Classification and Regression Tree (CRT) models classify children with clinical pneumonia based on radiological findings.

Abbreviations: CHI3L1, Chitinase 3-like-1; CRP, C-reactive protein; CRT, Classification and Regression Tree; CXR, chest x-ray; MC, misclassification cost; NLR, negative likelihood ratio; NPV, negative predictive value; PCT, procalcitonin; PLR, positive likelihood ratio; PPV, positive predictive value; sTie-2, soluble Tie-2; vWF, von Willebrand Factor.a Minimum number of cases was 10 per parent node (prior to split) and 5 per child node (following split). Misclassification cost is how many fold worse it is to misclassify the group of interest versus the other group. Pruning reduces overfitting by trimming trees down to simpler structures.b Generated using cross-validation with 10 sample folds.c x2 = model contains 2 different cut-points for a single biomarker.d Other infiltrates and normal CXR groups were combined into “non-end-point pneumonia.”e CRP and PCT were entered as categorical variables based on cut-points used in available point-of-care tests: CRP 40 μg/mL and PCT 0.5 ng/mL.Classification and Regression Tree (CRT) models classify children with clinical pneumonia based on radiological findings.