LAUSR

The objective of this study evaluated the accuracy and repeatability when determining cull sow body condition scores (BCS) evaluated by scorers utilizing digital images. Participants (n=6) were selected based on previous BCS scoring on live sows. Group standards were established utilizing the scores from two participants with extensive experience BCS sows. Other scorers were not provided training before they scored images and no “scorer calibration” activity occurred to ensure that each scorer was observing similar traits when applying BCS. Two separate groups of sows were scored and video images for each sow were collected and stored for evaluation. The images were recorded as a convenience sample from a cooperating sow abattoir. The cull sow video images were assigned a BCS using an 8-point scale (1,1+,2-,2,3,4,5,NS). Scores from the lower portion of the scale were classified (+ and -) to provide a more specific BCS difference evaluation among sows that were thin. When scorers evaluated the sows (n=165) from the first group, they scored the BC for each sow. Additionally, a random group of sows (n=40) images were selected to be scored a second time by the same scorer. Repeatability (inter- and intra- scorer), individual bias, group bias, and group deviation from standard’s mean BCS were calculated. Data from the first sow group showed that relative to the standard’s mean scores, participants overestimated BCS by .41 (P=0.0001). A similarly selected second set of cull sow images obtained from the same abattoir was recorded. The second group included more sow images (n=220) and a greater number images that were scored a second time(n=55) in order to calculate the repeatability values. The same measures described for group 1 were calculated. After data analysis from the second sow group, it was again noted that participants overestimated BCS by 0.1 (P=0.0009).