Clinical Response to Discordant Therapy in Third-Generation Cephalosporin-Resistant UTIs

This study was nested within a multisite retrospective study of children with third-generation cephalosporin-resistant UTIs presenting to an acute care setting of 1 of 5 children’s hospitals or a large managed care organization with 10 hospitals from July 1, 2012, to June 30, 2017. Sites were located in California (n = 3), Georgia (n = 1), Minnesota (n = 1), and Washington (n = 1) as detailed in Supplemental Table 4.

We defined UTI according to the AAP guidelines as having (1) pyuria on urinalysis or urine dipstick (≥5 white blood cells per high power field or any leukocyte esterase) and (2) urine culture with ≥50 000 colony-forming units (CFUs) per mL.³  Third-generation cephalosporin-resistant UTI was defined as a UTI with E coli or Klebsiella spp. nonsusceptible to ceftriaxone or cefotaxime. Urinalyses with microscopy, urine dipsticks, and urine cultures were performed per standard procedures of the participating sites’ clinical laboratories. Antibiotic susceptibilities were determined according to Clinical and Laboratory Standards Institute guidelines.¹⁷  Potential cases were identified by query of the microbiology laboratory database or electronic medical record at each site to identify children <18 years who presented to an acute care setting (emergency department [ED], general pediatrics and/or urgent care clinics, inpatient setting) and also met the following inclusion criteria: had a urinalysis and urine culture (obtained within 24 hours of admission if the patient was an inpatient) and had a urine specimen obtained via straight catheterization, clean catch, or suprapubic aspirate. If the collection method was not recorded, the case was included if there was no documentation of a Foley catheter or other type of urinary catheter.

Potential cases underwent chart review and were excluded if they were immunocompromised (known immunodeficiency, malignancy, receipt of immunosuppressants in the previous 6 months), had a history of urologic surgery other than circumcision, had a complex chronic condition (except for renal abnormalities),¹⁸  did not have documentation of a patient encounter or associated urinalysis or urine dipstick on the day of the urine culture, or had a UTI in the previous 30 days (to ensure that cases captured were index cases and not UTI relapses). Patients with complex chronic conditions were excluded because of concern that their chronic condition may make it difficult to assess their clinical response. Only 1 episode per patient was included, and repeat episodes were excluded. Finally, cases were excluded if they were (1) started on concordant antibiotics (antibiotics to which their urinary pathogen was susceptible) or were not started on antibiotics at the time of initial evaluation or (2) did not have follow-up documentation indicating contact between the health care provider and family after urine culture susceptibilities resulted, which could have included a phone follow-up, outpatient, ED, or inpatient note.

The target sample size for the parent study was 480 patients. For the 5 children’s hospitals, all cases who met inclusion criteria were reviewed. At the managed care organization, because of their large sample size and limited resources for chart review, we determined a priori to use simple random sampling to select ∼70% of cases meeting inclusion criteria for review to meet the target sample size.

Data were entered into a secure Research Electronic Data Capture tool hosted at Stanford University.¹⁹  The study was approved by each site’s institutional review board.

Escalation of care was defined as an ED visit or hospitalization for subjects treated as outpatients or transfer to the ICU (if already hospitalized) because of persistent symptoms attributable to the UTI while on discordant antibiotics before the return of urine culture results. Examining ED visits and hospitalizations has also been used in other studies analyzing outcomes such as treatment failure or adverse events in other conditions.^20,21  Follow-up visits to the primary care pediatrician or urgent care were not evaluated because they were thought to represent a lower level of illness severity than an ED visit or hospitalization and because differentiating routine clinic follow-up visits from “sick” visits for worsening symptoms is problematic. Return visits triggered exclusively by urine culture results in the absence of persistent symptoms were not categorized as escalation. Overall clinical response while on discordant antibiotics was assessed by physician chart review of follow-up documentation after the final urine culture results were available. Response was classified as “improved,” defined as complete or partial resolution of presenting signs and symptoms, or “not improved,” defined as persistence of the presenting signs and symptoms. Clinical response was assessed by a second reviewer at the same site in 34% of cases to determine interrater reliability (Κ = 0.79; 94% agreement). For patients who underwent repeat urine testing while on discordant antibiotics, we assessed resolution via improvement in pyuria and/or bacterial growth on urine culture.

