In swine production, safeguarding animal health is crucial, and combating diseases like porcine reproductive and respiratory syndrome (PRRS) virus is a central part of this effort. Vaccination remains the primary defense strategy, but determining the optimal dosage for the day-to-day realities of swine production is complex.
To help determine this, we collected closeout data and health-specific information from nursery and grow-finish sites in the Midwest that had experienced lateral PRRS virus breaks over a three-year period. Next, we used a visual mapping technique known as Directed Acyclic Graph (DAG) to help us see how various factors like the vaccine dose, vaccination type, timing of vaccination, other diseases and farm characteristics link to the health outcomes we cared about, such as mortality rates, cull rates, grade A and medicine costs. By laying out these connections, the DAG helped us spot the relationships between the variables, potential confounders and treatment outcome pathways and identify biases that affect our results.
With the insights from the DAG, we moved to perform Propensity Score Matching (PSM) for both the nursery and grow-finish datasets. This is a statistical causal inference technique that helped us balance groups on confounding factors to make them comparable except for the vaccine dose they received.
After we matched pigs based on their propensity scores, we applied linear mixed-effects models to assess differences between vaccination strategies (full vs. half-dose), incorporating the weights from the PSM and considering factors such as season, year, vaccine type, vaccination timing, farm status, other diseases and PRRS virus strain, with a random intercept being the flow. To see how the vaccine dose was related to pig health outcomes such as mortality, culling, grade-A pigs and medicine costs.