Is It Time to Stop Classifying Febrile Infants With Positive Urinalyses as High-Risk for Meningitis?

In this issue of Hospital Pediatrics, Young et al¹ present more evidence that a positive urinalysis result alone is not associated with a higher prevalence of bacterial meningitis in febrile infants 29 to 60 days of age. In this retrospective cohort study, 835 febrile infants underwent a full “sepsis workup,” including cerebrospinal fluid (CSF) testing, and the prevalence of bacterial meningitis was equivalent among infants with a positive versus a negative urinalysis (0.9% vs 1.0%, respectively). Furthermore, the authors separately analyzed 345 febrile infants 29 to 60 days of age with a positive urinalysis who were treated with antibiotics without CSF testing. None of these infants were subsequently diagnosed with bacterial meningitis or experienced an adverse outcome within 1 month.¹

To date, all of the criteria developed and commonly used for the risk stratification of febrile infants (Rochester, Philadelphia, Boston, Step-by-Step approach) include a positive urinalysis result as a parameter that classifies an infant as “not low-risk” for an invasive bacterial infection (ie, bacteremia and/or bacterial meningitis).^2–6 On 1 hand, this classification makes sense because the prevalence of bacteremia among infants with a positive urinalysis is ∼6% to 9%,^7–9 and these infants are theoretically at an increased risk for seeding the meninges leading to meningitis. However, with their data, Young et al¹ contradict this logic by showing that infants with a positive urinalysis are at no greater risk of meningitis than other febrile infants and that the urinalysis should not alter the pretest probability for meningitis in assessing the need for a lumbar puncture (LP) in this patient population. In other words, a positive urinalysis indicates that the infant likely has a febrile urinary tract infection (UTI) but does not itself inform the risk of meningitis.

Authors of several studies over the past decade have similarly attempted to dispel the notion that febrile infants 29 to 60 days of age with a positive urinalysis result are at high-risk for meningitis. Combined, researchers of these investigations have reported a prevalence of meningitis of 0% to 0.3% among infants aged 29 to 60 days with UTIs.^8,10,11 Although limited to retrospective studies, the body of evidence is still sufficiently large at this point to suggest that a positive urinalysis alone should not be used in the decision of whether to obtain an LP. Although Young et al¹ explored factors associated with bacterial meningitis in this current study, only 8 infants had bacterial meningitis thereby limiting the use of this exploratory analysis in clinical practice. A large multicenter investigation is needed to attain sufficient power to determine predictors of bacterial meningitis among febrile infants.

Historically, the goal of most risk stratification algorithms and clinical pathways for febrile infants has been to minimize errors of commission, such as unnecessary tests and treatment, while entirely eliminating errors of omission, such as delaying treatment of meningitis. Because of the fear of the potential sequelae of missing an infant with bacteremia and/or bacterial meningitis, we have tolerated an overwhelming majority of infants receiving unnecessary, invasive testing and treatment (ie, errors of commission). However, with a recent shift in culture centered on patient safety and value of care, clinicians are placing more emphasis on safely reducing these errors of commission as well.¹² On the basis of Young et al’s¹ data, performing an LP for every febrile infant with a positive urinalysis will only expose a majority of these infants to the risks of the procedure (transient hypoxemia, infection, epidermoid tumor, and bleeding)^13–16 without effectively reducing errors of omission, given the already exceedingly low prevalence of concomitant meningitis. An argument can be made that of the 3 infants with a positive urinalysis result who had meningitis as defined in this study, none truly had clear-cut meningitis. Two infants had traumatic LPs with relatively high red blood cells, low white blood cells, and minimal bacterial growth in the CSF (possibly reflecting traumatic LP and associated bacteremia). The third infant had an LP performed after antibiotics were administered and was diagnosed with meningitis on the basis of CSF glucose level of 37 without serum glucose for comparison. Although it was certainly reasonable to treat these infants as if they had meningitis, it is possible that none of them actually had bacterial meningitis.

We wonder what would have happened to these 3 infants if they never had an LP. Would they have suffered morbidity and mortality from the rare chance of missing, delaying the diagnosis of, or partially treating a case of meningitis? The fear of such outcomes seems to have driven the defensive practice of placing infants with a positive urinalysis in the high-risk for meningitis category. In the current study¹, the authors may alleviate some of these fears by reporting no negative outcomes for infants (even those with bacteremia) who did not have an LP in their workup. We cannot definitively conclude that there were no cases of meningitis in these infants; rather, the treatment of the UTI, with or without bacteremia, was sufficient to prevent any negative outcomes. Results from previous studies support that many febrile infants outside of academic institutions are managed without LPs and similarly have no adverse events or reported treatment failures.¹⁷ Therefore, omitting the LP does not necessarily mean you are ignoring a diagnosis of meningitis but that the LP may not have any effect on treatment decisions and outcomes in some groups of infants who are well-appearing and clinically improving on their treatment of a UTI.

As stated, 2 of the cases were difficult to diagnose because of traumatic LPs. Failed or traumatic LPs reportedly occur in 10% to 35% of attempts,¹⁸ leaving the CSF profile difficult or impossible to effectively analyze. What do we do in these situations? If we choose to assume that the patient does not have meningitis, it begs the question of why did we do the LP in the first place? If we choose to treat for meningitis because we cannot be certain that we have ruled it out, we have potentially sentenced the infant to 21 days of intravenous antibiotics.¹⁹ How do we further incorporate the data suggesting that some infants with a UTI may have a CSF pleocytosis in the absence of bacterial meningitis, especially if we decide to do the LP after antibiotics have been administered?²⁰ In choosing to do an LP for every infant with a positive urinalysis in an effort to rule out the exceedingly rare chance of concomitant meningitis, we will end up exposing a subset of infants to unnecessary, extended hospitalizations associated with prolonged intravenous antibiotic courses, other iatrogenic risks, and substantial health care costs.^21,22 In this current culture of balancing risk of omission and commission, we need to ask if these outcomes justify a routine LP for all infants with positive urinalyses.

So, should we regularly perform an LP when a febrile infant has a positive urinalysis? As Young et al¹ suggest, there is no possible universal recommendation that can perfectly balance errors of omission with commission. Given this lack of a perfect clinical recommendation, we believe the data can be used to help guide shared decision-making in determining whether to perform an LP on a case-by-case basis. The differences in individual clinician and family’s tolerance for risk, severity of illness presentation, patient characteristics, threshold for probability of disease, and overall goals of care should drive clinical decisions. Moving forward, further studies in which authors identify and validate factors associated with bacterial meningitis, guiding workup for truly high-risk infants, will be necessary to optimize the yield of our invasive testing. Based on the growing evidence base, a positive urinalysis alone is not one of those factors.