Bayesian estimation of diagnostic accuracy of somatic cell counts history and on-farm milk culture using Petrifilm® to identify quarters or cows that should be treated with antimicrobials in selective treatment protocols at dry off

Abstract

Bayesian latent class models were used to estimate the test accuracy (sensitivity (Se), specificity (Sp), and predictive values (NPV and PPV)) of cow-level somatic cell counts (SCC) data, quarter-level Petrifilm® on-farm milk culture, and quarter-level standard milk bacteriology for the identification of quarters that should possibly be treated with antimicrobials at dry off in dairy cows. Data of 282 cows from 9 dairy herds in Québec, Canada, with bulk tank SCC < 250,000 cells/mL were used. Estimated median herd-prevalence of infections that should be treated was 16.2 % (95 % credibility interval (CI): 11.0–22.7). Se and Sp estimates for quarter-milk culture using Petrifilm® were 82.2 % (95 %CI: 74.0–89.5) and 62.0 % (95 %CI: 58.6–65.6), respectively. Se and Sp for quarter-milk standard bacteriology were 67.4 % (95 %CI: 55.8–81.2) and 79.6 % (95 %CI: 76.4–83.0), respectively. Se and Sp of different SCC scenarios and thresholds were estimated. For first parity cows, using only the last Dairy Herd Improvement (DHI) test SCC with a threshold of 100,000 cells/mL appeared quite accurate, with Se, Sp, PPV, NPV and reduction of antimicrobial usage of 85.6 % (95 %CI: 69.6–95.6), 86.0 % (95 %CI: 80.0–91.7), 58.0 % (95 %CI: 42.3–74.2), 96.4 % (95 %CI: 91.3–99.0), and 75.3 % (95 %CI: 70.7–79.3), respectively. For cows of ≥ 2nd parity, using only the last DHI test SCC with a threshold of 200,000 cells/mL resulted in Se, Sp, PPV, NPV and reduction of antimicrobial usage of 75.3 % (95 %CI: 55.8–87.3), 84.0 % (95 %CI: 78.8–89.3), 47.2 % (95 %CI: 32.0–63.7), 94.7 % (95 %CI: 89.0–97.6), and of 77.0 % (95 %CI: 73.3–80.3), respectively. Adding quarter-level milk culture using Petrifilm® to cows identified as unhealthy using cow-level SCC data improved the test accuracy (mainly the PPV) and further reduced the use of antimicrobials. For instance, in ≥ 2nd parity cows, using only the last DHI SCC with a threshold of 200,000 cells/mL, adding a subsequent Petrifilm® test increased the reduction from 77.0 % (95 %CI: 73.3–80.3) to 89.5 % (95 %CI: 86.7–91.8). Considering the availability of SCC data, the easiness of using just the last DHI test, and the high NPV that could be achieved, producers may consider using just the last DHI test as a potential tool to identify cows that should be treated with antimicrobials at dry off. It may be used alone or in combination with quarter-level on-farm Petrifilm® milk culture on high SCC cows to further reduce the use of antimicrobials by identifying quarters that need to be treated.