Frequency of Treatment Failure of UTIs in Children With Congenital Urinary Tract Anomalies

We conducted a multicenter, retrospective cohort study at 6 different free-standing children’s hospitals located across the United States. We included children with CAKUT/NGB who presented to an emergency department (ED) from January 1, 2017 to December 31, 2018 and were diagnosed with UTI. We used International Classification of Diseases, 10th revision, Clinical Modification (ICD-10-CM)codes to identify children with CAKUT/NGB and UTI (Supplemental Table 3). Thus, all patients enrolled in this study had a clinician-diagnosed UTI. As there is no standardized definition for UTI in children with CAKUT/NGB (who are excluded from the American Academy of Pediatrics’ Clinical Practice Guideline on UTIs¹ ), we used a clinical diagnosis of UTI in this work. Data were collected via manual extraction of variables from medical charts and were entered into a centralized REDCap database.⁵  Institutional review board approvals were obtained at all participating sites.

We included children <18 years of age with either CAKUT/NGB who presented with fever (temperature ≧38°C either measured in the ED or reported by parents) or hypothermia (<36°C), were diagnosed with UTI as noted in the attending physician’s documentation and had a positive urine culture. Children with previous surgical repair of their lesions were included in the study. All urine culture specimens used for the diagnosis of UTI had to be collected by clean catch, catheterization, or suprapubic aspiration. We excluded children whose UTI was diagnosed with a bagged specimen as this would lead to uncertainty around our primary outcome of treatment failure. We excluded encounters where children had meningitis, children who had negative urine cultures, those with end-stage renal disease, UTI within the past 30 days, genitourinary surgery within the past 7 days, presence of a renal abscess, were discharged to a location other than the child’s home, and children transferred from another center with laboratory results not available in the electronic health record. We excluded children not discharged to home as subacute facilities can often diagnose and manage medical problems, such as UTIs, within their facility, affecting our ability to capture children with treatment failure. Children with renal abscesses were excluded as the presence of an abscess may affect both route and duration of antibiotics. We excluded children with a UTI within the past 30 days as children with a UTI within the past 30 days may receive differential treatment compared with those without a recent, prior UTI. For children with multiple encounters, each ED encounter and/or hospitalization was included if encounter-level eligibility criteria were met to more closely resemble real world circumstances.

We extracted several covariates for each encounter. Patient characteristics included the child’s sex, age in years, parent-reported race and ethnicity, and insurance status (public versus private or self-pay). We included race because it is a social, rather than biological, construct. Implicit bias may affect physician behavior, and therefore race may play a role in physicians’ treatment decisions.⁶  The majority of children in our cohort were categorized by primary lesion type. However, children with multiple lesions were included as a separate group if the specific combinations of anomalies occurred in 5 or more children within the cohort. We created a mixed category for children with other combinations of anomalies with <5 children. Additionally, to better understand the influence of NGB and obstructive pathology on outcomes, we created a binary variable for presence of neurogenic bladder, as well as a binary variable for the presence of obstructive lesions (ie, ureteropelvic junction obstruction, ureterovesical junction obstruction, ectopic ureter, posterior urethral valves, megaureter, ureterocele, and high-grade vesicoureteral reflux [VUR]). Illness-related covariates included multidrug resistant organisms (MDROs) and time to defervescence. We defined MDRO as an organism that was not susceptible to at least 3 of the following antibiotic classes: first and second generation cephalosporins, third and fourth generation cephalosporins, fluoroquinolones, sulfonamides, and aminoglycosides.⁷ 

Additionally, we extracted antibiotic treatment variables. We calculated duration of intravenous (IV) antibiotic treatment as follows: (last day of antibiotic − first day of antibiotic) +1. Using this calculation, encounters in which children received a single dose of IV antibiotics would be considered to have 1 day of IV antibiotics. We also extracted the individual IV antibiotic types administered and further classified antibiotic regimens by class.

