When young infants develop a fever during the first two months of life, clinical management decisions can diverge dramatically depending on which doctor evaluates them and which medical center receives them. A comprehensive population-based investigation published in Pediatric Research has dismantled the assumption that clinical variation in pediatric emergency care stems from a single uniform source. Analyzing statewide emergency department visits for infants aged eight to sixty days across New York, the research team demonstrated that different clinical choices are driven by entirely separate levels of the healthcare delivery system. While the decision to admit a febrile infant to the hospital is predominantly determined by individual physician-level factors, the decision to perform an invasive lumbar puncture is overwhelmingly governed by infant-level characteristics. This sharp divergence underscores that medical decision-making in acute pediatric care is not monolithic, revealing distinct targets for healthcare quality improvement.
Fever in infants aged sixty days or younger represents a frequent and clinically fraught presentation, accounting for up to one hundred thousand emergency department visits annually across the United States. Young infants with elevated body temperature face genuine risks of potentially fatal invasive bacterial infections, including bloodstream bacteremia and bacterial meningitis. Historically, clinical practice across emergency departments has exhibited wide variability in diagnostic evaluations and admissions, with disparities proving particularly pronounced among infants between twenty-nine and sixty days of life. Such extensive practice variation presents a major challenge to healthcare systems because it signifies that factors unrelated to clinical appropriateness can influence therapeutic pathways. Infants may consequently face unnecessary and painful diagnostic procedures, hospital-acquired infections, and disruptions to maternal breastfeeding from overtreatment, or alternatively suffer catastrophic missed diagnoses of invasive infections from undertesting.
To investigate the structural architecture behind these clinical discrepancies, the investigators designed a retrospective population-based cohort study utilizing records from the Statewide Planning and Research Cooperative System in New York State. The study cohort encompassed records from January 2012 through July 2021, capturing practice patterns established before the formal release of the American Academy of Pediatrics clinical practice guideline in 2021. From an initial identification of more than forty-one thousand febrile infants presenting to emergency departments, thirty-five thousand seven hundred fifty-six infants met the final inclusion criteria after excluding transfers, direct admissions, and encounters lacking clinician National Provider Identifier numbers. The analytical sample included preterm infants, patients diagnosed with concomitant bronchiolitis, and infants with underlying complex medical conditions. Across the study cohort, fifty-seven percent of infants were male, forty-four percent were covered by commercial insurance, and ten percent had a documented history of preterm birth.
The evaluation of these young patients took place across one hundred ninety-four hospital facilities and involved more than five thousand seven hundred unique treating physicians. Among the thirty-five thousand seven hundred fifty-six included infants, eighteen thousand one hundred twenty-four were admitted to the hospital, representing an overall hospitalization rate of 50.7 percent with a 95 percent confidence interval spanning 50.2 percent to 51.2 percent. Lumbar punctures were administered to twelve thousand nine hundred fifty-nine infants, yielding an overall procedure rate of 34.3 percent with a 95 percent confidence interval between 33.8 percent and 34.8 percent. Laboratory-confirmed invasive bacterial infections were identified in 2.4 percent of the overall cohort, with the highest burden documented in neonates aged eight to twenty-eight days at 3.4 percent. The vast majority of infants initially presented to general community hospitals rather than specialized pediatric facilities, an operational pattern that became increasingly frequent with advancing infant age.
Marked disparities emerged when examining patient care rates across different patient age brackets and across individual emergency facilities. Rates of both hospital admission and lumbar puncture exhibited an inverse relationship with infant maturity, steadily declining as age increased. For neonates aged eight to twenty-eight days, 65.6 percent were admitted and 48.9 percent underwent lumbar puncture, compared with 44.0 percent admitted and 30.6 percent undergoing spinal taps among older infants aged twenty-nine to sixty days. Facility-level disparities were pronounced: hospital-level admission proportions ranged from absolute zero to 73.5 percent across the overall cohort, and lumbar puncture frequencies ranged from zero to 56.6 percent. Post-hoc evaluations showed these patterns were not simply artifacts of management within the fourth week of life, as rates remained closely matched between infants aged eight to twenty-one days and those aged twenty-two to twenty-eight days.
