Impact of Early Oral Antibiotic Therapy in Infants With Bacteremic Urinary Tract Infections Free

We performed a retrospective cohort study involving 22 hospitals which make up a health care network in 2 western states. The study was approved with a waiver of informed consent by the institutional review board.

We identified all infants ≤ 90 days of age with gram-negative BSIs from the electronic data warehouse from January 1, 2002 through January 31, 2020. Infants were included if the pathogen identified in blood was also isolated from a urine culture collected within 24 hours of positive blood culture collection, with a minimum of 10 000 colony forming units (CFUs) from a catheterized urine specimen. After extraction of medical information from the electronic data warehouse, medical record review was conducted to confirm eligibility and to obtain more detailed patient information that was collected in a standardized REDCap database (Vanderbilt University). Infants with positive cerebrospinal fluid (CSF) cultures, pretreated CSF with white blood cell count >100, osteomyelitis, central lines, urologic stents, renal abscesses, ventriculoperitoneal shunts, and urinary tract infections occurring during a NICU stay were excluded. We also excluded infants who died during the same hospitalization. We did not exclude infants with preexisting urologic abnormalities or other complex medical conditions since at the age of our cohort many of these infants have undiagnosed vesicoureteral reflux (VUR) and potentially other urologic abnormalities. Although these factors may increase risk of recurrence, this is unlikely to impact optimal route of therapy.

We collected information on patient demographics, hospital caring for the patient, clinical variables (including fever duration and physical exam findings on presentation), laboratory data, microbiology results including sensitivities, IV antibiotic duration, total antibiotic duration, LOS, and reason for readmission and ED revisit within 30 days of discharge. Covariates of interest included sex, prematurity (< 37 weeks’ gestation), age (<30 days of age), preexisting complex medical problems, preexisting genitourinary conditions, ill-appearance, duration of fever, ICU stay, subsequent positive blood cultures, and pretreated lumbar punctures.

Patients were considered ill-appearing if the use of ‘lethargic,’ ‘toxic,’ ‘septic,’ or ‘sick or ill-appearing’ was mentioned in the first physician note for that admission or encounter. We identified patients who had other comorbidities or preexisting urologic conditions by reviewing the patient chart. For each individual patient, we also collected the highest and lowest vital sign and laboratory values for that hospitalization including heart rate, temperature, respiratory rate, white blood cell (WBC) count, and C-reactive protein (CRP) when data were available. We were most interested in the duration of fever which we classified dichotomously as continued fevers >38.0°C for >24 hours and >48 hours after the first positive blood culture. In regard to microbiology, we identified the number of consecutive positive blood cultures, the organisms identified, and the susceptibility results for each blood culture. IV antibiotic durations were calculated by the date of the first and last dose administered and reported as days of therapy with 1 day of therapy referring to the first calendar day an antibiotic was given. We calculated the total length of antibiotic treatment by reviewing the outpatient prescriptions generated as well as reviewing the discharge summary documentation. If a patient was discharged from the hospital with parenteral antibiotics via PICC line, this was included in the total IV antibiotic duration. We also identified those patients who were diagnosed with VUR within 3 months of discharge by reviewing available voiding cystourethrogram results. If a patient did not have a voiding cystourethrogram, diagnosis was labeled as ‘unknown.’

In 2014, the 289-bed, stand-alone children’s hospital (CH) standardized the care of infants with gram-negative BSI and UTI to receive oral antibiotic transition at discharge or earlier if appropriate oral options existed.¹⁰  Because the intervention was likely to increase the occurrence of early oral conversion at CH, we reported the rate of early oral conversion over time for CH and non-CH facilities.

The exposure of interest was early oral conversion. Patients were categorized into early oral conversion and long-course IV therapy. We defined early oral conversion as oral transition after 4 days or less of IV therapy. We selected the 4-days duration as the definition for several reasons, including the bimodal distribution of IV antibiotic duration in our cohort (Fig 1) and precedent in an adult study.¹¹ 

The primary endpoints were LOS in days of index admission and a composite of ED revisit or hospitalization with new or recurrent UTI within 30 days of discharge from the index hospitalization, described as 30-day revisit in the rest of the text. New UTI was defined as a new pathogen identified in urine culture and recurrent UTI was defined as same pathogen identified from new culture.

Patient demographics and outcomes were summarized using counts and percentages for categorical variables and medians and interquartile ranges (IQRs) for continuous variables. We compared variables between early oral conversion and long IV therapy groups using Wilcoxon rank tests for continuous variables because of distribution skew and χ² or Fisher’s exact tests for categorical variables.

