A framework for adaptive predictive values using individual symptoms of screening test users

Abstract Aim Screening tests are widely used for disease detection, but test users often struggle to interpret negative or positive test results correctly, in particular in the presence of symptoms. This study proposes a framework for adaptive predictive values (APV) that personalises test interpretation by incorporating individual symptoms. Subject and methods The APV framework integrates individual symptom data into the calculation of predictive values, modifying the individual’s prior probability of being diseased. The discriminatory power of different symptoms can be determined via Bayes Factors. The framework is illustrated by a web application (ShinyApp) for the estimation of predictive values for SARS-CoV-2 infection based on the presence of typical symptoms, by reusing the symptom profile data from the REACT-1 study, which allows users to adjust their test result interpretation. Results By incorporating individual symptoms, the APV framework personalises the predictive values of screening tests. The higher the discriminatory power of a symptom, the more the individualised risk estimations differ from standard values based on population-wide prior risks. The ShinyApp demonstrates how users can input their test result, test type, residence region, and recent symptoms to obtain a personalised interpretation of their infection risk. Conclusion The APV framework enhances the interpretation of screening test results by integrating personalised symptom data, improving risk estimation for individual test users. The approach can be extended beyond symptoms to include other individual characteristics.