Animal Waste, Water Quality and Human Health

96Animal Waste, Water Quality and Human Healthamounts of Salmonella per day. Furthermore, probabilistic models forcatchment-scale pathogen loading have been shown to be highly sensitive toestimates of shedding intensity (Dorner et al. 2004), underscoring theirimportance in load calculations.Accuracy of estimates of pathogen prevalence in livestock populations dependon sample size, test sensitivity and specificity, prevalence and intra-herdcorrelation of infection status. The wide variation in diagnostic test sensitivityused to detect these pathogens can result in widely different prevalence estimatesfor the same infected population. For example, the addition of immunomagneticbead separation prior to immunofluorescent microscopy (IMS-DFA) for detectinglow levels of Cryptosporidium parvum oocysts in bovine faeces generates a 50%probability of detection (DT 50 ) at 1.0 oocyst/g faeces and 90% probability ofdetection (DT 90 ) at 3.2 oocysts/g faeces (Figure 3.3; Pereira et al. 1999). Incontrast, immunofluorescent microscopy alone (DFA) generates a DT 50 at 200oocyst/g faeces and DT 90 at 630 oocysts/g faeces. Using DFA alone to surveyadult animals shedding low levels of oocysts will likely yield a downwardlybiasedestimate of the shedding prevalence. For example, a prevalence rate of7.1% compared to 0.6% (∼12-fold difference) of Cryptosporidium spp. sheddingwas estimated using IMS-DFA compared to DFA in adult cattle (Atwill et al.1998, 2003). However, using highly sensitive assays will detect more lowshedders in the population, increasing the positive prevalence rate but potentiallyreducing the calculated mean intensity of pathogen shedding per gram of faeces(Atwill et al. 2003). Quantitative microbial assays that are used to estimateshedding intensity need to be adjusted for the percent recovery of the assay,which is well established in the protozoan literature but less commonly so in thebacterial literature. The use of the geometric mean compared to the arithmeticmean for pathogen intensity data underestimates the actual pathogen loaddeposited in the environment; the downward bias of the geometric mean becomesincreasingly pronounced as the frequency distribution for intensity becomes moreright-skewed (e.g., protozoan parasite eggs in juvenile animals, with a smallpercentage of individuals shedding very high levels).Finally, the occurrence of many high priority waterborne zoonotic pathogens inlivestock populations is highly seasonal, due to seasonal patterns of parturitionor to unknown factors such as summer shedding of E. coli O157:H7 in beefcattle. Cryptosporidium spp. infection often increases after calving, lambing,foaling, and kidding occur. This suggests that accurate measures of prevalenceand intensity need to sample across the year and include the various age classesof livestock, with clear tabulation of each age class matched to pathogenoccurrence rather than generating a single estimate of prevalence rate andintensity for the entire livestock population (e.g., pooling data from neonates,