National Imaging and Antibiotic Practice Patterns in Children Presenting With Urinary Tract Infection to Community Practices: Little Impact From the American Academy of Pediatrics Guidelines?

Abstract

Our objective was to use community-based, national databases to evaluate diagnostic imaging and antibiotic prophylaxis practice patterns before and after the release of the 2011 American Academy of Pediatrics guidelines for acute febrile urinary tract infection. Using the National Ambulatory and Hospital Ambulatory Medical Care Surveys, urinary tract infection encounters were identified for patients aged 2 months to 18 years. Primary outcomes were utilization of antibiotics (as proxy for prophylaxis) and diagnostic imaging during encounters. Weighted multivariate logistic regression models were used to examine the association between time period (before and after 2011) and each of the primary outcomes. Among 8 588 035 weighted encounters, adjusting for covariates, there was insufficient evidence to suggest a difference between time periods for antibiotic utilization (odds ratio = 0.66, P = .12) or diagnostic imaging (odds ratio = 1.16, P = .56). Thus, we did not find evidence of changes in antibiotic utilization or diagnostic imaging practice patterns after the release of the 2011 American Academy of Pediatrics guidelines.

Introduction

Urinary tract infections (UTIs) are frequent in febrile children, occurring in nearly 5% of febrile infants younger than 12 months.¹ Up to 50% of these children have vesicoureteral reflux (VUR) and 55% have findings of pyelonephritis on 99m-Tc dimercaptosuccinic acid scans.^2,3 Patients with VUR were found to have significantly more renal scarring than patients without reflux.³ This scarring can lead to hypertension and chronic renal insufficiency later in life.

The American Academy of Pediatrics (AAP) released guidelines on the management of febrile UTI in children aged 2 to 24 months in 1999. These guidelines included 7 to 14 days of antibiotics, renal bladder ultrasound (RBUS), a strong consideration for either voiding cystourethrogram (VCUG) or radionucleotide cystogram (RNC), and antibiotic prophylaxis until imaging studies were completed.⁴ However, there was poor guideline adherence, with only 39% to 61% of appropriate patients undergoing VCUG and 44% of patients undergoing RBUS in accordance with guideline recommendations.^5–7 The AAP revised their guidelines in 2011; notable changes were recommendation against VCUG after a first febrile UTI.⁸ These guidelines were reaffirmed by the AAP in 2016.⁹

Previous studies have examined the relationship of diagnostic imaging and diagnosis of VUR in academic medical centers after the initiation of the 2011 guidelines. In a single-institution study, one group found a decrease in VCUG and RBUS in a cohort of emergency department patients after 2011.¹⁰ Garcia-Roig et al analyzed the Pediatric Health Information System, containing data from freestanding children’s hospitals, to determine if the AAP guideline affected the diagnosis and surgical management of VUR. They found that VCUG use decreased around the time of the publication of the guidelines and remained at this lower level through the remainder of the study period.¹¹

In this current study, we examined data from 2 ambulatory databases, the National Ambulatory Medical Care Survey (NAMCS) and the National Hospital Ambulatory Medical Care Survey (NHAMCS), to investigate community-based practice patterns of antibiotic utilization and diagnostic imaging in managing pediatric UTI after the 2011 AAP guidelines were published. We hypothesized that both antibiotic utilization and diagnostic imaging rates would decrease following the 2011 guidelines.

Methods

We performed a retrospective cohort study using encounter-level data from the nationally representative outpatient databases NAMCS and NHAMCS.¹² The databases are published annually by the National Center for Health Statistics. Data from years 2008 to 2010 and from 2012 to 2014 were included, while 2011 data were excluded to allow for a “burn-in” period during which the AAP recommendation were published. The NAMCS and NHAMCS databases are based on a sample of visits to community-based, ambulatory physicians and hospitals (outpatient departments, emergency departments, and ambulatory surgery centers), respectively. A 1-week period of patient visits is sampled from each provider for inclusion in the databases. NAMCS/NHAMCS are sampled at the provider rather than the patient level. Therefore, all observations were treated as unique encounters and any one patient could appear more than once in the dataset.

Patient encounters with (1) an International Classification of Diseases, Ninth Revision (ICD-9) diagnosis code for UTI (Appendix A) and (2) aged between 2 months and 18 years were included in the cohort. While the AAP guidelines were designed specifically for febrile UTI in patients between 2 and 24 months, it is our experience that many providers apply these guidelines to children older than 24 months.

