Imaging Practices and Implications in Young Infants With Urinary Tract Infection

Using the electronic health record, we identified infants aged ≤60 days with UTI who were hospitalized at our tertiary care children’s hospital from January 1, 2013 through May 7, 2018. The study was approved by the institutional review board; informed consent was waived for this retrospective study. We used encounter diagnosis codes to identify infants ≤60 days of age who had fever or urinary tract infection. Of these, we included only patients who were hospitalized (either on the acute care floor or ICU), had a urine culture, and had a final discharge diagnosis of UTI. Laboratory and/or microbiological evidence of UTI was not required for inclusion. Exclusion criteria included treatment of a separate source of infection requiring antibiotic therapy (eg, bacterial meningitis), chronic severe systemic disease (including neuromuscular disorder, congenital or acquired immunodeficiency, malignancy, quadriplegia or paraplegia, hydrocephalus or neurologic malformations, and organ transplant), infants in whom urinary tract foreign body exposure might have introduced infection (eg, indwelling genitourinary devices, genitourinary tract instrumentation, or surgical intervention within the preceding 7 days), or a diagnosis of hospital-acquired UTI.¹⁶  Data were included from the first admission and not subsequent readmissions, if applicable.

Variables were extracted from the chart by reviewing documentation in notes from the admission history and physical and the emergency medicine provider’s documentation (Appendix A in Supplemental Information). Four pediatric hospitalists performed chart reviews. Questions about data extraction were settled by group consensus; interrater reliability was not assessed. For the purposes of analysis, an undocumented subjective variable in the history or physical exam was considered a negative response. Vital signs were recorded based on the first data documented. Imaging both during and following admission was abstracted.

All inpatient RUS originally interpreted as abnormal were reread by a single study radiologist, and findings were characterized using the Society for Fetal Urology (SFU) criterion-based hydronephrosis grading system. This is a commonly used system for grading urinary tract dilation and is the accepted system at our institution, despite being inconsistently used.^17–19  The study radiologist recategorized ultrasounds originally interpreted as abnormal as having a minor abnormality if they were SFU grade 1–2 (not considered clinically significant) or if they had isolated pyelonephritis without a structural abnormality. The study radiologist classified ultrasounds as showing a significant abnormality if they met criteria for SFU grade 3–4 hydronephrosis, as these grades are commonly considered to be clinically significant in a postnatal ultrasound, or if they had other anatomic renal defects (eg, urinary tract duplication).²⁰ 

Associations between demographic and clinical characteristics and imaging outcomes were evaluated. Age was categorized as ≤28 days and 29–60 days; duration of fever was categorized as ≤48 hours and >48 hours; bacteremia was categorized as present or absent; ultrasound timing was categorized as ≤2 days and >2 days of hospitalization; and ill appearance on admission (defined as the presence of irritability, lethargy, poor capillary refill, or poor tone on exam) was categorized as present or absent. The demographics, clinical characteristics, and clinical outcome variables were summarized using descriptive statistics, which are frequencies and proportions for categorical variables, and medians and interquartile ranges (IQR) for continuous variables. The ability of NCB and CB inpatient ultrasound interpretation methods to predict findings on follow-up outpatient RUS and VCUG was assessed by calculating the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy ([true positives + true negatives]/total) for each inpatient test type and outpatient outcome. For this analysis, CB readings were categorized as normal or minor abnormality versus significant abnormality. Separate analyses were performed for outpatient VCUG categorized as normal versus any abnormality and normal or grade 1–2 versus high-grade (3–5) reflux. For study clinical characteristics association with imaging outcomes and for antibiotic prophylaxis association with imaging during admission, a χ² test or Fisher’s exact test was used for the statistical analysis. The McNemar’s test was used to compare sensitivities and specificities between 2 diagnosis tests. Statistical software SAS 9.4 was used for all the analyses. A P value <.05 was considered statistically significant.

