Back

Diagnosing COVID-19 infection: the danger of over-reliance on positive test results

Cohen, A. N.; Kessel, B.; Milgroom, M. G.

2020-09-28 epidemiology
10.1101/2020.04.26.20080911 medRxiv
Show abstract

Unlike previous epidemics, in addressing COVID-19 nearly all international health organizations and national health ministries have treated a single positive result from a PCR-based test as confirmation of infection, even in asymptomatic persons without any history of exposure. This is based on a widespread belief that positive results in these tests are highly reliable. However, data on PCR-based tests for similar viruses show that PCR-based testing produces enough false positive results to make positive results highly unreliable over a broad range of real-world scenarios. This has clinical and case management implications, and affects an array of epidemiological statistics, including the asymptomatic ratio, prevalence, and hospitalization and death rates. Steps should be taken to raise awareness of false positives, reduce their frequency, and mitigate their effects. In the interim, positive results in asymptomatic individuals that havent been confirmed by a second test should be considered suspect. Key messagesThe high specificities (usually 100%) reported in PCR-based tests for SARS-CoV-2 infection do not represent the real-world use of these tests, where contamination and human error produce significant rates of false positives. Widespread misunderstanding of these false positive rates affects an array of clinical, case management and health policy decisions. Similarly, health authorities guidance on interpreting test results is often wrong. Steps should be taken immediately to reduce the frequency and impacts of false positive results.

Matching journals

The top 11 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.