The primary outcomes were antibiotic utilization and diagnostic imaging rates. We relied on “utilization” and not “prophylaxis” because NAMCS/NHAMCS does not differentiate between prophylactic and treatment antibiotics. However, we figure that changes in antibiotic utilization may be used as a proxy for changes in antibiotic prophylaxis with the assumption that rates of antibiotic treatment for UTI did not change systematically during the study period, and since there were no systematic changes in the AAP guidelines for UTI antibiotic treatment, this is a reasonable assumption. Therefore, changes observed in antibiotic utilization may be attributed to changes in prophylaxis.

Antibiotic utilization was defined as the presence of at least one drug code for penicillin, cephalosporin, macrolide, tetracycline, nitrofurantoin, aminoglycoside, or trimethoprim-sulfamethoxazole at the encounter. Antibiotic utilization could represent either new (ie, administered or prescribed) or a continuation of a drug for a visit. NAMCS/NHAMCS collect information on drugs that were either ordered, supplied, administered, or continued during the visit. The survey, from which NAMCS/NHAMCS are derived, asks providers to document all drugs that are either over-the-counter, prescribed, dietary supplements, or allergy shots. Diagnostic imaging rates were dichotomized to be the presence of an ultrasound, X-ray, or “other imaging” during the encounter. In the databases, VCUGs are coded under “other imaging.”

Weighted descriptive statistics were utilized to describe the cohort.¹³ This accounted for the complex structure of NAMCS/NHAMCS and enabled national level inferences. Wald χ² was used for discrete variables and weighted analysis of variance was used for continuous variables to determine statistical significance between demographic characteristics. If the cohort did not have an observation in a subcategory of a variable, a Wald χ² could not be calculated and the Fisher’s exact test was used. Weighted logistic regression models were fit for both of our primary outcomes, antibiotic utilization and diagnostic imaging rates, adjusting for covariates. Both models were adjusted for patient age, sex, race/ethnicity, temperature, and insurance type.

We performed sensitivity analyses by analyzing 2 subsets of our cohort. First, we examined all encounters ages 2 to 24 months, that is, infant encounters, since the AAP guidelines were directed toward patients of this age. Second, we examined all encounters with a recorded temperature of at least 100.4°F or encounters with an ICD-9 diagnosis code for fever, which we defined as febrile UTI encounters. We repeated our weighted logistic regression models for antibiotic utilization and diagnostic imaging rates while adjusting for covariates for these 2 subsets of the overall cohort. For infant encounters, we did not adjust for age, but adjusted for sex, race/ethnicity, insurance type, and patient temperature. For febrile UTI encounters, we adjusted for age, sex, race/ethnicity, and insurance type, but not patient temperature since each encounter was classified as febrile. We were unable to conduct sensitivity analysis for children aged 2 to 24 months with febrile UTI due to an insufficient number of observations. Consistency of the results from these 4 models would ensure that we were not affecting the signal of the overall cohort due to these variables. All data analyses were performed using SAS (version 9.4, SAS Institute, Cary, NC) and R (version 3.3.3 for Windows) software.

Results

Our cohort consisted of 8 588 035 weighted encounters, of which 52% occurred before 2011 (Table 1). Of these encounters, 85.2% were female with a mean age of 8.7 years (standard error: 0.35). Fifty-seven percent were white non-Hispanic. A total of 51.9% of encounters before 2011 had private insurance compared with 40% after 2011 with private insurance. A total of 70.7% of encounters before 2011 had at least one antibiotic present compared with 66.4% after 2011. We observed that 13.1% of encounters had at least one imaging procedure before 2011 and this number increased to 18.7% after 2011. Specifically, there was a descriptive increase in both ultrasound (4.4% to 9.0%) and “other imaging” (10.3% to 13.5%) after 2011 (Appendix B). Notably, only 74.6% of encounters had a urinalysis performed.

Table 1.

Demographic Characteristics^a.

