• Setting: Parkland Memorial Hospital (PMH) is a public institution covering the population of Dallas County, Texas. The pregnant population is in large part covered by Medicaid funding. Annual deliveries have been 10,223-16,454 over the last 47 years. Race/ethnicity in VPT neonates was predominantly non-Hispanic Black initially and shifted progressively towards predominantly Hispanic. The Division of Neonatal-Perinatal Medicine at UT Southwestern Medical Center (UTSW) has been a member of the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) Neonatal Research Network (NRN) since its inception in 1986. Thus, the Division has contributed to the NRN-reported decreases in mortality associated with improved perinatal and neonatal therapy often due to rigorous multicenter trials. UTSW has also participated in the Vermont-Oxford Network (VON) since 2011.
• Study design and entry/exclusion criteria: This retrospective cohort study included VPT neonates born at PMH January 1977-June 2024 and without congenital anomalies or major discrepancy between birth weight and GA.
• Data included: Data included those in the PMH neonatal intensive care unit (NICU) database and comparisons with two multicenter databases: VON [neonates <1500 g birthweight, (very low birthweight, VLBW), 2011-2024] and NICHD NRN generic database [401-1000 g birthweight (extremely low birthweight, ELBW) or <29wks GA (extreme preterm, EPT), 2008-2024].
• Description of the PMH NICU database: The PMH NICU database was initiated in 1977 as part of the Maternal and Neonatal Data Acquisition, Transmission and Evaluation (MANDATE) project funded by the Robert Wood Johnson Foundation (1977-1982; Babata et al)². Except for a 10-month gap in 1985, data entry into the reformatted PMH NICU database was continued using many of the original data definitions until current. This database, currently in Microsoft Access 2007, has been validated multiple times. Many variables and algorithms used in the current study were further analyzed and validated. The NICU database was approved by UTSW Institutional Review Board as a quality improvement (QI) project. The NICU database has been used for annual presentation of neonatal procedures and outcomes to the Division of Neonatal-Perinatal Medicine, thereby identifying health care issues that should be addressed to improve outcomes.

Previous studies using the NICU database: The NICU database has been used as baseline for research and QI projects, many of which required merging with data from other internal (prenatal, resuscitation, follow-up) or external (VON, NRN) databases.^1,2 Among very-low-birthweight (VLBW, <1500g) neonates born 1977-2000, Kaiser et al. reported decreasing mortality over time and lowest mortality in Black females vs. other groups classified by sex, race and ethnicity.³ Data extracted from the NICU database have been used for serial studies in infants, children and preadolescents born prematurely (Reis, Wickland, Rosenfeld et al).^4-7 Data extracted from the NICU database into Research Electronic Data Capture (REDCap) electronic data capture tools have been used for recent studies in adults born prematurely, i.e., the Cardiac Growth After PrEterm Birth study (CAPE study) (Goss et al).⁸

