OBJECTIVE: We aimed to predict the risk of complicated appendicitis in children, constructing a risk-based prediction tool with the optimal combination of sensitivity and specificity outcomes. PATIENTS AND METHODS: This is a prospective study on a random sample of children with acute appendicitis who underwent appendectomy. Clinical examination, history, routine laboratory tests, Alvarado and pediatric appendicitis scores, operative and histopathological findings were taken into consideration. The predictive ability of the outcome variables was assessed by the Receiver Operating Characteristics (ROC) analysis. The overall predictive ability and determination of the best cut-off value (the higher sum of sensitivity plus specificity) were calculated. A Classification and Regression Tree (CRT) was used to create a multi-level classification algorithm. The model was set to predict the outcome of complicated appendicitis, considering as potential predictors the demographic characteristics, the clinical findings, and the outcome parameters. RESULTS: The various combinations of clinical and laboratory parameters did not improve their overall diagnostic ability. However, the CRT analysis resulted in a short classification algorithm based on the Pediatric appendicitis score, neutrophils percentage and the CRP. This model yielded a significantly better predictive ability than all the other combinations of the outcome parameters. The application of the model would predict complicated appendicitis with 90% sensitivity and 78.6% specificity. CONCLUSIONS: The constructed predictive model may be a useful tool for daily practical use by the clinician, especially in areas where modern diagnostic imaging facilities are absent or not always available. Clinical evaluation and close follow-up remain the more accurate preoperative method to decide the performance and timing of appendectomy.
OBJECTIVE: We aimed to predict the risk of complicated appendicitis in children, constructing a risk-based prediction tool with the optimal combination of sensitivity and specificity outcomes. PATIENTS AND METHODS: This is a prospective study on a random sample of children with acute appendicitis who underwent appendectomy. Clinical examination, history, routine laboratory tests, Alvarado and pediatric appendicitis scores, operative and histopathological findings were taken into consideration. The predictive ability of the outcome variables was assessed by the Receiver Operating Characteristics (ROC) analysis. The overall predictive ability and determination of the best cut-off value (the higher sum of sensitivity plus specificity) were calculated. A Classification and Regression Tree (CRT) was used to create a multi-level classification algorithm. The model was set to predict the outcome of complicated appendicitis, considering as potential predictors the demographic characteristics, the clinical findings, and the outcome parameters. RESULTS: The various combinations of clinical and laboratory parameters did not improve their overall diagnostic ability. However, the CRT analysis resulted in a short classification algorithm based on the Pediatric appendicitis score, neutrophils percentage and the CRP. This model yielded a significantly better predictive ability than all the other combinations of the outcome parameters. The application of the model would predict complicated appendicitis with 90% sensitivity and 78.6% specificity. CONCLUSIONS: The constructed predictive model may be a useful tool for daily practical use by the clinician, especially in areas where modern diagnostic imaging facilities are absent or not always available. Clinical evaluation and close follow-up remain the more accurate preoperative method to decide the performance and timing of appendectomy.