Antibiotic Resistance of Urinary Tract Infection Recurrences in a Large Integrated US Healthcare System

Abstract

Urinary tract infections (UTIs) are among the most common bacterial infections in the United States (US), primarily affecting women [1]. In the US, half of women experience ≥1 UTI by the age of 35 years, and 1 in 3 women experiences ≥1 UTI requiring antibiotic treatment in her lifetime, with 11% experiencing ≥1 UTI per year [2, 3]. UTI incidence in the US increased by 52% between 1998 and 2011, resulting in 400 000 hospitalizations with an estimated cost of $2.8 billion in 2011 [4]. UTIs have also been increasing in outpatient settings [5] and cause substantial negative impact on activity and health-related quality of life [6].

Uropathogenic Escherichia coli is the most common pathogen causing UTIs. Escherichia coli accounts for up to 80% of uncomplicated UTIs (uUTIs), which typically affect otherwise healthy individuals without underlying structural abnormalities or instrumentation [7–9], and approximately a quarter of complicated UTIs (cUTIs) [10], which occur in the presence of structural abnormalities or instrumentation compromising the urinary tract [11–13]. Other gram-negative pathogens such as Klebsiella pneumoniae, Proteus mirabilis, and Pseudomonas aeruginosa and gram-positive pathogens such as Staphylococcus spp, Enterococcus faecalis, and group B Streptococcus can also cause UTIs [7, 9]. Following an initial uUTI, subsequent UTIs are common and may be caused by the same or different pathogens.

Antibiotic treatment for UTI is often empiric without urine culture or susceptibility testing [14, 15], and is usually based on national guidelines and local resistance profiles [15]. The Infectious Diseases Society of America recommends trimethoprim-sulfamethoxazole (TMP-SMX), nitrofurantoin, or fosfomycin as first-line therapy [14]. UTIs also have a high recurrence rate, leading to increased use of antibiotics and antibiotic resistance [16, 17]. Antibiotic resistance in uropathogens has been increasing [18], resulting in increased risk of treatment failure, patient morbidity, healthcare costs, and use of broad-spectrum antibiotics, further exacerbating antibiotic resistance [7, 18, 19].

Given the continued burden of UTIs, current data on repeated occurrence of UTI are needed. Also, despite global increases in multidrug resistance (MDR) [20–22], recent data on antibiotic resistance of uropathogens are lacking, especially antibiotic resistance patterns for those patients experiencing UTI recurrences. Leveraging electronic health records (EHR) from a cohort of individuals with uUTI at Kaiser Permanente Southern California (KPSC), we described the antibiotic resistance of subsequent uUTIs and cUTIs (from 2016–2021) following an initial uUTI (from 2016–2020).

METHODS

Study Setting

This study was conducted at KPSC, a large, integrated healthcare organization serving >4.8 million residents of Southern California, closely mirroring the sociodemographically diverse Southern California population [23]. KPSC's EHR captures all aspects of members' health information, including demographic/socioeconomic characteristics, diagnoses, test results, and medications from all care settings. Standard laboratory tests from specimens collected throughout the region are processed in the 2 centralized regional reference laboratories. Documentation from care received outside KPSC is submitted for reimbursement and integrated into the EHR. This study was approved by the KPSC institutional review board, with a waiver for informed consent.

Study Population and Data Collection

This study leveraged data from a retrospective cohort study of adults with index uUTI during 2016–2020 who were followed through 2021 for subsequent UTI events [5, 24]. To assemble the prior cohort, we first identified all UTI events from EHR by (1) UTI diagnosis code with an antibiotic prescription within ±3 days, or (2) positive urine culture with an antibiotic prescription within ±3 days, or (3) positive urine culture with a diagnosis code within ±7 days (Supplementary Table 1); positive cultures without either a diagnosis code or an antibiotic prescription were excluded as potential asymptomatic bacteriuria. UTIs occurring within 30 days of each other were counted as a single event [5, 24]. Among these, UTIs were then defined as cUTI using a definition adapted from Carreno et al, which defines cUTI based on select International Classification of Diseases, Tenth Revision urologic infection codes, structural abnormalities/obstruction, and procedures (Supplementary Table 1) [25]. All UTIs that were not defined as cUTI were defined as uUTI. The first uUTI during 2016–2020 was defined as the index uUTI. We excluded individuals with <1 year of continuous KPSC membership and those with any UTI in the year prior to index date (date of the index uUTI).

