Objectives To determine predictors of parenteral antibiotic duration and the association between parenteral treatment duration and relapses in infants <3 months with bacteraemic urinary tract infection (UTI).
Design Multicentre retrospective cohort study.
Setting Eleven healthcare institutions across the USA.
Patients Infants <3 months of age with bacteraemic UTI, defined as the same pathogenic organism isolated from blood and urine.
Main outcome measures Duration of parenteral antibiotic therapy, relapsed UTI within 30 days.
Results The mean (±SD) duration of parenteral antibiotics for the 251 included infants was 7.8 days (±4 days), with considerable variability between institutions (mean range 5.5–12 days). Independent predictors of the duration of parenteral antibiotic therapy included (coefficient, 95% CI): age (−0.2 days, −0.3 days to −0.08 days, for each week older), year treated (−0.2 days, −0.4 to −0.03 days for each subsequent calendar year), male gender (0.9 days, 0.01 to 1.8 days), a positive repeat blood culture during acute treatment (3.5 days, 1.2–5.9 days) and a non-Escherichia coli organism (2.2 days, 0.8–3.6 days). No infants had a relapsed bacteraemic UTI. Six infants (2.4%) had a relapsed UTI (without bacteraemia). The duration of parenteral antibiotics did not differ between infants with and without a relapse (8.2 vs 7.8 days, p=0.81).
Conclusions Parenteral antibiotic treatment duration in young infants with bacteraemic UTI was variable and only minimally explained by measurable patient factors. Relapses were rare and were not associated with treatment duration. Shorter parenteral courses may be appropriate in some infants.
- General Paediatrics
- Infectious Diseases
Statistics from Altmetric.com
What is already known on this topic
Detection of bacteraemia in infants with urinary tract infection often drives extended hospitalisations and parenteral antibiotic courses.
No published recommendations exist to guide treatment duration.
What this study adds
Infants receive variable parenteral antibiotic durations, which are explained only minimally by clinical factors.
Relapses are rare and not associated with the duration of parenteral antibiotics.
Bacteraemia is detected in 3–17% of infants with urinary tract infections (UTI),1–8 and bacteraemic UTI now represents the most common source of bacteraemia in young infants.7 ,9 ,10 The clinical significance of bacteraemia with UTI is uncertain.1 ,11 ,12 Although no published guidelines exist on the management of bacteraemic UTI, the available literature suggests that detection of bacteraemia leads to longer hospitalisations and durations of parenteral (intravenous or intramuscular) antibiotics in infants with UTI.6 ,8 ,11–13
Using data from multiple institutions on infants <3 months of age with bacteraemic UTI, we sought to (1) determine predictors of duration of parenteral antibiotic therapy and (2) determine the association between duration of parenteral treatment and relapse within 30 days.
This retrospective investigation involved 20 hospitals at 11 healthcare institutions around the USA (table 1). The central site for the study was Santa Clara Valley Medical Center, in San Jose, California. Each participating site had pre-existing databases from prior and/or ongoing investigations that included data on infants<3 months with bacteraemic UTI. Eight institutions were involved in a Pediatric Research in the Inpatient Setting (http://www.prisnetwork.org/) investigation on the epidemiology of bacteraemia in young infants.9 Two institutions had previously completed single-centre investigations on the management and outcomes of infants with bacteraemic UTI.12 ,14 Finally, an investigation from Kaiser Permanente Northern California involving 10 different Kaiser hospitals evaluated the epidemiology of bacteraemia in young infants.10 For the current investigation, all infants <3 months of age with bacteraemic UTI identified from these initial databases underwent a second chart review using a protocol specifically designed for this study.
Infants were included if they were <3 months of age and had the same pathogenic organism isolated from the blood and urine, regardless of the urine culture colony count, urinalysis findings or indication for the cultures. Infants were excluded prior to data submission to the central site if they had major comorbidities (defined as neuromuscular conditions such as spina bifida, prior urological surgery other than circumcision or immunodeficiency), if they were initially managed in the intensive care unit, or if they had indwelling urinary or central venous catheters at the time the cultures were drawn. Infants were excluded after data submission if they received prolonged parenteral treatment for other reasons (meningitis, osteomyelitis), if data on parenteral treatment were missing or if transfer out of the study site occurred during the hospitalisation.
