LAUSR

Despite longstanding efforts to improve newborn care, the ability of neonatology and the public to monitor and influence the quality, outcomes, and efficiency of care has been mixed. On the bright side, the Vermont-Oxford Network has pioneered the comparative measurement of care processes and outcomes primarily for very low-birth-weight newborns through benchmarking, research, and quality improvement initiatives. Statewide perinatal collaborative efforts, most notably the California Perinatal Quality Care Collaborative and the California Children’s Services (CPQCC/ CCS), have expanded the range of newborn care data and provided further stimulus to improvement activities. While these efforts remain important, the growth of neonatal intensive care unit (NICU) care has overtaken these data used by these initiatives, leaving the care and outcomes of most newborns in the dark. Currently, to our knowledge, no entity is responsible or has the means to monitor medical care for the total birth cohort in the United States, and only limited data sets are available at the state level. It is a striking truth that in 2012, more than threequarters of US NICU admissions were heavier than the usual 1500 g threshold (ie, very low birth weight) for data reporting in Vermont-Oxford Network member units, and the most common NICU admission was a normal-birth-weight newborn.1 The CPQCC/CCS captures data for almost all California newborns admitted to NICUs, including those heavier than 1500 g with high illness acuity. While the data collection is minimal for the 88% of NICU admitted newborns with lower illness acuity, there is just enough information to underscore the critical importance of shedding light on the care of larger and less premature newborns. In this issue of JAMA Pediatrics, Schulman et al2 report near population–based NICU admission rates in California, with a focus on those with 34 or more weeks’ gestation, representing 97% of all 2015 California births and 79% of all NICU admissions. The investigators found that inborn admission rates varied markedly across hospitals, with high rates in both regional-level and nonregional-level units (eg, community and intermediate, respectively). By itself, this is not surprising— levels of high illness acuity in newborns varied across hospitals just as much. What is striking is that hospitals with higher NICU admission rates had a lower proportion of admitted newborns with high illness acuity. This troubling finding suggests that the illness acuity level for NICU admissions differs from one hospital to the next, raising the possibility that it is not adequately known which newborns will benefit from NICU care and/or that a great deal of NICU care is provided needlessly. Equally concerning is that it has taken many years for these patterns of care to become visible. In the bright light of the Vermont-Oxford Network and CPQCC/CCS efforts, it seems that we have forgotten the less ill newborns that now dominate NICU care and the severe data limitations that hinder further research. The methods used in the study by Schulman et al2 provide a glimpse of the difficulties this causes. If asked as a hypothetical—design a study to measure the relationship between inborn newborn illness acuity and NICU admission rates across hospitals in California— one would choose a birth cohort design that includes all newborns. The data set would include maternal and newborn characteristics up to the point of care initiation, recognizing that diagnostic labeling can be biased by the quantity and type of medical care received by the newborn.3,4 Neonatal intensive care unit admission rates could then be compared across hospitals for gestational age and illness acuity levels. Health outcomes and postdischarge use associated with the varying NICU admission rates could also be assessed. Instead, the paucity of data available to CPQCC/CCS forced Schulman et al2 to use a narrow set of acuity measures, falling short of the full range of reasonable indications for NICU care, but they did include “early bacterial sepsis,” which may be biased in its diagnostic assignment. These limitations do not threaten the conclusion that admission criteria vary irrationally across hospitals, but the study cannot provide insight into the magnitude of the problem, the characteristics of newborns affected, or the consequences of the practice patterns. Knowledge of the full scope of neonatal care will require additional care and outcome measures for a wider set of births than exists today. Although they often fall short of all live births, the population-based newborn data sets include the US birthlinked death file, state all-payer claims data, and state Medicaid files. The birth-linked death file includes the entire US birth cohort with generally reliable information about each newborn and its mother, including infant deaths in the first year of life. Importantly, the file has an indicator for NICU admission for those states using the 2003 revised birth certificate form, but little other information about medical care is provided. Geographic identifiers are limited to counties. An alternative data set is all-payer claims data files. These are available to researchers for a few states. The files include both hospital and professional services from birth to beyond the initial hospital care episode. Medicaid is another Related article Opinion