The study population herein consists of a subcohort of individuals from the prior index uUTI cohort, who had a culture-positive index uUTI with an antibiotic prescription and/or a diagnosis code (Supplementary Figure 1). Among this study cohort, we examined index uUTIs (UTI 1), and all subsequent culture-confirmed UTIs (uUTI and cUTI) through 2021. Antibiotic susceptibility data on aminoglycosides, carbapenems, cephalosporins, fluoroquinolones, fosfomycin, nitrofurantoin, penicillins, TMP-SMX, and others were obtained by routine testing of positive urine cultures at KPSC laboratories using VITEK 2 antimicrobial susceptibility cards; this testing is conducted as part of standard of care in the 2 centralized KPSC regional reference laboratories, and results are integrated into the EHR.

Analysis

We described the demographic and clinical characteristics of those with an index uUTI by the number of subsequent UTIs, including sex, age, race/ethnicity, healthcare utilization, and Charlson comorbidity score. We examined the number of index uUTIs and subsequent UTIs, as well as the median time between each UTI, by pathogen identified at index uUTI. We also described the proportion of initial and subsequent UTIs with antibiotic nonsusceptibility (intermediate and resistant) by antibiotic class. We used the Cochran-Armitage trend test to assess for trend across index UTI and subsequent UTIs for each pathogen, as well as each nonsusceptible antibiotic class. Additionally, we reported prior antibiotic exposure (antibiotics dispensed for any reason in the 12 months prior to index date). Finally, we reported the proportion of UTI isolates with antibiotic nonsusceptibility by the number of nonsusceptible antibiotic classes, and for each UTI (for UTI 1, 2, 3, etc) during the study period, including MDR (intermediate or resistant to ≥3 antibiotic classes) patterns. All analyses were conducted using SAS version 9.4 and Enterprise Guide version 8.2 software (SAS Institute, Cary, North Carolina).

RESULTS

We identified 777 817 adults with ≥1 UTI event during 2015–2021, with ≥1-year KPSC membership prior to index date (Supplementary Figure 1), of whom 475 013 had ≥1 uUTI during 2016–2020. After applying exclusion criteria, the final cohort included 148 994 individuals who had a culture-confirmed index uUTI between 1 January 2016 and 31 December 2020.

Characteristics of the UTI Cohort

The overall cohort was 88% female and 44% Hispanic, with a mean age of 50 years (standard deviation, 19 years) (Table 1). The cohort included 111 784 individuals who experienced 1 uUTI only, 31 754 with 2–3 UTIs, 4203 with 4–5 UTIs, and 1253 with ≥6 UTIs during the study period. The proportion that was female and the mean age were higher for each subsequent UTI. Those with 1 uUTI only were most commonly Hispanic (45%) followed by White (34%), and those with ≥6 UTIs were most commonly White (48%) and Hispanic (39%), reflecting the race/ethnicity distribution of the older KPSC members. Individuals with 1 uUTI only were most commonly covered by commercial insurance (66%), then Medicare (23%), while those with ≥6 UTIs were most commonly covered by Medicare (57%), then commercial insurance (36%). Charlson scores, number of outpatient/emergency department/inpatient visits, and the number of antibiotics dispensed in the year prior to index UTI were higher for individuals with more subsequent UTIs experienced during the study period.

Characteristics of UTIs

Overall, the proportion of individuals with urine culture obtained and those testing positive at each UTI remained largely unchanged from index uUTI through UTI 6 (Supplementary Table 2). Of the 148 994 individuals with an index uUTI, 66 607 (45%) had a subsequent UTI (diagnosis code with antibiotic prescription, or positive culture with diagnosis code or antibiotic prescription), of which 28 007 (42% of 66 607 and 19% of 148 994) were culture-confirmed with susceptibility results. Among the 28 007 UTI 2 meeting inclusion criteria for our study, the median time between index uUTI and UTI 2 was 300 days (interquartile range [IQR], 126–627 days), and 7250 (26%) developed an additional UTI (UTI 3) after a median 209 days (IQR, 93–428 days) (Supplementary Table 3). The median number of days between UTI 3 and 4, UTI 4 and 5, and UTI 5 and 6 was 160 (IQR, 77.5–341), 140 (IQR, 74–288), and 126 (IQR, 65–247), respectively. Duration between each culture-confirmed UTI was similar when stratified by pathogen isolated at index uUTI. cUTIs accounted for 11% of all subsequent UTIs.

