Follow-Up Blood Cultures in Young Infants With Bacteremic Urinary Tract Infections

We performed a retrospective cohort study of infants with Gram-negative bacteremic UTI. We extracted data from 22 hospitals that make up Intermountain Healthcare. Patients were identified by using a microbiology database with manual chart verification. We included infants <90 days of age with concomitant Gram-negative bacteremia and UTI with the same pathogen between January 2002 and January 2020. Infants with culture proven meningitis, osteomyelitis, central line infection, and infections occurring during NICU hospitalization were excluded. The following variables were collected: age, gestation, clinical appearance on presentation, duration of fever, time of initial and repeat blood culture collection, time to blood culture positivity (TTP), number of consecutive positive blood cultures, vesicoureteral reflux (VUR) diagnosis within 3 months after discharge, length of stay (LOS), duration of IV antibiotics, and readmission within 30 days of discharge.

We defined a UTI as a positive urine culture with 50 000 CFU/mL obtained by catheterization.³  We defined TTP as time from culture collection to initial positive blood culture. We defined a positive FUBC as any positive blood culture that was collected between 8 and 96 hours after initial blood culture. Although authors in some studies have defined a positive FUBC as a culture obtained at least 24 hours from initial culture, we included cultures obtained before that time frame because our minimum TTP was 8 hours and only 1 of 154 of FUBCs was obtained before initial blood culture positivity, indicating that these cultures were likely collected in response to the first positive blood culture. We defined contaminants as FUBCs obtained between 8 and 96 hours after initial blood culture that grew a different pathogen that was consistent with commensal organisms per Centers for Disease Control and Prevention definitions.⁹  Clinical appearance was ascertained by manual chart review and infants were categorized as “well appearing” or “not well appearing” on the basis of documentation by the physician. Infants were defined as febrile if temperature was >38.0°C within 24 hours of the first blood culture collection. Temperature in the nursing record was ascertained by either measurement via axillary or rectal thermometer on admission and/or arrival to the ED. When nursing records were unavailable, we collected temperature from the physician documentation.

Patients were assigned to 1 of 3 groups: negative FUBC, positive FUBC, and no FUBC. The primary end point was hospital LOS. Secondary end points included number of FUBCs performed, rate of infants receiving early oral switch defined as ≤4 days of IV antibiotics, IV length of treatment, and all-cause 30-day ED revisit or readmission.

Blood culture practices and FUBCs are summarized by using descriptive statistics. Central tendencies were reported as median and interquartile ranges (IQRs) for continuous variables and as proportion and size for categorical variables. Between group comparisons were performed by using Fisher’s exact test for binomial variables and Kruskal–Wallis rank sum test followed by pairwise comparisons by using Wilcoxon rank test with continuity correction for continuous variables with a 2-sided P < .05 considered statistically significant. A Bonferroni correction was used to account for multiple comparisons for pairwise binomial variables with a 2-sided P < .017 considered statistically significant. All statistical analyses were performed by using R software version 3.5.1 (R Foundation, Vienna, Austria).

We included 174 infants with Gram-negative bacteremic UTI. A total of 153 (88%) patients had at least 1 FUBC. Positive FUBCs occurred in 14 of 153 infants (9%), and 4 of 153 (3%) were contaminants (3 coagulase negative Staphylococcus species and 1 Kocuria species). Repeat FUBCs were obtained for all infants with contaminated blood cultures. In our cohort, 11 FUBCs were drawn to detect 1 true-positive FUBC. The minimum and median TTP for all patients were 8 hours and 14 hours (IQR 12–17), respectively. The median time to FUBC collection from initial positive blood culture in all patients was 19 hours (IQR 15–26). Only 1 of 153 (0.7%) of all FUBCs was collected before the initial positive blood culture and the critical laboratory callback performed, and that patient subsequently had negative FUBC. Median time from initial blood culture becoming positive to FUBC collection was 3.7 hours (IQR 1.7–10.2). FUBCs were collected <24 hours from the initial culture in 102 of the 153 (67%) infants with FUBCs. In addition, we also evaluated the percentage of patients who received a FUBC by year to ensure that practice had not dramatically changed overtime given our small number of patients with positive FUBCs. This remained fairly unchanged (Fig 1).

On this basis of our previously described definitions, 135 infants had negative FUBCs, 18 infants had positive FUBCs and 21 patients had no FUBCs. Clinical characteristics between the groups during hospital course were similar apart from organism distribution, TTP, and VUR diagnosis (Table 1). Infants with positive FUBCs were more likely to have Enterobacter species compared with infants with negative FUBCs (17% vs 1%; P = .012) whereas infants with negative FUBCs were more likely to have Escherichia coli compared with infants with positive FUBCs (95% vs 78%; P = .027). There was a slightly longer time to positive culture in infants without FUBCs (19; IQR 15,21) compared with infants with negative FUBCs (14; IQR 12–17; P = .002). Additionally, although unknown at the initial hospitalization, infants with positive FUBCs were more likely to have VUR diagnosed within 3 months of discharge than infants with negative FUBCs (73% vs 38%; P = .012) and no FUBCs (73% vs 13%; P = .002).

