Estimating Degree of Mastitis from Time-Series Measurements in Milk: A Test of a Model Based on Lactate Dehydrogenase Measurements

An example of an individual cow mastitis risk (MR) profile (○; right y-axis) relative to weeks in milk. Peaks in MR (e.g., at wk 5.5, 10.5, 12.3, and 12.8) are generated by rapid increases in the model indicator, smoothed lactate dehydrogenase (LDH) amount (μmol/ min; left y-axis), as shown by the solid line with no symbols. The raw, unsmoothed, values of the inline indicator LDH amount are shown as solid circles. This cow received veterinary treatment for mastitis on DIM 73 and 90 (indicated by arrow,▾) but not on DIM 39 (wk 5.5) when both MR and SCC showed a peak. The profile of SCC is shown on the left y-axis (+, count × 10⁻⁵); milk yield (▵;, kg/d) is also shown on the left y-axis.

Mastitis risk (MR) relative to days from veterinary treatment of mastitic cows (○) and healthy controls (+), where health definitions are based solely on the presence or absence of veterinary treatments. Panel A shows median values and panel B shows the 75 and 25% quartiles of the distribution of MR values. Mastitis risk was generated by a model based on changes in the level and slope of milk lactate dehydrogenase amount.

Mastitis risk (MR) relative to days from veterinary treatment of true mastitic (○) and true healthy controls (+), where health definitions are based on the presence or absence of veterinary treatments supplemented by changes in SCC (true healthy controls had no veterinary treatment and no peak in SCC; true mastitic cases had a veterinary treatment and a peak in SCC). Panel A shows median values and panel B shows the 75 and 25% quartiles of the distribution of MR values. Mastitis risk was generated by a model based on changes in the level and slope of milk lactate dehydrogenase amount.

The average difference in mastitis risk profile between true mastitic and true healthy cows. Days on which the difference was significant are indicated by 1, 2, or 3 vertically aligned asterisks (*) for P<0.05, P<0.01, and P<0.001, respectively. Mastitis risk was generated by a model based on changes in the level and slope of milk lactate dehydrogenase amount. True healthy cows had no veterinary treatment and no peak in SCC; true mastitic cows had a veterinary treatment and a peak in SCC.

The median values of mastitis risk (MR) time-aligned according to peak in mastitis risk rather than time of veterinary treatment (mastitic cows: —○—; healthy cows: —+—).The likelihood of the median MR value of the mastitic cows belonging to the healthy population is shown by the dashed line (on the same scale as MR). Mastitis risk was generated by a model based on changes in the level and slope of milk lactate dehydrogenase amount. True healthy cows had no veterinary treatment and no peak in SCC; true mastitic cows had a veterinary treatment and a peak in SCC.

Time profiles of A) SCC, B) milk yield, C) indicator-based risk, and D) additional risk factor for true mastitic (—○—) and true healthy (—+—) cows. Profiles are time-aligned to the mastitis risk peak. The indicator-based risk was generated from changes in the level and slope of milk lactate dehydrogenase amount. The additional risk factor was generated from other animal- and herd-related factors available on-farm (DIM, udder characteristics, milking duration, and disease history). True healthy cows had no veterinary treatment and no peak in SCC; true mastitic cows had a veterinary treatment and a peak in SCC.

The residual mastitis risk (MR) relative to DIM. The residuals were derived from a regression of MR risk on SCC. The horizontal lines indicate residual MR levels of 0.7 and –0.3. Mastitis risk was generated by a model based on changes in the level and slope of milk lactate dehydrogenase amount.

A schematic illustration of the inappropriateness of specificity/sensitivity calculations when the condition being measured is on a continuous scale. The shaded area denotes the observed relationship between true degree of infection and the degree of infection estimated by a given detection system; in this example, the error of estimation is constant over the range of degree of infection. Using a classification threshold of 0.7 for specificity/sensitivity calculations, estimates can be classified as true negatives (TN), false negatives (FN), true positives (TP), and false positives (FP). As the true degree of infection approaches the threshold, the proportion of estimates judged to be false increases dramatically. This does not reflect the fact that the error of estimation is constant over the full range of true degree of infection. The arrows denote the size of the error necessary for a given estimate to be misclassified (as a false positive); this clearly depends upon the true degree of infection.