Hierarchical and spatial analyses of pneumonia-lesion prevalence at slaughter in New Zealand lambs

We recorded lesions of moderate-to-severe pneumonia (≥10% lung surface area affected; “pneumonia”) in 1,899,556 lambs submitted to three New Zealand abbatoirs between December 2000 and September 2001. The average prevalence of pneumonia ranged between 7 and 13%, by abbatoir. We ran a two-level mixed-effects binomial logistic-regression model with the prevalence of pneumonia as the outcome, and adjusting for abbatoir and month. The intracluster correlations for batch (slaughter lambs from the same farm sent at the same time) and farm were 31.3 and 12.4%, respectively. (We also noted threefold differences in odds across abbatoirs, and >30-fold differences among slaughter months.) Case flocks (those in the upper quintile of pneumonia prevalence) generally were not clustered in the spatial incidence-risk analysis (after adjusting for flock-level and batch level effects, and as compared to flocks in the lower two quintiles). We therefore concluded that the risk of moderate-to-severe pneumonia-lesion prevalence detectable at slaughter of lambs was determined at the flock and batch level, rather than at the spatial level.