One hundred ninety-three subjects were included in this study.¹⁵  Sixty-nine infants (35.8%) were ≤28 days. Participants were 54.4% male; 59.6% were White, 18.7% Black, and 21.7% other or unspecified race. One hundred eighty infants (93%) had a renal ultrasound while inpatient (Table 1). One hundred fourteen (63%) of the inpatient RUS were initially interpreted as abnormal by the radiology group using the NCB system. After applying the criterion-based SFU grading system and recategorizing the imaging results, 85 ultrasounds (75% of the total initially read as abnormal) were determined to have a minor abnormality, and 29 (25%) were categorized as having a significant abnormality (Fig 1).

Of the 85 infants with a CB minor abnormality, 25 (29%) did not have follow-up imaging. Of the 60 infants who had some type of follow-up imaging, 15 (18% of the total infants with a CB minor abnormality) studies were abnormal, and 45 (53%) were normal. Forty-five of these 60 infants had a follow-up VCUG performed, of which 7 (16%) showed low-grade reflux and 6 (13%) showed high-grade reflux or other serious abnormality. Thirty-one of the 60 infants had a repeat RUS performed, of which 1 (3%) was abnormal (Fig 1).

Of the 29 infants with a CB significant abnormality on initial ultrasound, 3 (10%) did not have follow-up imaging. Of the 26 who had some type of follow-up imaging, 20 (69% of the total infants with a CB significant abnormality) were abnormal and 6 (21%) were normal. All 26 had a follow-up VCUG performed, of which 10 (38%) were normal, 2 (8%) showed low-grade reflux, and 14 (54%) showed high-grade reflux or another serious abnormality. Thirteen of the 26 infants also had a repeat RUS in addition to VCUG in follow-up, of which 7 (54%) were abnormal (Fig 1).

Results of the univariable analysis with clinical characteristics and imaging outcomes are shown in Table 2. Infants with a non-E. coli pathogen were more likely to have a CB significant abnormality on ultrasound compared with infants with only a CB minor abnormality (P = .007). None of the other clinical characteristics examined were significantly associated with imaging findings.

Results of the univariable analysis with inpatient imaging findings based on NCB reads and clinical outcomes are shown in Table 3 and Supplemental Table 7. There were no significant statistical associations between normal or abnormal imaging and any of the relevant clinical outcomes, except for antibiotic prophylaxis on discharge (P = .0006). Normal and abnormal imaging reads using a NCB system were not associated with any follow-up imaging outcomes. When the criterion-based SFU system was applied to compare the 3 groups, a significant abnormality was associated with overall abnormal follow-up imaging (P < .0001), abnormal follow-up renal ultrasound (P = .0004), high-grade reflux on VCUG (P = .0047), and discharge antibiotic prophylaxis (P < .0001). There was no statistically significant difference between normal and minor abnormalities on inpatient imaging (Table 4 and Supplemental Table 8).

Inpatient NCB RUS interpretations had higher sensitivity than CB readings for predicting abnormalities on follow-up outpatient RUS or VCUG (94% to 100% compared with 52% to 88%), however specificity was lower (5% to 13% versus 79% to 90%) and test accuracy was <50% for all NCB measured follow-up studies (Table 5). Likewise, the PPV for NCB abnormalities was low at 18% to 41% with a higher NPV at 75% to 100%. CB abnormalities had a reasonably high NPV (72% to 97%), better PPV (54% to 77%), and higher accuracy (68% to 85%) (Table 5).

Recommendations made by physicians on the discharge summary for follow-up imaging are shown in Table 6. About 9% of physicians recommended follow-up imaging if the infant had a normal inpatient renal ultrasound, 51% if there was a minor abnormality, and 69% if there was a significant abnormality.

