Serologic testing for porcine parvovirus was used to illustrate the difficulties of interpreting results of an imperfect test performed on a sample from a herd. Herd seroprevalence and true prevalence levels were classified as low (< 40%), medium (40-79.9%) or high(greater than or equal to 80%). Results obtained from simulating herd-level testing, assuming various test sensitivities and specificities and number of animals tested, often deviated from what would be expected based on individual-level test results. Herd-level test results were evaluated as herd-level predictive values (HPV) and herd-level sensitivities (HSE) for each of the three seroprevalence and true prevalence categories, respectively. HPV responded differently, as test sensitivity and specificity and number of animals tested changed, depending on the herd seroprevalence category. An increase in sensitivity resulted in an increased HPV (low) and HPV (medium), but a reduced HPV (high). An increase in specificity resulted in an increased HPV (medium) and HPV (high), but a decreased HPV (low). As the number of animals tested increased, the HPV (low) and HPV (high) increased, while HPV (medium) only had a minor or inapparent change, especially when the sensitivity was less than or equal to 90%. HSE also responded differently to changes in test sensitivity and specificity and number of animals tested, depending on the herd true prevalence category. As sensitivity increased, HSE (low) and HSE (medium) decreased, while HSE (high) increased. On the other hand, as specificity increased, HSE (low) and HSE (medium) increased, while HSE (high) decreased. Increasing the number of animals tested resulted in increases in HSE (low) and HSE (medium), but a decrease in HSE (high), especially when the test sensitivity was less than or equal to 90%. Findings from this study illustrate the difficulties associated with extrapolating individual animal test results to the herd level. Also, it emphasizes the importance oi knowing the herd-level seroprevalence and true prevalence in the population prior to estimating herd-level predictive values and sensitivities, respectively.