From our target sample size of 480 for the parent study, we predicted that 75% would be started on discordant therapy with 80% follow-up and result in 288 patients for this study. This would allow us to report a 95% confidence interval (CI) of ±3% for a 5% escalation rate and a 95% CI of ±5% for an 80% clinical improvement rate. The proportion with care escalation was calculated from the study population. The proportions with improved clinical response and symptom improvement were calculated from the number of patients with this information recorded in follow-up documentation. Proportions were compared by using the χ² test, and nonparametric continuous variables were evaluated by using the Wilcoxon rank-sum test. We calculated 95% binomial exact CIs for proportions calculated. Analysis was performed by using Stata version 15 (Stata Corp, College Station, TX).

A total of 1259 children presented to an acute care setting with urine cultures growing ≥50 000 CFU/mL of E coli or Klebsiella spp. nonsusceptible to ceftriaxone or cefotaxime, and 777 children were excluded (Fig 1). The most common reason for exclusion was a complex chronic condition followed by a history of urologic surgery and immunocompromised status. Of the remaining 482 children with third-generation cephalosporin-resistant UTIs, 399 (82.8%) were started on discordant antibiotic therapy. From this cohort, 316 (79.2%) children had follow-up documentation of urine culture results while on discordant antibiotics.

For the 316 included children, the median age was 2.4 years (interquartile range [IQR] 0.6–6.5), and 247 (78.2%) were girls (Table 1). Most children (249 of 316; 78.8%) received a complete urinalysis with microscopy, and 67 (21.2%) received a urine dipstick only. The most common presenting symptom was fever. Of the 44 (13.9%) patients hospitalized at initial presentation, 20 (45%) were <2 months of age. Most children (285 of 316; 90.2%) were started empirically on a cephalosporin, with 205 (64.9%) receiving a first-generation cephalosporin (Table 2).

Patients excluded because of lack of follow-up documentation were older and more likely to present with dysuria. They were less likely to present with fever or poor oral intake or be initially hospitalized than included patients (Supplemental Table 5).

Escalation of care occurred in 7 of 316 children (2.2%; 95% CI 0.8%–4.5%) (Table 3), with an age range of 6 to 15 months. All presented with fever and were started empirically on a cephalosporin. Four children had an ED visit without hospitalization, and 3 children were hospitalized because of persistent symptoms. No children who were initially hospitalized were transferred to the ICU while on discordant therapy.

Data on overall clinical response were available for 230 of 316 children (72.8%). Median time to follow-up was 3 days (IQR 2–3). A total of 192 of 230 children (83.5%; 95% CI 78.0%–88.0%) had overall clinical improvement while on discordant antibiotics. Improvement by symptom is shown in Fig 2. Among children who had clinical response recorded, the escalation rate was 7 of 230 (3.0%; 95% CI 1.2%–6.1%).

Of the 86 children who had documentation of follow-up contact without a clinical response recorded, 85 had phone follow-up notes and 1 had a secure message sent, generally to document family notification of results and new prescriptions. Compared with patients with clinical response information recorded, patients without this information were older, more likely to have dysuria, and less likely to have fever and be initially hospitalized (Supplemental Table 5).

Nineteen children had either repeat urinalysis (16 children) or urine dipstick (3 children) while on discordant therapy. Ten children (53%) had resolution of pyuria, and 6 (32%) had improvement of pyuria. Three patients (16%) did not have improvement in pyuria; 1 had no change, and 2 had pyuria that worsened. Seventeen patients had a repeat urine culture while on discordant therapy. Eleven (65%) had a negative urine culture, and 3 (18%) grew the same pathogen on repeat culture with improvement in the colony count. The colony count remained unchanged in 2 patients and increased in 1.