Data collection and variables
Bacteraemic UTIs at each site were originally identified using existing microbiology databases. Patients were considered ill-appearing if there was specific physician documentation of being ‘lethargic’, ‘toxic’, ‘septic’, ‘sick’ or ‘ill-appearing’ on presentation. Length of stay (LOS) was an integer for each patient, determined by the number of calendar days separating the admission and discharge dates. Duration of parenteral antibiotic therapy (intravenous or intramuscular) was calculated as the number of calendar days separating the first and last parenteral dose for the index infection. An additional day was added for infants who were given daily intramuscular ceftriaxone, given the long half-life of this antibiotic. For example, a child who received intramuscular ceftriaxone on two successive mornings was recorded as having received 2 days of parenteral antibiotics.
Relapsed bacteraemic UTI and relapsed UTI without bacteraemia (from here on referred to simply as ‘relapsed UTI’) were defined as a second episode caused by the same pathogenic organism within 30 days of the initial episode, and readmission was defined as a re-hospitalisation for UTI and/or bacteraemia within 30 days. Clinical deterioration was defined as developing meningitis during acute treatment or within 30 days or needing transfer to an intensive care unit during acute treatment.
The assembled cohort was limited to the number of total available cases at each institution. To determine the predictors of the duration of parenteral antibiotic therapy, all potential predictor variables (table 3) were included in a linear fixed-effects regression model. Fixed-effects regression adjusts for clustering by institution with indicator variables and is most appropriate when the research focus is the pooled within-cluster effects of relatively few clusters—such as our 11 institutions.15
To estimate this clustering effect and how it was affected by the predictors in our model, we calculated the intraclass coefficient (ICC) of both the empty model with no predictors and the full model with all predictors. This ICC reflects the proportion of variance in duration of parental antibiotic therapy due to institutions. The reduction in ICC due to the addition of predictors (from the empty to full models) reflects the degree to which the clustering effect of institution was due to patient-level clinical predictors.15 We calculated CIs for these ICCs with a bias-corrected and accelerated parametric bootstrap with 2000 replications.16
Both continuous predictor variables (year of blood culture and age in weeks) were grand-mean centred. Although abnormal imaging results were analysed as a predictor of UTI relapse, abnormal imaging was not analysed as a predictor of treatment duration given that some studies were obtained after discharge. We used graphical methods to check assumptions of the fixed-effects model: a normal probability plot of the standardised residuals; a scatter plot of standardised residuals against fitted values; Lowess smoothing of standardised residuals against each continuous predictor; and box plots of standardised residuals grouped by each binary predictor.
Parenteral antibiotic duration was analysed both as an outcome variable and as a predictor variable (of relapse). Because relapse within 30 days was a rare outcome, only bivariate (and not multivariable) analyses were performed to analyse the association between clinical variables, including parenteral treatment duration and the primary outcome.17 For bivariate analysis, Fisher's exact test or χ2 test was used, as appropriate, for categorical variables, and the Wilcoxon rank sum test or Students’ t test was used, as appropriate, for continuous variables. All analyses were performed using STATA V.13 (StataCorp, College Station, Texas, USA).
A total of 276 infants with bacteraemic UTI were identified, from which 25 infants were excluded: 5 had missing data or were transferred out, 19 were treated for meningitis (7 with cerebrospinal fluid (CSF) cultures that were positive for the same organism as the blood and urine, 6 with CSF pleocytosis ranging from 15 to 870 cells/mm3 and 6 with no CSF obtained but who were nonetheless treated empirically), and one was diagnosed with osteomyelitis and treated parenterally for 43 days. Exclusions for other criteria had occurred at each site prior to data submission and were not quantified. Baseline characteristics of the remaining 251 patients are presented in table 2. Escherichia coli was the most common organism (89.6% of cases).
The mean±SD duration of parenteral antibiotics was 7.8±4 days, and the mean LOS was 7.4±4 days. There was substantial variability in the duration of parenteral antibiotics (figures 1 and 2), with the clustering effect of institutions accounting for a third of it (ICC 0.33, 95% CI 0.18 to 0.43) in the empty model. In the full model, the ICC dropped minimally (0.31, 95% CI 0.17 to 0.37), suggesting that clustering was not due to the included patient-level predictors. The association between clinical variables and the duration of parenteral antibiotics is reported in table 3.
