Sensitivity, specificity and predictive probability values of serum agglutination test titres for the diagnosis of Salmonella Dublin culture-positive bovine abortion and stillbirth

The objective of this study was to determine the diagnostic value of maternal serology for the diagnosis of Salmonella Dublin bovine abortion and stillbirth. A retrospective, unmatched, case–control study was carried out using twenty year's data (1989–2009) from bovine foetal submissions to an Irish government veterinary laboratory. Cases (n = 214) were defined as submissions with a S. Dublin culture-positive foetus from a S. Dublin unvaccinated dam where results of maternal S. Dublin serology were available. Controls (n = 415) were defined as submissions where an alternative diagnosis other than S. Dublin was made in a foetus from an S. Dublin unvaccinated dam where the results of maternal S. Dublin serology were available. A logistic regression model was fitted to the data: the dichotomous dependent variable was the S. Dublin foetal culture result, and the independent variables were the maternal serum agglutination test (SAT) titre results. Salmonella serology correctly classified 87% of S. Dublin culture-positive foetuses at a predicted probability threshold of 0.44 (cut-off at which sensitivity and specificity are at a maximum, J = 0.67). The sensitivity of the SAT at the same threshold was 73.8% (95% CI: 67.4%–79.5%), and the specificity was 93.2% (95% CI: 90.3%–95.4%). The positive and negative predictive values were 84.9% (95% CI: 79.3%–88.6%) and 87.3% (95% CI: 83.5%–91.3%), respectively. This study illustrates that the use of predicted probability values, rather than the traditional arbitrary breakpoints of negative, inconclusive and positive, increases the diagnostic value of the maternal SAT. Veterinary laboratory diagnosticians and veterinary practitioners can recover from the test results, information previously categorized, particularly from those results declared to be inconclusive.