The outcome of interest was treatment failure, defined as an ED encounter or hospitalization for UTI with the same pathogen within 30 days of the index encounter. The 30-day interval was chosen similar to prior studies³  to allow capture of return encounters for treatment failure that may occur after completion of the antibiotic course.

All analyses were conducted in R v4.1.2⁸  or Stata 13.1. Analysis was performed at the level of each individual encounter to the ED and/or hospitalization. Demographic and patient characteristics, stratified by treatment outcome (success versus failure), were calculated as mean±SD for continuous variables (age), or total number (N) and percentage (%) for categorical variables (all other measures). Crude differences between the 2 outcome groups (ie, without controlling for covariates or correcting for multiple testing) were conducted using independent t tests for continuous variables and χ-square (or Fisher’s exact, as appropriate) tests of difference for categorical variables.

Of 2014 encounters, 482 encounters from 404 unique children met criteria for inclusion in this study, of which 29 (6.0%) had treatment failure. The most common reasons for exclusion were the absence of fever or hypothermia (n = 646), UTI in the prior 30 days (n = 299), nonincluded lesion (n = 259), and no UTI documented in the electronic health record (n = 193) (Fig 1). The types of anatomic anomalies seen in children excluded for UTI in the past 30 days are listed in Supplemental Table 5. Urine culture results are shown in Supplemental Table 6.

In univariate analyses, there were no difference in patient demographics between those with and without treatment failure. (Table 1) More encounters in the treatment failure group had bacteremia than in the nontreatment failure group (35.7% vs 12.6%, P = .04). There were no other statistically significant differences in aspects of the illness course, including the need for intensive care, the presence of MDROs, or discharge home from the ED. (Table 1)

There was significant heterogeneity in the types of CAKUT in children with treatment failure. (Table 2). Within all children with treatment failure, the most common 3 lesions were NGB (27.5%), mixed (20.7%), and duplex collecting system with high-grade VUR (10.3%). Within specific lesion types, children with posterior urethral valves and those with duplex collecting systems and high-grade VUR had the highest proportions of children with treatment failure (20% for each group), followed by children in the “mixed” category, which had a rate of treatment failure of 15.0% (Table 2).

There was no difference in duration of either IV antibiotics with efficacy against the uropathogen or total duration of antibiotics between children with and without treatment failure (duration IV antibiotics: 3.4±2.5 days, 3.5±2.8 days, P = .87; total duration: 10.2±3.8 days, 10.8±4.0 days, P = .39) (Fig 2). Similarly, there was no statistically significant difference in the rate of treatment failure between children who did and did not receive any IV antibiotics (4.6% vs 2.8%, P = .73). There was also not a statistically significant difference in IV antibiotic regimens between children with and without treatment failure (P = .34, Supplemental Table 4). Among children with treatment failure, 6 children had an index UTI because of an MDRO isolate. Of these 6 patients, 2 had antibiotics changed during the index encounter because of the resistance of the pathogen. Overall, of the 46 children with MDROs in the cohort, 7 had an antibiotic change because of resistance.

Thirty three patients total had bacteremia, of which 5 (15.2%) had treatment failure. Sixty-six percent of patients with bacteremia had imaging of the kidneys. Twenty nine of the 33 patients with bacteremia were admitted to the hospital and received a mean of 6.6 (SD 3.1) days of IV antibiotics. The 2 admitted children with treatment failure received IV antibiotics for 9 and 3 days during the index admission. Three of the 5 bacteremic children with treatment failure were discharged from the hospital from the ED from the index encounter, 2 of which were subsequently called back to the ED after the blood culture became positive. On repeat presentation to the ED, all 5 children with treatment failure were febrile and 3 (60%) presented in compensated shock, as determined by heart rate and blood pressure measurements. Of the 3 patients in compensated shock, 1 child re-presented 3 days after initial discharge because of worsened symptoms and was found to have never filled the prescribed antibiotics. Another child re-presented 26 days after the index encounter and had an asymptomatic period between the 2 UTIs. The last patient re-presented 4 days after the index encounter with dehydration and was found to have acute kidney injury.

