Body condition score accuracy and repeatability from evaluation of cull sow digital images at a midwestern harvest facility

Abstract

The objective of this study was to assess accuracy, repeatability, and reproducibility when evaluating cull sow body condition score (BCS) using a digital medium. Selected digital images were sourced from recorded images of sows brought to a Midwest abattoir. Digital images were collected on 2 separate occasions. Each sample, grouped by capture date, represented a unique scoring session. Scorers (n = 6) with experience at assessing sow body condition used images to assign a BCS to each sow using a 7-point BCS scale. Using applied values, scores were adjusted to the standard 5-point scale. Because scorers assessed individual sows’ body condition from recorded video images, the mode score (BCSMode) was calculated for each sow and considered the gold standard. Scoring distributions, mode scores, individual bias, percent agreement with BCSMode, percent interobserver agreement, Spearman correlations evaluating scorer agreement, repeatability, and reproducibility were calculated. Scorer bias from the pooled cull sow groups (n = 386 total available video images from 2 separate collections) ranged from −0.25 (±0.5) to 0.51 (±0.9). Spearman correlation coefficients for cull sow BCS measured on all sows for all scores during the second scoring round were lower than those observed in the first round. Additionally, it was observed that repeatability estimates improved from round 1 and round 2 (Rd 1 = 0.74 and Rd 2 = 0.76), and reproducibility slightly decreased between round 1 and round 2 (Rd 1 = 0.52 and Rd 2 = 0.47). These repeatability and reproducibility changes demonstrate that as experience level increases, scorers begin to develop their interpretation of the scale used to assess body condition. In turn, they become more repeatable within themselves but may differ from other scorers. The ability for scorers to accurately identify low-BCS sows could serve as a cumulative lifetime welfare indicator where harvest facilities could provide valuable BCS feedback on individual sow and group average basis. The accuracy, repeatability, and reproducibility reported in this study suggest that digital images are an effective medium to assess cull sow BCS.