Simulating interferon-γ assay testing patterns in Irish bovine tuberculosis breakdowns

Abstract Since 2015, national policy has evolved to substantially increase the use of the interferon-γ assay to diagnose bovine tuberculosis in Irish cattle. To help evaluate the effectiveness of this strategy, we developed and validated a predictive model of interferon-γ test implementation. We describe testing patterns between 2005 and 2022 (n = 349,072 interferon-γ tested cattle in 6422 breakdowns) and have built predictive models based on 88,199 interferon-γ tested cattle in 1407 breakdowns ending between 2015 and 2020. Using logistic regression, we modelled the probability that a herd-level interferon-γ test event would follow an eligible herd-level comparative intradermal tuberculin (skin) test. For herd-level interferon-γ test events, we used zero-truncated negative binomial regression to model the count of cattle tested. The final multivariable models performed better than null models in predicting testing practices. The probability of eligible herd-level skin tests being followed by interferon-γ testing varied spatially. Testing increased in breeding herds, with increasing year of breakdown end and with skin test case count. It decreased with increasing herd size and count of previous interferon-γ tests in the same breakdown. Count of cattle sampled within interferon-γ herd-level test events varied with herd type (dairy, mixed, beef-breeding, store and fattener in descending order) and spatially. It increased with year of breakdown end, herd size and skin case count. We demonstrate that prediction of veterinary decision making can be enhanced using our hurdle model with routinely collected data. This can be incorporated into operational mechanistic models and enhances understanding of programme implementation.