Characteristic	Before 2011, N = 4 474 893	After 2011, N = 4 113 142	Total, N = 8 588 035
Sex
Female	3 744 152 (83.7%)	3 568 771 (86.8%)	7 312 923 (85.2%)
Male	730 741 (16.3%)	544 371 (13.2%)	1 275 112 (14.9%)
Age
Mean (SE)	9.11 (0.54)	8.16 (0.46)	8.65 (0.35)
Race/ethnicity
Hispanic	1 044 846 (23.4%)	974 448 (23.7%)	2 019 294 (23.5%)
Non-Hispanic black	673 557 (15.1%)	564 877 (13.7%)	1 238 434 (14.4%)
Non-Hispanic other	201 149 (4.5%)	224 306 (5.5%)	425 455 (5.0%)
Non-Hispanic white	2 555 341 (57.1%)	2 349 511 (57.1%)	4 904 852 (57.1%)
Insurance type
Medicare/Medicaid	1 659 774 (37.1%)	1 823 846 (44.3%)	3 483 620 (40.6%)
Other/missing	490 823 (11.0%)	644 362 (15.7%)	1 135 185 (13.2%)
Private	2 324 296 (51.9%)	1 644 934 (40.0%)	3 969 230 (46.2%)
Temperature (°F)
Mean (SE)	98.57 (0.09)	98.79 (0.10)	98.67 (0.07)
At least one antibiotic present	3 165 241 (70.7%)	2 729 391 (66.4%)	5 894 632 (68.6%)
At least one imaging procedure performed	587 778 (13.1%)	770 456 (18.7%)	1 358 234 (15.8%)
Urinalysis	3 399 186 (76.0%)	3 011 018 (73.2%)	6 410 204 (74.6%)

Abbreviations: SE, standard error; NAMCS, National Ambulatory Medical Care Survey; NHAMCS, National Hospital Ambulatory Medical Care Survey.

^aNational estimates of patient characteristics and outcomes of interest for full cohort stratified by year—2008–2010 and 2012–2014.

NAMCS and NHAMCS survey design was used to obtain national estimates.

After adjusting for covariates, there was insufficient evidence to suggest a difference in the odds between time periods and antibiotic utilization rates (odds ratio [OR] = 0.66, 95% confidence interval [CI] = 0.40–1.12, P = .12; Table 2). Similarly, there was insufficient evidence of a difference in the odds between time periods and diagnostic imaging rates (OR = 1.16, 95% CI = 0.71–1.91, P = .56).

Table 2.

Primary Outcomes^a.

Outcome	OR (95% CI)	P
At least one antibiotic present	0.66 (0.40–1.12)	.12
At least one diagnostic imaging procedure performed	1.16 (0.71–1.91)	.56

Abbreviations: OR, odds ratio; CI, confidence interval.

^aResults of multivariate logistic regression modeling for the 2 primary outcomes. OR given represents the indicator of a visit being before 2011 or after 2011, where before 2011 is the reference group. Model for full cohort is adjusted for age, sex, race/ethnicity, insurance type, and patient temperature (°F).

The sensitivity analyses of the 2 subsets were consistent with our overall results (Table 3). There was a descriptive decrease in antibiotic utilization rate in both subsets, including a decrease from 88.3% to 72.2% for patients in the febrile UTI cohort (Appendix B). However, we found insufficient evidence to suggest a difference in the odds between time periods and antibiotic utilization rates (OR = 0.52, 95% CI = 0.18–1.55, P = .24), and time periods and diagnostic imaging rates (OR = 1.22, 95% CI = 0.38–3.92, P = .73) for infant encounters. We found significant evidence to suggest a difference in the odds between time periods and antibiotic utilization rates (OR = 0.28, 95% CI = 0.08–0.98, P = .046) for febrile UTI encounters. Finally, there was insufficient evidence of a difference in odds between time periods and diagnostic imaging (OR = 1.65, 95% CI = 0.64–4.30, P = .30) for febrile UTI encounters.

Table 3.

Sensitivity Analyses^a.

Outcome	Cohort	OR (95% CI)b	P
At least one antibiotic present	Infantc	0.52 (0.18–1.55)	.24
	Febrile UTId	0.28 (0.08–0.98)	.046
At least one imaging procedure performed	Infantc	1.22 (0.38–3.92)	.73
	Febrile UTId	1.65 (0.64–4.30)	.30

Abbreviations: OR, odds ratio; CI, confidence interval; UTI, urinary tract infection; NAMCS, National Ambulatory Medical Care Survey; NHAMCS, National Hospital Ambulatory Medical Care Survey.

^aNational estimates of patient characteristics and outcomes of interest for sensitivity analyses stratified by year—2008–2010 and 2012–2014. NAMCS and NHAMCS survey design was used to obtain national estimates.

^bResults of weighted logistic regression modeling for the 2 primary outcomes. OR given represents the indicator of a visit being before 2011 or after 2011, where before 2011 is the reference group.

