A few points about the 2 common diagnostic #tests & #COVID19
The nasal-swab test which looks for viral particles & an acute infection
The antibody test for a past-COVID19 infection

We need to consider three things to determine how useful a 'diagnostic test' is for a person

(I stress a person and not population).

1/ The sensitivity of the test (ability of a test to pick up true positives e.g. the person actually has or had COVID19 and also has a positive test result).

2/ The specificity of the test (ability of the test to pick up true negatives).
3/ The prevalence of the disease in the community.

All three impact what a 'positive result' means for a person.

Let's consider these two diagnostic tests:

The test for viral particles with a nasal swab (COVID-19 RT-PCR) test:
(sens = 70%; spec = 95%)

The new serology (antibody) test announced by Roche:
(sens = 100%; spec = 99.8%)

a) Viral swab test:

See the attached images. The sensitivity & specificity of the test is known (70% & 95%). The three models below, assume a prevalence of 20% and 50% and 90%.
- 20% prevalence (or 20% likelihood of acute COVID19 based on symptoms): A positive test would mean

the likelihood of COVID19 is 78% (=post-test likelihood of disease). The counter-point to this- with a 20% background prevalence- is that 22% of people who 'test positive' do not have COVID19. That means that testing a lower prevalence group with a nasal swab- will falsely tell

over 1:5 people that they tested positive (when they don't have COVID19)- that could impact how they behave & expose themselves in the coming weeks/ months.
- 30% prevalence model: A positive test would mean the likelihood of COVID19 is 93% (=post-test likelihood of disease). The

counter-point to this- with a 50% background prevalence- is that 7% of people who 'test positive' do not have COVID19.
- 90% prevalence model: A positive test would mean the likelihood of COVID19 is 99% (=post-test likelihood of disease).

a)Serology (antibody test):
See the attached images. The sensitivity and specificity of the test is possibly known (100% & 99.8%). The two models below, assume a prevalence of 5% and 20%.

- 5% prevalence model: A positive test would mean the likelihood of past COVID19 is 96% (=post-test likelihood of disease). That means the 1:20 people will be told they had a past COVID19 infection, when they did not.
- 20% prevalence model: The positive predictive value is 99%

Like all tests, the importance of a 'positive' finding varies by baseline prevalence and no test is 100% accurate.Phrases like '100% accurate' don't convey what's behind the test. Read the wonderful @drjessicawatson below for more information &numbers: https://www.bmj.com/content/369/bmj.m1808