Detecting intramammary infection at the end of lactation in dairy cows.

To reduce antimicrobial use, infusion of antimicrobials into only infected cows at the end of lactation (selective dry cow therapy) is preferable to infusion of every cow with antimicrobials. Use of selective dry cow antimicrobial therapy requires differentiation of probably infected from uninfected cows to enable treatment allocation. Milk somatic cell count (SCC) has been used to distinguish between cows with and without intramammary infection (IMI). However, SCC may be influenced by milk yield, stage of lactation, breed, and herd-level variables such as prevalence of infection. Cut points for SCC, to distinguish between cows with and without an IMI, may need to differ between cow age groups and breeds, or among herds. This study evaluated associations between SCC and major pathogen IMI in one or more quarters of 2,606 cows from 36 herds in 4 regions of New Zealand. In the last week of lactation, cows selected at random had milk samples collected from each quarter, and the teat-end condition and hygiene of the udder were scored. Herd- and cow-level data including age, breed, milk volume, and SCC at each production were recorded, and bulk tank milk SCC and volume of milk shipped were collated. At cow level, the association between average, maximum, and last cow-composite SCC, and presence of a major pathogen IMI in one or more quarters of cows, was examined using receiver operator curves. Predictive logistic regression models were then developed that included potential effect modifiers such as age, milk yield, and bulk tank milk SCC. The population average prevalence of major pathogen IMI was 7.2% of cows (95% confidence interval = 5.9-8.6), and this varied significantly between herds. The average, maximum, and last cow-composite SCC of lactation were all predictive of presence of a major pathogen IMI and did not differ in their ability to discriminate infected from uninfected cows. However, the optimal cut points for the last SCC, the maximum SCC, and average SCC were 108, 152, and 105 × 1,000 cells/mL, respectively. Inclusion of age, bulk tank SCC, and history of clinical mastitis improved overall model fit. However, inclusion of these variables did not improve the discriminatory power of maximum cow-composite SCC used alone. We conclude that cow-composite SCC on its own resulted in sensitivities and specificities of between 0.76 and 0.86, and 0.71 to 0.80, respectively, for determination of presence of major pattern IMI, and the predictive value was not improved by addition of other predictor variables.