UNSTRUCTURED
In the recent pandemic of COVID-19, an almost-daily practice of medical diagnostic tests was introduced to the public. In spite of its urgency and necessity to detect the infected, it has also caused avoidable distress and lack of rightful precaution in case of false results, due to its binary outcome presentation or equivocal approach to assess the risk of an individual by a regional prevalence. A better way to interpret medical test results and educate the lay population is clearly warranted. We propose a concept to personalise test outcomes using a hierarchical Bayesian model to integrate personal information of test users into the calculation of the predictive value of diagnostic tests. The purpose of the model is to obtain a realistic and personalised estimation of risk. In a case study, we show an effective improvement in predictive value in SARS-CoV-2 diagnostic tests by integrating symptom information from the test users. The necessity of data on the symptom distribution in the COVID-negative sample is highlighted. We created a risk calculator for public use to help interpret individual SARS-CoV-2 diagnostic tests. We outline the advantage of this model and the outlook in its application, and call for regular surveillance to prepare its deployment.