Predictability by somatic cell counts related to prevalence of intrammary infection within herds

Somatic cells were counted and bacteria identified for milk samples from 719 lactating dairy cows in 12 commercial herds. These pooled data were used to look at the accuracy of alternative thresholds of somatic cell counts as indicators of intramammary infection. Sensitivity, specificity, and predictability positive and negative at alternative cell count thresholds were calculated. There was an increase of incidence of infection associated with increased cell count. Raising the cell count threshold increased predictability positive and specificity and decreased sensitivity and predictability negative. Also, the effect of prevalence of infection on predictabilities was examined for the pooled data on sensitivity and specificity at a cell count threshold of 400,000 cells/ml. As prevalence of infection increased, predictability of a positive test result also increased, but predictability negative decreased. Differences in accuracy of somatic cell count as a predictor of infection at varying prevalences of infection suggest that test results should be interpreted for individual herds.