Accuracy of a discriminant analysis model for prediction of coliform mastitis in dairy cows and a comparison with clinical prediction.

We tested an equation, which had been developed previously using discriminant analysis, for predicting whether a cow has coliform mastitis. Variables indicating a high probability of coliform infection included history of previous mastitis in the affected quarter, weakness, clear or white color of milk, water consistency of the milk, swelling of the udder, lack of previous mastitis in other quarters, lack of palpable udder abscesses, and a high body temperature. Application of this predictive equation to 114 cows with mastitis to determine if they would have coliform organisms cultured from the affected quarters resulted in an accuracy of 71% (sensitivity = 0.42, specificity = 0.85), compared to an accuracy of 62% (sensitivity = .64, specificity = .61) for cowside prediction by the attending clinicians. Changing the cutoff score of the discriminant rule so that the sensitivity of the discriminant prediction was similar to that of the clinicians yielded an accuracy of 64% (sensitivity = .64, specificity = .64).