Effective disease surveillance relies on accurate pathogen testing and robust prevalence estimates. Diagnostic specificity (DSp), the probability that an uninfected animal tests negative, is high when false positives are low. Diagnostic sensitivity (DSe) is the probability an infected animal tests positive; higher DSe means fewer false negatives. However, sensitivity and false negatives are harder to estimate without a "gold standard", an assay that can detect between 90 - 100% of true positive infections. Occupancy estimation of infection prevalence offers one solution by allowing for imperfect detection of the pathogen. Testing potentially infected tissues multiple times allows for the use of a Bayesian multistate occupancy model to estimate the probability of pathogen infection in tissues [Formula: see text] and detection probabilities [Formula: see text] for different assays. Using [Formula: see text] and [Formula: see text] from the posterior distribution, the conditional probability of detecting the pathogen can be modeled, allowing for the calculation of DSe. Renibacterium salmoninarum is a bacterial pathogen causing bacterial kidney disease among salmonid species and was the model pathogen we used to train our model. The current testing standard for salmonids combines initial screening for antibodies using direct fluorescent antibody test (DFAT) with polymerase chain reaction (PCR) confirmation to detect R. salmoninarum. However, detection of R. salmoninarum still varies between species, tissues, and assays. Here, a multi-state occupancy model was used to estimate detection probability among individual and dual kidney/liver infections with DFAT and qPCR in fish with an unknown infection status. Both assays produced false negatives, but qPCR had fewer than DFAT and a higher DSe. Infection state was often misclassified, but multiple surveys per individual or combining tissues for testing improved DSe for both assays.