As more and more people are exposed to COVID-19 and effective vaccines are made available online, the prevalence of anti-SARS-CoV-2 antibodies in the population will increase, making positive individual test results more reliable. However, only molecular (PCR) or antigenic tests, as well as the patient`s risk of exposure and symptoms, should be used to identify active COVID-19 infections.4 However, a negative Sofia SARS-2 FIA antigen result does not exclude active COVID-19 infection. If patients have positive exposure and/or symptoms, confirm negative Results of Sofia SARS-2 FIA antigen with a molecular follow-up test.17 Fig. 5 simulates a screening test in a low prevalence setting where truth negatives in the field are significantly more common than ground truth positives. An example of such a scenario can be found in cervical cancer screening using Pap smear cytology, where significant rates of false positives (cell abnormalities of unknown importance) can be expected and positive test results do not necessarily give high confidence in the presence of high-level disease [28,29]. In general, a known uncertainty corresponds to a precisely expected rate of misclassification. However, for diagnosis in a particular cohort of patients, an observer does not know for sure which patients were misdiagnosed or even how many were misclassified. For example, if a test is used for binary classification and we know that a negative call for a test is 95% accurate, then for each patient classified as negative, there is a 5% chance that the patient is actually positive. The random nature of the uncertainty means that for every 100 patients in a study who were rated negative by the test, we expect 5% to be misclassified, but it may be that in the study, 10 are misclassified or none (although both alternatives are relatively unlikely). The misclassification rate of a study can be estimated in several ways, para. B example by testing samples repeatedly, comparing with another test with probably higher accuracy, or assuming an expected error rate from other sources of information. Additional simulations have shown that it is extremely unlikely that even a perfect test will achieve very high performance in a diagnostic evaluation study, even if there is little uncertainty in the comparison device against which the test is evaluated.
As shown, for example, by S7 Supporting Information (“Very High Performance Tests”), a modest classification error rate of 5% in the comparative test leads to rejection of the perfect diagnostic test with a probability of more than 99.999% if 99% OF PPA (sensitivity) or NPA (specificity) is required in a diagnostic evaluation study. Thus, a specific numerical requirement for test performance, in particular a very high performance requirement such as 99% PPA (sensitivity), can only be meaningfully discussed if all classification uncertainties in a study are excluded or characterized and the measured test performance is interpreted in relation to the theoretical limits of comparative uncertainty. Abbreviations: ROC, operating characteristics of the receiver; ASC, area below the ROC curve; CI, confidence interval; LRTI, lower respiratory tract infection; NAP, negative percentage of match; net present value, negative predictive value; PPP, positive percentage of match; PPV, positive predictive value; ROC, operating characteristics of the receiver; RPD, retrospective diagnosis of the physician; SIRS, Systemic Inflammatory Response Syndrome Negative Percent Agreement (NPA): The percentage of negative comparison calls that are labeled negative by the test to be evaluated. This value is calculated in the same way as the specificity. However, the NPA is used instead of specificity to recognize the fact that, due to the uncertain comparison, this measure should not be interpreted in such a way as to accurately reflect the measure that specificity presupposes. .