To rigorously disentangle the specific origins of these treatment differences, the researchers implemented a three-level multilevel logistic regression model employing random effects for individual physicians and hospitals. This variance component framework partitioned the observed variation into infant-, physician-, and hospital-level factors while controlling for relevant covariates at each hierarchical level. Covariates incorporated into the adjusted models included infant demographic parameters, clinical diagnoses, and health service area characteristics; physician age, sex, clinical experience, and board certification specialty; and hospital-level annual encounter volume, urban-rural classification, local invasive bacterial infection proportions, and status as a general versus children’s hospital. Because severity of illness scores derived from administrative groups demonstrated strong collinearity with admission, illness severity was exclusively included as an adjustment covariate for the lumbar puncture outcome.
The adjusted multilevel models revealed a profound structural divergence between the two care decisions. Physician-level attributes emerged as the primary determinant of hospitalization, explaining 59.3 percent of the adjusted variance, compared with 64.2 percent in unadjusted models. Conversely, infant-level factors contributed 81.9 percent of the adjusted variance in lumbar puncture decisions, paralleling an 82.9 percent contribution in unadjusted calculations. In age-stratified analyses, infant characteristics accounted for less than fifteen percent of hospitalization variance in both younger and older infants, whereas infant characteristics continued to account for over eighty percent of lumbar puncture variation across both age cohorts. For both outcomes, variance stemming from unobserved infant factors declined as infant age increased, whereas the variance attributed to observed infant-level factors remained stable across the evaluated age spectrum.
The study also tested two formal hypotheses regarding institutional contributions to practice variation using bootstrapped 95 percent confidence intervals across ten thousand iterations. The investigators first hypothesized that hospital-level factors would explain less than ten percent of variance for both decisions. However, empirical findings demonstrated that hospital-level factors accounted for 20.9 percent of total variance in hospitalizations, with a 95 percent confidence interval from 13.2 percent to 28.6 percent, directly refuting the hypothesis for admissions. For lumbar puncture procedures, hospital-level factors accounted for 12.2 percent of variance, though the wide confidence interval ranging from 1.9 percent to 22.4 percent crossed below ten percent. The second hypothesis posited that hospital-level variance would be significantly greater in infants aged twenty-nine to sixty days compared to neonates aged eight to twenty-eight days; this hypothesis was also rejected, as age-stratified hospital-level variance showed no statistically significant differences between age groups.
Analysis of post-encounter trajectories revealed telling dynamics regarding institutional practice culture and patient revisits within seven days of the initial encounter. The authors observed no significant correlation between a hospital’s baseline admission rate and its overall rate of seven-day emergency revisits, yielding a coefficient of determination of zero. However, when evaluating revisits that specifically resulted in subsequent hospital admission, a strong positive association emerged: facilities with higher initial admission rates demonstrated significantly elevated rates of revisits requiring hospitalization, with a coefficient of determination of 0.26. Similar positive correlations appeared in age-stratified analyses for both neonates and older infants. In alignment with Wennberg’s small area variation framework, these findings indicate that hospitals maintain consistent internal practice styles or facility personalities, such that institutions with lower thresholds for admitting febrile infants at primary encounters consistently apply lower thresholds upon return.
The authors highlighted several limitations inherent to their study design, chief among them the reliance on administrative billing data lacking real-time physiologic variables, serum biomarkers, or physical examination nuance, which likely constitutes a substantial portion of unobserved infant-level variance. Additionally, reliance on National Provider Identifier linkages may not always capture the exact clinician executing specific management decisions, and billing code omissions remain possible. Nonetheless, the findings establish that hospitalization decisions are heavily swayed by individual physician tendencies, whereas lumbar punctures remain closely tied to patient-specific features. Because hospital-level factors consistently account for ten percent or more of variation in both outcomes, targeted health-system interventions focused on hospital protocols and standardized clinical decision support present meaningful opportunities to reduce arbitrary practice variation and optimize care for vulnerable febrile infants.
Subject of Research: Sources of practice variation in the clinical management of febrile infants
Article Title: Sources of practice variation in febrile infants aged 8–60 days: a multi-level analysis
Article References: Yaeger, J. P., Singh, S., Love, T., & Fiscella, K. A. (2026). Sources of practice variation in febrile infants aged 8–60 days: a multi-level analysis. Pediatric Research. https://doi.org/10.1038/s41390-026-05603-2
Image Credits: AI Generated
DOI: 10.1038/s41390-026-05603-2
Keywords: Pediatrics, Febrile Infants, Emergency Medicine, Practice Variation, Hospitalization, Lumbar Puncture, Multilevel Analysis, Healthcare Quality, Sources, practice, variation, febrile