Here we describe our findings from one of the largest cohort studies of children with CAKUT/NGB to date and report that 6.0% of these children with CAKUT/NGB who had a clinical diagnosis of UTI had treatment failure. In this foundational work, we also present granular data on individual CAKUT/NGB lesion types, types of antibiotics, and durations of IV antibiotics administered in children with and without treatment failure. Further, we identified associations between bacteremia and treatment failure of UTIs in this population of children with CAKUT/NGB. Altogether, the findings of this work can be used to inform future studies to optimize UTI treatment in children with CAKUT/NGB.

We identified a frequency of treatment failure of 6.0% in this cohort of children with CAKUT/NGB, which is a lower frequency than expected. Sparse research exists to describe treatment failure in the CAKUT/NGB population. We found no statistically significant difference between antibiotic regimens between children with and without treatment failure. When examining the CAKUT lesion type, we did not find a statistically significant difference in children with obstructive lesions or neurogenic bladder in children with and without treatment failure. Further, there were no statistically significant differences in individual CAKUT lesion types in children with and without treatment failure, although this analysis was limited by small numbers of children with individual lesion types. One potential reason for the lack of difference is because of the heterogenous categories of lesions included in the study population. Although we had 6 sites in this study, we had small numbers of children in several categories of lesions. These low numbers highlight the small proportion of children affected by some CAKUT lesions, underscoring the difficulty in obtaining data to guide treatment decisions. However, other categories of lesions, including neurogenic bladder and high-grade VUR, were well-represented within the cohort, providing benchmarks for the rate of treatment failure for UTI for children in these groups. To better aid in identifying risk factors for treatment failure, further large studies should be powered to find differences in treatment failure among more homogenous populations of children with obstructive lesions or neurogenic bladder.

We did not find a difference in duration of IV antibiotics between children with and without treatment failure. Although there are limited studies in the literature focusing on risk factors for treatment failure in children with a full spectrum of CAKUT/NGB anomalies, several studies exist that include a small proportion of children with anomalies within the general study population. Most existing studies examine the risk of treatment failure in context with the duration of IV antibiotics or total duration of antibiotics administered. One multicenter retrospective cohort study examined outcomes in infants <60 days of age with bacteremic UTIs and found no difference in 30-day recurrence of UTI between children treated with short-course (≦7 days) or long course (>7 days) of IV antibiotics.⁹  Of note, this study included both children with known genitourinary anomalies as well as those with complex chronic conditions. The reported rate of treatment failure in this study was 5%, which is only slightly lower than our reported rate of 6.2%.⁹  A more recent study examined retrospective data to compare outcomes of children between 6 months and 18 years with a clinical diagnosis of pyelonephritis who were either treated with short (≦9 day) or long courses of antibiotics.⁴  These authors did not find a difference in treatment failure between children treated with short versus long antibiotics, but did find that children between 4 and 13 years of age had higher odds of treatment failure than children <4 years or >13 years. Of note, this study did include a subgroup of children with urological anomalies and reported a higher rate of treatment failure in children with urological anomalies compared with those without.⁴  Neither of these studies identified an association between IV antibiotic duration or total antibiotic duration and treatment failure, which is consistent with our results. However, differences in the study design, populations, and outcomes preclude direct comparison of these existing studies in the literature to our findings.

It is interesting to note that we did not find an association between MDRO status and treatment failure. This contradicts several reports in the literature. One study of children treated with empirical antibiotics in the ED for UTI found that children with an extended-spectrum β-lactamase producing bacteria had increased odds of 30-day all-cause return visits to ED. The authors noted that of the children who returned to the ED, 66% had symptoms similar to their initial presentation, suggesting that at least a proportion of these children had treatment failure.¹⁰  Another study of 185 infants found that children with a UTI because of an extended-spectrum β-lactamase-producing organism had a higher rate of treatment failure, further supporting this association.¹¹  It is possible that as heightened awareness of MDROs grow, especially in children with CAKUT/NGB, providers are more proactive around changing antibiotics once an MDRO result returns, thus ensuring adequate coverage and decreasing the risk of treatment failure. It is also possible that the empiric antibiotics given to children with CAKUT/NGB are broader than those given to children without CAKUT/NGB, as only a few children in our cohort with MDROs required a change in antibiotics because of the resistance pattern of the isolate.

