Analysis of the predictive value of clinical and sonographic variables in children with suspected acute appendicitis using decision tree algorithms: Decision tree analysis paediatric appendicitis

Introduction We evaluated the predictive value of sonographic and clinical variables in the diagnosis of acute appendicitis in children using decision tree modelling. Methods Data pooled from two prior studies in the same population of children referred for ultrasound examination of suspected acute appendicitis. Ultrasound and clinical variables were collated and compared with patient records. Decision tree algorithms were used to model data to identify highly discriminatory variables. Receiver operative characteristic (ROC) curve analyses were performed on different models of appendiceal diameter criteria. Results There were 687 examinations included. Diameter modelling indicated that categorical assessment – below 6 mm as negative, diameters between 6 and 8 mm equivocal for appendicitis and above 8 mm positive – was more accurate (AUROC = 0.921) than the most accurate binary cut‐off (7 mm, AUROC = 0.886). Decision tree analysis supported categorical diameter criteria, demonstrated that the presence of echogenic mesentery was an important variable and showed that common blood test results can be complementary discriminators of ultrasound findings. Discussion The use of a binary appendiceal diameter cut‐off was less accurate than a three‐category model. Absence of peri‐appendiceal mesentery inflammation is an important negative predictor of appendicitis in children, even without direct visualisation of the appendix.