A Novel CART-Driven Decision Tree Combining NLR and CRP for Early Prognostication of Severe Acute Pancreatitis: A Prospective Vietnamese Cohort Study.

Abstract

Background: Severe acute pancreatitis (SAP) is a life-threatening condition requiring early risk stratification. While the Bedside Index for Severity in Acute Pancreatitis (BISAP) is widely used, its reliance on complex parameters limits its applicability in resource-constrained settings. This study introduces a decision tree model based on Classification and Regression Tree (CART) analysis, utilizing Neutrophil-to-Lymphocyte Ratio (NLR) and C-reactive Protein (CRP), as a simpler alternative for early SAP prediction.

Methods: In a prospective cohort of 340 patients at National Hospital, Vietnam (November 2022-September 2023), NLR, CRP, and BISAP scores were assessed upon admission. CART analysis was used to develop a decision tree, and model performance was compared with BISAP using receiver operating characteristic (ROC) curves, decision curve analysis (DCA).

Results: The CART model identified NLR ≥11.4 and CRP ≥173.3 mg/L as optimal thresholds for SAP prediction. The model achieved an area under the curve (AUC) 0.866 in the validation cohort, statistically comparable to BISAP (AUC = 0.900, p = 0.286). The model demonstrated high sensitivity (90.9%), specificity (84.5%), and accuracy (86.25%), confirming its robustness. DCA highlighted similar clinical benefits with BISAP, but the CART-based model offered greater simplicity, making it ideal for resource-limited settings.

Conclusion: The CART-derived decision tree using NLR and CRP provides an accessible and reliable tool for early SAP prediction. With performance comparable to BISAP but requiring fewer resources, this model supports rapid, evidence-based decision-making in clinical practice.