We anticipated that there would be inherent differences between patients who received early oral conversion and long IV therapy. We hypothesized that patients continuing long IV therapy would be more likely to be sick-appearing, more likely to be younger, more likely to have persistent bacteremia, more likely to be seen at non-CH facilities, more likely to be admitted before 2014, and more likely to have VUR. Therefore, we calculated propensity scores using multivariable logistic regression model in which the dependent variable was short-course IV therapy; the independent variables were significant in univariable analyses at a level of 0.10. Matched weights were constructed which assign higher weight to subjects with greater equipoise, ie, subjects who have similar probabilities of achieving long- versus short-course IV therapy.¹²  We checked that standardized differences were <0.2 after matched weighting for all variables included in the propensity score model, an indication that covariate balance has been achieved between exposure groups.¹³  Outcomes models comparing early oral transition to long IV therapy were built using survey-weighted regression with our matched weights implemented with svyglm from the survey package in R.¹⁴ 

We conducted noninferiority tests using univariable survey-weighted regression with an inverse-Gaussian distribution for LOS because of distribution skew and logistic regression for 30-day revisit. The noninferiority margin was set a priori on the basis of clinical judgement of no more than a 10% increase in LOS and 30-day revisit rate, ie, using a noninferiority margin of 1.1. Hypotheses were tested using a 1-sided test with a significance level of 0.05 which is equivalent to comparing the upper bound of a 2-sided 90% confidence interval (CI) with our noninferiority margin.¹⁵  Regression coefficients (β) for both outcomes were exponentiated (e^β), resulting in ratios or relative effects for LOS and odds ratios (ORs) for 30-day revisit. We were limited to a univariable logistic regression model for 30-day revisit because of a low event rate. Statistical analyses were implemented by using R version 3.6.0 (Vienna, Austria).

We identified 245 children ≤ 90 days of age with gram-negative pathogens in both blood and urine. The following patients met exclusion criteria and were not included in analyses: 35 children managed in NICUs, 16 patients with primary bacterial infections other than UTI (eg, meningitis, osteoarticular infections, etc.), 4 patients with pretreated CSF and CSF WBC counts >100/µl, 1 patient with a ventriculoperitoneal shunt, 2 patients with ureteral stents, 1 patient with a renal abscess, 7 with urine counts <10 000 CFU/mL, 1 patient with a polymicrobial blood stream infection, and 4 patients with inadequate discharge documentation.

We included 174 patients, including 5 patients with urine CFUs/mL between 10 000 and 50 000 (4 with urinalysis positive for leukocyte esterase ≤ and/or >5 WBC/high-power field and 1 without urinalysis performed). Urine LE was present (trace or more) in 100% of the 163 infants with LE tested. Urine nitrites were present in 78 (47%) of the 167 infants tested, bacteria were observed on microscopy in 146 (92%) of 159 infants tested, and > 5 WBC/high-power field were observed on microscopy in 143 of 159 infants tested (90%). The median duration of IV antibiotic therapy was 8 days (IQR 3, 14) (Fig 1).

Of the cohort, 73 patients received early oral conversion and 101 received long IV therapy. Patients receiving long IV therapy were younger, more likely to have prolonged bacteremia, less likely to be seen at CH, and more likely to have Grade 4 or 5 VUR diagnosed up to 3 months after date of bacteremia (Table 1). Higher rates of early oral conversion occurred in later admission years (Fig 2). Covariates in univariable analysis associated with likelihood of early oral conversion included first seen at CH post-2014 intervention, ethnicity, days of life at bacteremia, calendar year of admission, grade of VUR, and persistent bacteremia. Absolute standardized differences both unadjusted and after weighting are available in Table 2.

The most frequently received outpatient antibiotics for patients receiving early oral conversion were cefdinir (18 of 73; 25%), cephalexin (17 of 73; 23%), and amoxicillin (13 of 73; 18%). The most frequently received outpatient antibiotics for patients receiving long IV therapy were ceftriaxone (59 of 101; 58%) and other IV cephalosporins (12 of 101; 12%) including cefazolin, cefuroxime, cefepime, ceftazidime, and cefotaxime (Table 3). Twelve (12 of 101; 12%) of the patients receiving long-course IV therapy completed treatment inpatient; all were well-appearing at admission, and 9 (9 of 12, 75%) of these patients had no fevers beyond 24 hours, and 11 (11 of 12, 92%) had no fevers beyond 48 hours. Oral antibiotics were used to complete a portion of the total antibiotic duration in 15 (15 of 101, 15%) infants receiving long IV therapy.

In simple univariate analysis, the median LOS in infants receiving early oral conversion was 2 days (IQR 2,3) compared to 4 (IQR 3,5) days in the long IV course group (P < .001). In the weighted analysis, early oral conversion was noninferior to long-course IV therapy with a 57% shorter LOS (e^β = 0.43; 90% CI: 0.35, 0.53, P < .001).

