Prediction Model with External Validation for Early Detection of Postoperative Pediatric Chylothorax

Earlier diagnosis of chylothorax following pediatric cardiac surgery is associated with decreased duration of chylothorax. Pleural fluid testing is used to diagnosis chylothorax which may delay detection in patients who are not enterally fed at time of chylothorax onset. Our aim was to develop and externally validate a prediction model to detect chylothorax earlier than pleural fluid testing in pediatric patients following cardiac surgery. A multivariable logistic regression model was developed to detect chylothorax using a stepwise approach. The model was developed using data from patients < 18 years following cardiac surgery from Primary Children's Hospital, a tertiary-care academic center, between 2017 and 2020. External validation used a contemporary cohort (n = 171) from Lucille Packard Children's Hospital. A total of 763 encounters (735 patients) were analyzed, of which 72 had chylothorax. The final variables selected were chest tube output (CTO) the day after sternal closure (dichotomized at 15.6 mL/kg/day, and as a continuous variable) and delayed sternal closure. The highest odds of chylothorax were associated with CTO on post-sternal closure day 1 > 15.6 mL/kg/day (odds ratio 11.3, 95% CI 6,3, 21.3). The c-statistic for the internal and external validation datasets using the dichotomized CTO variable were 0.78 (95% CI 0.73, 0.82) and 0.84 (95% CI, 0.78, 0.9) and performance improved when using CTO as a continuous variable (OR 0.84, CI: 95% CI 0.80, 0.87). Using the models described, chylothorax after pediatric cardiac surgery may be detected earlier and without reliance on enteral feeds.