Test, test, test for COVID-19 antibodies: the importance of sensitivity, specificity and predictive powers

N Kumleben,R Bhopal,T Czypionka,L Gruer,R Kock,J Stebbing,F.L Stigler

doi:10.1016/j.puhe.2020.06.006

Abstract

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) antibody tests of varying specificity and sensitivity are now available. For informing individuals whether they have had coronavirus disease 2019 (COVID-19), they need to be very accurate. For measuring population prevalence of past infection, the numbers of false positives and negatives need to be roughly equal.With a series of worked examples for a notional population of 100,000 people, we show that even test systems with a high specificity can yield a large number of false positive results, especially where the population prevalence is low. For example, at a true population prevalence of 5%, using a test with 99% sensitivity and specificity, 16% of positive results will be false and thus 950 people will be incorrectly informed they have had the infection. Further confirmatory testing may be needed.Giving false reassurance on which personal or societal decisions might be based could be harmful for individuals, undermine public confidence and foster further outbreaks.

Highlights

To help reverse the current lockdowns while suppressing COVID-19 rates, we need to identify who currently has the infection and who has had it and recovered
As RT-PCR testing to detect current infection has been recently discussed in detail,[1] we focus in this paper on antibody tests
We show here how to measure the test’s accuracy and how this changes along with the prevalence of disease (12 tables showing the results with varying sensitivity, specificity and population prevalence of 1%, 5%, 10%, and 20% are available in the Supplementary File)

Summary

Introduction

To help reverse the current lockdowns while suppressing COVID-19 rates, we need to identify who currently has the infection and who has had it and recovered. We show here how to measure the test’s accuracy and how this changes along with the prevalence of disease (12 tables showing the results with varying sensitivity, specificity and population prevalence of 1%, 5%, 10%, and 20% are available in the Supplementary File). Sensitivity (A/A+C) is the proportion of people with a disease who, when tested, receive a positive test result.

Results

Conclusion