81 - Heterogeneous patients, services, and risk-adjusted outcomes by neonatal intensive care unit (NICU) acuity across the gestational age and birth weight spectrum

Publication Number: 3077.81

Nicolas P. Goldstein Novick, Children's Hospital of Philadelphia, Rose Valley, PA, United States; Joshua Radack, Childrens Hospital of Philadelphia, Boston, MA, United States; Scott A. Lorch, The Children's Hospital of Philadelphia, University of Pennsylvania, Philadelphia, PA, United States; Timothy Nelin, Childrens Hospital of Philadelphia, Philadelphia, PA, United States; Elizabeth Salazar, Children's Hospital of Philadelphia, Philadelphia, PA, United States; Sara Handley, Childrens Hospital of Philadelphia, Philadelphia, PA, United States

Background: Current AAP guidelines categorize all level III NICUs together. We previously found marked heterogeneity within level III NICUs in very preterm/very low birth weight (VPT/VLBW) volume and services, identifying five distinct types with latent class analysis (LCA). How older/larger infants fare overall and in comparison across these acuity classes and within level IV NICUs is unknown.

Objective: To examine heterogeneity across level III-IV NICUs and compare outcomes for VPT/VLBW and older/larger infants across NICU acuity classes.

Design/Methods: We performed a retrospective cohort study of level III-IV NICUs using linked vital statistics and discharge data from hospital births (≥10/year) in FL, MI, OR, PA, and SC (2010-2020). We generated two cohorts, 1) VPT/VLBW: infants with GA < 32 wk and/or birth weight (BW) < 1500 g, 2) Older/Larger-High Acuity (HA): infants with GA ≥32 wk and BW ≥1500 g with length of stay (LOS) >5 days and/or any invasive ventilation. At the NICU-year level, we analyzed patient characteristics and services. Across the five LCA-derived level III acuity and level IV hospitals, we compared characteristics, services, and unadjusted counts of mortality, sepsis, and extreme LOS (>90th percentile by GA week) in the full and stratified cohorts. We compared risk-adjusted rates of these outcomes across NICU acuity with Poisson regression with year and state fixed effects and standard errors clustered by hospital.

Results: We analyzed 1708 NICU-years from 181 facilities, with 2121 median annual hospital births (IQR 1515, 3164) and 231 NICU admissions (144, 391). Median average GA was 34.6 weeks (33.8, 35.3) and BW was 2395 g (2245, 2568). Across NICU acuity, there was significant variation in births, NICU admissions, GA, BW, LOS, invasive ventilation, surgery, and major congenital anomalies (Table 1). Unadjusted annual counts of death, sepsis, and extreme LOS increased with acuity; the VPT/VLBW cohort had higher mortality but lower sepsis and extreme LOS compared to the Older/Larger-HA cohort (Table 2). Risk-adjusted mortality rates were lower with rising acuity in the VPT/VLBW cohort but higher in the two highest acuity classes in the Older/Larger-HA cohort; across both cohorts, sepsis rates were highest in the moderate acuity classes (3b-3d) and extreme LOS rates increased with acuity (Figure 1).

Conclusion(s): NICU populations are heterogeneous and risk-adjusted mortality and morbidity rates vary across NICU acuity class. Broad levels of care may mask important differences within and across ranges of GA and BW. More granular, data-based levels may improve benchmarking and regulation.

Table 1. Characteristics of NICUs and their patient populations across five latent class analysis-derived Level III NICU classes and level IV NICUs
• 1 Small-for-gestational age defined as < 10th percentile on the Fenton growth chart
• 2 Macrosomia defined as a birthweight > 4,500 grams
• All median comparisons were significant at the <0.001 level with the Kruskal-Wallis equality-of-population rank test; all count comparisons were significant at the <0.001 level with the Pearson Chi-Square test
• Per AAP guidelines, Level III NICUs provide comprehensive care for very preterm/critically ill newborns including sustained life support, advanced ventilation, invasive monitoring, peri-operative care, and access to pediatric subspecialists. Level IV NICUs additionally must offer complex neonatal surgery (including cardiothoracic surgery), ECMO, and function as regional referral centers.
• Abbreviations: LCA (latent class analysis), IQR (interquartile range), NICU (neonatal intensive care unit)

Table 2. Unadjusted annual mean counts of mortality, sepsis, and extreme length of stay across five latent-class analysis derived Level III classes and level IV NICUs
• All pairwise comparison t-tests were significant at the 5% level with the following exceptions: 1 3d versus 3e for in-hospital death in the full sample, 2 3d versus 3e for in-hospital death in the VPT/VLBW cohort, 3 3a versus 3b for in-hospital death in the Older/Larger-High Acuity cohort, 4 3e versus 4 for sepsis in the VPT/VLBW cohort, 5 3b versus 3c versus 4 for sepsis in the Older/Larger-High Acuity cohort, 6 3d versus 4 for extreme length of stay in the full cohort, 7 3d versus 4 for extreme length of stay in the Older/Larger-High Acuity cohort
• * Extreme length of stay defined as > 90th percentile within gestational age at birth week for the full cohort
• Counts of hospitals within each LCA class add to more than total hospital count because LCA class sometimes varied over time within hospital
• Abbreviations: LCA (latent class analysis), CI (confidence interval), NICU (neonatal intensive care unit), VPT/VLBW (very preterm/very low birthweight)

Figure 1. Forest plot of risk-adjusted incidence risk ratios of in-hospital death, sepsis, and extreme length of stay across NICU acuity in the full, very preterm/very low birthweight, and older/larger-high acuity cohorts
• Models: Poisson regression with birthyear and state fixed effects, cluster-robust standard errors for hospital, and adjusted for the following: average gestational age, proportion of small-for-gestational-age, number of infants born with major congenital anomalies, and number of patients born with macrosomia
• Extreme length of stay defined as > 90th percentile within gestational age at birth week for the full cohort
• Abbreviations: VPT/VLBW (very preterm/very low birthweight), NICU (neonatal intensive care unit)