Measuring the impact of sow farm outbreaks with PRRS virus on the downstream mortality using causal inference methods.

Abstract

Porcine reproductive and respiratory syndrome virus (PRRSV) remains a significant challenge to the swine industry, resulting in substantial productivity and, consequently, economic losses. This study aimed to quantify the impact of PRRSV outbreaks in sow farms on nursery mortality using causal inference methods. The study design followed a retrospective observational approach, where PRRSV epidemic status in source sow farms was the exposure, and nursery mortality (percentage of dead pigs in the first 60 days post-weaning) was the outcome. Causal inference techniques were employed to estimate the effect of the exposure (PRRSV epidemic status) on the outcome (nursery mortality). Data from a Midwestern US swine production system comprising 2,592 lots of pigs, representing approximately 5 million pigs marketed between January 2021 and December 2022, were analyzed. A causal diagram was constructed to visualize the relationship between PRRSV epidemic exposure and nursery mortality, while controlling for potential confounding factors including season, average parity at farrow, and sow farm Mycoplasma hyopneumoniae status. Four analytical approaches were employed: univariate and multivariable regression models, propensity score matching, and a doubly robust method. The results indicated that PRRSV epidemic lots had higher nursery mortality compared to non-epidemic lots, regardless of the modeling approach used. The doubly robust method provided the most accurate estimates, offering lower mortality differences and narrower confidence intervals. This study demonstrated the application of causal inference methods on swine data to measure the impact of PRRSV on swine nursery mortality, which is an approach commonly used in other epidemiology areas but not well explored in veterinary epidemiology. The findings highlight the importance of employing causal inference models in veterinary epidemiology to improve the accuracy of disease impact assessments in field conditions, with potential applications in studying other pathogens or disease-related factors in livestock production.