^cAdjusted for sex, race/ethnicity, insurance type, and patient temperature (°F).

^dAdjusted for age, sex, race/ethnicity, and insurance type.

Discussion

The 2011 AAP guidelines for children aged 2 to 24 months with febrile UTI substantially altered the recommended management for these patients. Prior to 2011, routine VCUG following a febrile UTI was recommended practice; afterward, this practice was discouraged. Our study aimed to evaluate the diagnostic imaging and antibiotic utilization practice patterns after the new guidelines were published by examining 3-year periods before and after guideline release using a national database. Based on previous work that examined imaging practice patterns since 2011 and the expectation that guidelines carry weight in physician practice, we hypothesized that the incidence of both diagnostic imaging and antibiotic utilization rates would decline after 2011 in accordance with the new guidelines.^10,11

However, we did not find evidence of a statistical difference in antibiotic utilization rates before and after the 2011 guidelines. While not statistically significant, there was a slight decline in antibiotic utilization coinciding with the 2011 recommendation against routine antibiotic prophylaxis. This decline may coincide with decreasing antibiotic prophylaxis, given our assumption that antibiotic treatment patterns remain unchanged during our study period. Unfortunately, our database did not allow us to distinguish specifically between antibiotics ordered for prophylaxis versus those ordered for treatment. Any decrease that we observed in antibiotics might be reasonably attributed to changes in prophylaxis practice patterns with the assumption that antibiotic treatment patterns remained static or increased during the time period of our study.

Interestingly, less than 75% of febrile UTI encounters after 2011 were treated with an antibiotic. More surprisingly, only half of infant UTI encounters had antibiotics before or after the new guidelines. In cases of both febrile and afebrile UTI, guidelines recommend some duration of antibiotic treatment.¹⁴ Our data suggested that providers may be undertreating pediatric UTI. While we have no reason to suspect that treatment practice patterns would substantially decrease after 2011, we cannot exclude this possibility. Indeed, the updated 2011 AAP guideline stating that initiating oral antibiotic treatment is as effective as parenteral treatment might have actually increased antibiotic treatment, dampening the change we would observe from a decrease in prophylaxis.⁸ Regardless, changes in prophylaxis practice patterns cannot explain why such a small proportion of UTI encounters both before and after the 2011 guidelines utilized an antibiotic.

Similar to antibiotic utilization, we found no evidence of a difference in diagnostic imaging patterns before and after the 2011 guidelines. Instead, we observed descriptive increases in diagnostic imaging rates after 2011. Interestingly, ultrasound was rarely utilized in either time period, while “other imaging” was often used in febrile UTI. Our findings suggest that physicians may only rarely obtain the appropriate routine diagnostic imaging for febrile UTI, despite guideline recommendations. There is evidence that this pattern in practice has existed for at least a decade across guideline revisions.^5–7

Our results do not replicate previous findings that suggested decreased use of diagnostic imaging after 2011.^10,11 There are several possible explanations for these differences in results. The first concerns the physicians sampled. Previous studies, one from a single institution and one from a national sample, largely included patients associated with freestanding children’s hospitals.^10,11 In contrast, NAMCS/NHAMCS samples from a wide range of outpatient encounters at children’s hospitals, emergency departments, and demographically diverse outpatient clinics in the community. While pediatric providers at dedicated children’s hospitals seem likely to remain up-to-date with new guidelines, it is possible that community providers may be less aware of or willing to adapt practice in response to new guidelines. In one study of general practitioners, 35% reported difficulty changing their routines and habits following guideline updates, despite a strong belief that following guidelines improves patient care.¹⁵ A lack of guideline adherence may be especially pronounced in family practice and general emergency medicine providers who have less experience caring for the pediatric population. Inclusion of these potentially less guideline-compliant providers in our sample may explain our results.

Furthermore, there is skepticism among some providers who disagree with the updated AAP guidelines.¹⁶ Some providers doubt ultrasound as a reliable VUR screening test, arguing that RBUS is inadequately sensitive to detect and grade VUR (level C evidence).^10,17 This distrust may explain the low utilization of ultrasound we observed for febrile UTI especially. In terms of VCUG, most physicians express concern for radiation exposure and patient discomfort, which at least in part contributed to the 2011 guidelines. However, some debate the quality of B-level meta-analysis–based evidence that suggested antibiotic prophylaxis was ineffective in preventing recurrent UTIs in patients with grades I to IV VUR.^8,16 These data also influenced the AAP to conclude a lack of evidence in the protective role of prophylactic antibiotic prophylaxis, even in the face of the 2014 RIVUR trial justifying reaffirmation of the guidelines in 2016.¹⁸ Our results add to the ongoing dialogue in the debate of managing UTI and VUR.

