LAUSR

Body temperature (BT) is widely used to evaluate health and heat load status in cattle. Despite its importance, studies vary in how BT is measured and in the biological interpretation of the results. Costs, practicality, labor, and welfare concerns can affect how BT is measured, including frequency of measurement and the type of device used. Inaccurate BT outcomes may have implications for cattle welfare; for example, animals may only receive treatment when fever is identified. Our objectives were (1) to compare measurement of vaginal temperature (VT) using relatively small, inexpensive, and low-accuracy loggers (±0.5 to ±1°C, iButton range; Embedded Data Systems, Lawrenceburg, KY) to a high-accuracy logger (±0.1°C; StarOddi, Gardabaer, Iceland), and (2) to evaluate how different BT sampling strategies correspond to 24-h VT in lactating dairy cows. To address the first objective, VT data from 54 cows were recorded every 45 min for 12 d/cow, on average, using 2 different types of temperature loggers (StarOddi DST centi-T and iButton DS1921H or DS1922L) attached to a shortened, hormone-free controlled internal drug release insert. Average VT obtained from both loggers were compared using mixed models and regression analyses. In addition, we tested the consistency of the low-accuracy loggers in detecting cows with elevated BT using the kappa coefficient of concordance. To address the second objective, VT data from 20 cows were recorded every min for 9 to 11 d/cow using StarOddi loggers. Using these data, we estimated average VT using 11 sampling strategies (every 5, 10, 15, 30, 45, 60, and 120 min, 1×/d recorded in the morning or afternoon, 2×/d, or 3×/d). Estimates and observed means were compared using linear regression. Compared with StarOddi loggers, the iButtons either underestimated (H model: 38.7 vs. 38.0 ± 0.06°C) or overestimated VT (L model: 38.7 vs. 39.2 ± 0.04°C). When considering elevated or fever VT thresholds, iButtons did not correctly classify animals; kappa coefficients of concordance were ≤0.35. Measuring VT as often as every 120 min resulted in more accurate estimates compared with strategies that recorded it once to thrice per day. These results indicate that the type of device (i.e., data logger) and sampling strategies affect BT outcomes and that these decisions affect the interpretation of BT data.