We report associations between bacteremia and treatment failure. In our cohort, bacteremia was present in 11.8% of children who had a blood culture drawn and was associated with treatment failure. Our rate of bacteremia is higher than rates of bacteremia reported in either infants (between 4%–7%)^12,13  or all children (5.6%) with UTI.¹⁴  As urologic anomalies are associated with bacteremic UTIs,^15,16  this higher rate of bacteremia in our cohort is not surprising. The difference between rates of bacteremia in children with and without treatment failure in our study is also not surprising: 2 of 5 (40%) children in our cohort with bacteremia and treatment failure were discharged from the hospital from the ED and were called back to the ED because of the positive blood culture results. Therefore, as this group was discharged before culture results, children discharged from the ED had an asymmetric and possibly higher risk of admission than children admitted to the hospital, potentially affecting our results. However, all children with bacteremia and treatment failure were febrile and many were in compensated shock on their repeat encounter, making it possible that treatment failure occurred in these children because of inadequate antibiotic treatment and poor source control of their UTIs given their complex anatomy. Larger studies are needed to ascertain what aspects of this complex anatomy or illness could put these children at higher risk for severe illness and treatment failure.

There are several strengths and limitations to this work. Our study is one of the largest studies to evaluate for risk factors for UTI treatment failure in children with CAKUT/NGB. Although limitations exist inherent to a retrospective design, we were able to use manual extraction to obtain clinical data at a granular level. Although we did include multiple centers in this work, all the included sites are tertiary or quaternary care children’s hospitals. Thus, results may not be generalizable to children seen in other settings. In addition, the prevalence of treatment failure in our full cohort was lower than anticipated, limiting both our ability to identify risk factors associated with treatment failure and to optimally adjust for the nonindependence of our data. Larger studies are needed to confirm our findings and explore associations between duration of IV antibiotics and treatment failure, especially in children with CAKUT and bacteremic UTIs. The lack of a standardized definition of UTI in children with complex genitourinary systems, as well as the inability to control for practice patterns of individual clinicians, is another limitation. Further, our definition of UTI was a UTI diagnosed by the attending physician: thus, some children may be falsely identified as having a UTI. However, our study cohort represents real-world circumstances of children diagnosed with UTIs by their treating physicians, who may not have a guideline to assist with diagnosis in this population. We were also unable to collect data on children who sought care outside of each healthcare system, and thus may have missed some cases of treatment failure, although integrated electronic medical systems may have lessened this limitation. Further, we were unable to determine whether antibiotics were changed after discharge from the ED and were unable to include antibiotic discordance with antimicrobial sensitivities in the analysis. An additional limitation is that we included children with surgical repair of their lesions, which may affect their risk of treatment failure. Further, as children younger than 3 months of age may not always have clinically-apparent temperature instability, we may have excluded a disproportionate number of young infants. Lastly, a number of children were excluded from this study because of UTIs within the prior 30 days and some of these UTIs could represent treatment failure in the outpatient setting. Future studies should encompass outpatient management of UTIs to garner the full spectrum of illness and risk factors for treatment failure in this population.

Treatment failure occurred in 6.0% of children with CAKUT/NGB in our cohort. We did not find a difference in duration of antibiotics between children with and without treatment failure on a univariate analysis. However, we did find a difference in the frequency of bacteremia, suggesting that a small proportion of children with higher infectious burden and bacteremia are at risk for treatment failure. Further investigation is needed to determine specific anatomic, patient and illness characteristics associated with increased risk of bacteremia in children with CAKUT/NGB.

Members of the UTI in CAKUT study group include: Cynthia Abou Zeid and Gemma Beltran.