The median hospital LOS in days was longer in patients with positive FUBCs (4; IQR 3–5) compared with infants with negative FUBCs (3; IQR 2–4; P = .043) and infants without FUBCs (3; IQR 1–3; P = .025; Table 2). Early oral switch occurred in 4 infants (22%) with positive FUBCs, 54 infants (40%) with negative FUBCs, and 15 infants (71%) without FUBCs (P = .01). The median duration of IV therapy was 14 days (IQR 6–15) in the positive FUBC group, 10 days (IQR 3–14) in the negative FUBC group and 4 days (IQR 2–6) in the group without FUBCs (P = .003; Table 2).

The composite 30-day revisit end point occurred in 4 infants with negative FUBC, 0 infants with positive FUBC, and 1 infant without FUBC (P = .724). Of the 5 patients with 30-day revisits, 2 had recurrent UTIs with same organism (E coli) and 2 had UTIs with Enterobacter and recurrent UTIs with different organisms (Enterococcus and Stenotrophomonas). All of these patients had recurrence after treatment as well as diagnosis of VUR. None had recurrent episodes of bacteremia or ICU stay. The final readmission was due to a FUBC collected for unclear reasons 7 days after initial positive blood culture. The FUBC grew Viridans Streptococci which prompted readmission and repeat FUBC.

Despite our small sample size, our findings corroborate that positive FUBCs are rare in infants with bacteremic UTIs^2–4,6  and that outcomes are good in patients regardless of FUBCs and their results. Only 14 of 153 (9%) of patients had positive FUBCs, and none of those had a readmission within 30 days or required ICU level of care. The difference in median LOS was 2 days longer in infants with positive FUBCs; however, these infants did not have higher incidence of persistent fever or ill appearance. These findings suggest that clinicians wait for a specified time period after collecting a FUBC to ensure it is negative before discharge, regardless of symptom resolution. Thus, FUBCs seem to restart the clock for discharge readiness and prolong LOS. Infants with negative FUBCs were more likely to have E coli and infants with positive FUBCs were more likely to have Enterobacter species. Whether identification of Enterobacter species is an independent risk factor for persistent bacteremia would require future study. Patients diagnosed with VUR were more likely to have positive FUBCs than patients with negative and no FUBCs; however, this diagnosis is not usually known in this patient population at time of presentation, making this less useful information for clinicians in our opinion.

Our data suggest that healthy infants with E coli as the causative organism are at low at risk for having a positive FUBC, and outcomes are comparable even when the FUBC is positive. Therefore, we propose that these criteria could be used to identify low-risk patients with bacteremic UTI that do not need a FUBC. Although some institutions may not have identification of the pathogen within 24 hours of blood culture collection, the use of molecular diagnostics to help identify pathogens is rapidly increasing.¹⁰  At our institution, for example, rapid diagnostics identify the pathogen in virtually all cases which allows us to identify most pathogens before 24 hours. Nonetheless, repeat blood cultures are likely unnecessary in most cases. In the case that positive FUBCs demonstrate improved outcomes, targeted blood cultures for patients with non–E coli organisms could be of higher yield. Furthermore, FUBCs inevitably lead to increased contamination and false-positive results. As previously stated, of the 18 patients with a positive FUBC, 4 infants grew a commensal organism consistent with a contaminant. Each of these infants had repeat FUBCs performed. Additionally, 1 patient was readmitted for evaluation of a contaminated FUBC collected for unknown reasons immediately before discharge. In previous larger studies, blood culture contamination rates as high as 11% have been reported, also showing a clear association with increased cost, LOS, unnecessary testing, and prolonged IV antibiotic therapy.^7,8  Consequently, FUBCs in infants with bacteremic UTI not only provide unclear clinical benefit, but they are also associated with increased costs and unnecessary interventions.

Should clinicians choose to continue to obtain FUBCs in infants with bacteremic UTI, our study provides some interesting opportunities to standardize the process. In our population, 67% of infants had FUBCs drawn before 24 hours from first blood culture with median time to FUBC being 19 hours. As previously mentioned, these were likely obtained in response to a positive blood culture. Allowing 24 hours of antibiotic exposure before FUBC collection may decrease the rate of positive FUBCs and represent episodes of true persistent bacteremia more accurately. For this reason, we have made the recommendation to obtain FUBCs after 24 hours of antibiotics at our institution if the clinician feels that a FUBC is beneficial; however, our data support that documentation of clearance is not routinely needed.

Our study has some limitations. First, our sample size may not be large enough to detect more subtle differences in clinical characteristics and outcomes between those patients with positive versus negative FUBCs. Likewise, we were unable to address risk factors for positive FUBCs due to the low incidence of such events. Second, we were not able to capture revisit to outpatient clinics for UTI recurrence or revisits to hospitals outside of intermountain system, however, these were likely a small number as intermountain is the principal health care system caring for infants in Utah. Finally, given that we did not break down our analysis by years, it is possible that there could have been some subtle changes over time in IV antibiotic duration and LOS regardless of FUBC collection given our large time frame.

With this study, we found that persistent bacteremia in infants with bacteremic UTI is a rare occurrence and outcomes are good in patients with bacteremic UTI regardless of FUBCs and their results. Infants with positive FUBCs received prolonged courses of IV antibiotics and had an increased LOS. FUBCs also increase the risk of contaminants which lead to increased cost and unnecessary interventions. Although further studies would be needed to determine if FUBCs improve outcomes, our data suggests that the routine practice of obtaining FUBCs is likely unnecessary and can drive unnecessary interventions, especially in infants with known E coli infections. If FUBCs are shown to improve outcomes, a more targeted approach would be higher yield, such as obtaining FUBCs in infants with non–E coli infections.