• Primary outcomes: In the current study,¹ in-hospital mortality among VPT neonates decreased from 47% in 1977 to 6% in 2024 in association with successive changes in prenatal and neonatal clinical care. Mortality decreased 97-98% by 1986 in neonates born at 30-32 weeks GA (born 8-10 weeks before expected date of birth) and 92-93% by 1991 in neonates born at 27-30 weeks GA (born 10-13 weeks before expected date of birth). However, mortality in neonates born at 22-23 weeks GA (born 17-18 weeks early) remained >95% until 2010 and decreased 50% from 2016 to 2024. Notably, racial-ethnic differences in in-hospital mortality among VPT neonates disappeared at PMH in 2009.
• Comparison with other centers: In multivariable analyses, in-hospital mortality at PMH was not significantly different from other centers in NRN or from VON.
• Analysis of changes in racial-ethnic differences: Factors that may explain the lower mortality in Black neonates in Kaiser’s study³ include (1) their high frequency of maternal insurance reported in the present study and (2) less respiratory distress syndrome than in White neonates. Disappearance of racial-ethnic differences in in-hospital mortality among VPT neonates in 2009 may be explained at least in part by using standardized processes of care for all racial-ethnic and socioeconomic patient categories and by decreasing respiratory distress syndrome-related mortality.
• Limitations: The study was powered only for the primary outcomes; all other analyses should be considered exploratory. Because of the observational design without analysis of difference-in-differences, analyses show associations between interventions and outcomes but not causality. Since data were derived from a single center, results may not apply to other centers. For implementation of interventions that were not immediately documented in the NICU database, only temporal associations could be assessed. Several types of bias were addressed by appropriate analyses. However, lack of data on fetal mortality and stillbirths may bias analysis of relationship of mortality with maternal or delivery room interventions and may cause left truncation of early neonatal mortality. Potential confounding from unmeasured individual characteristics and interventions and changes in medical practice cannot be excluded.
• Strengths: This study has many strengths, including a large sample size (n=10,882, average 227 per year), long duration (47yrs), and inclusion of well validated data with very few missing data points. Moreover, several analytical methods were used to minimize bias. For several interventions and processes of care that were documented in the NICU database, multivariable analysis was used to assess independent associations with either lower or higher mortality. In addition, different analyses yielded similar results in trends and racial/ethnic comparisons. Importantly, multiple guidelines were followed, including REporting of studies Conducted using Observational Routinely collected health Data (RECORD), prognostic model development and unbalanced data. All multivariable models had excellent metrics, and the best model was internally validated.
• Summary: Long-term assessment of local changes in clinical care permits annual analysis of their impact on mortality/morbidity in comparison with neonatal/perinatal trends.
• Future directions of the NICU database include converting program codes and data from Microsoft Excel (about to phase out) into another platform that will allow integrating with artificial intelligence (Babata et al).² This may facilitate updating codes, extracting electronic data from electronic health records, linking with medical devices, laboratory systems and other databases, and developing and applying machine learning-derived predictive models. Ongoing validation will be performed, as reported in the current study.

References

1. Brion LP, Rosenfeld CR, Burchfield PJ, Babata K, Brown LS, Jaleel M. Changes in mortality in very preterm neonates in a single institution in Dallas, Texas, 1977-2024: a cohort study. Pediatr Res. 2026 May 23. doi: 10.1038/s41390-026-05106-0. Epub ahead of print. PMID: 42177309.
2. Babata K, Rosenfeld CR, Jaleel M, Burchfield PJ, Oren MS, Albert R, Steven Brown L, Chalak L, Brion LP. A validated NICU database: recounting 50 years of clinical growth, quality improvement and research. Pediatr Res. 2025 Jun;97(7):2224-2234. doi: 10.1038/s41390-024-03624-3. Epub 2024 Oct 21. PMID: 39433962.
3. Kaiser JR, Tilford JM, Simpson PM, Salhab WA, Rosenfeld CR. Hospital survival of very-low-birth-weight neonates from 1977 to 2000. J Perinatol. 2004 Jun;24(6):343-50. doi: 10.1038/sj.jp.7211113. PMID: 15116138.
4. Reis JD, Tolentino-Plata K, Caraig M, Heyne R, Rosenfeld CR, Brown LS, Brion LP. Double-blinded randomized controlled trial of optimizing nutrition in preterm very low birth weight infants: Bayley scores at 18-38 months of age. J Perinatol. 2023 Jan;43(1):81-85. doi: 10.1038/s41372-022-01572-6. Epub 2022 Dec 6. PMID: 36473929.
5. Reis JD, Heyne R, Rosenfeld CR, Caraig M, Brown LS, Burchfield PJ, Lair CS, Petrosyan E, Jabob T, Nelson DB, Brion LP. Follow-up of a randomized trial optimizing neonatal nutrition in preterm very low birthweight infants: growth, serum adipokines, renal function and blood pressure. J Perinatol. 2024 Jan;44(1):78-86. doi: 10.1038/s41372-023-01821-2. Epub 2023 Nov 14. PMID: 37964083.
6. Wickland J, Steven Brown L, Blanco V, Heyne R, Turer C, Rosenfeld CR. Persistent high blood pressure and renal dysfunction in preterm infants during childhood. Pediatr Res. 2023 Jan;93(1):217-225. doi: 10.1038/s41390-022-02083-y. Epub 2022 Apr 28. PMID: 35484228.
7. Rosenfeld CR, Steven Brown L, Blanco V, Heyne R. Relationship between serum lipids/adipokines and obesity/hypertension in very-preterm infants during early childhood and preadolescence. Pediatr Res. 2025 Dec 11. doi: 10.1038/s41390-025-04578-w. Epub ahead of print. PMID: 41381874.
8. Goss K. Goss Lab. https://labs.utsouthwestern.edu/goss-lab/patients-and-parents-0, accessed 05/11/2026.