Length of Stay and Complications Associated With Febrile Infants <90 Days of Age Hospitalized in the United States, 2000–2012

Discussion

Herein, we provide the first description of long-term national trends regarding certain characteristics of hospitalization for young febrile infants <90 days of age with suspected SBI. In this large US population including both community- and academic-based institutions, we found no clinically relevant change in the hospital LOS over the study period. Our hypothesis was that LOS in these infants might slowly decrease over time because of clinicians’ acknowledgment of ongoing changes in SBI epidemiology and increasing use of rapid viral diagnostic tests. In particular, the former reflects the experience that identifying a specific bacterial pathogen in blood or cerebrospinal fluid is an increasingly uncommon event.¹⁴ Since 2000, the decline in the incidence of SBI due to previous group B Streptococcus, Listeria monocytogenes, and pneumococcus has accelerated,^11–13 and recent data have been used to note the value of rapid viral testing toward less overall diagnostic testing and therapy in those with suspected SBI.^2,10,24,25 With our data, we suggest that clinicians’ acceptance or awareness of such information is incomplete and/or is impacted by a lack of availability of viral diagnostics and/or perhaps some other factors that may encourage more conservative practice such as personal experience, practice–group dynamics, and other impacts of peer-to-peer practice or medicolegal concerns.²⁶

Previous studies of the hospitalization of young febrile infants have been focused on the likelihood of admission,^27,28 antimicrobial use,²⁹ and outcomes,³⁰ with relatively little or no attention to LOS and its associated predictors. We found certain patient characteristics to be associated with longer LOS including race and/or ethnicity, geographic location, and insurance type. It is clear, however, that LOS will also depend generally on specific clinical issues and nonmedical issues (eg, family dynamics), variables for which we did not have data. Regardless of the age group or condition for which LOS is studied, it is important to consider correlated outcomes at the practitioner, hospital, or other level.³¹ Nonetheless, relatively few empirical data exist regarding predictors of LOS for young infants admitted for nonsurgical conditions. Others have reported that prematurity and concomitant comorbidities were associated with prolonged LOS for infants <6 months of age with UTI,³² and the specific infecting pathogen was associated with an increased LOS in children <2 years of age with bronchiolitis.³³ We are not aware of any such study of infants and LOS, for any medical condition, performed on a large longitudinal data set such as we report here.

We found that hospital LOS did not change for the fever group or the viral respiratory group. Most infants in the fever group had an LOS of ≤2 days, generally consistent with the “48 hours rule out sepsis” workup, but nearly one-third had an LOS ≥3 days, and 8% had an LOS for ≥4 days, which is well beyond when >99% of blood culture results become positive.³⁴ It is unclear why those infants with uncomplicated fever without an associated infection diagnosis had prolonged LOS. It may be that the patients had diagnoses or psychosocial family issues that were not captured with ICD-9-CM codes. It is also possible that variations in coding practice at the hospital level classified a true 48-hour admission as 3 days.

We detected >350 complications of care in our cohort, and among those with a specified cause, nearly 80% were related to receipt of medication. We believe this to be a significant underestimate insofar as ours is a retrospective cohort study using administrative data, and systematic underdocumentation of adverse event-related ICD-9-CM codes is well described.^35,36 Importantly, we found that infants with complications had a longer LOS. Authors of 2 previous small single-center studies of hospitalized young infants noted 20% to 29% of subjects developed a complication, and in 1 study, 60% were preventable.^7,8 Moreover, a survey of parents of febrile infants hospitalized for sepsis evaluation revealed negative psychosocial consequences, including severe stress, disruption of breastfeeding, and continued perceptions that the child was ill despite no SBI found.³⁷ Infants hospitalized unnecessarily to “rule out sepsis” or who remain in the hospital for prolonged periods clearly present risks to the child and family of potentially severe complications and may create an economic burden as well.

We found that the likelihood of a complication of care increased with age, independent of the presence of an SI and concomitant chronic comorbidity. There are no other published data on this topic. Why this relationship is present is unclear, but for example, we speculate that perhaps older infants are less likely to be swaddled and therefore move their extremities more freely resulting in intravenous catheter dislodgement. As well, they may be more likely to receive empirical vancomycin because the likelihood of potential penicillin-resistant pneumococcal disease is greater in infants 60 to 90 days compared with those <60 days of age.³⁸ As well, these infants are therefore susceptible to potential nephrotoxicity associated with vancomycin and/or untoward events associated with monitoring drug levels. We did not have access to data to explicitly explore these hypotheses.

The initial decision to admit a febrile infant <90 days of age can be difficult and must balance the risks of missing an SBI with unnecessary admission and overtreatment. A report by the Agency for Healthcare Research and Quality did not provide conclusive guidance,² and no consensus exists on the appropriate management in the outpatient setting.^6,39–41 Much data have been used to contribute to a summary estimate of 7% to 11% of febrile infants <90 days predicted to have an SBI.¹ Researchers of recent data from the Northern California Kaiser Permanente system note that among those with SBI, ∼92% have UTI, 6% have bacteremia, and <0.5% have meningitis.⁹ Authors of previous work have provided the following useful information to guide clinicians toward minimizing hospitalization and LOS: (1) febrile infants with a confirmed viral respiratory infection have a much lower likelihood of concomitant SBI,^25,42,43 (2) evidence-based protocols minimize LOS,²⁴ and (3) in the outpatient private-practice setting, individualized clinical judgment compared with strict adherence to guidelines lessens the likelihood of admission without worsening outcomes.³⁰

We acknowledge the methodological limitations of ICD-9-CM code data; their robustness is also variable across pediatric age group and specific diagnoses.^20,22,44 Most germane to our work is that of Aronson et al²¹, in which variability in the accuracy of such codes for “fever” was noted among hospitalized young infants by ascertainment method and by institution. Specifically, KID LOS data are not available in hours nor does KID allow for the consideration of “observation stays” of <24 hours. In addition, we could not determine provider reasoning or other qualitative information that might reveal discharge decision-making and its potential impact on LOS. We do not have data beyond 2012 because certain state data, compared with our study period, are not included in the 2015 KID and/or SID data release (KID is made available only every 3 years). Regardless, given the fairly constant data trends we observed over our long study period, we believe the likelihood of a meaningful change (in 2015) is low especially given that no new practice guideline has been published in the interim. Our data set does, however, include the largest study population on the topic, includes diverse hospital types from a well-dispersed geography, and a weighting method allowing for greater generalizability. We also identified predictors for the likelihood of prolonged hospital LOS, which has not been well described in this age group. Furthermore, our 13-year study period enabled an understanding of the potential impact of clinical factors germane to the management decision-making for these febrile infants.

We have shown that hospital LOS for young, otherwise uncomplicated febrile infants did not substantively change from 2000 to 2012, despite decreasing risk for bacteremia and meningitis and maturation and acceptance of viral diagnostic testing. As well, important harms were identified. With these observations, we support the presumption that clinical integration of contemporaneous SBI epidemiology and more aggressive viral testing for inpatients have not been maximized in the inpatient setting. Targeted efforts used to raise awareness and implications of these issues in the care of inpatients, in concert with evidence-based guidelines for the approach to such infants in the emergency department setting may minimize LOS and complications associated with prolonged hospitalization. More data on the rationale behind the decision to admit and later discharge young febrile infants, prospective data collection on the incidence of complications related to procedures performed in the hospital, and the associated economic implications to hospital systems and families would be informative.