Appendiceal diameter as a predictor of appendicitis in children: improved diagnosis with three diagnostic categories derived from a logistic predictive model

Abstract

Objectives: To develop and assess the performance of a diameter-based logistic predictive model and a derived 3-category interpretive scheme for the sonographic diagnosis of paediatric appendicitis.

Methods: Appendiceal diameters were extracted from reports of ultrasound examinations in children and young adults. Data were used to generate a logistic predictive model which was used to define negative, equivocal and positive interpretive categories. Diagnostic performance of the derived 3-category interpretive scheme was compared with simulated binary interpretive schemes.

Results: Six hundred forty-one appendix ultrasound reports were reviewed with appendicitis present in 181 (28.2 %). Cut-off diameters based on the logistic predictive model were ≤6 mm = normal, >6 mm-8 mm = equivocal and >8 mm = positive with appendicitis present in 2.6 % (11/428), 64.9 % (72/111) and 96.1 % (98/102) of cases in each group. These cut-offs conferred 97.2 % accuracy with 17.3 % (111/641) of cases considered equivocal. Of the binary cut-offs, a 6 mm cut-off performed best with 91.6 % accuracy. AIC analysis favoured the logistic model over the binary model for prediction of appendicitis.

Conclusions: A 3-category interpretive scheme based on a logistic predictive model provides higher accuracy in the diagnosis of appendicitis than traditional binary diameter cut-offs. Inclusion of an equivocal interpretive category more accurately reflects the probability distribution of prediction of appendicitis by ultrasound.

Key points: • Three diameter categories outperform a 6-mm cut-off to diagnose appendicitis • Three categories allow more confident exclusion of appendicitis • Three categories allow more confident diagnosis of appendicitis • Three categories more accurately reflect the probability of appendicitis by ultrasound.