Overall, E coli was the most common pathogen identified for index and all subsequent UTIs (Table 2). At index uUTI, E coli was identified in 117 955 (79%) individuals, followed by Klebsiella (10 192 [7%]). From the index uUTI to UTI 6, the proportion of E coli UTI decreased from 79% to 73% (P for trend <.001) while the proportion of Klebsiella spp UTI increased from 7% to 11% (P for trend <.001).

Prior Antibiotic Exposure

In the year prior to the index uUTI, the most common antibiotic class dispensed for any reason was penicillins (16%), followed by first-generation cephalosporins (13%) and fluoroquinolones (5%) (Figure 1, Supplementary Table 4). In the year prior to each subsequent UTI, first-generation cephalosporins and fluoroquinolones were the most commonly dispensed, followed by penicillins and nitrofurantoin.

Figure 1.

Proportion of individuals with culture-confirmed urinary tract infection (UTI) with antibiotics dispensed in 12 months prior to each UTI (all pathogens), Kaiser Permanente Southern California, 2015–2020.

Open in new tabDownload slide

Antibiotic Nonsusceptibility

Among the index uUTIs, 84 929 (57%) were nonsusceptible to ≥1 antibiotic class, and 18 595 (13%) to ≥3 antibiotic classes (Table 3; shown for E coli UTI only in Supplementary Table 5). For index uUTIs, nonsusceptibility to penicillins (48%) and TMP-SMX (20%) was most common, followed by nitrofurantoin (15%) and first-generation cephalosporins (9%).

The proportion of isolates with antibiotic nonsusceptibility increased for subsequent UTIs (P for trend <.001); at UTI 6, 65% were nonsusceptible to ≥1 antibiotic class and 20% were nonsusceptible to ≥3 antibiotic classes (Figure 2, Supplementary Table 6). The proportion of isolates nonsusceptible to penicillins, TMP-SMX, nitrofurantoin, fluroquinolones, cephalosporins, aminoglycosides, and carbapenems increased for subsequent UTIs (P for trend <.001). Among all isolates nonsusceptible to ≥1 antibiotic class, 8% were “intermediate” and 92% were “resistant” (Supplementary Table 7).

Figure 2.

Proportion of nonsusceptible isolates, by the number of nonsusceptible antibiotic classes by urinary tract infection (UTI) event (all pathogens combined), 2016–2020.

Open in new tabDownload slide

Among antibiotics commonly used for UTIs (penicillins, nitrofurantoin, TMP-SMX, fluoroquinolones, aminoglycosides, and first-generation cephalosporins), nonsusceptibility to 1 antibiotic class only was observed in 24% of index uUTIs (Figure 3). Nonsusceptibility to 2 antibiotic classes was observed in 22%. MDR was observed in 13% of index uUTIs (nonsusceptibility to 3 antibiotic classes [9%], 4 antibiotic classes [3%], 5 antibiotic classes [1%], ≥6 antibiotic classes [<1.0%]); 42% of all isolates from index uUTIs were susceptible to all included antibiotic classes. The most common nonsusceptibility patterns included nonsusceptibility to penicillins and TMP-SMZ (9%), and to penicillins and nitrofurantoin (6%). MDR patterns including penicillins was common (12%), and MDR patterns involving penicillins and TMP-SMX plus ≥1 antibiotic class accounted for 9% of all observed resistance patterns.

Figure 3.

Antibiotic nonacceptability patterns of culture-confirmed urinary tract infections (UTIs), 2016–2020. Rows with index UTI (UTI 1) count <50 not shown. Only up to sixth UTI (UTI 6) shown due to low counts for subsequent UTIs. Abbreviations: AGS, aminoglycoside; CSP, cephalosporin; FLQ, fluoroquinolone; NFT, nitrofurantoin; PNC, penicillin; T-S, trimethoprim and sulfamethoxazole; UTI, urinary tract infection.

