A dichotomous variable was created to indicate the presence or absence of an increase
in cows’ RT. A total of 6 different temperature ranges
above a cow’s individual baseline were examined, with
0.8°C being chosen as an appropriate deviation to indicate a fever when considering specificity and sensitivity.
Temperature deviations less than 0.8°C produced an
excess of false positives, with those above 0.8°C producing an excess of false negatives (Table 1). When a
deviation of 0.8°C (1.4°F) above baseline was observed,
a “temperature increase alert” was recorded as 1 (yes),
otherwise a 0 was recorded indicating no temperature
increase. Increases in RT were evaluated for 2 different timeframes: first the authors assessed associations
between core body temperature increases within 7 d
before the health event (and for 7 d among control
cows), and second for a timeframe of 4 d preceding the
health event (and 4 d among control cows). These timeframes were chosen to ensure that an adequate amount
of time elapsed before disease diagnosis to identify any
temperature deviations from baseline. For example,
although 86% of mastitis cases had an increased RT
within 2 d before diagnosis, only 65% of the pneumonia
cases had an increased RT during this period, with the
remaining 35% of pneumonia cases having an increased
RT on d 3 or 4 before diagnosis. For this reason, the 4-d
period preceding a health event was chosen as a more
acceptable timeframe for inclusion.
Data generated by the TSRB in cows enrolled in the
study and gathered by the TempTrack software were
not available to the farm personnel during the first 60
DIM. All cows with a health event were identified by
farm personnel as showing clinical signs corresponding
to the diseases examined in the study. As farm personnel had no knowledge of temperature readings recorded
by the TSRB, RT were not used to identify cows with
health events.