Only six infants (2.4%, 95% CI 0.8% to 5.1%) had relapsed UTI caused by the same organism, four of whom were readmitted to the hospital. No infants had a relapsed bacteraemic UTI, four infants (1.6%) had a recurrent UTI within 30 days caused by a different organism (all Enterococcus spp.) and six infants (2.4%) were readmitted to the hospital. There was no difference in the duration of parenteral antibiotics between infants with and without relapsed UTI caused by the same organism (8.2 vs 7.8 days, p=0.81). Voiding cystourethrogram (VCUG) abnormalities were more common in infants with a relapsed UTI caused by the same organism (100% vs 32%, p=0.001). All of the VCUG abnormalities in the infants with relapsed UTI caused by the same organism involved vesicoureteral reflux, ranging from grades I to V. Otherwise, there were no significant differences between infants with and without relapsed UTI caused by the same organism.
All six readmissions went to the paediatric ward; four readmitted infants had UTI caused by the same organism, one infant had UTI caused by a different organism and one infant had bacteraemia caused by the same organism (Enterobacter cloacae) but a negative urine culture. This infant had been treated initially with 11 days of parenteral antibiotics. None of the included infants developed clinical deterioration.
One or more repeat blood cultures were obtained in 86.9% of patients, and three or more were obtained in 8% of patients. Repeat cultures were positive (same organism as initial infection) in 11/315 (3.5%) cultures and in 8/218 (3.7%) infants who had repeat blood cultures obtained. Only 2/8 infants (0.9% of all infants with repeat cultures) had a positive repeat culture despite being afebrile for at least 24 h prior to the repeat culture being drawn.
This investigation represents the largest study to date to analyse parenteral antibiotic duration and outcomes of young infants with bacteraemic UTI. Infants with this condition were provided durations of parenteral antibiotic that were quite variable, and this variability was only partially explained by factors relating to the infant and his or her clinical course. None of the 251 infants had a relapsed bacteraemic UTI. Few infants had relapsed UTIs, and no babies deteriorated during treatment. The only clinical variable associated with relapsed UTI was an abnormal VCUG, suggesting that relapsed UTIs are more related to underlying genitourinary abnormalities rather than insufficient treatment of the initial UTI.
We chose to analyse the association between parenteral antibiotic duration and relapsed UTI because we anticipated that other outcomes such as relapsed bacteraemic UTI and/or clinical deterioration would be extremely rare events (indeed, none of these outcomes occurred in any of the included infants), and that relapsed UTI would represent the best available indicator of possible undertreatment of the index infection. While information on renal scarring might have provided additional information regarding outcomes, dimercaptosuccinic acid (DMSA) scans were rarely done in our cohort. However, multiple prior trials have demonstrated no association between the route or duration of antibiotic therapy and future scarring in paediatric patients with UTI.2 ,18–21 Although these trials did not focus specifically on young infants or bacteraemic UTI, there is no published evidence demonstrating that young age or bacteraemia increase the risk of renal scarring in children with UTI. In fact, in one trial, 0/13 (0%) patients with bacteraemic UTI had evidence of renal scarring on a 6-month follow-up DMSA scan compared with 26/288 (10%) of the children with UTI and a negative blood culture.2 Children in that trial received either 0 or 3 days of intravenous antibiotics.
There are no available published recommendations on the duration of parenteral treatment for bacteraemic UTI. Data on parenteral antibiotic use can be gleaned from existing studies and suggest that prolonged parenteral antibiotic courses offer no clear benefit. However, these studies are limited by small numbers, lack of inclusion in some cases of infants <1 month or lack of focus on clinical outcomes.2 ,6 ,8 ,11–13 The largest investigation, by Honkinen et al,11 involved 134 children with bacteraemic UTI (66% of whom were <3 months) and reported a mean duration of intravenous antibiotics of 6.3 days, 2.5 days longer on average than the duration provided to age-matched and gender-matched children with non-bacteraemic UTI. Although the authors report that ‘all patients made an uneventful recovery’, there is no mention of relapses. The absence of published guidelines or solid evidence on this issue likely explains some of the variability seen in our study.22
According to our data, there were five independent predictors of parenteral antibiotic duration. Increased age, year of blood culture and female gender were associated with a shorter duration, but these effects were small (−0.2 days for each additional week of age, −0.2 days for each additional calendar year and −0.9 days for female gender). A non-E. coli organism and a positive repeat blood culture during acute treatment had a larger impact on duration (average of an additional 2.2 and 3.5 days, respectively), but only 34 infants (13.5%) had one or both of these risk factors. Some variability can also be explained by different institutional practices, as evidenced by the ICC remaining above 0.3 even after addition of patient-level predictors. The average duration of parenteral antibiotic in the institution with the longest mean was over twice as long as the institution with the shortest mean (figure 2). Such inter-institutional variability for intravenous antibiotic duration has been described previously for UTI13 and osteomyelitis23 and suggests that there is opportunity to standardise care while potentially reducing waste and harm.
