed (as opposed to any patient with uUTI). Although the data provide coverage of the US, the distribution of participating centers at a regional level may under-represent certain areas and specific patient subgroups, and results may not be generalizable to the entire female US population. The study was designed to include isolates from outpatients without a hospitalization within 24 h of urine culture; however, the possibility of hospitalization to a hospital not affiliated with the BD Insights Research Database could not be ruled out. Finally, the prevalence of AMR to fosfomycin was not assessed due to its limited use in the US, the requirement of a laborious reference standard method (agar dilution) that is not commonly used in diagnostic microbiology [31], and its omission from commercial susceptibility panels in the US.