P0352 / #1996: PREDICTIVE MODEL FOR CHYLOTHORAX AFTER CARDIAC SURGERY

Abstract

Aims & Objectives: Chylothorax occurs in 2.8-3.9% of children post-cardiac surgery and is associated with increased length of stay and mortality. The median time to diagnosis is reported as 4-8.5 days postoperatively. We hypothesize that early diagnosis without dependence on initiation of enteral feeds will lead to earlier resolution. We propose a predictive chylothorax model for earlier detection than traditional pleural fluid testing (PFT). Methods: A prospective cohort of 405 patients <18 years old (415 encounters) post-cardiac surgery at a single tertiary-care academic center between 2016 and 2019 was studied. Exclusion criteria were mechanical circulatory support and chylothorax diagnosis within post-chest closure day (PCD) 1. Multivariate regression models optimized AUC and selected from postoperative chest tube output (CTO), age, race, genetic syndrome, surgical weight, STAT category, single/two ventricle repair with/without arch reconstruction, operative times, prior surgery, and delayed sternal closure (DSC). Results: PFT confirmed chylothorax in 37 encounters (9.2%). The final model included DSC and identified CTO of 15.7 mL/kg on PCD1 as having important predictive value with AUC 0.65. Patients with CTO >15.7 mL/kg on PCD1 were more likely to have chylothorax with a model sensitivity of 85%, specificity of 68%, positive predictive value of 31% and accuracy of 84%. The negative predictive value of CTO <15.7 mL/kg on PCD1 was 96%. DSC was a significant risk factor (Table).Conclusions: Chest tube output on PCD1 may be an early predictor of chylothorax allowing for diagnosis prior to the initiation of enteral feeds. Early diagnosis has potential to decrease duration of chylothorax and improve related outcomes.