Urinary Tract Infections in Young Febrile Infants: Title and subTitle BreakIs Selective Testing Acceptable?

IN THIS ISSUE of the ARCHIVES, Newman et al¹ present data from the Pediatric Research in Office Settings' (PROS) Febrile Infant Study, one of several articles we look forward to seeing from this remarkable data set. The article addresses urinary tract infections (UTIs) in febrile infants younger than 3 months who were seen in office practices.

As one of their study questions, the authors ask, Do practitioners follow published practice guidelines, referring specifically to routine urine testing for UTI in the evaluation of febrile infants? Of the 3066 infants enrolled in the PROS Febrile Infant Study, only 1775 had their urine tested. Therefore, the answer is no. The conclusion that practitioners do not follow guidelines is not unexpected given the results of studies cited by the authors, but Newman and colleagues suggest a more provocative question: Should practitioners be following published guidelines in this situation, or is what they are currently doing working just fine? That question generates others, all of which deserve consideration.

Are the published guidelines for the evaluation of fever in infants in this age group based on evidence? Yes. They are based on many studies conducted in several large pediatric emergency departments (EDs). Do the guidelines apply to office practices? This important question underscores the importance of the American Academy of Pediatrics' (Elk Grove Village, Ill) PROS network. The authors note that only 7% of the practices participating in this study are located in inner cities, certainly different from the locations of the EDs in which the published information was generated. Are demographic differences likely to affect the incidence of UTI and the yield of routine testing? Although white infants appear to have a higher rate of UTI than African American or Hispanic infants (among girls, at least), the effect of this difference on suburban practice should be an increased awareness of UTI and rate of testing, not a reduction. The major difference in the settings may be the mindset of the clinician. An ED orientation generally considers individuals to be potentially more ill than they appear, in contrast to an office orientation, which considers children rarely to be seriously ill; the former puts more emphasis on testing, the latter on reassurance and follow-up. A family that comes to the office is presumed to be able to maintain continuity and come back should there be a lack of improvement. In only 3.9% of cases were the PROS study practitioners unable to follow the course of this illness to its symptomatic conclusion. Such is not the case in EDs, and this may justify testing in the ED setting as a way of identifying which patients need to be actively pursued. The value of follow-up has more than a psychological effect: Newman and colleagues identified that the longer the duration of fever, the higher the rate of UTI. In office practice, it is presumed that there will be a second chance to make the diagnosis if the symptoms continue.

Why not test all febrile infants? The authors recognize that a major obstacle to urine testing is obtaining the specimen. One quarter of the specimens were collected in a urine bag. Published data suggest that 85% of the time, a positive culture result in a bag-collected specimen is likely to be false-positive.² However, Newman and colleagues did not find an excess of positive urine cultures among infants whose urine was collected in a bag, and suggest that the urine bags performed adequately in this study. Factors such as the time between voiding and the removal of the bag from the perineum and between specimen collection and refrigeration or processing are important, and these steps may be carried out in a more timely fashion in the office setting than in the environments where the published studies were conducted. Specimens collected in a bag are suitable for urinalysis but the literature suggests caution in interpreting urinalysis results in young infants. Crain and Gershel³ presented data indicating that in the first 8 weeks of life, a positive urinalysis does not necessarily indicate UTI and a negative urinalysis does not rule out UTI. In the PROS study, a urinalysis was performed on specimens from 1652 infants; a culture was performed on specimens from 1608. One or the other was performed on specimens from 1775 infants. These numbers indicate that practitioners decided not to culture the urine of some of the infants after urinalysis, presumably trusting a negative urinalysis to rule out UTI. To better assess the performance of urinalysis and bag urine, we need a more detailed evaluation of the data by urinalysis component and method of specimen collection, which the authors have begun (T. B. Newman, MD, oral communication, October, 2001).

In addition to the "hassle factors" related to obtaining a urine specimen (including time, effort, and cost), another obstacle to testing all febrile infants is the clinical feedback to practitioners that selective testing appears to work, making routine testing seem unnecessary and excessive. Is selective testing, in fact, an acceptable strategy? Selective testing could be considered acceptable only if the clinical risks associated with missing the diagnosis are low. As the authors point out, the high rate of UTI-associated bacteremia in the first month of life argues against selective testing for infants in this age group. For infants older than 1 month, the authors consider symptoms to be the major issue associated with UTI. The notion that UTI in infants is of concern primarily because of symptoms is contrary to decades of concern about UTI being a marker for underlying abnormalities that, when coupled with UTI, may impair renal function and produce hypertension decades later. Conventional wisdom considers the opportunity to perform imaging studies to be a benefit of identifying a UTI in a young febrile infant. The authors propose that the risk to infants in this study who did not receive urine testing was low. Although they provide compelling short-term outcome data, only a follow-up study could confirm that the risk is as low as suggested. Nuclear scans following the treatment of a UTI in febrile infants demonstrate scarring in 9.6% to 30% of cases.⁴ Scarring has been associated with hypertension and reduced renal function, but as Hoberman et al⁴ note, such correlations were based on scars demonstrable by intravenous pyelography rather than more sensitive nuclear scans. Newman and colleagues may be correct that with selective testing, the practitioners identified all of the infants with UTI that required diagnosis, treatment with antibiotics, and evaluation with imaging studies, but such a conclusion remains a hypothesis. What makes their hypothesis so provocative is that conventional wisdom and the published guidelines are also based on hypotheses and inferences rather than on direct data. There is no longitudinal study following treated and untreated infants with febrile UTIs to determine the rates and extent of scarring and correlate those outcomes with later blood pressure level and renal function. Moreover, such a study is unlikely to be conducted. As Newman and Maisels⁵ state in the context of a different clinical problem, "If guidelines are to be evidence-based, they must recommend what the evidence actually supports, rather than being overly conservative." These authors challenged us to take a "kinder, gentler approach" to the evaluation and treatment of jaundice in the term newborn,⁶ and Newman and colleagues again challenge conventional wisdom. It is a healthy challenge.

If selective testing is to be considered, which infants should be tested? There are identifiable risk factors, particularly being in the first month of life (because of associated bacteremia), being female, or being an uncircumcised boy. The higher the fever and the longer its duration, the greater the likelihood of UTI and the more urine testing should be considered. Other factors appear to ameliorate risk, such as respiratory disease. The authors identify that Hispanic infants seem to have a lower rate of UTI than white infants, but the practical significance of this observation is mitigated by the low circumcision rate in Hispanic boys; most Hispanic infants are in the 2 groups at increased risk: girls and uncircumcised boys.

The results of this study are provocative, and we await additional articles from the PROS Febrile Infant Study. The study was conducted in office practices, whereas the previous studies on which practice guidelines were based were conducted in EDs. As the Evidence-Based Medicine Working Group headed at McMaster University (Hamilton, Ontario) has stressed, readers of medical literature need to ask the following question: "Can the results be applied to my patient care?"⁷ The extent to which data generated in one setting apply to the other is an important factor and is one of many reasons to cheer for the PROS network. The challenge to conventional wisdom in the current article gives us another reason to cheer for Dr Newman and his colleagues.