322 - Occurrence of Intermittent Hypoxemias and Bradycardias Events in Neonates Prior to Discharge from the Neonatal Intensive Care Unit

Publication Number: 4316.322

Sarah T. Wu, McGill University Faculty of Medicine and Health Sciences, Pointe-Claire, PQ, Canada; Jasmine Chapdelaine, McGill University Faculty of Medicine and Health Sciences, Montreal, PQ, Canada; Ana Gabriela Saavedra Ruiz, RI-MUHC, Laval, PQ, Canada; Xavier Morin-Duchesne, Montreal Children's Hospital, Montreal, PQ, Canada; Robert Kearney, McGill University Faculty of Medicine and Health Sciences, Montreal, PQ, Canada; Guilherme Santanna, McGill University Health Center, MONTREAL, PQ, Canada; Wissam Shalish, McGill University Health Center, Montreal, PQ, Canada

Background: Neonates commonly experience episodes of intermittent hypoxemia (IH) and bradycardia during NICU stay. Discharge decisions often rely on the absence of nurse-recorded events requiring intervention, though decisions prompting intervention are unclear.

Objective: (1) Describe the occurrence of IH and bradycardia events from nursing records and bedside monitor data in neonates before NICU discharge; (2) Identify features of IH and bradycardia events prompting intervention.

Design/Methods: This retrospective study included infants discharged from the Montreal Children's Hospital NICU between Feb 2024 and Feb 2025. Demographic and peri-discharge data were obtained from medical records. IH (oxygen saturation, SpO₂ < 90%) and bradycardia (heart rate, HR < 100bpm) requiring intervention (tactile stimulation) were extracted from nursing logs. Continuous SpO₂ and HR data from bedside monitors were analyzed to identify all IH and bradycardia events, then linked to nursing logs to classify them as self-resolving or requiring intervention. Median (IQR) event frequency and percentage of time spent below thresholds were compared across preterm and term groups. Nadirs, magnitude drops, and total durations of events with or without intervention were compared and optimal thresholds were derived with corresponding diagnostic accuracy metrics.

Results: Among 199 infants, 10,989 IH and 54,668 bradycardia events were identified from the bedside monitor, of which 28 and 68 required intervention per nursing log, respectively. Event frequencies were stable in the week before discharge, with medians of 42.0 [21.2-96.2] IH events < 90% and 14.0 [1.0-142.5] bradycardias < 100bpm per patient, and 0.57% [0.24-1.14] time < 90% SpO₂ and 0.04% [0.02-0.08] time < 80% SpO₂ (Fig. 1). Event patterns were similar across preterm and term groups, except term infants had more HR < 100bpm events. Compared to self-resolving events, those requiring intervention had lower nadirs and larger drops from baseline for IH ( < 90% and < 80%) and bradycardia ( < 100bpm and < 80bpm) but similar total durations (Fig. 2). An SpO₂ drop > 10% and HR drop > 23bpm predicted IH < 80% and bradycardia < 80bpm with balanced accuracies of 73% and 86%. In contrast, using absolute nadir values (SpO2 < 63% and HR < 73bpm) yielded lower balanced accuracies of 45% and 35%, respectively (Table 1).

Conclusion(s): IH and bradycardias persist until discharge and most resolve spontaneously, with substantial overlap in event features compared to those that prompt intervention. Future work should focus on developing objective, monitor-based tools to guide event intervention decisions.

Daily distribution of IH and bradycardia event frequency and time < 90% and < 80% SpO₂ during the week preceding discharge in preterm and term infants
Daily distribution of events over the week preceding discharge for preterm (N= 63, median PMA = 38.0 wks [36.9 - 40.4]) and term infants (N = 136, median PMA = 40.0 wks [39.2-41.0]). (A) the number of IH events (SpO₂ < 90%), (B) the number of bradycardia events (HR < 100 bpm), (C) the percentage of time spent with SpO₂ < 90%, and (D) the percentage of time spent with SpO₂ < 80%.

Comparison of the distribution of metrics for events with and without interventions
(A) IH events with SpO2 < 90% (B) IH events with SpO2 < 80% (C) bradycardia events with HR < 100 bpm, and (D) bradycardia events with HR < 80 bpm. The lowest value refers to the minimum SpO₂ or HR reached during an event. Drop magnitude is the difference between the last normal value (SpO₂ ≥ 90% or ≥ 80%; HR ≥100 bpm or ≥80 bpm, respectively) and the lowest value. Fall duration is the elapsed time from the last normal value to the previous occurrence of the lowest value, and recovery duration is the elapsed time from the first occurrence of the lowest value to the first subsequent normal value. Total event duration spans from the last normal value before the fall to the first normal value after recovery.

Diagnostic accuracy metrics for predicting interventions across different event types
Thresholds indicate the optimal cutoff for each event, determined through Youden's J characteristic. Performance metrics include sensitivity (true positive rate), specificity (true negative rate), positive predictive value (PPV), negative predictive value (NPV), and balanced accuracy (average of sensitivity and specificity). Event types include IH with SpO₂ < 90% and < 80%, and bradycardia with HR < 100 bpm and < 80 bpm.