Assessment of the accuracy of diagnostic tests in patients with suspected acute appendicitis using ROC analysis

Acute appendicitis happens in 7% of people during their lives, but there is no consensus about criteria for emergency appendectomy. TI1is study was carried out to determine the appropriate combination of diagnostic tests and variables with high accuracy.

All case records of patients attending the Imam teaching hospital for abdominal tenderness and suspected appendicitis during year 2001 to 2004 were reviewed. We collected data on all tests and post operating reports. Data were analyzed performing logistic regression analysis and Receiver Operating Characteristic (ROC) curve.

A total of 577 cases were studied. Considering Area under the Curve (AUC) the best variables combination were percent of poly morpho nuclear (PMN %), sex and vomiting. This combination had a high accuracy for correct decision for appendectomy in suspected patients with abdominal tenderness (AUC = 77.3 Cl 95% = 70-84 %). Optimal cut-off point for white blood cell {WBC) count, PMN % and degree of fever, were 7.5 (1000 /mic. L), 65% and 37.7°" (oral temperature). Even with determining the best combination of conventional appendicitis diagnostic tests, there was not a good agreement with post operation pathology results (Kappa index= 61%, SD= 15%).

To reduce the false positive diagnosis (inappropriate selection for appendectomy) and false negative (discharged patients who need emergency appendectomy), we recommend, at least in high-risk groups (Children, young women and senile patients), more diagnostic tests be performed.