EVIDENCE-BASED ANSWER
Clinical judgment remains the gold standard for hospital admission of infants with bronchiolitis, and it cannot be replaced by objective criteria (strength of recommendation [SOR]: B, based on prospective and retrospective cohort and retrospective case-control studies). Oxygen saturation (SaO2) is the most consistent clinical predictor of deterioration, though different investigators vary cutoffs from 90% to 95% SaO2 and the vast majority of infants with saturations in this range do well (SOR: B, based upon prospective cohort studies).
CLINICAL COMMENTARY
The key is being able to identify a “sick” child
Mike Polizzotto, MD
Rockford Family Medicine Residency Program, Rockford, Ill
As a medical student I was taught that one of the most important skills I could develop is the ability to look at a child and know whether he or she is “really sick” or “not so sick.” In determining which patients with bronchiolitis I admit to the hospital, I look at findings such as age (<3 months), medical history, oxygen saturation, and respiratory rate and effort. I also evaluate less tangible data, including the parents’ level of comfort taking their child home, and the number of visits they have already made to the emergency department or clinic for this same problem. For me, a pulse oximetry reading of 93% (or any other individual finding) does not mandate admission. Although one might hope for more objective evidence upon which to base decisions, those of us who are comfortable using this type of gestalt will find the results of this inquiry reassuring.
Evidence summary
Bronchiolitis is the most common diagnosis among hospitalized infants aged <1 year in the US.1 It is usually mild with a self-limited course. A 1997 study following 1113 healthy full-term infants through 20 consecutive winter seasons showed a 5% hospitalization rate of all infants with a positive respiratory syncytial virus (RSV) cult\ure (hospitalization rates with other pathogens were not reported), confirming the mild nature of most cases of bronchiolitis.2 RSV accounts for 50% to 80% of bronchiolitis, along with other pathogens such as parainfluenza virus, influenza virus, and human metapneumovirus. A recent analysis of Centers for Disease Control and Prevention data from 1979 to 1997 showed that an average of 95 children died annually in the US from bronchiolitis, and 77% of these were aged <1 year (median age at death was 3 months).3
When discrete measures such as vital signs and scoring scales for respiratory distress are compared, infants who have mild disease courses are very similar to those who subsequently have a more severe illness. This, combined with the low incidence of serious illness, lowers the predictive value of any single clinical criteria for hospital admission (including oxygen saturation, respiratory rate, apparent respiratory distress, and day of illness) to the degree that no objective criteria are useful to make a decision for or against hospitalization.4,5
In 2 good-quality retrospective case-control studies, which enrolled infants with milder disease discharged from emergency departments, no infants returned with illness severe enough to require admission to an intensive care unit (ICU).5,6 No criteria were found that could predict subsequent severe course and need for admission.
One good-quality prospective study that enrolled 213 infants, younger than 13 months and presenting with bronchiolitis as outpatients found that physician impression of appearance was a better predictor of severe illness than numeric scoring systems such as the Yale Observation Scale or the Clinical Asthma Score.7 Pulse oximetry (<95%), prematurity (<34 weeks gestational age), respiratory rate >70/minute, atelectasis, “ill” or “toxic” appearance, and age <3 months were associated with more severe illness (defined as inability to remain alert and active or well hydrated throughout their illness). Oxygen saturation (SaO2) <95% was the most objective predictor of severity (positive predictive value=87%; negative predictive value=73%). The study population was more ill than what is typical in outpatient settings (42% required admission and 11% required mechanical ventilation); therefore the positive predictive value would be lower in a milder, more typical outpatient population.