The benefits of continued intravenous antibiotic administration and hospitalisation in an infant who has recovered clinically are largely theoretical and were not evident in our data, but the risks are more readily apparent. Peripheral intravenous catheters are a leading source of pain and distress for hospitalised children,24 and infiltrates can cause serious complications.25 If peripherally inserted central catheters are placed, complications occur in ∼20% of cases and are even more common in infants.26 Hospitalised infants are at risk for nosocomial infections such as seasonal viruses.27 Finally, hospitalisation costs are substantial—in 2003, the mean cost of infection-related hospitalisations in infants (mean LOS 3.5 days) was $4000.28 A decrease in the duration of parenteral therapy from, say, 14 days to 3 days (two common durations in our study) could lead to a significant reduction in resource use and iatrogenic harm.
Most (86.9%) children had repeat blood cultures obtained during acute treatment. The high percentage of repeat cultures suggests most practitioners believe that a positive blood culture warrants documentation of clearance. However, repeat cultures were positive in only 3.7% of patients, and only two infants (0.9% of all infants who had repeat blood cultures obtained) were afebrile during the 24 h period prior to the positive repeat blood culture being drawn. Therefore, the overall yield of repeat blood cultures is low, and clinical parameters such as defervescence may be useful in considering whether or not to repeat a blood culture. Detection of a positive culture, though uncommon, was associated with prolonged parenteral antibiotics. However, it is unknown whether this change in management affected clinical outcomes.
This study has several limitations. First, it is possible that infants may have been treated for relapses elsewhere, which would lead to an underestimation of the risk of relapse. Second, severity of illness can be difficult to capture in a retrospective chart review, and unmeasured variables such as time to defervescence, socio-economic factors or concerns surrounding parental reliability or medication compliance may have affected the duration of parenteral therapy. However, the predilection for certain numbers of days of therapy, such as 3, 7, 10 and 14 days (figure 1), suggests that practitioners often choose fixed durations rather than response to therapy to guide decisions on duration. Third, we did not collect data on oral antibiotic duration or the use of prophylactic antibiotics, though the standard practice at all of the participating institutions is to provide a minimum of 7–10 days total antibiotics.
Finally, we are unable to conclude that infants who received long parenteral courses would have fared as well with shorter courses. While a randomised trial would likely provide important insight on the appropriate treatment course, the sample size required to demonstrate a small but potentially meaningful difference would be very large (eg, nearly 1000 patients to demonstrate a statistically significant difference between a 5% vs 10% relapse risk). In the absence of such a trial, this relatively large retrospective study should lead us to question the use of prolonged courses of parenteral antibiotics in all patients with bacteraemic UTI.
In infants <3 months with bacteraemic UTI, the duration of parenteral antibiotic therapy is variable and only partially explained by clinical characteristics. Relapse is rare and is not associated with the duration of parenteral antibiotic therapy, suggesting that shorter courses may be safe and appropriate for generally healthy infants with bacteraemic UTI who have recovered clinically.
ARS had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. None of the authors has any relevant financial interests, activities, relationships or affiliations. We thank Brian P. Lucas, MD, MS, for his assistance with methodology and statistical analysis. We thank the Pediatric Research in the Inpatient Setting (PRIS) network (http://www.prisnetwork.org/) for playing an instrumental role in the formation of this collaboration and for their ongoing dedication to research on hospitalised children.
Twitter Follow Clifford Chen at @pedshospitalist and Alan Schroeder @safelydoingless
Contributors ARS conceptualised and designed the initial study, coordinated data collection from all of the sites, carried out the analysis and interpretation of data, drafted the manuscript and approved the final manuscript as submitted. MWS, PWC, HKR and TLG conceptualised and designed the initial study, coordinated or co-coordinated data collection at each of their respective sites, reviewed and revised the manuscript, and approved the final manuscript as submitted. EAB coordinated data collection for all of the participating Pediatric Research in the Inpatient Setting (PRIS) sites, reviewed and revised the manuscript, and approved the final manuscript as submitted. MB-S, CNC, JF, VL, RCE, KEJ, MM and KEW coordinated data collection at each of their respective sites, reviewed and revised the manuscript, and approved the final manuscript as submitted.
Competing interests None declared.
Ethics approval Institutional Review Board from all participating institutions.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement Deidentified unpublished data are available upon request from ARS.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.