Standardising new approaches for validation of diagnostic assays and estimation of measurement uncertainty in veterinary diagnostic testing

Download Full Version MS Word format [243kb]

Axel Colling

Summary

Diagnostic tests are important for population based veterinary medicine including research, surveillance, certification of freedom from disease, prevalence estimation, epidemiologic studies, risk management and modelling of infectious diseases. Little attention has been given to the post-analytical phase of the diagnostic process e.g. data management, analysis and interpretation. The purpose of the diagnostic testing and the structure and availability of the data determine appropriate methods for data analysis.

It is recognized that epidemiological considerations play an important role in diagnostic testing. A positive predictive value can be seen as a function of prevalence for example, an error is often made in assuming that an assay with 99 per cent specificity and sensitivity will generate one false positive or one false negative for every 100 animals in a population.  However, the disease prevalence directly affects the interpretation of positive test results. For example, if the infection status of the population is one per 1000 animals, then for every 1000 tests on that population, there are likely to be 11 positive results. Of these results, ten may be false positive and one a true positive.  Hence, only around nine per cent of the positive test results will accurately predict the infection status of the animal. This means the positive test results from this population will be incorrect 91 per cent of the time (Fig 1).