Thirty-day revisits occurred in 3 (4.1%) patients receiving early oral conversion and 2 (2.0%) patients receiving long IV courses (P = .65). Of these revisits for UTI, new UTI occurred in 1 patient receiving early oral conversion, and 1 patient receiving long IV course; both patients were initially infected with Enterobacter cloacae, and subsequent urine cultures were positive for Enterococcus faecalis in 1 child and Stenotrophomonas maltophilia in the other. Recurrent UTI occurred in 2 (2.7%) patients receiving early oral conversion and 1 (1%) in the long IV course (P = .57). All recurrent infections were Escherichia coli. In the weighted analysis, the OR of 30-day revisit in the early oral conversion was 4.22 (90% CI: 0.83, 21.34, P = .91) compared to long-course IV therapy. Thus, we were not able to show noninferiority of early oral conversion compared with long-course IV therapy in terms of 30-day revisit rates. None of the revisits were associated with ICU readmission, bacteremia, meningitis, mechanical ventilation, or vasopressor support.

Early oral conversion in infants with gram-negative BSI and UTI was associated with a 50% shorter LOS compared to prolonged IV therapy. Overall, 30-day revisits for UTI were infrequent in both groups; however, we were unable to demonstrate noninferiority of early oral conversion compared with prolonged IV antibiotics.

Our observation that prolonged IV administration was associated with longer durations of hospital LOS is similar to that observed in previous studies of UTI in infants.^1,3  Early oral conversion has been associated with decreased hospital LOS in other diseases such as community acquired pneumonia¹⁶  and pediatric bone and joint infections.¹⁷  It is plausible that this association is causative. Given the increased burden of OPAT compared to oral alternatives,⁹  families may opt to continue care in hospital. OPAT therapy also requires coordination with insurance and home health companies before discharge, which takes time and is often not available on weekends, whereas oral prescriptions are available conveniently at local pharmacies open 24 hours daily 7 days a week. It is possible that unmeasured confounding variables such as ability to tolerate feeds may have influenced LOS. However, the following argue that intended IV duration likely contributed to the hospital LOS: the majority of our cohort were clinically well-appearing febrile infants; the durations of fever were not different between treatment groups; there was a bimodal distribution of the duration of intravenous antibiotics; and the rate of early oral conversion increased with time.

Although we were unable to determine noninferiority, rates of recurrence and new UTIs were low in both groups of patients. A recently published living review of treatment in infants < 2 months with UTI performed a subgroup analysis in those with bacteremia.¹⁸  They identified 54 patients receiving IV duration of ≤ 3 days compared to 304 receiving > 3 days with no difference in UTI recurrence observed between the 2 groups in random effects model (OR: 0.91; 95% CI: 0.22 to 3.8). Larger sample sizes are needed to test for differences in 30-day revisits. Early oral therapy may still be acceptable to reduce hospital LOS, prevent PICC complications, decrease costs, and decrease impact on caregiver quality of life.¹⁷  Risk of recurrence or reinfection has been linked to vesicoureteral reflux grade,^19,20  which may be a more important driver of 30-day outcomes than the choice of antibiotic route.

Our findings have implications for the treatment of well-appearing febrile infants. All 163 infants with gram-negative bacteremia and UTI in our cohort with urine LE tested had a positive LE result. The near-perfect sensitivity of the urinalysis in infants with gram-negative bacteremic UTI has been observed in previous studies.^21,22  The high sensitivity of UA in this setting coupled with the low rate of 30-day revisits in infants with early oral transition may provide confidence for clinicians to follow AAP febrile infant guidelines including early oral conversion, earlier discharge, and home observation for select well-appearing febrile infants.²³ 

Our study has limitations. Few 30-day revisits occurred in our population, limiting the power to detect differences between early oral conversion and long IV therapy. Adherence to therapy was unable to be measured and may have contributed to 30-day revisit rates. There was potential for misclassification of infants if recurrent or new UTIs were managed in the outpatient setting outside of our integrated health care network; however, given the age and previous history of bacteremia, patients would be more likely to present to acute care settings. Classification of recurrent UTIs was based on finding the same pathogen in a new urine culture. It is possible that some recurrences were truly new infections. Including infants with preexisting urologic abnormalities may have changed the likelihood of 30-day revisit; however, VUR and other abnormalities are often unknown at this age and were also included in the cohort. We did not evaluate nor describe the use of prophylactic antimicrobial agents used to prevent recurrent infections in infants with high-grade VUR.

This study suggests that early oral conversion for infants with gram-negative bacteremia and UTI results in decreased hospital LOS. Recurrence is rare. Prospective studies would be helpful to corroborate these findings.

This investigation was supported by the University of Utah Population Health Research (PHR) Foundation, with funding in part from the National Center for Advancing Translational Sciences of the National Institutes of Health under award UL1TR002538. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.