Open in new tabDownload slide

For subsequent UTIs, nonsusceptibility patterns were similar, with the most common nonsusceptibility to 2 antibiotic classes including penicillins and TMP-SMZ (7%–9%) and penicillins and nitrofurantoin (7%–9%). However, the prevalence of nonsusceptibility gradually increased for subsequent UTIs. For UTI 6, nonsusceptibility to 1 antibiotic class only was observed in 17%, and nonsusceptibility to 2 antibiotic classes in 19%. Nonsusceptibility to 3, 4, 5, and 6 antibiotic classes was observed in 15%, 7%, 2%, and 1% of UTI 6 isolates, respectively (not counting patterns with isolate counts <5); 30% were susceptible to all antibiotic classes.

DISCUSSION

In this study, among a large cohort of adults with culture-confirmed index uUTI during 2016–2020, we characterized uropathogens and patterns of antibiotic nonsusceptibility for index and subsequent culture-confirmed UTIs. Escherichia coli was the most common uropathogen overall, with UTIs due to other pathogens including Klebsiella and Proteus mirabilis increasingly common at subsequent events. About 1 in 5 individuals with an index uUTI developed a subsequent UTI within 10 months, although the time between the index and subsequent UTIs varied widely. Among those with ≥1 subsequent UTI, 1 in 4 developed an additional UTI after about 7 months of the second UTI. Penicillins and TMP-SMX were most commonly responsible for both single and multiple class antibiotic nonsusceptibility. For subsequent UTIs, nonsusceptibility patterns were similar, but the proportion of nonsusceptible isolates gradually increased.

Escherichia coli has been frequently reported as the dominant pathogen for both uUTIs and cUTIs, with Enterococcus spp and Candida spp substantially more common in cUTIs [16]. Studies have reported that after an initial UTI, 27% of women experienced recurrence within 6 months, and 3% had a second recurrence within the same period of time [26, 27]. In a study in the primary care setting, 53% of women ≥55 years of age and 36% of younger women reported a recurrence within 1 year of an initial UTI [28]. More recent data from a multicenter Dutch nursing home study reported recurrence by the same E coli strain after several months up to over 400 days after the initial UTI [29]. We observed that among individuals with a culture-confirmed index uUTI, about 34% had ≥2 subsequent UTIs during the study period. After the index UTI, about 20% experienced a second UTI after about 10 months, with additional UTIs occurring in 4- to 7-month intervals in about 25% of those who experienced a second UTI. Long time intervals between UTIs may be because we only included culture-confirmed UTIs, and not UTIs identified only by diagnosis codes with antibiotics. The long intervals may also be partially explained by some strains of uropathogens colonizing the bladder for longer periods or recolonization by the same or a different strain [29].

Since the 1940s, various antibiotics have been recommended for treatment of UTIs [30]. However, growing antibiotic resistance has contributed to recurrence and chronicity of UTIs, making UTI treatment increasingly complicated. A substantial increase in UTIs resistant to penicillins and TMP-SMX has been reported [31, 32]. Co-resistance to ampicillin and TMP-SMX, and MDR involving ampicillin, TMP-SMX, and ciprofloxacin have also been commonly reported in the US [33–35]. US surveillance data from 2000 reported MDR in 7% of E coli, including resistance to ampicillin (98%), TMP-SMX (93%), cephalosporin (87%), ciprofloxacin (39%), and nitrofurantoin (8%) among the resistant isolates [33]. More recently, a study from Italy of >30 000 E coli urine isolates found that resistance to commonly prescribed antibiotics increased significantly between 2000 and 2019 with isolates collected from 2015 to 2019 having 30% resistance to ciprofloxacin, 24% to TMP-SMX, and >10% to third-generation cephalosporins with 17% MDR and 12% extended-spectrum β-lactamase (ESBL) producers [36]. Another study evaluating MDR among expanded-spectrum cephalosporin-resistant Enterobacterales (ESCR-E) found that 91% of ESCR-E uropathogenic E coli (UPEC) causing UTI had an ESBL phenotype. Furthermore, ESCR-E UPEC with ESBL was 1.7 times more likely to be MDR than non-ESBL isolates [37].

