Accuracy of clinical variables in the identification of radiographically proven constipation in children

Study objective: To determine whether clinical variables accurately identify children with radiographically proven constipation. Methods: Prospective, cross sectional case series of children 2-12 years of age with abdominal pain (AP) requiring radiographic evaluation. Constipation was defined radiographically as the presence of fecal material throughout the colon. The presence of other pathology was noted. The pediatric emergency department (ED) physicians recorded a comprehensive history and physical examination and a provisional diagnosis was made. Radiographs were initially interpreted by the pediatric ED attending physicians; the official interpretation was later provided by a single board certified pediatric radiologist who was blinded to the ED interpretation. A discriminant analysis was performed to identify variables that could best discriminate between patients with, and without, radiographically proven constipation. Results: In total 251 patients were enrolled over a 12 month period. Four variables were noted to be more common in constipated patients: a history of normal or hard stools, absence of rebound tenderness, presence of tenderness in the left lower quadrant and stool in the rectal vault on exam. Stool present on rectal exam was the best discriminator between patients with and without constipation. The discriminant analysis model had a sensitivity of 77%, specificity of 35% and a negative predictive value of 55%. Conclusion: No clinical variable, either as a single variable or in a model, accurately identified patients with abdominal pain and radiographically proven constipation.