Efficacy of statistical process control procedures to identify deviations in continuously measured physiological and behavioral variables in beef heifers resulting from an experimentally combined viral–bacterial challenge

Authors:
Open Access
Publisher website
Google Scholar
Abstract
The objective of this experiment was to determine if statistical process control (SPC) procedures coupled with remote continuous data collection could accurately differentiate between animals experimentally inoculated with a viral-bacterial (VB) challenge or phosphate buffer solution (PBS). Crossbred heifers (N = 38; BW = 230 ± 16.4 kg) were randomly assigned to treatments by initial weight, ADG, BHV-1 and MH serum titers. Feeding behavior, DMI, animal activity and rumen temperature were continuously monitored remotely prior to and following VB challenge. VB-challenged heifers exhibited decreased (P < 0.01) ADG and DMI, as well as increased (P < 0.01) neutrophils and rumen temperature consistent with a bovine respiratory disease (BRD) infection. However, none of the heifers displayed overt clinical signs of disease. Shewhart and cumulative summation (CUSUM) charts were evaluated, with sensitivity and specificity computed on the VB-challenged heifers (n = 19), and PBS-challenged heifers (n = 19) respectively, accuracy was determined as the average of sensitivity and specificity. To address the diurnal nature of rumen temperature responses, summary statistics (Mean, minimum, maximum) were computed for daily quartiles (6-h intervals), and these quartile temperature models were evaluated separately. In the Shewhart analysis, DMI was the most accurate (95%) at deciphering between PBS- or VB-challenged heifers, followed by rumen temperature (94%) collected in the 2 nd and 3 rd quartiles. Rest was most the accurate accelerometer-based traits (89%), and meal duration (87%) and bunk visit (BV) frequency (82%) were the most accurate feeding behavior traits. Rumen temperature collected in the 3 rd quartile signaled the earliest (2.5 d) of all the variables monitored with the Shewhart, followed by BV frequency (2.8 d), meal duration (2.8 d), DMI (3.0 d) and rest (4.0 d). Rumen temperature and DMI and remained the most accurate variables in the CUSUM at 80 and 79%, respectively. Meal duration (58%), BV frequency (71%) and rest (74%) were less accurate when monitored with the CUSUM analysis. Furthermore, signal day was greater for DMI, rumen temperature and meal duration (4.4, 5.0 and 3.7 d, respectively) in the CUSUM compared to Shewhart analysis. These results indicate that Shewhart and CUSUM charts can effectively identify deviations in feeding behavior, activity and rumen temperature patterns for the purpose of detecting sub-clinical BRD in beef cattle.

This publication has 43 references indexed in Scilit: