Comparison of modeled sampling strategies for estimation of dairy herd lameness prevalence and cow-level variables associated with lameness

Abstract

Monitoring herd lameness prevalence has utility for dairy producers and veterinarians in their efforts to reduce lameness, for animal welfare assessment programs, and for researchers. Locomotion scoring is a method used to quantify lameness and calculate prevalence. Because of the time necessary to locomotion score each cow in large dairy herds, a sampling strategy to determine herd lameness prevalence that allows scoring of fewer cows would be useful. Such a sampling strategy must be validated for accuracy compared with the lameness prevalence when all cows in a herd are locomotion scored. The purpose of this study was to assess 3 previously suggested methods of estimating lameness prevalence by strategic sampling of dairy herds. Sampling strategies tested included (1) sampling a calculated number of cows in the middle third of the milking parlor exit order for each pen, (2) sampling a calculated number of cows weighted across pens and distributed evenly within each pen, and (3) sampling all cows in the high production, low production, and hospital pens. Lactating cows on 5 dairy farms in Washington and Oregon (n=4,422) were locomotion scored using a 5-point scale to determine herd-level lameness prevalence (percentage with locomotion score ≥3). Milking parlor exit order, order in headlocks at the feed bunk within each pen, and breed were recorded for each cow. The number of days in lactation, milk production, and parity were collected from farm computer records. Pen grouping strategy for each farm was obtained by interview with farm management. Sampling strategies were modeled using the locomotion score data set for each herd. Estimates of lameness prevalence obtained from the milking parlor exit order sample and the sample distributed across pens were within 5 percentage points of the whole herd prevalence. The third strategy estimated the lameness prevalence within 5 percentage points on 4 farms, but overestimated prevalence on 1 farm. Pen-level prevalence obtained by locomotion score of all cows in the pen was variable and not reliably predictive of herd-level prevalence. Cows of Holstein breed, parity >1, and exiting the milking parlor in the last 20% of the pen had greater odds of lameness compared with other breeds, parities, and milking parlor exit order groups in a multivariate analysis. This study indicates that the sampling strategies using the middle of milking parlor exit order and a calculated sample distributed across the herd may be used to obtain an estimate of herd lameness prevalence.

Effect of merging levels of locomotion scores for dairy cows on intra- and interrater reliability and agreement

“…Differences from other studies for intrarater reliability and agreement may be explained by the different levels of experience of the raters participating in different experiments (March et al, 2007;Gibbons et al, 2012) and the practical conditions in which the locomotion scoring was performed: scoring from video in the present study versus live scoring in other studies (O'Callaghan et al, 2003;Thomsen et al, 2008). Results obtained in the present experiment were similar to others reported in the literature for interrater reliability (Thomsen et al, 2008;Hoffman et al, 2013) and interrater agreement (Winckler and Willen, 2001;Katsoulos and Christodoulopoulos, 2009) for similar 5-level locomotion scores. Merging levels had no effect on the overall intra-and interrater reliability for most combinations, with some exceptions for the 3-level scale combination 1(234)5 and 2-level scale combinations 1(2345) and (1234)5, which presented κw and κ estimates lower than the acceptance threshold.…”

“…One factor is related to the description of the lameness status of cows. In the literature, the 2-level combination (12)(345) is the most used to classify cow as nonlame (levels 1 and 2) and lame (levels 3, 4, and 5; Winckler and Willen, 2001;Katsoulos and Christodoulopoulos, 2009;Hoffman et al, 2013). The 2-level combination (123)(45) is also commonly used to classify cows as lame (levels 4 and 5; Bicalho et al, 2007a,b;Ito et al, 2010).…”

Manual and automatic locomotion scoring systems in dairy cows: A review

“…In most MLSSs, the threshold to classify a cow as lame was when the locomotion score exceeded the middle level of the scale (e.g. locomotion score ≥ 3 in five-level scales) (Winckler and Willen, 2001;Channon et al, 2009;Chapinal et al, 2009;Hoffman et al, 2013). An alternative approach to classify a cow as lame was when two of the five gait and posture traits scored ≥ 3 on a five-level scale (O'Callaghan et al, 2003).…”

Sequential sampling: a novel method in farm animal welfare assessment

“…Based on this association, an alternative to random sampling has been proposed by Main et al (2010) who suggest locomotion scoring up to 100 cows from the middle of the milking order. This scheme has since been found to be able to provide accurate herd level prevalence when also applied to large herds with multiple groups (Hoffman et al, 2013). In general, though, the question of how a random sample may be achieved on farm can to some extent be considered to be a logistical one, and is common to most sampling schemes.…”