Using a standardized, CB SFU grading system to classify ultrasounds initially read as abnormal as either minor or significant abnormalities resulted in recategorization of three-quarters of the ultrasounds originally interpreted as abnormal to the category minor abnormality. This CB grading of inpatient RUS allowed more accurate identification of patients with abnormalities on follow-up renal imaging. Although a NCB abnormal result was not associated with abnormal follow-up imaging, a significant CB abnormality was strongly associated with abnormal follow-up imaging. Similarly, there was no significant difference between imaging follow-up outcomes when comparing an infant with a CB minor abnormality to one with a normal inpatient ultrasound. Therefore, delineating abnormalities using the SFU grading system allowed us to better identify infants who would go on to have an abnormal urinary tract on follow-up imaging.

The NCB abnormal ultrasounds included ultrasounds with both CB significant and minor abnormalities, so it is not surprising that the sensitivity was higher for all NCB results and the specificity was higher for all CB results. The NPV of the CB ultrasound reads was similar to NCB reads for predicting abnormal follow-up imaging (72% to 97% vs 75% to 100%). In contrast, the NCB abnormal reads were not particularly informative about the infant’s urinary tract, with a poor PPV (18% to 41% vs 54% to 77%) and low specificity (5% to 13% vs 79% to 90%) compared with CB abnormalities. Most importantly, there was no statistically significant association with the main outcome of interest, abnormal follow-up urinary tract imaging, with an NCB abnormal ultrasound read, whereas a CB abnormal read was significantly associated with abnormal follow-up imaging.

One-third of the infants with CB minor abnormalities in our study did not receive any follow-up imaging. Although some of those infants may have been lost to follow-up for logistical reasons or may have obtained imaging outside of our system, we suspect that many did not get any follow-up because there are no guidelines for management of these minor abnormalities, leaving providers to determine the need for further imaging on an individual basis. The data from our hospitalist group supports this notion, as only half of physicians had any recommendation for follow-up imaging at discharge for infants with minor abnormalities on their inpatient ultrasounds. In comparison, two-thirds of hospitalists recommended follow-up imaging in infants with significant abnormalities, suggesting that they were more confident in stating a follow-up plan for those patients.

Almost half of the infants with follow-up imaging had a repeat ultrasound done. However, although ultrasonography is less invasive and does not expose patients to radiation, it has relatively poor sensitivity for VUR.¹  Arguments against renal ultrasound as a way to identify VUR in infants with a first-time UTI include the relatively poor sensitivity and specificity of the modality, the dynamic nature of reflux precluding its consistent detection, and the limited effect of sonographic findings on clinical management, especially in an era in which the utility of antibiotic prophylaxis is being heavily questioned.²¹  Of the infants with CB minor abnormalities on their original ultrasounds, only 1 of the repeat RUS revealed any pathology. The majority of infants with CB significant abnormalities on initial RUS had persistent abnormalities on repeat imaging. The NPV of a CB ultrasound result for predicting follow-up RUS findings was excellent (97%), suggesting that repeating ultrasounds in infants with CB “normal or minor abnormality” on inpatient RUS does not provide further diagnostic value and should not be pursued.

Although the NPV of a CB ultrasound interpretation for predicting high-grade reflux on follow-up VCUG was quite high (87%), about 13% of patients with a minor abnormality on CB inpatient RUS interpretation showed high-grade reflux on follow-up VCUG, so we feel most comfortable concluding that VCUG would be appropriate in follow-up of even a minor (SFU grade 1–2) abnormality on inpatient RUS. Larger studies will need to be done to better define the appropriate follow-up for infants with minor abnormalities, as our sample size was too small to lead to a definitive recommendation, but we hope that our findings stimulate interest in this question.

Several clinical variables have been associated with abnormal ultrasounds in the setting of UTI. Non-E. coli pathogens have been shown to be present more commonly in infants with abnormal urinary tracts, especially in those with high-grade VUR.^5,6,10,22  Similarly, bacteremia is reported to be more common in infants with abnormal urinary tract anatomy.²²  It is debated whether rates of urinary tract abnormality in younger infants with UTI are higher than in older infants.^5,6,9,23  We examined these clinical factors along with age, duration of fever (as a proxy for response to therapy), ultrasound timing, and ill appearance using univariate analysis and found that only the presence of a non-E. coli pathogen was associated with CB significant abnormalities of the urinary tract. The lack of association with other factors in our findings may be related to our smaller sample size compared with other studies.

