Latency and chronicity of paratuberculosis makes diagnosis challenging even in adult cattle. Often, "optimisation" of sensitivity of diagnostic tests for chronic diseases is done by selection of a population containing many late-stage animals. Actually, random selection of animals with subsequent correction through various population characteristics must be considered epidemiologically correct. The purpose of this study was to estimate sensitivity and specificity of faecal culture (FC) and ELISA for specific animal groups through stratification by cow characteristics and alternative disease definitions. First, the sensitivities and specificities of an ELISA and faecal culture were estimated using a no-gold standard method and the overall accuracy of the ELISA was described. Second, the probability of being ELISA-positive was determined for cows grouped by parity and stage of lactation. Information was then combined to increase the interpretability of the accuracy of both faecal culture and ELISA. The probability of ELISA-positivity varied significantly by parity and stage of lactation. In general, the probability of being ELISA positive was 2-4 times higher in second and higher lactation than in first lactation. For cows with low levels of antibodies, the sensitivity of FC was 0.20-0.30 in first parity cows. In higher parity cows the sensitivity of FC was 0.45-0.55. For cows with high levels of antibodies, the sensitivity of FC increased to >0.80. The likelihood of providing a correct diagnosis was on average 0.89 using an uncorrected ELISA-value. Correction by parity increased the likelihood to 0.92 for first parity cows and decreased the likelihood to 0.86 for second and higher parity cows. Thus, the probability of being ELISA-positive is higher in higher parities, but the ELISA response is apparently purer in the first parity. It is concluded that inclusion of cow and population characteristics can improve the interpretation of diagnostic tests.