453 - Predicting tracheostomy using ventilator support at 36 and 44 weeks in infants with severe bronchopulmonary dysplasia

Poster Number: 453
Publication Number: 453.431

Jonathan C. Levin, Boston Children's Hospital, Boston, MA, United States; Xuenan Wang, Brown University School of Public Health, Providence, RI, United States; Kristen T. Leeman, Boston Children's Hospital, Boston, MA, United States; Matthew Kielt, Nationwide Children's Hospital, Columbus, OH, United States; Lisa Grady, Women & Infants Hospital of Rhode Island, Providence, RI, United States; Leif D. Nelin, Nationwide Children's Hospital, Coumbus, OH, United States; Christopher H. Schmid, Brown University, Providence, RI, United States; Robin L. McKinney, Hasbro Children's Hospital at Rhode Island Hospital, Providence, RI, United States

Background: Early identification of infants with severe bronchopulmonary dysplasia (sBPD) that will require tracheostomy may be beneficial, as earlier placement may aid with growth, participation in developmental care, and can have implications for counseling families.

Objective: To develop prediction models for tracheostomy or death prior to discharge using clinical data at 36 and 44 weeks post-menstrual age (PMA).

Design/Methods: Multicenter, retrospective case-control study across 9 centers of the BPD Collaborative. We included infants with severe BPD (2001 NHLBI definition) born ≤ 32 weeks gestational age who were admitted at a Collaborative center at 36 and/or 44 weeks PMA. Infants transferred prior to determination of discharge respiratory support were excluded. Logistic regression models were developed using baseline demographics as well as ventilator, weight, and medication data at 36 and 44 weeks PMA using forward selection with 10-fold cross validation. Two-way interactions were considered. Missing data were handled by multiple imputation. The outcome of interest was tracheostomy or death prior to discharge.

Results: Of 999 infants included, 183 (18%) underwent tracheostomy placement or died prior to hospital discharge. Infants in the tracheostomy/death group were more likely to be SGA (33% v 18%, p< 0.001), and had lower birthweight (755g v 817g, p=0.009). The 36 week model (n=969) included level of ventilation support (invasive or non-invasive ventilation), inspired oxygen, and use of systemic steroids, (Table 1).  The model performed well, with an area under the curve (AUC) of 0.86 (95% CI 0.83 – 0.89). The 44 week model (n=572) included level of ventilation support, PEEP, use of systemic steroids, and use of pulmonary hypertensive medications (iNO or sildenafil) (Table 2). It performed well with an AUC of 0.90 (95% CI 0.88 – 0.93).Conclusion(s): Models using ventilator and medication data at 36 and 44 weeks both perform well to predict the outcome of tracheostomy/death among a population of infants with sBPD cared for at BPD Collaborative centers. This tool may help to more quickly identify infants likely to require tracheostomy and help with counseling families.
36 week PMA regression modelLogistic regression model with outcome of tracheostomy or death prior to discharge. The overall performance was area under the curve of 0.86 (95% CI 0.83-0.89). The model was developed using baseline demographics as well as ventilator, weight, and medication data at 36 weeks PMA with 10-fold cross validation.
44 week PMA regression modelLogistic regression model with outcome of tracheostomy or death prior to discharge. The overall performance was area under the curve of 0.90 (95% CI 0.88-0.93). The model was developed using baseline demographics as well as ventilator, weight, and medication data at 44 weeks PMA with 10-fold cross validation.