A multivariate analysis of childhood abdominal pain in Trinidad

This is a multivariate analysis of the data recorded in assessing 1158 consecutive admissions presenting to a children's surgical ward with acute abdominal pain. There were 56 binary variables available for entry into the analysis. A statistical software package was used to perform a stepwise discriminant analysis on the data. The program selected 18 variables as having discriminating power in assigning patients to the six diagnostic groups. In order of discriminating power these were, mainly, a positive urine culture, the bowel history, the findings on rectal examination, the location of abdominal tenderness, the presence of a mass, and the white cell count. Lesser discriminating potential was assigned to the presence of dehydration; fluid levels on erect abdominal films, a rise in temperature, an increased pulse rate, the presence of urinary symptoms, and the general appearance of the child. Use of these data led to an overall correct classification of 80.7% of cases. It is concluded that these variables should be included in the assessment of children with acute abdominal pain.