These data should be considered within the context of their limitations. Given the retrospective database design, there is a possibility of miscoding and the use of billing-oriented ICD-9 codes. Additionally, since providers may in some cases diagnose and treat acute UTI based on clinical suspicion without confirmatory testing, we cannot verify that all patients included indeed had a UTI. Since we were unable to track patients longitudinally, we were also unable to determine if an encounter was a patient’s first UTI. Whether an encounter was for a first or recurrent UTI influences if an antibiotic was intended for treatment or prophylaxis. Furthermore, this also alters what diagnostic imaging would be appropriate (RBUS for first, VCUG for recurrent). Counting recurrent encounters as first encounters might result in an underestimate of RBUS and overestimate of “other imaging” like VCUG. However, in many instances, whether at the emergency department or with a new provider, physicians with incomplete medical records are unlikely to know for certain a child’s UTI history, forcing them to treat these UTIs as first UTIs. Thus, this limitation reflects the reality of uncertainty in practice and may provide more reliable information about patterns in the community.

Our dataset did not include a primary indication for either diagnostic imaging or antibiotic utilization. It is possible that imaging ordered for other reasons during the same encounter, unrelated to UTI, might be falsely attributed to a UTI diagnosis in our analysis. While such instances are likely rare, the rates of imaging for indications other than UTI should be no different before and after 2011. It is also possible that a documented antibiotic was either new (ie, prescribed and/or administered during the visit) or a continuation from previous treatment. However, distribution of new compared with previously prescribed antibiotics should be no different before and after 2011. Therefore, this should not influence the validity of our analysis. Additionally, we were unable to distinguish VCUG utilization since this procedure is not directly reported in NAMCS/NHAMCS. Instead there is an “other” diagnostic imaging variable where this procedure is likely reported. As such, we avoided reporting VCUG imaging rates.

Last, the majority of the patients in our cohort fall outside the strict target population of the AAP guidelines, who are patients with febrile UTI aged between 2 and 24 months. While we hoped to include sensitivity analysis for this group, there was an insufficient sample size to perform multivariate analysis with meaningful estimates for visits fitting these criteria. However, it is our experience that many clinicians interpret the AAP guidelines as a recommendation for pediatric patient older than 24 months. Nevertheless, the AAP guidelines target a specific patient population, and physicians may not employ the same clinical practice in treating UTIs for these older pediatric patients.

Conclusions

Using a nationally representative sample of ambulatory encounters, we found no evidence of changes in diagnostic imaging or antibiotic utilization practice patterns for the management of acute pediatric UTI after the introduction of the 2011 AAP UTI guidelines. However, we observed descriptive decreases in antibiotic utilization and descriptive increases in imaging. Low overall imaging rates may suggest low guideline compliance, even after 2011.

Appendix A.

ICD-9 Inclusion Criteria^a.

ICD-9 Code	Description
599.0	Urinary tract infection, site not specified
590	Infections of the kidney
595	Cystitis
597	Urethritis not sexually transmitted

Abbreviation: ICD-9, International Classification of Diseases, Ninth Revision.

^aEncounters were included in the study who had any of the above-mentioned primary diagnosis ICD-9 codes for urinary tract infection.

Appendix B.

Descriptive Imaging and Antibiotic Characteristics^a.

	Overall		Infant		Febrile UTI
Characteristic	Before 2011	After 2011	Before 2011	After 2011	Before 2011	After 2011
Ultrasound	4.4%	9.0%	7.3%	14.5%	1.2%	6.1%
Other imaging	10.3%	13.5%	18.5%	16.0%	33.5%	41.3%
Antibiotic	70.7%	66.4%	51.2%	40.7%	88.3%	72.2%

Abbreviations: UTI, urinary tract infection; NAMCS, National Ambulatory Medical Care Survey; NHAMCS, National Hospital Ambulatory Medical Care Survey.

^aNational estimates of patient characteristics and outcomes of interest for full cohort and sensitivity analyses stratified by year—2008–2010 and 2012–2014. NAMCS and NHAMCS survey design was used to obtain national estimates.