Similar to prior findings, in our data, the most common nonsusceptibility patterns included nonsusceptibility to penicillins and TMP-SMZ (9%), nonsusceptibility to penicillins and nitrofurantoin (6%), and MDR involving penicillins (9%). Although nitrofurantoin resistance remained low in most settings since 2000 (∼2%) [24, 38], we observed a high prevalence of nonsusceptibility to nitrofurantoin (>14%). Increased use of nitrofurantoin among the recommended first-line antibiotics for UTI may help explain the high nonsusceptibility. The change in susceptibility may also be explained by change in causative bacteria (eg, decreasing proportion with E coli and increasing proportion with Klebsiella). Also, our definition of “nonsusceptibility” included “intermediate.” Of all isolates nonsusceptible to ≥1 antibiotic class, about 9% were “intermediate.” Among isolates nonsusceptible to nitrofurantoin, about half were “intermediate.” As such, varying definitions of antibiotic resistance used in previous studies (eg, definition of “nonsusceptibility” with or without including “intermediate”) may partially explain the difference.

Less is known about antibiotic resistance of recurrent UTI (rUTI). Studies have shown that most recurrences have been associated with specific antibiotic resistance traits and sequence types [39]. A study of rUTIs in nursing home residents showed that the recurrences were caused by both the same and different strains of E coli or other uropathogens [29]. In our data on UTI recurrences, we observed that the prevalence of both single antibiotic nonsusceptibility and MDR gradually increased with subsequent UTIs. Nonsusceptibility to any antibiotic class increased from 57% at index uUTI to 65% at UTI 6. More importantly, MDR increased from 13% at index uUTI to 20% at UTI 6. Nonsusceptibility to each antibiotic class increased for most tested antibiotic classes including penicillins, TMP-SMX, nitrofurantoin, fluroquinolones, and cephalosporins. Such increases are likely due to increased exposure to antibiotics used to treat previous UTIs. Even short courses of antibiotics impact the microbial flora of the perineal and gastrointestinal tract, the major sources of uropathogens, often selecting for resistant pathogens and increasing the risk of subsequent UTIs due to resistant organisms [7].

Our study has several limitations. First, we did not have sequence data for different strains of uropathogens. Therefore, for cases in which subsequent UTIs were caused by the same pathogen, we could not determine whether the subsequent UTIs were due to relapse of the same strain or a new infection by a different strain of the same uropathogen. Second, while we observed an increase in nonsusceptibility with subsequent UTIs, some may have been cases of asymptomatic bacteriuria; however, we only included positive cultures with a diagnosis code and/or an antibiotic prescription that were considered UTIs by KPSC clinicians. Third, we evaluated nonsusceptibility patterns at the level of antibiotic class, grouping some antibiotic classes with a range of antimicrobial coverage into 1 class (ie, penicillins). Antibiotic drugs within the same antibiotic class typically share the same mechanisms of resistance; thus, our findings on most antibiotic classes can be applied broadly in the clinical setting. We also did not evaluate dose and duration of antibiotics, which could contribute to nonsusceptibility. Fourth, although UTI symptoms usually prompt seeking of medical care, our cohort may have missed some cases that were not medically attended. Also, because the volume of clinical visits and laboratory testing substantially decreased during the coronavirus disease 2019 pandemic, sample size was small in 2020. Data from this period may also have captured more severe cases who were more likely to seek care and have a culture ordered. In addition, health-seeking behaviors that potentially have an impact on both development of antibiotic resistance and UTIs may be different in those who have a urine culture and those who do not. By focusing our study on individuals with a urine culture, our patient population may have been more likely than the general population to have had recent UTI or history of antibiotic-resistant UTI, which could overestimate resistance. Last, rUTIs are often defined as ≥3 UTI events within 1 year or ≥2 UTI events in the last 6 months [15]. We did not adopt a formal definition for rUTI because our goal was to examine characteristics of all culture-confirmed UTI occurring during the study period.

Increasing antibiotic resistance limits antibiotic options for outpatient treatment of UTIs and renders antibiotic treatment less effective. As repeated UTIs are a major cause of antibiotic resistance, a better understanding of uropathogens responsible for recurrence is needed. Importantly, given that antibiotic therapy for UTI is empiric and uropathogens increasingly demonstrate nonsusceptibility, continuously updated data on susceptibility patterns are critical to guide appropriate antibiotic prescribing as well as to inform interventions to prevent UTI recurrence. As the degree of resistance to the most frequently prescribed antibiotics for UTI evolves over time and may vary by region, it is also important to continuously monitor the local prevalence of uropathogens and their susceptibility to guide appropriate antibiotic use.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

Notes

Acknowledgments. The authors thank Pooja M. Patel, MPH, and Rossy F. Perez for their assistance in drafting the manuscript.

Financial support. This study was supported by GlaxoSmithKline (GSK).

References

Author notes