236 - Improving Prediction of BPD Severity Using Bedside Monitor and Alarm Data in Extremely Low Gestational Age Newborns

Publication Number: 236.5457

Ananta Wadhwa, University of Rochester School of Medicine and Dentistry, Rochester, NY, United States; Jack Chang, University of Rochester, Rochester, NY, United States; Irina Prelipcean, University of Rochester School of Medicine and Dentistry, Rochester, NY, United States; Jiebo Luo, University of Rochester, Rochester, NY, United States; Xing Qiu, University of Rochester, Rochester, NY, United States; Colby Day, University of Rochester School of Medicine and Dentistry, Rochester, NY, United States; Igor Khodak, Golisano Children's Hospital at The University of Rochester Medical Center, Rochester, NY, United States; Nancy E.. Robertson, University of Rochester Medical Center -Clinical and Translational Science Institute, Rochester, NY, United States; Charles Rosenberg, University of Rochester School of Medicine and Dentistry, Rochester, NY, United States; Divya Chhabra, University of Rochester School of Medicine and Dentistry, Rochester, NY, United States; Andrew M. Dylag, University of Rochester, Rochester, NY, United States

Background: Extremely Low Gestational Age Newborns (ELGANs, born less than 28 weeks’ gestation) experience intermittent hypoxemia (IH) events that are associated with poor respiratory and neurodevelopmental outcomes.

Objective: To determine if IH event frequency and duration is associated with worse Bronchopulmonary Dysplasia (BPD) severity.

Design/Methods: We established a time-series ELGAN database (N=351) containing real-time data views from the electronic medical record (EMR) and Q1 minute bedside vital sign monitors. IH was defined as a “SpO2 low” event triggering an alarm lasting between 10 and 300 seconds. Oxygen saturation histograms were calculated for a subset (N = 49) of infants using Q1 minute monitor data and SpO2% cutoffs: ( < 80%, 81-85%, 86-90%, 91-95%, >96%). BPD severity was determined by the Neonatal Research Network 2019 definition among survivors at 36 weeks corrected gestational age. Summary variables of IH frequency and duration were tabulated over the first 28 postnatal days and fed into multivariable regression models along with variables from the published Neonatal Research Network BPD Calculator. Model performance was evaluated using AUROC, AIC, and BIC.

Results: IH frequency was increased among any infants with any BPD compared to no BPD (p < 0.0001, Table 1), but did not stratify by BPD severity (Figure 1). IH duration and SpO2 < 85%, however, increased with each grade of BPD severity (p < 0.0001, Table 1) and stratified by BPD severity (Figure 2). The AUROC when the variables in the NRN regression model were matched to our single-center cohort was 0.747 at DOL 28, comparable with published performance. When IH frequency and duration were added to the model, AUROC improved to 0.773 and the AIC and BIC statistics lowered from 636 to 619 and 776 to 769 respectively, suggesting improved model fit.

Conclusion(s): IH frequency can distinguish between infants with and without BPD, but not BPD severity. IH duration and SpO2 < 85% aligns with increasing BPD severity. Taken together, these data suggest that infants with worse BPD severity experience longer, deeper desaturations. In addition, incorporating IH duration and frequency to BPD regression models improves performance and may distinguish infants on different BPD trajectories. Future models can leverage bedside monitor and alarm parameters to elucidate BPD trajectories earlier in the NICU hospitalization.

Intermittent Hypoxemia frequency by BPD severity and Day of Life
Median (line) and IQR (shaded area) of IH Events by Day of Life and BPD Severity. BPD Levels: Black - 0 | Blue - 1 | Green - 2 | Red – 3

Intermittent Hypoxemia duration by BPD severity and Day of Life
Median (line) and IQR (shaded area) of IH Duration of Day of Life and BPD Grade (Up to Day 28). BPD Levels: Black - 0 | Blue - 1 | Green - 2 | Red – 3

Summary Statistics for Intermittent Hypoxemia Frequency, Duration, and time spent with SpO2 < 85% by BPD Severity