MACHINE LEARNING MODELS TO PREDICT FAILURE IN ADMINISTRATION OF INTRAPLEURAL TPA/DNASE IN PATIENTS WITH COMPLICATED PARAPNEUMONIC EFFUSIONS/EMPYEMA: A MULTICENTER STUDY

TOPIC: Disorders of the Pleura TYPE: Original Investigations PURPOSE: Intrapleural fibrinolytic therapy (IPFT) with tissue plasminogen activator (tPA) and deoxyribonuclease (DNase) has been shown to reduce need for surgical intervention complicated pleural effusion/empyema (CPE/empyema). For patients in whom tPA/DNase is likely fail, however, receipt this may simply delay inevitable. The goal study was identify risk factors failure combined intrapleural therapy. METHODS: We performed a multicenter, retrospective chart review who received treatment CPE/empyema. Clinical variables included demographic data, radiographic parameters at time diagnosis, results from fluid analysis. compared four different machine learning classifiers (L1-penalized logistic regression, support vector machine, XGBoost LightGBM) by multiple bootstrap-validated metrics, including F-beta (with recall weighted twice as heavily precision) area under receiver operator characteristic (AUC) curve. Model hyperparameters were tuned using adaptive optimization (tree-parzan estimator), variable importance estimated random permutation rank 19 candidate clinical respect their ability predict Data randomly split into training test sets an 80:20 ratio. Training data further same ratio hyperparameter tuning. whole pipeline repeated 200 times order derive distributions evaluation metrics. RESULTS: 466 participants five institutions across United States. All IPFT management effusions/empyema. Resolution CPE/empyema achieved 78% (n=365) cases. Support (SVM) best 57%, followed (F-beta 47%), LGBM 46%). Although average performance L1-penalized regression adequate (50% F-beta), it highly unstable ranging 0 100%. Of predictors failure, all models exceptL1 agreed that single most important predictor whether abscess or necrotizing pneumonia present, presence thickening protein pleural. (L1 ranked important.) CONCLUSIONS: Our analysis based on large, multicenter database demonstrated abscess/necrotizing helps triage fail be indicated. These warrant investigation validation prospective study. CLINICAL IMPLICATIONS: Simple findings can used fail. two helpful expedite infection. DISCLOSURES: No relevant relationships Akshu Balwan, source=Web Response Billie Bixby, Consultant relationship olympus medical inc Please note: $1-$1000 Christopher Gilbert, Response, value=Consulting fee verathon $1001 - $5000 AstraZenaca $5001 $20000 no disclosure file Jed Gordon; Danai Khemasuwan, Chakravarthy Reddy; Trinidad Sanchez, Samira Shojaee, Jeff Sorensen, Candice Wilshire,