Thirty-two physical examination and laboratory variables were recorded
during examination of 165 horses admitted for acute abdominal disease
. Univariate analyses were performed to determine which of the variabl
es were significantly different between horses that lived or died. Ste
pwise logistic regression was performed to identify variables with the
best predictive value. Four variables (heart rate, peritoneal fluid t
otal protein concentration, blood lactate concentration, and abnormal
mucous membrane) remained significant when entered into the model. His
tograms for each significant variable were used to set ''cutting-point
s,'' establishing categories that were made into a table of assigned v
alues from which a Colic Severity Score (CSS) for each horse was calcu
lated. Seventy-one horses in a second group were used to validate the
scoring chart. Case mortality rate was similar in both groups (20.6% i
n development group versus 21.1% in validation group). All horses with
a CSS > 7 died, whereas 75% of those with a score of less than or equ
al to 7 lived. For the validation group, use of the scoring table yiel
ded a positive predictive value of 100%, negative predictive value of
91.8%, sensitivity of 66.7%, and specificity of 100%. The overall accu
racy of the CSS was 93%. The CSS is a rapid and accurate method for pr
edicting survival in cases of equine acute abdominal disease. (C)Copyr
ight 1995 by The American College of Veterinary Surgeons