TY - JOUR
T1 - Liver fluke in beef cattle – Impact on production efficiency and associated greenhouse gas emissions estimated using causal inference methods
AU - Jonsson, N. N.
AU - MacLeod, M.
AU - Hayward, A.
AU - McNeilly, T.
AU - Ferguson, K. D.
AU - Skuce, P. J.
PY - 2022/3
Y1 - 2022/3
N2 - We aimed to estimate 1) the marginal effect of liver fluke (Fasciola hepatica) infection on productivity of Scottish beef cattle, and 2) the associated greenhouse gas emissions intensity (GHG EI). Data comprised 240,065 abattoir records from NE Scotland from 2014 to 2017, including the presence or absence of lesions typical of liver fluke in the liver at the time of slaughter, from which we inferred liver fluke infection status. The retrospective analysis of abattoir records to estimate marginal effects of an exposure is complicated by the multi-dimensional, clustered nature of the datasets, which result in confounding of potential causal factors with the exposure. Causal inference methods are required to identify and correct for variation in background exposure. We constructed directed acyclic graphs (DAGs) of observed variables, including the potential confounders, breed, sex, breeder, finisher, season of birth and year of birth. We then applied inverse probability weighting (IPW) to adjust for variation among exposure risk and applied a doubly robust generalized linear model (DRGLM) to the weighted observations to estimate the marginal effect of fluke on the growth rate of animals and total days from birth until slaughter. We compared these estimates with the results of linear mixed effects (LME) models with the same variables, treating breeder and producer as random effects. To estimate GHG EI, we applied IPCC tier-2 type GHG calculations to the marginal effects estimated from IPW with DRGLM. The IPW with DRGLM model estimated that animals with active fluke lesions (adult fluke seen on postmortem inspection) gained 17 (95 % CI 12–22) g/d less saleable beef than animals with no lesions and no visible fluke. Animals with active fluke lesions were 11 (95 % CI 6.5–15) d older at slaughter weight than animals with no lesions. Animals with historic lesions in which there was scarring of the liver but in which no adult fluke were seen showed a wide variation in effect estimates, consistent with some misclassification. The effect estimates from LME models suggested slightly lower effects of fluke on growth rate and days to slaughter but with overlapping 95 % confidence intervals. Calculation of the associated GHG emissions suggest the EI of meat from a herd with no fluke is approximately 1.5 % lower than the same herd with fluke. Sustainably controlling liver fluke would have additional production benefits not included in this estimate and could therefore have a much greater impact on GHG EI in practice than demonstrated here.
AB - We aimed to estimate 1) the marginal effect of liver fluke (Fasciola hepatica) infection on productivity of Scottish beef cattle, and 2) the associated greenhouse gas emissions intensity (GHG EI). Data comprised 240,065 abattoir records from NE Scotland from 2014 to 2017, including the presence or absence of lesions typical of liver fluke in the liver at the time of slaughter, from which we inferred liver fluke infection status. The retrospective analysis of abattoir records to estimate marginal effects of an exposure is complicated by the multi-dimensional, clustered nature of the datasets, which result in confounding of potential causal factors with the exposure. Causal inference methods are required to identify and correct for variation in background exposure. We constructed directed acyclic graphs (DAGs) of observed variables, including the potential confounders, breed, sex, breeder, finisher, season of birth and year of birth. We then applied inverse probability weighting (IPW) to adjust for variation among exposure risk and applied a doubly robust generalized linear model (DRGLM) to the weighted observations to estimate the marginal effect of fluke on the growth rate of animals and total days from birth until slaughter. We compared these estimates with the results of linear mixed effects (LME) models with the same variables, treating breeder and producer as random effects. To estimate GHG EI, we applied IPCC tier-2 type GHG calculations to the marginal effects estimated from IPW with DRGLM. The IPW with DRGLM model estimated that animals with active fluke lesions (adult fluke seen on postmortem inspection) gained 17 (95 % CI 12–22) g/d less saleable beef than animals with no lesions and no visible fluke. Animals with active fluke lesions were 11 (95 % CI 6.5–15) d older at slaughter weight than animals with no lesions. Animals with historic lesions in which there was scarring of the liver but in which no adult fluke were seen showed a wide variation in effect estimates, consistent with some misclassification. The effect estimates from LME models suggested slightly lower effects of fluke on growth rate and days to slaughter but with overlapping 95 % confidence intervals. Calculation of the associated GHG emissions suggest the EI of meat from a herd with no fluke is approximately 1.5 % lower than the same herd with fluke. Sustainably controlling liver fluke would have additional production benefits not included in this estimate and could therefore have a much greater impact on GHG EI in practice than demonstrated here.
KW - Abattoir
KW - Cattle
KW - Greenhouse gas emissions
KW - Liver fluke
KW - Production impact
UR - http://www.scopus.com/inward/record.url?scp=85123082399&partnerID=8YFLogxK
U2 - 10.1016/j.prevetmed.2022.105579
DO - 10.1016/j.prevetmed.2022.105579
M3 - Article
C2 - 35066320
AN - SCOPUS:85123082399
SN - 0167-5877
VL - 200
JO - Preventive Veterinary Medicine
JF - Preventive Veterinary Medicine
M1 - 105579
ER -