Observer agreement on pen level prevalence of clinical signs in finishing pigs.

Four observers performed a standard clinical examination of finisher pigs in two commercial finisher herds. In herd 1,600 finisher pigs in 44 pens were examined. The observers assessed clinical signs of lameness, umbilical hernia and tail bite according to a standardized procedure. The prevalence of the clinical signs was estimated at the pen level. The procedure was repeated after 3 months in another herd, where 730 finisher pigs in 69 pens were examined. The agreement between observer pairs was assessed using prevalence-adjusted bias-adjusted kappa (PABAK) and proportionate-agreement estimates (Ppo) and Pneg). Observer bias was present despite training and standardization of the participating observers. The highest pen level agreement for the observer pairs was found for pens that had one or more pigs with tail bite (PABAK = 0.82-1.00) and umbilical hernia (PABAK = 0.77-1.00). The agreement was fair-to-moderate for identification of pens holding one or more lame pigs (PABAK = 0.27-0.7). In general, the average agreement of observer pairs on absence of clinical signs (Pneg) was higher than for presence (Ppos). The observer bias varied between observer pairs and with the clinical signs.