In this multisite retrospective study of 316 children with third-generation cephalosporin-resistant UTIs started on discordant antibiotic therapy, we found that few children required escalation of care and most experienced initial clinical improvement. Furthermore, in the small group of children that underwent repeat urine testing while on discordant therapy, most had resolution or improvement in pyuria and sterilization of their urine cultures. Our findings suggest that current empiric regimens for UTI as informed by local susceptibility patterns are reasonable while awaiting urine culture results. Additionally, given that these patients initially received what would generally be considered inadequate treatment, our findings may provide some insight into the natural history of UTIs and/or trigger further investigation into the relationship between in vitro urine culture susceptibilities and in vivo clinical response to treatment.

Our results are similar to smaller studies from outside the United States in whichchildren with ESBL UTIs were examined. A multicenter study from France¹⁶  and a single-center study from Japan¹⁴  (with 301 and 55 patients, respectively) reported that the time to apyrexia was similar for children with ESBL UTIs who received concordant versus discordant empiric antibiotic therapy. However, most (71%) UTIs in the study from France did not meet AAP UTI diagnostic criteria, and the only clinical response examined was fever resolution. A study of 28 pediatric inpatients in Greece with ESBL UTIs revealed a 94% and 95% clinical and microbiologic response rate, respectively, to empiric antibiotics, although 86% received discordant therapy, compared to 100% clinical and microbiologic response in 56 children with non-ESBL UTIs started on concordant therapy.¹⁵  In contrast, for 154 children with ESBL UTIs in Turkey (90% of whom received discordant antibiotics), 44% and 45% had clinical and microbiologic response, respectively, to empiric therapy compared to 95% with clinical and/or microbiologic response in the non-ESBL UTI group with 5% of patients on discordant therapy.⁸  The reasons for these discrepant response rates are not clear but may relate to differences in the patient population or definition of clinical and/or microbiologic response. Lastly, although not limited to ESBL UTIs, 1 US-based study of hospitalized children with UTIs revealed that fever duration did not differ between 178 patients receiving concordant therapy and 22 receiving discordant therapy.²² 

There are several possible explanations for the low rate of care escalation and the high rate of overall clinical improvement seen in children on discordant antibiotics. First, despite the lack of in vitro susceptibility, antibiotics may achieve sufficient concentrations in the urine and renal parenchyma to achieve a clinical response.²³  Microbiologic breakpoints are usually established to be relevant at all sites of infection, except for cerebrospinal fluid, and many antibiotics reach higher concentrations in the urine than in plasma.²³  Second, some UTIs may resolve spontaneously without antibiotic treatment, similar to otitis media and streptococcal pharyngitis.^24–26  In a study of 3066 febrile infants <3 months old, of the 807 who did not undergo urine testing or initial antibiotic treatment and had follow-up care, 61 (7.6%) were predicted to have a UTI, but only 2 were ultimately diagnosed with 1, suggesting that many UTIs improve without treatment.²⁷  Third, there is ongoing debate regarding the definition of UTI,²⁸  and despite using commonly accepted diagnostic criteria, a small proportion of our patients may not have had a true UTI.

