Session: Neo-Perinatal Health Care Delivery 4: Epidemiology/Health Services Research
271 - Analysis of Factors Associated with Mortality to Discharge in Very-Preterm Infants (VPT, ≤32 Weeks’ Gestation) in a Public County Hospital: 46-Year Retrospective Cohort Study.
Assistant Professor University of Texas Southwestern Medical School Dallas, Texas, United States
Background: Parkland Health and Hospital System (PHHS), a minority-serving community hospital, delivers >12,000 infants annually, with ~1,400 admissions to the NICU. We utilized a well validated NICU database extending from 1977-2022 to investigate factors associated with mortality to discharge in VPT infants. Objective: To assess factors associated with mortality to discharge in VPT infants, excluding congenital anomalies. Design/Methods: Retrospective cohort study conducted at PHHS from 1977-2022. Gestational age (GA) was determined by the best obstetrical estimate. Pediatric estimate of GA was used if there was no obstetric estimate or the difference was >1wk from obstetrical estimate during the first 2 epochs and >2wks during subsequent epochs. Size for GA was assessed using Olsen’s curves. Stepwise logistic regression analyses were used to identify significant factors associated with mortality to discharge. Since many additional variables were available as of 2009, two models were developed: one for the 1977-2022 cohort and the other for 2009-2022. The best models were selected to optimize area under the curve (AUC) (accuracy) and the Youden’s index (sensitivity and specificity). Only models with nonsignificant Hosmer-Lemeshow test (i.e., fitting the data) were retained. Results: Among 10,394 VPT infants born 1977-2022, 1495 died before discharge (Table 1). The best model was built using Epoch, GA, small for GA, sex, category (Hispanic, Non-Hispanic Black and Non-Hispanic White) and 5-min Apgar scores (0-3 vs 4-7 and 8–10). This model had an AUC=0.897 and a Youden’s index 0.640.
Among 3244 VPT infants born in 2009-2022, 300 died before discharge (Table 2). The best model was built using GA, birthweight, sex, and antenatal steroids (ANS), whose use was expanded in 2015. There was no evidence for disparity. This model had an AUC=0.963 and Youden’s index 0.824.
Conclusion(s): In the 46-year model, mortality of VPT infants to discharge was associated with GA, size for age, sex and birthweight, sex, Non-Hispanic Blacks or White and 5-min Apgar scores. In the 14-year model, mortality before discharge was associated with GA, birthweight, ANS and sex. These data underscore the value of adding ANS to the care of VPT infants. IMG_7386.jpeg
