Classification tree for identifying ineffective breathing pattern in children with acute respiratory infection

Abstract

The objective of the study was to verify defining characteristics with greater predictive power to aid in the classification of ineffective breathing pattern using classification trees in children with acute respiratory infections. A cross-sectional study was carried out in two pediatric hospitals with 249 children with acute respiratory infection. For data collection, a specific instrument developed for the study was used. Three induction algorithms were used to generate the trees: Chi-square Automatic Interaction Detection, Classification and Regression Trees, and Quick, Unbiased, Efficient Statistical Tree. Three trees were constructed to aid in the identification of ineffective breathing pattern. The classification trees generated present probabilities conditional to the occurrence of the diagnosis associated with dyspnea and changes in respiratory depth. Ineffective breathing pattern was present in 65.5% of the sample. Thus, the probability of occurrence of this diagnosis in children with acute respiratory infection was 100% with the presence of dyspnea and changes in respiratory depth.