These findings have implications for empiric antibiotic therapy for suspected UTI. Given the increasing prevalence of third-generation cephalosporin-resistant isolates,²  there has been discussion regarding the need for broader empiric therapy.^6–9,12  First-generation cephalosporins are a common empiric option and were used by the majority of patients treated with oral antibiotics in this study. If empiric therapy were to include coverage for these resistant organisms, the most common oral options would be amoxicillin-clavulanate, ciprofloxacin, or trimethoprim-sulfamethoxazole, which would have implications for antimicrobial stewardship and the risk of drug-associated adverse events.¹³  Although some studies suggest that a 48-hour delay in the initiation of antimicrobial therapy in febrile UTI may increase the risk of renal scarring,²⁹  authors of a meta-analysis found no association.³⁰  Given the uncertain nature of the association between delayed treatment and renal scarring and the fact that discordant antibiotics may achieve some clinical response, the risk of renal scarring may be less of a concern. The low rate of care escalation and high rate of clinical improvement while on discordant antibiotics suggests that for most patients, it would be reasonable to continue current empiric antibiotic practices until urine culture sensitivities return. However, the choice of empiric antibiotic regimen should take into account local antibiotic susceptibilities and patient characteristics because there may be some patients whose clinical and microbiologic history would favor broader empiric therapy, such as patients with severe symptoms and/or multiple risk factors for resistant organisms.

This study has several strengths. To our knowledge, it is the first US-based study using a large, multisite cohort to investigate the initial response of children with third-generation cephalosporin-resistant UTIs started on discordant antibiotic therapy. We used the AAP definition of UTI, which required both pyuria and a positive urine culture result for diagnosis to decrease the likelihood of misclassification. We also used an objective outcome, escalation of care, to assess clinical worsening, and our outcome of overall clinical response was assessed by 2 reviewers in a subset of charts with a 94% agreement rate.

This study also has several limitations. First, we did not establish a control group of patients with UTI treated with concordant antibiotics because these patients are typically not contacted for follow-up if treated as outpatients. However, our overall rate of care escalation was low and comparable to studies in which patients were likely started on concordant antibiotics.^31,32  Schnadower et al³²  reported that 51 of 1842 (2.8%) infants 29 to 60 days old with a febrile UTI had an adverse event (eg, death, shock, ICU admission) when treated with antibiotics. Dayan et al³¹  reported that 0 of 128 infants <60 days of age admitted to an acute care unit for UTI required an ICU transfer. Although there are limited data regarding care escalation of older children started on concordant antibiotics, these studies would suggest that it is generally uncommon. Second, we did not include revisits to the primary care pediatrician or urgent care as escalation because we sought to identify children who were ill enough to warrant an ED visit or hospitalization and because differentiating routine follow-up visits from sick visits is problematic. Therefore, we did not capture visits that may have occurred in outpatient clinics. Third, there may be selection bias because of missing follow-up or clinical response information. Patients without follow-up documented when culture results returned were not included in the study. Although it is possible that these patients could have presented to other institutions if they worsened, characteristics of those without follow-up documentation suggest that they probably had less severe illness than the study cohort. Within the study cohort, some patients did not have clinical response information recorded. Characteristics of this group also suggest that they may have had less severe illness than those with clinical response data, and it would be reasonable to infer that omission of clinical response information in the follow-up documentation likely signifies that the patients had improved. Therefore, the total cohort of patients on discordant antibiotics may have had a lower care escalation rate and higher clinical improvement rate than we found in our study cohort with complete data. Fourth, we excluded patients with a history of urologic surgery, immunosuppression, and complex chronic conditions; therefore, the results cannot be applied to this population.

In a large cohort of children with third-generation cephalosporin-resistant UTIs started empirically on discordant antibiotics, we found that the rate of care escalation was low, and most children experienced initial clinical improvement. These results suggest that current empiric regimens are reasonable even with concerns of increasing antimicrobial resistance. Future research should prospectively evaluate the in vitro and clinical and microbiologic response to discordant therapy in children with UTIs resistant to third-generation cephalosporins to better define antibiotic susceptibility thresholds and determine ideal antibiotic regimens.

We thank Gomathi Krishnan, PhD, for her valuable assistance in data extraction and Christopher Russell, MD, for serving as a second reviewer for 1 of the sites.

AAP

American Academy of Pediatrics

CFU

colony-forming unit

CI

confidence interval

ED

emergency department

ESBL

extended-spectrum β-lactamase

IQR

interquartile range

UTI

urinary tract infection