Use of antibiotic prophylaxis following UTI is highly debated. Rates of prophylaxis in children under 24 months admitted with a first-time UTI have been reported to be 15%.²³  Antibiotic prophylaxis is currently recommended by the American Urological Association for all infants <1 year of age with VUR of any grade and a history of febrile UTI, given the greater morbidity from recurrent UTI in this specific population.²⁴  Several large studies have shown antibiotic prophylaxis to be of benefit in prevention of UTI, with a risk reduction of UTI recurrence of up to 50% (number-needed-to-treat ranging from 8–14) but no effect on hospitalization rates or subsequent renal scarring.^25,26  The degree of reflux may certainly play a role, as antibiotic prophylaxis in children with lower-grade VUR has been shown in some studies not to prevent UTI recurrence.^27,28  Using CB ultrasound readings with SFU grades could possibly allow for better targeted antibiotic prophylaxis, which had a relatively high use in our population. Of those discharged on prophylaxis who had inpatient imaging, 14% and 38%, respectively, had normal findings or only minor CB abnormalities. The lack of use of a CB grading system for ultrasounds may have contributed to excess antibiotic prophylaxis in these infants, although our study was not designed to answer this question definitively.

The SFU system developed in 1993 is 1 of several grading systems designed to more quantitatively describe ultrasound abnormalities in the fetal or neonatal kidney by classifying degrees of renal collecting system dilation.¹⁷  Compared with other grading systems (eg, the Urinary Tract Dilation classification), the SFU system is thought to over diagnose pathology; SFU grade is 1 considered to be normal in most cases with Urinary Tract Dilation classification.^18,29  Although it is the most commonly used system, there is much variability with only 19% of institutions reporting use of a standardized system.¹⁹  Between urologists and radiologists, there are significant differences: 80% of urologists report utilizing the SFU system, compared with only 36% of radiologists.^19,28  The majority of radiologists (66%) prefer using descriptive reads, as is the case in our institution, although the large majority (97%) report that having a unified system would be helpful.²⁸  To our knowledge, the clinical consequences of these varied approaches have not previously been reported. We hope that our findings lead to larger studies to investigate how the use of a standardized system can more specifically characterize abnormalities and improve the management of infants admitted with UTI.

There are several limitations to our study. This was a single-center study at an academic institution, which limits the generalizability of the imaging practices and indications for prophylactic antibiotic use. In addition, our sample size was small. Some of the variables were dependent on documentation practices and, therefore, may have been inaccurate. Because a strict cutoff for colony-forming units and the presence of leukocyte esterase for defining UTI are debated and because we were most interested in the practices of providers after diagnosis of a UTI, we did not limit our cohort with regards to urinalysis or microbiologic data, and it is possible that some of the infants did not have a UTI. The imaging studies were reviewed by only 1 independent radiologist. It is possible that some of the follow-up imaging was done outside of our system and was therefore not accounted for in our study.

The majority of young infants admitted with UTI had a renal ultrasound while hospitalized, and nearly two-thirds of these were initially identified as having a urinary tract abnormality. The application of a CB system of ultrasound evaluation allowed classification of many of these abnormalities as minor. Although neither a NCB abnormality nor a CB minor abnormality was associated with abnormal follow-up imaging, a CB significant abnormality was strongly associated with abnormal follow-up imaging and high-grade reflux on VCUG. A standardized approach to the interpretation of inpatient RUS could limit the need for follow-up RUS, optimize the use of follow-up VCUG, and possibly reduce overuse of antibiotic prophylaxis.

We thank Stacey Hay and Kim Gajewski for database design and build, Brett Bordini for assistance with revisions of the manuscript, Kelsey Porada for assistance in preparation of the manuscript.